f
<
&
-------
ACKNOWLEDGMENTS
The staff of the Office of Ground Water and Drinking Water (OGWDW) of the U.S.
Environmental Protection Agency (EPA) would like to thank the following groups and individuals
for their assistance in completing this report. Work on this report was directed by Rebecca K. Allen
and Patricia S. Hall of OGWDW. The final report was prepared for EPA by Industrial Economics,
Incorporated (IEc); Lisa Robinson managed the project with assistance from James Neumann,
Michel Woodard Ohly, Maura Flight and other IEc staff.
The EPA staff would like the thank the peer reviewers who commented on earlier versions
of this report and contributed significantly to its development. Dr. Thomas DeLeire of the
University of Chicago, Dr. Winston Harrington of Resources for the Future, and Dr. Scott Ramsey
of the Fred Hutchinson Cancer Research Center commented on a March 2001 draft of the report
prepared by International Consultants Incorporated. Dr. A. Myrick Freeman III of Bowdoin
College, Dr. James K. Hammitt of Harvard University, and Dr. W. Douglass Shaw of the University
of Nevada commented on a revised version of this report prepared in October 2002 by IEc. In
addition, several EPA staff provided comments on drafts of the report and substantially influenced
its contents, including John B. Bennett, Joel Corona, John Powers, Mahesh Podar, and Daniel
Schmelling of EPA's Office of Water, and Chris Dockins, Robin Jenkins, and Bill O'Neil of EPA's
National Center for Environmental Economics.
11
-------
TABLE OF CONTENTS
ACKNOWLEDGMENTS	i
EXECUTIVE SUMMARY	iv
1.0 INTRODUCTION	1-1
1.1	Valuation Framework 	1-2
1.1.1	Fatal Risk Valuation 	1-3
1.1.2	Nonfatal Risk Valuation 	1-4
1.1.3	Relationship of Willingness to Pay to Cost of Illness	1-5
1.2	Application to Regulatory Analysis	1-8
1.2.1	Data Needs for Regulatory Analysis	1-8
1.2.2	Data Provided by the Cost of Illness Literature	1-9
2.0 THEORETICAL AND EMPIRICAL APPROACHES	2-1
2.1	Methods for Valuing Lost Productivity	2-1
2.1.1	Human Capital Method 	2-2
2.1.2	Friction Cost Method	2-3
2.1.3	Relationship of Productivity Losses to Social Welfare Losses	2-5
2.2	Methods for Valuing Social Welfare Losses	2-7
2.2.1	Willingness to Pay for Risk Reductions 	2-8
2.2.2	Willingness to Pay for Transportation
and Recreational Opportunities	2-9
2.3	Implications for Valuing Illness-Related Time Losses	2-13
2.3.1	Time as a Commodity and a Resource 	2-14
2.3.2	The Labor-Leisure Trade Off	2-16
2.3.3	Conclusions 	2-18
3.0 CALCULATING THE VALUE OF LOST TIME	3-1
3.1	Valuing Types of Time Use	3-1
3.1.1	Market Work Time 	3-3
3.1.2	Nonmarket Work Time 	3-5
3.1.3	Leisure Time 	3-7
3.1.4	Sleep Time	3-9
3.2	Valuing Time for Individuals Not Engaged in Paid Work 	3-10
3.2.1	Children	3-10
3.2.2	Elderly, Unemployed, or Out of the Labor Force 	3-12
in
-------
TABLE OF CONTENTS
(continued)
3.3. Determining the Extent of Time Losses 	3-13
3.3.1	Types of Time Losses 	3-14
3.3.2	Acute vs. Chronic Illness	3-16
3.3.3	Caregivers 	3-18
3.4	Example of Calculations 	3-19
3.5	Limitations and Assessment of Uncertainty 	3-25
REFERENCES	Ref-1
Appendix A: EXAMPLES FROM THE EMPIRICAL LITERATURE	 A-l
Appendix B: DATA SOURCES	B-l
B.l Time Loss Data 	B-l
B.2 Employment and Wage Data	B-5
B.3 Survival Statistics 	B-10
iv
-------
EXECUTIVE SUMMARY
In the 1996 amendments to the Safe Drinking Water Act, Congress increased the role of
benefit-cost analysis in determining appropriate standards for drinking water contaminants. As a
result, the staff of the Office of Ground Water and Drinking Water in the U.S. Environmental
Protection Agency (EPA) have undertaken a number of projects to improve the approaches used in
these analyses. This report focuses on one aspect of the assessment of human health-related
benefits: valuing time losses associated with illness. These time losses result when ill individuals
cannot engage in their normal work and leisure activities or when participation in these activities
is less productive or less enjoyable than usual.
Valuation of time losses is useful in those benefit-cost analyses where EPA relies on cost of
illness estimates to value risk reductions for nonfatal illnesses. The preferred approach for valuing
these benefits is to determine individual willingness to pay for the risk reductions of concern. For
fatal risks, EPA has a well-established framework for implementing the willingness to pay approach.
For nonfatal risks, EPA often lacks empirical estimates of willingness to pay for the specific health
risks of concern. In such cases, analysts may at times rely on estimates of the medical costs of
illness. However, these estimates may understate the value of risk reductions because they exclude
consideration of many attributes of the illness — such as avoiding the associated activity restrictions
as well as related pain and suffering.
In these cases, analysts may wish to add an estimate of the value of time losses to the
estimates of the medical costs of illness to come closer to a complete accounting of the effects of
the health risks of concern. While accounting for time losses does not fully capture the value
associated with pain and suffering, it increases the extent to which activity restrictions are included
in the benefits estimates. Traditionally, cost of illness studies have used the human capital approach
to estimate the indirect costs of illness. The human capital approach focuses on lost productivity,
including both time lost from paid work and, in some cases, from unpaid work (such as housework
and volunteer activities). However, this approach excludes consideration of the effects of illness on
nonwork activities and hence is an incomplete measure of the social welfare costs of activity
restrictions.
To address this deficiency, EPA has developed an approach for valuing time losses that more
broadly considers the impact of illness on all types of activities. This approach is based on
consideration of the opportunity costs of time, consistent with the framework underlying the use
of willingness to pay estimates to value health-related benefits. For paid work, this approach
assumes that total compensation (wages and benefits) is equal to the employers' valuation of the
workers output. Hence if a worker is absent due to illness, society loses this output and the time
losses can be valued using compensation data. For time spent on nonmarket work and leisure
activities, this approach assumes that an individual will engage in such unpaid activities only if, at
the margin, the value of these activities is greater than the wage that he or she can earn if employed.
Thus after-tax wages can be used to estimate the value of nonwork time.
When used in benefit-cost analysis, this approach has several limitations that can be
v
-------
addressed in a variety of ways. In particular, this report suggests that analysts may wish to assign
a "zero" value to losses associated with sleep time, due to problems related to estimating the impact
of illness on sleep and to determining the dollar value of this time. In addition, valuing time losses
is difficult for segments of the population that do not engage in paid work, such as children, retirees,
and others out of the labor force. Finally, imperfections in the labor market may also mean that
wage and benefit rates over- or understate the value of time losses. At minimum, analysts will need
to discuss the impacts of these limitations qualitatively; related uncertainties also can be addressed
by sensitivity analysis or probabilistic assessment.
vi
-------
1.0 INTRODUCTION
In 1996, Congress amended the Safe Drinking Water Act (SDWA) to increase the
consideration of benefit-cost analysis in regulatory decision-making. Prior to the amendments, the
U.S. Environmental Protection Agency (EPA) was required to establish Maximum Contaminant
Levels (MCLs) for drinking water contaminants at the lowest feasible level, regardless of the
relative benefits and costs associated with achieving these levels. Under the 1996 amendments, the
EPA Administrator may, at his or her discretion, establish less stringent MCLs if the costs of
achieving the lowest feasible level exceed its benefits. Hence developing improved estimates of
regulatory benefits, including the value of reducing risks to human health, has become a priority for
EPA's Office of Ground Water and Drinking Water.1 This report addresses one aspect of the
valuation of risks to human health: determining the value of time losses associated with nonfatal
illnesses.
Health risk reductions have many attributes that individuals value. For example, decreasing
health risks diminishes the expenditures on disease prevention and treatment; the need to divert time
from normal activities to bed rest, doctor's visits, or hospitalization; and the experience of pain and
suffering. In general, economists prefer to use measures of value that address the combined effects
of all of these attributes; i.e., that estimate the total value of the decrease in the risk of incurring a
particular health effect. Such an approach avoids problems related to sorting out the contributions
of different attributes and potential double-counting. However, this holistic approach is not always
possible in benefit-cost analysis, given the available empirical research, and at times analysts may
find it necessary to separately assess different components of the value of health risk reductions.
The preferred methodology for valuing human health-related benefits is to apply estimates
of individual willingness to pay for risk reductions. EPA has well-established methods for
implementing this methodology for fatal risks. For nonfatal risks, EPA often lacks empirical
estimates of willingness to pay for the specific health effects of concern. In these cases, analysts
may at times build values by estimating the medical costs of illness from the health economics
literature and adding the value of lost time based on the approach discussed in this report. Such an
approach may understate total willingness to pay to avoid the health risk of concern because it may
not fully address the value of avoided pain and suffering as well as other attributes of the risk
reduction.
The types of time losses discussed in this report are sometimes referred to as indirect costs
or lost productivity. However, this report takes a more comprehensive approach than most of the
indirect cost literature. That literature focuses on the effects of time usage on the production of
goods and services, whereas this report is focused more broadly on the welfare effects of time use —
considering the value of leisure as well as work time to the individual and society. This report
'Drinking water regulations may result in a variety of other benefits, including aesthetic
improvements (e.g., better water taste, odor, or color), reduced materials damages (e.g., decreased pipe
corrosion), or ecological improvements (e.g., due to source water protection measures).
1-1
-------
generally uses the term "time losses" rather than "indirect costs" or "lost productivity" to emphasize
this different focus.
1.1 VALUATION FRAMEWORK
The practice of benefits assessment at EPA is based on the theory of welfare economics. The
following sections briefly describe EPA's approach to valuation to provide context for this report,
and discuss the relationship of cost of illness estimates to this framework.2
When determining the value of benefits such as those resulting from drinking water
regulations, welfare economists begin with the assumption that individuals derive utility (or a sense
of satisfaction or well-being) from the goods and services they consume. Individuals can maintain
the same level of utility while trading off different bundles of goods and services and their
willingness to make these trade offs can be measured in dollar terms. In so doing, individuals seek
to maximize the utility they receive from the goods and services they consume given relevant budget
constraints. This approach recognizes that, in choosing which activities to pursue, individuals are
subject to a variety of economic and social constraints. These constraints interact with personal
preferences to influence how individuals allocate their time among various types of activities —
including activities that earn income (which then can be expended on goods and services or
invested) and activities that are not income-generating.
The dollar value of the benefits of drinking water regulations would be most directly
measured by determining the change in income (or compensation) that has the same effect on utility
as the regulatory requirements. In practice, economists generally rely on the concept of opportunity
costs in valuing both costs and benefits. The opportunity cost approach recognizes that, because
resources are limited, any decision to use resources for one purpose means that they cannot be used
for other purposes. Hence the value of the resource can be determined based on the value of its next
best use.
Willingness to pay is the approach applied to measure opportunity costs most often in
regulatory analysis. For example, in its guidance on the practice of regulatory benefit-cost analysis
in the Federal government, the Office of Management and Budget (OMB) notes that "[t]he principle
of "willingness to pay" (WTP) captures the notion of opportunity cost by measuring what individuals
are willing to forego to enjoy a particular benefit."3 Individual willingness to pay represents the
2EPA's approach to valuation is discussed in detail in: U.S. Environmental Protection Agency,
Guidelines for Preparing Economic Analyses, September 2000, EPA 240-R-00-003, Chapter 7. For a more
detailed and technical explanation of the underlying theory and methods, see: Freeman, A Myrick III, The
Measurement of Environmental and Resource Values: Theory and Methods, Second Edition, Washington,
D.C.: Resources for the Future, 2003.
3As OMB notes, willingness to accept compensation can also provide a "valid measure of opportunity
costs" under some conditions, but is rarely used to due practical problems related to its empirical
1-2
-------
maximum amount of money an individual would voluntarily exchange to obtain an improvement;
e.g., in drinking water quality or in health status.
Willingness to pay is a different concept than cost or price. Cost generally refers to the
resources needed to produce a good or service; it may not measure the full value of the good or
service to consumers. Price is determined by the interactions of suppliers and consumers in the
marketplace. An individual's willingness to pay may exceed the current price, in which case he or
she benefits from the fact that the market price is less than he or she is willing to pay. If price
instead exceeds willingness to pay, the individual would presumably choose to not purchase the
good. The amount by which willingness to pay exceeds price is referred to as consumer surplus by
economists, and aggregate changes in this difference (i.e., across all consumers) can be used to
measure the dollar value of the social welfare effects of government policies. For example,
consumers generally benefit from price decreases because the decrease means that, for those
purchasing the good, willingness to pay will exceed the price by a larger amount.
1.1.1 Fatal Risk Valuation
EPA has developed standard practices for applying the concepts of welfare economics to
value changes in fatal risks, based on estimates of the "value of statistical life." The value of
statistical life does not refer to identifiable lives, but instead to small reductions in mortality risks
throughout a population. A "statistical" life hence can be thought of as the sum of small individual
risk reductions across an entire exposed population. For example, if 100,000 people would each
experience a reduction of 1/100,000 in their risk of premature death as the result of a regulation, the
regulation would "save" one statistical life (i.e., 100,000 * 1/100,000). If each member of the
population of 100,000 was willing to pay $60 for this risk reduction, the corresponding value of a
statistical life would be $6 million (i.e., $60 * 100,000). Value of statistical life estimates are
appropriate only for small changes in risk; they do not value saving an individual life.
EPA's approach for valuing fatal risks is described in its Guidelinesfor Preparing Economic
Analysis,4 This approach has been subject to substantial peer review and implemented in numerous
regulatory analyses. It was originally derived in the early 1990s from evaluation of the then-
available empirical research, which yielded 26 estimates from value of statistical life studies suitable
for consideration in the context of environmental policy analyses. Of these estimates, 21 are derived
from studies that assess the increase in wages that workers demand for riskier jobs and five that are
based on contingent valuation surveys. The best estimates from these studies (in 2000 dollars) range
from $0.8 million to $17.5 million per statistical life saved, with a mean of $6.3 million. Analysts
often adjust these values to reflect differences between the scenarios considered in the studies and
the scenarios associated with a particular regulation, such as the growth in real income over time.
measurement. U.S. Office of Management and Budget, Regulatory Analysis (Circular A-4), September 2003,
p. 18.
4U.S. Environmental Protection Agency (September 2000), pp. 87 - 94.
1-3
-------
Given the uncertainty in these estimates, analysts generally present a range of values for fatal risk
reductions, based on sensitivity analysis or probabilistic modeling.
EPA is currently conducting a number of projects to improve the approach for valuing fatal
risks. While the approach will continue to evolve as new research is completed, its basic framework
is reasonably well established and generally accepted for use in EPA analyses. Hence this report
does not address the valuation of fatal risks in any detail. Rather, it is concerned with issues related
to valuing nonfatal risks, as introduced below.
1.1.2 Nonfatal Risk Valuation
As noted earlier, EPA's preferred approach for valuing health risk reductions is to determine
individual willingness to pay for related improvements.5 These estimates are generally developed
through stated preference methods (e.g., using surveys to collect information on reported values) or
revealed preference methods (e.g., using data on related behaviors — such as purchases of home
water filters — to estimate these values). Unfortunately, such studies have been completed for only
a small subset of the illnesses associated with exposure to drinking water contaminants, and
completing additional studies often requires more substantial time and resources than are available
given regulatory deadlines and budgetary constraints.
As a result, analysts often transfer benefits estimates from existing studies rather than
conduct new primary research. Benefit transfer involves reviewing the relevant valuation literature,
selecting studies that address effects similar to those addressed by the regulations, and applying the
estimates from these studies to the regulatory analysis. Key issues in conducting these transfers
include ensuring that the studies used are of reasonable quality (e.g., adhere to best practices for the
particular type of research) and are applicable to the policy of concern (e.g., consider similar health
effects and similar populations). In some cases, it may be possible to adjust the primary research
results to address differences between the study scenario and the regulatory scenario.
Such benefit transfers may not be appropriate for some health effects or may result in highly
uncertain estimates. For example, the primary research literature may not include willingness to pay
estimates for health effects that are sufficiently similar to the risk reductions resulting from the
regulation, or the available studies may not be of adequate quality for use in regulatory analysis.
Uncertainties stemming from the benefit transfer process may lead analysts to be interested in
presenting more than one estimate of related values, or in using a different approach for valuation.
Faced with these uncertainties, EPA analysts may choose to consider estimates from the cost
5EPA's Guidelines for Preparing Economic Analyses provide more information on the valuation of
nonfatal risks and on the use of benefit transfers. See: U.S. Environmental Protection Agency (September
2000), pp. 85-87 and 94-98.
1-4
-------
of illness literature.6 This literature is extensive and includes estimates for a wide range of health
effects. It generally focuses on estimating expenditures associated with medical treatment, including
doctor's visits, hospitalization, medications, and other medical goods and services. Some cost of
illness studies also consider indirect costs, particularly the work time lost due to doctor's visits,
hospitalization, and (in some cases) disability. However, as discussed below, this approach does not
provide an ideal measure of value from a welfare economics perspective.
1.1.3 Relationship of Willingness to Pay to Cost of Illness
Although cost of illness estimates may be used for valuation in cases where willingness to
pay estimates are not available, these estimates are not the preferred measure of value from a welfare
economics perspective for a variety of reasons. First, they address a health outcome that differs
significantly from that associated with drinking water or other environmental regulations. Such
regulations lead to future changes in risk; e.g., a rulemaking might provide a 1/100,000 decrease in
an individual's risk of incurring a particular disease by decreasing future contaminant levels. Cost
of illness estimates instead reflect incurred costs to an individual and society for specific cases of
diagnosed illnesses. Hence the scenarios assessed in cost of illness studies differ in both the
perspective (past vs. future) and the level of certainty (diagnosed cases vs. risk of incidence) from
the scenarios of regulatory concern. Medical treatment also may not return an ill individual to his
or her original health state, whereas regulatory action may allow some individuals to avoid the
illness entirely.
Second, cost of illness estimates usually exclude several types of impacts that are important
from a social welfare perspective, including averting or defensive expenditures, indirect costs, and
pain and suffering. Averting or defensive expenditures refer to costs that individuals incur to avoid
illness, such as the use of water filters. To the extent that individuals discontinue these practices as
6Health economists often advocate an alternative approach ~ the use of quality adjusted life years
(QALYs) or other types of health utility indices. This approach generally involves ranking the impacts of
different health effects, then using this ranking to determine the relative value of a year in poor health (due
to a specific health effect) compared to a year in good health. Traditionally, EPA has not applied this
approach in benefit-cost analyses due to concerns about its inconsistency with welfare economics theory and
the concept of willingness to pay. (For a discussion of related issues, see, for example, U.S. Environmental
Protection Agency, Benefits and Costs of the Clean Air Act, 1990-2020: Revised Analytical Plan for EPA's
Second Prospective Analysis, prepared by Industrial Economics, Incorporated, May 2003, pp. 8-11 to 8-12,
and Advisory Council on Clean Air Compliance Analysis, Review of the Revised Analytical Plan for EPA's
Second Prospective Analysis-Benefits and Costs of the Clean Air Act 1990-2020, prepared for the U.S.
Environmental Protection Agency, May 2004, Chapter 11 and Appendix F.) EPA is now considering the use
of these methods in cost-effectiveness analysis in response to new OMB guidance. As requested by OMB,
an Institute of Medicine expert panel is now addressing several issues related to this approach (Office of
Management and Budget, September 2003, p. 13).
1-5
-------
the result of a regulation, related savings can be counted as a benefit of the rule.7 However, these
practices are often motivated by a range of concerns, not all of which may be affected by a particular
regulation. For example, individuals may purchase water filters due to their concerns about a
number of contaminants as well as their desire to improve water taste. A regulation that addresses
only a few contaminants is not likely to noticeably affect the purchase of filters by these individuals.
Hence, in the context of regulatory analysis, the exclusion of averting expenditures from cost of
illness estimates may be of lesser concern than the other considerations raised in this section.
Cost of illness estimates also often exclude indirect costs. Conceptually, these indirect costs
include any type of impact of illness on the economy other than direct medical expenses; however,
this term is most often applied in practice to the consideration of productivity losses. Some cost of
illness studies incorporate consideration of the loss in productivity that results when ill individuals
are unable to engage in their normal schedule of paid work; a few also consider the lost productivity
associated with unpaid work activities, such as housekeeping or volunteer efforts. These studies
suggest that indirect costs can be significant. For example, a study of all chronic conditions in the
U.S. indicated that the indirect costs of morbidity (time lost from paid work and housekeeping)
totaled $72.9 billion in 1990.8 These costs were approximately 17 percent of the medical costs,
which were estimated at $425.2 billion in the same year.9 Other studies (several of which are
referenced later in this report) report indirect costs that equal or exceed the direct costs for certain
illnesses.
Finally, cost of illness estimates do not address the value of avoiding the pain and suffering
associated with illness.10 Studies comparing cost of illness and willingness to pay estimates for a
variety of health effects suggest that the exclusion of pain and suffering and other impacts leads to
cost of illness estimates that understate willingness to pay by a significant amount.11 The degree of
7Care must be taken to ensure that the assumptions regarding defensive behaviors are consistent
between the risk assessment and the benefits valuation portions of the benefit-cost analysis. If averting
behaviors are taken into account in the risk assessment (e.g., because the analysis is based on data from a
population that undertakes such activities), the resulting risk estimates will be lower than if the risk
assessment considered the risks that result when no averting actions are undertaken. In the latter case, adding
an estimate of decreased defensive expenditures to the value of the risk reductions could overstate actual
benefits.
8Hoffman, C., D. Rice, andH. Sung, "Persons with Chronic Conditions: Their Prevalence and Costs,"
Journal of the American Medical Association, Vol. 276, No. 18, November 1996, pp. 1473-1479.
9Medical cost figure includes costs for fatal and nonfatal conditions.
10 For a summary of relevant studies, see: U. S. Environmental Protection Agency, Handbookfor Non-
cancer Health Effects Valuation, December 2000, Appendix B.
nIn some cases, medical expenditures may be greater than individual willingness to pay if third
parties (i.e., insurance) finance treatments that individuals would not be willing to fund more directly from
their own income.
1-6
-------
understatement varies depending on the characteristics of the health effect, the types of costs
considered, and the nature of the study design.12 Available comparisons result in willingness to pay
to cost of illness ratios ranging from about a factor of two to a factor of 79; most of the ratios are
between three and six. In other words, in these studies the costs of illness are typically one-third to
one-sixth of the willingness to pay estimates.
This result is consistent with theoretical work comparing willingness to pay and cost of
illness approaches. For example, Harrington and Portney show that the medical costs of illness plus
the value of lost earnings and defensive expenditures are likely to be less than willingness to pay for
a change in pollution-induced illness under most conditions, due to the exclusion of the value of lost
leisure time and of avoiding the disutility (e.g., pain and suffering) associated with illness.13 They
conclude that the cost of illness approach (including medical costs and lost wages) can be used as
a lower bound estimate of the true value of benefits in most cases.
In this context, it is important to note that even empirical estimates of individual willingness
to pay may understate societal values, in particular if the methods used in the study do not address
costs borne by third parties or altruistic concerns about the health of others.14 The issue of third
party costs is addressed in the current OMB guidance on valuing morbidity impacts, which
recommends that these values include both: "(1) the private demand for prevention of the nonfatal
health effect, to be represented by the preferences of the target population at risk, and (2) the net
financial externalities associated with poor health such as net changes in public medical costs and
any net changes in economic production that are not experienced by the target population.1,15 Later
in the guidance, OMB also notes that economic analyses should include monetary values of "gains
or losses of time in work, leisure, and/or commuting/travel settings" if they are significant.16
Despite its limitations, EPA analysts may, at times, rely on the medical cost of illness
literature for valuation when willingness to pay estimates are lacking or are subject to enough
uncertainty that comparison to other measures is desirable. As discussed above, such estimates are
likely to understate true willingness to pay in most cases. The goal of this report is to address one
12For example, these studies consider health effects that differ in severity and duration, use cost of
illness estimates that vary in the extent to which they include forgone earnings or insurance-paid expenses,
and estimate willingness to pay using different contingent valuation and averting behavior methods ~ the
contribution of each of these factors to the overall degree of understatement found in the comparison studies
is uncertain.
13Harrington, W. and P. Portney, "Valuing the Benefits of Health and Safety Regulations," Journal
of Urban Economics, Vol. 22, 1987, pp. 101- 112.
14Inclusion of altruism can lead to double-counting if not carefully addressed; see U.S. Environmental
Protection Agency (September 2003), p. 92.
15U.S. Office of Management and Budget (September 2003), pp. 28 - 29.
16U.S. Office of Management and Budget (September 2003), p. 36.
1-7
-------
of the several factors that contribute to this understatement — the need to consider the effects of
illness on activity restrictions (i.e., time losses).
1.2 APPLICATION TO REGULATORY ANALYSIS
The purpose of this report is to develop estimates of the value of lost time that can be added
to estimates of the medical costs of illness in to develop improved measures of the welfare effects
of reducing nonfatal health risks. To date, researchers largely have been interested in assessing the
costs of illness as part of the determination of the relative cost-effectiveness of alternative resource
allocations, both at the individual patient level (e.g., evaluating treatment options) and at the
institutional or national level (e.g., evaluating options for investing health care or research dollars).17
As such, these studies often are not directly suited for use in benefits valuation for drinking water
or other regulations.
In contrast, EPA regulatory analyses focus on estimating the total national costs and benefits
of various options for regulating a particular contaminant (e.g., arsenic) or group of contaminants
(e.g., microbial pathogens), as well as the distribution of the costs and benefits across different sub-
populations of interest. Hence they generally require information on lifetime costs per case, whereas
cost of illness researchers may be more interested in the costs associated with different treatments
for a particular symptom or in national illness-related expenditures for a particular year. The
following sections explore how the differences between these two types of analyses affect the choice
of data and methodologies used to estimate the value of time losses.
1.2.1 Data Needs for Regulatory Analysis
The benefits portions of EPA regulatory analyses generally start with an assessment of the
human health risks attributable to each regulatory option, considering each major health effect
associated with the contaminant of concern.18 For example, an analysis of the effects of arsenic
exposure may separately address lung and bladder cancer, as well as ischemic heart disease and
diabetes or other relevant health endpoints. Estimated post-regulatory risk levels are compared to
a "no action" baseline — under which analysts predict future risks in the absence of the regulatory
changes under consideration. The risk changes attributable to each regulatory option are then
summed across the affected population to determine the number of statistical cases avoided
17For a discussion of best practices from this perspective, see: Gold, M.R., J.E. Siegel, L.B. Russell,
and M.C. Weinstein (eds.), Cost-Effectiveness in Health and Medicine, Oxford: Oxford University Press,
1996.
18EPA provides guidance on risk assessment in a number of other documents and hence issues related
to this part of the analysis are not addressed in detail in this report.
1-8
-------
nationally.19 Depending on the characteristics of the rulemaking, the available health science
research, and other factors, these estimates may be disaggregated by gender and/or by age group
affected, by type of water system, by geographic location, or by other categories of concern.
A simplified example of this process is as follows. Risk assessors may estimate that a
specific MCL (e.g., of 10 |ig/L) will reduce the annual average individual risk of incurring a
particular type of kidney disease by 1/10,000, compared to the baseline risk prediction. If this risk
reduction is experienced by a total population of 50,000 persons, the number of statistical cases
avoided would be five cases (1/10,000 * 50,000) per year.20 Risk assessors may also note that about
half of these cases would be fatal, and that the fatalities averted by the regulation would primarily
be among elderly members of the population. Furthermore, uncertainty analysis may indicate that
the number of cases averted may be understated or overstated by a factor of four.
The outcome of this risk analysis is usually an estimate of the number of fatal and nonfatal
statistical cases avoided, possibly disaggregated by age group or other key attributes. This estimate
is often reported as a range or probability distribution that reflects uncertainties in the analysis. The
estimate of cases avoided then becomes the input for the valuation of related benefits. Hence
economists working on benefit-cost analyses are generally interested in "per statistical case" values
that can be applied to these risk estimates.
1.2.1 Data Provided by the Cost of Illness Literature
In contrast to the need for "per statistical case" estimates in regulatory benefit-cost analysis,
much of the cost of illness literature has been developed to support decisions on medical treatment
or on the allocation of health care resources. These studies may be based on the "prevalence" or
"incidence" of the conditions of concern. Prevalence-based studies assess the costs of all cases
existing in a given year, whereas incidence-based studies consider costs per new case diagnosed.
Either method can be used to estimate direct (medical) and/or indirect (e.g., time) costs.21
19Cost of illness studies often address all the cases of the illness found in the population studied
regardless of cause. In contrast, regulatory analyses usually address only the risks associated with the
contaminants of concern. Because most environmental contaminants are only one of many potential causal
factors contributing to the overall incidence of adverse health effects, regulatory analyses usually address a
smaller number of cases than would a cost of illness study focused on the same population. Cases associated
with environmental contaminants may also have different characteristics than cases associated with other
causes; e.g., may be more or less severe or affect different age groups.
20This type of analysis is more difficult for noncancer health effects than for most cancers, due to the
availability of generally accepted practices for cancer dose-response assessment. Developing such practices
for noncancer health effects is an active area of research because the lack of dose-response functions severely
limits the quantification of most noncancer health endpoints in many benefit-cost analyses.
21Examples of these studies are provided in Appendix A as well as in footnoted references provided
throughout this report.
1-9
-------
To illustrate this difference, consider a hypothetical study designed to estimate the costs
associated with a specific illness in a given year (e.g., 1998). Statistics characterizing the prevalence
of the illness would estimate the costs for all individuals suffering from that illness in that year,
regardless of the year of diagnosis. The use of incidence statistics, on the other hand, would yield
an estimate of the costs over the entire period assessed (discounted back to the base year — 1998),
by those individuals who were first diagnosed with the illness in that year.
Data sets characterizing the incidence and prevalence of an illness for a given year will
contain overlapping observations but will yield different results. For example, an individual who
contracts the disease in February 1998 will be represented by both incidence and prevalence data
for 1998. However, a person who contracts the disease in November 1997 and is still sick in 1998
will only appear in the prevalence sample for 1998.
The decision to use prevalence or incidence statistics depends on the goal of a particular
analysis and the availability of data. Any transfer of a benefit estimate from the original study to a
different context is generally accompanied by careful consideration of the quality of the study and
the similarity of the health endpoints and populations considered.22 In a case where both
prevalence- and incidence-based studies are available and suitable for transfer, incidence-based
values are generally preferred for regulatory benefit-cost analysis. Such studies provide "per case"
values that can be easily linked to the results of the risk assessment (i.e., the number of statistical
cases avoided) as described above.
The available incidence-based studies may, however, have shortcomings that must be
considered when used in regulatory analysis. For example, for some illnesses, the available studies
may not track the full course of the illness. Ideally, the cost estimate would cover the complete time
period over which the health effect is experienced, which may range from a few days (e.g., for
certain gastrointestinal ailments) to the remainder of the individual's lifetime (e.g., for illnesses such
as diabetes where the symptoms can be treated but the illness can not be "cured"). However, the
available research may consider a shorter time frame (e.g., one, five, or 10 years) determined largely
by the available data.
In addition, for some illnesses, the available incidence-based studies may provide
disaggregate data that cannot directly be used to develop "per case" estimates. Such studies may
track costs separately for individuals undergoing different types of treatment or experiencing
different types of illness-related events, rather than providing estimates of the aggregate costs of all
treatments and all events associated with a typical case. For example, such studies may track the
costs associated with angina or heart attacks separately, without providing estimates of the
likelihood that a typical patient with ischemic heart disease would experience events of each type.
In addition, some studies may not separate the costs of fatal cases from the costs for nonfatal cases.
22Guidance on conducting benefit transfers is provided in: U.S. Environmental Protection Agency
(September 2000), pp. 85-87.
1-10
-------
To address many of these concerns, EPA developed "per case" cost of illness estimates for
several health effects, separating costs for survivors and non-survivors.23 These estimates do not
include indirect costs, however. Hence analysts may wish to add the value of time losses (using the
approach discussed in the subsequent chapters of this report) to these cost of illness estimates for
a more complete accounting of the impacts of illness. For illnesses not covered in the EPA
handbook, analysts will need to review the health economics literature to determine how to best
estimate the medical costs of illness.24
The remainder of this document focuses on approaches for estimating the per case value of
the time losses associated with nonfatal health effects. It is concerned primarily with time losses
for persons living with the disease and their caregivers. Although health economists sometimes use
estimates of lost lifetime earnings to value premature mortality, EPA instead values mortality by
applying willingness to pay estimates from the wage-risk and contingent valuation literature as
discussed earlier in this chapter. The approach discussed in this report may, however, be useful in
assessing time losses due to morbidity prior to death as well as due to nonfatal health effects.
The next chapter first reviews the empirical literature that addresses the value of time in a
number of different contexts, including studies of nonfatal health risks, transportation options, and
recreational opportunities. It then describes the theoretical underpinnings of these studies and the
relationship of this theory to the valuation of illness-related losses. Chapter 3 then describes the
approach for valuing time diverted from paid or unpaid work, leisure, or sleep in more detail,
provides an example of the suggested approach, and summarizes related limitations and the
assessment of uncertainty. These chapters are followed by a list of references used in developing
the report. The report concludes with two appendices: Appendix A summarizes selected examples
of time valuation from the indirect cost literature, while Appendix B discusses data sources that can
be used to conduct the analyses described in this report.
23U.S. Environmental Protection Agency, Cost of Illness Handbook, February 2001.
24EPA has not yet developed general guidance for working with the available cost of illness literature,
beyond the approaches applied to the particular studies used in the Cost of Illness Handbook. Analysts will
need to assess the quality of individual studies, consider their suitability for use in regulatory analyses, and
determine how to best convert prevalence estimates, or incidence-based per treatment or per event estimates,
to estimates of lifetime costs per case. In addition, because EPA is interested in using these studies primarily
for valuing nonfatal health effects, approaches for separating out the costs of fatal versus nonfatal cases may
be needed.
1-11
-------
2.0 THEORETICAL AND EMPIRICAL APPROACHES
Review of the available empirical literature suggests that few researchers have fully
addressed the value of time losses due to illness in a social welfare context. Researchers concerned
with the availability of goods and services focus on the loss in productivity that results when ill
individuals are unable to engage in normal work activities. These studies do not fully capture the
effect of time losses on individual or social welfare, because they do not address the impacts on time
normally used for leisure or other uncompensated activities nor do they address the utility (or
satisfaction) potentially associated with work beyond its effect on production.
Researchers who are instead concerned with valuation within a social welfare context
generally do not provide separate or comprehensive values for all types of time losses. While many
studies of willingness to pay for risk reductions consider the effect of illness on normal activities,
they do not provide a value for time losses that is separate from the value of other attributes of the
illness. Several researchers have estimated the value of time spent traveling in the context of studies
of transportation options and recreation opportunities. However, these estimates focus on the
allocation of time to specific types of activities; not on its allocation to the full range of work and
leisure activities that could be affected by illness.
While the available empirical research is of limited usefulness in providing specific dollar
values that can be applied to illness-related losses, these studies involve substantial research into the
factors that can affect these values. Hence these studies, and the theoretical framework that supports
them, have a number of interesting implications for the valuation of both work and nonwork time.
This chapter reviews these studies and their advantages and limitations in the context of
valuing time losses due to illness. It then discusses the key tenets of economic theory that underlie
these studies and their implications. The chapter concludes by introducing an approach that uses
compensation data to estimate the opportunity costs associated with illness-related time losses,
which is discussed in more detail in the following chapter.
2.1 METHODS FOR VALUING LOST PRODUCTIVITY
Most of the available empirical studies of the effects of illness on time use are concerned
with the potential loss in productivity; i.e., the decrease in the goods or services produced as a result
of disability or absenteeism from work. The most common approach for valuing this lost
productivity is the human capital approach, which uses compensation data to measure the value of
a worker's product. An alternative is the friction cost method, which focuses on the loss in
productivity that occurs prior to replacement of absent workers. Both methods provide an
incomplete accounting of the value of time losses from a social welfare perspective because they
exclude consideration of the utility an individual gains from involvement in both work and nonwork
activities. The main differences between the two methods are the assumptions they make regarding
the functioning of the labor market, particularly regarding the impact of unemployment. Both
approaches are briefly summarized below.
2-1
-------
2.1.1 Human Capital Method
The most frequently used approach for estimating the value of time losses due to illness is
the human capital approach, which is concerned with the effects of illness on the production of
goods and services. (A number of examples of this approach are referenced later in this report, and
selected studies are reviewed in Appendix A.) This well-established approach focuses on individual
productivity, defined as output over time, and assumes that workers are paid the value of their
marginal product. As a result, worker compensation can be used to estimate the costs associated
with absenteeism or reduced output. Underlying this approach are a number of standard theoretical
assumptions regarding the functioning of labor markets, particularly that: (1) these markets are
competitive; (2) firms seek to maximize profits; and (3) unemployment is insignificant. The last
assumption is relevant because, for example, if an absent worker is replaced by a worker who would
be otherwise employed at a different job, the net national change in output is still equivalent to the
effects of having one less worker. If instead an absent worker is replaced by an unemployed
individual, the net change in output may be small and temporary, resulting largely from the time
needed for the transition to a new worker.
The human capital approach is closely linked to the approaches used to estimate the medical
costs of illness. In both cases, researchers are interested in the effects of related costs — in terms of
dollar expenditures (medical costs) or decreased production of goods and services (lost work time) —
on the wealth of the nation. Hence a number of cost of illness studies, including several discussed
later in this report, use the human capital approach to estimate the indirect costs of illness.
Applications of the human capital approach typically focus on paid work time, because
compensation for this work provides a straightforward measure of the marginal value of this time
to society. However, several analysts have extended this approach to encompass unpaid productive
work; e.g., in the household or as volunteers. Although individuals engaged in nonmarket labor are
not paid a wage, many observers consider such activities productive because they could be
performed by a professional in return for compensation. To estimate the value of time associated
with nonmarket labor, a "wage rate" must be derived, generally based on information on market
rates for similar activities (e.g., for paid domestic workers).
The human capital approach is generally not used to value lost nonwork (leisure or sleep)
time because researchers assume that related productivity losses are reflected at least in part in lost
earnings; e.g., an illness-related decrease in rest or relaxation would manifest itself in part in
decreased productivity during work hours. In addition, the effects of illness on individuals who do
not engage in productive market or nonmarket work are not considered. However, some human
capital studies consider time losses during childhood to the extent that they affect future earnings;
for example, a child who misses a substantial amount of schooling due to illness may earn less as
an adult.
Implementation of the human capital approach for valuing productivity losses is relatively
straightforward. It relies on fairly simple calculations that involve well-defined variables and use
2-2
-------
established, readily available data sources of known quality. The literature includes completed
studies for a number of illnesses, and analysts can construct their own estimates relatively easily.
As a result, this approach can be applied across a wide range of illnesses.
2.1.2 Friction Cost Method
The friction cost approach is a comparatively new method for assessing productivity losses
due to illness.1 This approach was developed largely to address the mitigating effects of
unemployment and other market conditions on productivity losses. It assumes that productivity will
decrease temporarily while the employer implements measures to replace the ill individual, rather
than over the full course of the illness. Proponents of this approach note that, when the labor market
is not at full employment, it is possible to replace affected workers (after a period of adaptation), by:
(1) hiring qualified unemployed individuals; (2) utilizing existing labor reserves within the firm; or
(3) reallocating employees to different functions and postponing non-urgent tasks. The "friction
period" is defined as the time it takes to find and train a new employee or reallocate duties among
existing employees.
Implementation of the friction cost method is relatively complex because it requires
estimates of the length and frequency of illness-related friction periods as well as the loss in
productivity that accrues during these periods.2 (The costs associated with finding replacement
workers, including advertising and interviewing candidates, may also be included.) Generally, the
duration of job vacancies is used to estimate the friction period; this vacancy rate presumably
decreases as unemployment increases and reflects the efficiency of the labor market.3
One example of this approach is a study by Hutubessy et al. that addresses the indirect costs
of back pain in the Netherlands in 1991.4 This study is of interest both because it assesses friction
costs for a single, nonfatal condition and because it compares the effects of using the human capital
approach to the friction cost approach. In the human capital portion of the study, the researchers
1 Koopmanschap. M.A., F.F.H. Rutten, B.M. van Ineveld, and L. van Roijen, "The Friction Cost
Method for Measuring Indirect Costs of Diseas q," Journal of Health Economics, Vol. 4, 1995, pp. 171-189.
2Koopmanschap, M.A., and F.F.H. Rutten, "A Practical Guide for Calculating Indirect Costs of
Disease," PharmacoEconomics, Vol. 10, No. 5, 1996, pp. 460-456.
3Researchers also argue that, by affecting the labor costs per unit of output and hence competitiveness
in the world market, absenteeism can have macroeconomic impacts. Because the reductions in the risks of
nonfatal illnesses attributable to drinking water regulations are not large enough to have noticeable
macroeconomic impacts, such impacts are not discussed in this report. See: Koopmanschap, M.A., and F.F.H.
Rutten, "Indirect Costs in Economic Studies," PharmacoEconomics, Vol. 4, No. 6, 1993, pp. 446-454.
4 Hutubessy, R.C.W., M.W. van Tulder, H. Vondeling, and L.M. Bouter, "Indirect Costs of Back Pain
in the Netherlands: a Comparison of the Human Capital Method with the Friction Cost Method," Pain, Vol.
80, 1999, pp. 201-207.
2-3
-------
multiply the number of sick days attributable to back pain by the mean costs (paid wages or
disability pension) per day. In the friction cost portion, they both limit the number of sick days
included to the friction period (estimated as equal to the average vacancy duration) and add in
consideration of the estimated elasticity of work time vs. work output.5 They find that the human
capital approach leads to estimates over three times larger than the friction cost approach.
This result is to be expected because the friction cost method estimates productivity losses
over a shorter time period than does the human capital approach. However, the friction cost
approach could significantly understate the effects of illness on output if replacement of ill workers
diverts other workers from productive tasks. For short-term or relatively mild conditions,
productivity losses will depend on the extent to which any loss attributable to illness is
counterbalanced by greater-than-usual productivity by co-workers or by the ill worker upon his or
her return to work. For conditions with more severe or longer term impacts, net national losses will
depend on the extent to which ill workers are be replaced by someone who is currently unemployed
(compensating at least in part for the productivity loss), or by moving someone from another job
(shifting the loss from one location to another). In either case, researchers must address difficult
questions regarding the net effects of illness on productivity, which often require understanding
conditions at the level of the individual firm.
The assumption that ill workers will be replaced by unemployed or otherwise underutilized
individuals may not be true in economies with low unemployment rates or in industries that require
highly specialized skills. In addition, productivity may not return to its previous level if the ill
individual was in a position where factors that are not easily offset by training (such as tenure) have
a significant effect on output. Thus while the friction cost method suggests that the human capital
approach may overstate the effects of illness on productivity, the extent of overstatement is uncertain
and will vary depending on the nature of the illness, the characteristics of the job, and the conditions
in labor market at the time when the absence occurs. Finally, because the friction cost method does
not account for the opportunity costs of working (i.e., the trade-off between labor and leisure time)
and does not assign value to labor beyond the friction period, it is an incomplete measure of the
value of lost work time.6
The friction cost method is less well established than the human capital approach and its
5This elasticity may represent less than a one-to-one correspondence; e.g., one day of absence from
work may lead to less than a one day loss in output, if other workers produce at higher than normal rates to
compensate for the absent worker. In the case of the Netherlands in 1991, Hutubessy et al. assume a elasticity
of 0.8 for annual labor time vs. labor productivity.
6For discussion of these and other concerns regarding the consistency of the friction cost approach
with neoclassical welfare economic theory, see: Johannesson, M. and G. Karlsson, "The Friction Cost
Method: A Comment," Journal of Health Economics, Vol. 16, 1997, pp. 249 - 255. For commentary on the
use of the friction cost approach in cost-effectiveness analysis, see: Weinstein, M.C., Siegel, J.E., et. al,
"Productivity Costs, Time Costs and Health-Related Quality of Life: A Response to the Erasmus Group,"
Health Economics, Vol. 6, 1997, pp. 505-510.
2-4
-------
implementation in the context of regulatory analysis would present a variety of challenges. Friction
cost estimates are available for only a few health effects, and many of the available studies address
countries other than the U. S.7 In addition, these studies usually address losses within a given year.
It would be difficult to convert these annual estimates to the types of estimates of lifetime costs per
case required for regulatory analysis, especially because the level of unemployment (and hence the
likelihood that an ill worker will be replaced with an unemployed individual) is likely to vary over
time.
Like the human capital method, the friction cost approach focuses on productive work and
does not address other types of activities. While proponents of the friction cost approach believe
that it could be used to estimate the effects of illness on reduced productivity while working, on
unpaid productive activities (such as household tasks or volunteer work), and on leisure time, they
have not yet developed an approach for quantifying friction-related costs in these areas.8 In addition,
using this method to assess the future impacts of illness (such as in the case of childhood illnesses
that affect adult productivity) presents difficulties due to the need to predict unemployment levels
as well as the duration and costs of friction periods over time. Finally, while both the human capital
and friction cost methods exclude consideration of the impact of illness on the individual utility
associated with work and nonwork activities (beyond the impact on productivity), the friction cost
method also does not address the effects of lost compensation on the ill individual and his or her
family.
2.1.3 Relationship of Productivity Losses to Social Welfare Losses
The human capital and friction cost approaches to valuing lost productivity differ in terms
of the underlying assumptions regarding the operation of labor markets and the value of labor as
well as in terms of practical application. However, the most significant deficiency of both
approaches is their narrow focus on production and the exclusion of other aspects of social welfare.
The goal of EPA regulatory analyses is to develop as complete an accounting as possible of
the social welfare impacts of alternative policies. Both of the methods discussed above fall short
of this goal because they focus on goods and services and do not consider other factors that affect
individual well-being. As discussed by A. Myrick Freeman III in his seminal work on benefits
valuation:
The economic concept of value employed here has its foundation in neoclassical welfare
economics. The basic premises of welfare economics are that the purpose of economic
7For example, a recent search of the MEDLINE bibliographic database identified 11 studies that
apply the friction cost method, many of which were completed in countries other than the U.S. and address
symptoms (such as neck pain) or health effects (such as mental illness) not usually associated with reductions
in drinking water contamination
8Koopmanschap and Rutten (1993).
2-5
-------
activity is to increase the well-being of individuals who make up society, and that each
individual is the bestjudge of how well off he or she is in a given situation. Each individual's
welfare depends not only on that individual's consumption ofprivate goods and of goods
and services produced by the government, but also on the quantities and qualities each
receives of nonmarket goods and service flows...9
In this context, analysts are concerned with the effect of illness on foregone market
production (paid work), foregone nonmarket production (e.g., volunteer or household activities), and
any additional diminished utility (or sense of well-being) associated with both work and nonwork
activities (including leisure and sleep time). Analysts are also concerned with the impacts of illness
on other individuals such as dependent children or unpaid caregivers.10
In contrast, as discussed above, most of the work on valuing time losses has been completed
in the context of estimating the market impacts of illness, focusing on the ill individual. In this
context, individuals are viewed as mechanisms of production. The analyst is generally not concerned
with other (nonmarket) factors affecting individuals' sense of well-being, except, of course, if they
affect market productivity. For example, in his discussion of the use of the human capital approach
to value premature mortality, Freeman notes:
The human capital approach is fundamentally at odds with the individualistic perspective
of welfare economics and the theory of value. By in effect asking what the individual is
worth to society, the human capital approach ignores the individual's own well being,
preferences, and WTP [willingness to pay]. It defines the social worth of an individual in
a narrow way, that is, as the individual's market productivity, thereby ignoring the value of
that person's health and well-being to loved ones."
Similar concerns have been voiced regarding the friction cost method, which takes an even
narrower view of the value of lost work time. For example, Johannesson andKarlsson note that "the
friction cost approach is based on implausible assumptions not supported by neoclassical economic
theory." Economic theory suggests that "[a] firm can be expected to hire labour until the marginal
cost of labour equals the marginal value of the products produced by the worker. If a worker is
absent this would represent a marginal loss of labour, whose value for the firm equals the gross
income of the worker.1,12 Furthermore, the authors note the friction cost approach does not take into
account the loss in leisure time that accrues when short term absences are offset by greater-than-
9Freeman, A.M. Ill, The Measurement of Environmental and Resource Values: Theory and Methods,
Second Edition, Washington, D.C.: Resources for the Future, 2003, p. 7.
10Paid care is included in the medical cost component of the analysis and hence is not included in the
valuation of time losses.
"Freeman (2003), p. 302.
12Johannesson and Karlsson (1997).
2-6
-------
normal productivity by others or by the ill worker upon his or her return. For long term absences,
they note that, after the friction period, the price (i.e., opportunity cost) of work time is set close to
zero, which is inconsistent with both economic theory and empirical evidence.
2.2 METHODS FOR VALUING SOCIAL WELFARE LOSSES
The generally preferred method for valuing social welfare losses due to illness would be to
rely on estimates of individual willingness to pay, as discussed earlier in this report.13 Such
estimates could directly address the limitations of the human capital and friction cost methods
described above by providing a broader measure of the impact of illness on individual welfare. For
goods and services that are not directly bought and sold in the market place, economists generally
rely on stated or revealed preference methods to estimate these values. Stated preference methods
typically involve asking individuals what they would be willing to pay, whereas revealed preference
methods use information on the price of market goods to estimate willingness to pay for related
nonmarket goods.
While it would be possible to use revealed or stated preference methods to value illness-
related time losses, few, if any, such studies have been completed. As discussed below, most
empirical work provides a total value for all attributes of an illness, without separating out the value
of time losses.14 These studies suggest, however, that illness-related activity restrictions are an
important influence on individual willingness to pay for risk reductions.
The value of time has been studied extensively, however, in the literature on willingness to
pay for transportation and recreation options. While this literature focuses somewhat narrowly on
particular types of time use rather than the full range of leisure and work activities potentially
affected by illness, the underlying theory and the results of related empirical studies have a number
of implications for the valuation of illness-related losses, as described in the following sections.
13Estimates of willingness to accept compensation may be appropriate in some cases, but are rarely
applied in practice, as discussed in U.S. Office of Management and Budget, Regulatory Analysis (Circular
A-4), September 2003, p. 18 and U.S. Environmental Protection Agency, Guidelines for Preparing Economic
Analyses, September 2000, EPA 240-00-003, p. 60. Freeman (2003) provides detailed discussion of related
theoretical and practical issues.
14More information on these methods is available in U.S. Environmental Protection Agency
(September 2000) and Freeman (2003).
2-7
-------
2.2.1 Willingness to Pay for Risk Reductions
A number of researchers have used stated or revealed preference methods to assess the value
of risk reductions without separating out the value of related time losses. This holistic approach is
undertaken largely because it is the preferred approach to valuation. As discussed in Chapter 1 of
this report, separate valuation of time losses is needed only when the existing literature does not
provide suitable estimates of willingness to pay for the risk reductions of concern and analysts lack
the time or resources necessary to undertake new willingness to pay studies.
Available research suggests that the effect of illness on time usage is an important
component of individual willingness to pay for risk reduction. Several willingness to pay studies
include lost time explicitly in the valuation scenario.15 For example, in a survey of willingness to
pay for a program to reduce bad asthma days, Rowe and Chestnut elicited information on the effects
of asthma on work, school, chores, and leisure activities.16 However, the authors did not ask
respondents to report their willingness to pay to avoid these time losses separately; rather, they asked
respondents to report their total willingness to pay for the program.
Researchers also have found that time losses are a key factor affecting the magnitude of
individual willingness to pay for health risk reductions. For example, Magat, Viscusi, and Huber
found that "must restrict recreational activity" was one of the consequences of nerve disease to
which respondents were most adverse.17 Even for minor symptoms (coughing spells, stuffed up
sinuses, etc.), Berger et al. found that a small percentage of respondents ranked loss of work at
home or loss of recreation as the most important reason for their value for relief of symptoms.18 In
their study of angina, Chestnut et al. found that "subjects said that the most bothersome effects of
a worsening of their condition ...would be decreased ability to do desired activities (recreation,
chores, or work), and pain or discomfort."19
15The cited studies generally use stated preference methods. Revealed preference methods that
consider averting behavior or defensive expenditures also implicitly involve the valuation of time, since
engaging in these behaviors (e.g., installing and maintaining a water filter) has a time cost as well as a
monetary cost.
16Rowe, R.D. and L.G. Chestnut, Valuing Changes in Morbidity: WTP vs. COIMeasures, prepared
for the U.S. Environmental Protection Agency and the California Air Resources Board, undated.
17Magat, W.A., W. K. Viscusi, and J. Huber, The Death Risk Lottery Metric for Valuing Health Risks:
Applications to Cancer and Nerve Disease, prepared for the U.S. Environmental Protection Agency, 1992,
Table 4.
18Berger, M. C., G. C. Blomquist, D. Kenkel, G. S. Tolley, "Valuing Changes in Health Risks: A
Comparison of Alternative Measures," The Southern Economic Journal, Vol. 53, 1987, pp. 977-984.
19Chestnut, L.G., S.D. Colome, L.R. Keller. W.E. Lambert, etal., Heart Disease Patients Averting
Behavior, Costs of Illness, and Willingness to Pay to Avoid Angina Episodes, EPA Report 230-10-88-042,
October 1988, p. 4.
2-8
-------
In addition, researchers have used stated preference methods to assess the value of restricted
activity days, which are generally defined as time periods when individuals find that their activities
are more limited than normal. This approach is closely related to the concept of lost time, but (as
applied in the available literature) generally results in a value for all of the restrictions associated
with a particular illness, rather than in a value of time losses that could be adjusted for application
across different illnesses.
The restricted activity day approach is often used in the air pollution context. For example,
researchers have linked self-reported data on activity restrictions to data on air pollution levels to
determine the effects of this pollution on normal activities. They then apply estimates of willingness
to pay for avoiding these restrictions (derived from stated preference studies of minor respiratory
symptoms) to the estimates of the number of activity restricted days to determine their value.20
2.2.2 Willingness to Pay for Transportation and Recreational Options
In its simplest form, neoclassical economic theory suggests that individual swill allocate time
between paid work and other activities so that, at the margin, the value of paid work is equal to the
value of uncompensated activities. If individuals receive no utility from work (beyond the effects
of income on consumption), then wage rate can be used to estimate the opportunity cost, or shadow
price, of time. Research on the value of travel time, within the context of evaluating transportation
and recreation options, thus often uses the wage rate as a reference point for comparing results
across studies. Researchers are not always consistent in the basis for this comparison (e.g., in
whether they compare their findings to individual or household income and in whether they include
or exclude taxes and benefits). Such comparisons indicate, however, that the empirical results of
these studies vary greatly.
Travel time has been studied extensively by researchers interested in assessing different
transportation options.21 Some of these studies focus on the relative value of different modes of
transport (e.g., airplane, car, walking) taking into account comfort and convenience as well as speed
and related economic costs.22 Other studies focus more on the value of each time increment spent
in travel for different purposes, such as business or leisure activities.
The variation in the resulting values is evident in the U.S. Department of Transportation's
20See, for example, the approach applied in: Abt Associates, Final Heavy Duty Engine/Diesel Fuel
Rule: Air Quality Estimation, Selected Health and Welfare Effects Methods, and Benefits Results, prepared
for the U.S. Environmental Protection Agency, December 2001.
21For a review of these studies and the underlying theory, see: Small, K., Urban Transportation
Economics, Luxembourg: Harwood Academic Publishers, 1992, pp. 36-45.
22MVA Consultancy et al., "Research Into the Value of Time," in R. Layard, R. and Glaister, S.( eds.),
Cost-Benefit Analysis, Second Edition, Cambridge: Cambridge University Press, 1994, pp. 235-272.
2-9
-------
(DOT's) recommendations for valuing changes in travel time attributable to its programs, which are
based on detailed review of the literature.23 For business travel (during paid work hours), DOT
indicates that a plausible range is 80 to 120 percent of the total hourly compensation rate (including
wages and benefits), and recommends a best estimate of 100 percent. For personal travel
(commuting, shopping, recreation, etc.), DOT suggests a plausible range of 35 to 90 percent of pre-
tax wages, with a best estimate of 50 to 70 percent depending on whether the travel is local or
intercity. These ranges reflect, in part, the fact that travel has both desirable and undesirable
attributes which may have counterbalancing effects in determining total value. For example, a trip
may include time spent on relaxing scenic routes as well as in stressful urban congestion. Business
travelers may devote some of their travel time to productive work, or may find that their productivity
suffers if a travel delay leads them to miss meetings or other work activities.
Travel and related uses of leisure time also have been studied extensively in the context of
valuing recreation opportunities; e.g., the availability of public lands for activities such as fishing
or hiking.24 The fundamental assumption is that the value of a recreational opportunity is at least
as great as the value of what one is willing to give up (e.g., the opportunity costs of money and time
expenditures) in order to participate in related activities.25 For example, a simple travel cost model
may use market data and survey information to determine the money costs (e.g., fuel, tolls, and
access fees) and the time costs (e.g., spent traveling and on-site) to make inferences about individual
willingness to pay. Some models, such as random utility models (sometimes referred to as discrete
choice models), consider environmental quality variables as well as travel costs that affect an
individual's choice between different recreational sites. While the money costs considered in these
studies are often observable, the value of time is more difficult to measure and there is no consensus
regarding how it is best valued. Further, disagreement exists regarding the valuation of time spent
in travel verses time spent on-site.
Several studies of recreation opportunities value travel time at some fraction of the wage rate,
frequently one-third. The use of one-third of the wage rate is somewhat arbitrary and appears to
have its origins in some of the early transportation literature.26 For example, many economists
23U.S. Department of Transportation, Departmental Guidance for the Valuation of Travel Time in
Economic Analysis (Memorandum from F.E. Kruesi), April 1997, and U.S. Department of Transportation,
Revised Departmental Guidance, Valuation of Travel Time in Economic Analysis, (Memorandum from E.
H. Frankel), February 2003.
24For more detailed discussion of approaches for valuing recreational opportunities, see U.S.
Environmental Protection Agency (September 2000), pp. 73 - 74, and Freeman (2003), Chapter 13.
25As noted by Wilman, under this approach "[i]t is assumed that a travel-cost increase would be
viewed by the recreationalist as being equivalent to a fee increase for an on-site visit." Wilman, E.A., "The
Value of Time in Recreation Benefit Studies," Journal of Environmental Economics and Management, Vol.
7, 1980, pp. 272- 286.
26Shaw, W. D., and P. Feather, "Possibilities for Including the Opportunity Cost of Time in
Recreation Demand Systems," Land Economics, Vol. 75, No. 4, 1999, pp. 592-602.
2-10
-------
reference a 1976 study by Cesario that examines the question of how to value time in recreational
studies.27 This study reviews the then-available literature, which focused almost exclusively on the
value of commuting time and included several studies of the choice between private and public
transport options. Cesario indicates that "[t]hese early studies were plagued by the usual
methodological problems besetting research into any new area of inquiry." However, he goes on
to note that "it may be tentatively concluded.. .that the value of nonwork travel is between one-fourth
and one-half of the wage rate" for the average individual and trip. In his own analysis of recreation
opportunities in several parks, Cesario then uses one-third of the wage rate to represent the results
of his literature review, while noting that this value "is arbitrary."
Although some researchers subsequently used this value of one-third the wage rate in their
own studies, a number of others began to more directly investigate the value of time in the context
of recreational valuation.28 The results of these studies vary greatly, due to differences in the
modeling approaches used (and the simplifying assumptions they incorporate) as well as in the data
sources and types of activities considered. For example, in a 1981 study of sportfishing, McConnell
and Strand directly examine the value of travel time as a proportion of the wage rate.29 Their simple
travel cost model assumes that individuals have flexible work hours and that the ratio of the
opportunity cost of time to income is constant. They find that, for the typical angler included in their
survey, travel time is valued at approximately 60 percent of hourly income.
A study published in 1983 by Smith, Desvousges and McGivney then examined the
assumption that the opportunity cost of time could be represented as a fixed percentage of the wage
rate.30 The researchers rely on a survey that collected relatively detailed socioeconomic information
on respondents, but that did not report their wage rates. Hence the researchers predict wage rates
for each individual, based on each respondent's reported characteristics and wage data for similar
individuals from the Current Population Survey, using a hedonic model. Hedonic approaches are
statistical methods for imputing missing values from available data, and are often useful when
individual wage rates are unknown — either because wage data were not collected or because the
researcher is interested in estimating a wage (or a shadow price of time) for individuals not in the
labor force.
27Cesario, F.J., "Value of Time in Recreation Benefit Studies," Land Economics, Vol. 52, No. 1,
February 1976, pp. 32-41.
28Shaw and Feather have published a number of articles (cited throughout this section) that review
these and related studies as well as their implications for further research. See, for example, Shaw, W. D.,
"Searching for the Opportunity Cost of an Individual's Time," Land Economics, Vol. 68, No. 1, 1992, pp.
107-115.
29McConnell, K.E., and I. Strand, "Measuring the Cost of Time in Recreation Demand Analysis: An
Application to Sportfishing," American Journal of Agricultural Economics, Vol. 63, 1981, pp. 153-156.
30Smith, V. K., W. H. Desvousges, and M. P. McGivney, "The Opportunity Cost of Travel Time in
Recreation Demand Models," Land Economics, Vol. 59, No. 3, 1983, pp. 259-278.
2-11
-------
The Smith, Desvousges and McGivney study uses these hedonic results to estimate the value
of travel and on-site time for 43 water-based recreation sites. They experiment with different model
specifications that (1) include or exclude the value of on-site time, and (2) estimate the opportunity
cost of travel time based on the predicted wage or one-third of this amount. They find that including
the value of on-site time improves the model estimates for about half of the sites. Their tests of
alternative values for travel time indicate that neither the full wage rate nor one-third the wage rate
is "unambiguously superior to the other as an approach for approximating the opportunity cost" of
travel time. They note that these results may largely reflect the impact of missing data (e.g., on the
flexibility of work hours), and suggest that additional research is needed to better estimate these
values.
Other researchers subsequently examined the effects of work hour flexibility on the value
of leisure time, and found that these values vary depending on labor market status. For example,
in a 1987 study, Bockstael, Strand, and Hanemann examine the effects of fixed work hours on the
valuation of time.31 Their study is based on a survey of sportfishers in Southern California and
considers the value of both travel and on-site time. The researchers find that, for individuals with
flexible work hours, the average opportunity cost of time was about equal to their wage rate. In
contrast, for individuals with fixed work schedules, the opportunity cost of time was about 3.5 times
the wage rate.
A more recent (1999) study by Feather and Shaw further examines the impact of inflexible
work hours. The researchers first estimate the value of leisure time (referred to as the shadow wage)
for each individual based on work force status and other attributes, and then use a random utility
model to estimate the value of river recreation.32 They build on a theoretic model developed by
Heckman that considers whether an individual is underemployed (working fewer hours than desired)
or overemployed (working more hours than desired) due to the prevalence of jobs with fixed work
hours.33 Of the employed individuals (who include both hourly and salaried employees), roughly
one-third reported flexible hours, one-third were overemployed, and one-third were underemployed.
The researchers find that the shadow wage or opportunity cost of time exceeds the market wage in
cases where an individual is working more hours than he or she would prefer, but is less than the
market wage in cases where an individual is working fewer hours than desired. Where work hours
are flexible, the opportunity cost was reasonably close to the market wage rate.
31Bockstael, N. E., I. E. Strand, and W. M. Hanemann, "Time and the Recreational Demand Model,"
American Journal of Agricultural Economics, Vol. 69, 1987, pp. 293-202.
32The authors compare the results of this approach with models that estimate the value of time as a
fixed fraction of the wage rate or that apply a hedonic approach. Feather, P. and W.D. Shaw, "Estimating
the Cost of Leisure Time for Recreation Demand Models," Journal of Environmental Economics and
Management, Vol. 38, 1999, pp. 49-65; and, Feather, P. and W.D. Shaw, "The Demand for Leisure Time in
the Presence of Constrained Work Hours," Economic Inquiry, Vol. 38, No. 4, October 2000, pp. 651-661.
33Heckman, J., "Shadow Prices, Market Wages, and Labor Supply," Econometrica, Vol. 42, No. 4,
July 1975, pp. 679-694, as cited in Feather and Shaw (1999).
2-12
-------
The variation in individual valuation of travel costs is explored further using a somewhat
different approach in Englin and Shonkwiler's 1995 research on the value of boating, angling, and
swimming trips to freshwater recreation sites.34 To reflect the assumption that the different values
that individuals assign to travel costs are unobservable, they build an econometric model that
includes travel costs as a latent (unobserved) variable, and include data on various indicators of its
value. They find that time spent traveling is valued at approximately 40 percent of the wage rate.
These studies are concerned with particular types of time use, such as travel, rather than the
diverse types of activities potentially affected by illness. They suggest, however, that empirical
measures of the value of time can vary greatly depending on the modeling approach used and the
context within which it is studied, highlighting the complexities inherent in this endeavor. In each
of the studies cited above, the researchers acknowledge related difficulties and explore some of the
factors (such as the availability of flexible work hours) affecting estimation of these values. Despite
these difficulties, the development of these studies provides a number of insights — both theoretic
and practical, that are broadly relevant to the valuation of time losses. The following section
discusses these implications.
2.3 IMPLICATIONS FOR VALUING ILLNESS-RELATED TIME LOSSES
The empirical literature provides relatively little information on the dollar value of time
losses due to illness that can be used directly in regulatory analysis. This literature tends to focus
narrowly on particular types of time use (e.g., productivity or travel) or yields estimates (e.g., of the
overall value of risk reduction) that combine consideration of activity restrictions with consideration
of other impacts. This literature is, however, a rich source of information on the factors that
influence the value of time, that is useful in considering how to best value time losses in cases where
more complete measures of willingness to pay for risk reductions are not available. The pursuit of
the types of studies discussed earlier has led to a number of developments in the theoretical models
considered in time valuation that have wide-ranging implications.
As noted in Chapter 1, the practice of benefit-cost analysis is based in welfare economic
theory. The following sections review certain aspects of that theory that have particular relevance
to the valuation of time. This discussion is not intended to be a comprehensive review of the
extensive literature related to this topic; rather, it provides a brief overview of key considerations
as well as references for those interested in more information.
The first section discusses the complex role that time plays in individual welfare
maximization as both a "commodity" and a "resource." The second section describes the valuation
of leisure time based on individual decisions regarding paid work. Each section begins by
summarizing the key related tenets of the simple neoclassical economic model and then briefly
34Englin, Jeffrey, and J.S. Shonkwiler, "Modeling Recreation Demand in the Presence of
Unobservable Travel Costs: Toward a Travel Price Model," Journal of Environmental Economics and
Management, 1995, Vol. 29, No. 3, pp. 368-377.
2-13
-------
discusses the theoretical and empirical research on the effects of relaxing some of these assumptions.
The final section then introduces an approach for using compensation data to value the opportunity
costs of time use, based on the framework that results from the research summarized previously in
this chapter.
2.3.1 Time as a Commodity and as a Resource
The basic neoclassical economic model assumes that individuals will consume a mix of
goods and services that maximizes their utility (or sense of satisfaction or well-being), subject to
their income constraint. In the simplest form of this model, time does not enter the utility function
directly, although income clearly depends on the amount of time spent working (as well as on other
factors such as education and the availability of investment assets). In addition, the consumption
of goods and services also requires the use of time; e.g., to eat a meal, go to the movies, or
participate in other activities.
A number of researchers have explored the implications of incorporating time more directly
into this model, in particular, by considering both the "commodity" value and "resource" value of
time. The commodity value of time is associated with the level of utility, or pleasure, one gains
while participating in an activity (i.e., from consumption of goods or services). The resource value
of time is related to its scarcity — because the total number of hours per day is fixed, time saved in
one activity can be used to engage in other activities which may generate a higher or lower level of
utility.
One frequently cited early model that formalizes these relationships was developed by
DeSerpa, who expands the neoclassical model to include three essential features:
...(1) utility is a function not only of commodities but also of the time allocated to them; (2)
the individual's decision is subject to two resource constraints, a money constraint and a
time constraint; and (3) the decision to consume a specified amount of any commodity
requires that some minimum amount of time be allocated to it, but the individual may spend
more time in that activity if he so desires.35
Under this model, the difference between the commodity value (time in its current use) and the
scarcity or resource value (the value of time in its an alternative use — which may be higher or
lower) represents the value of time saved. A positive value of time saved suggests that it could be
devoted to an alternative activity of greater value to the individual.
DeSerpa notes that one implication of this model is the need to differentiate between time
spent in various types of activities. Economists normally define leisure as time spent on activities
35DeSerpa, A. C., "A Theory of the Economics of Time," The Economic Journal, Vol. 81, No. 324,
December 1971, pp. 828-846.
2-14
-------
other than work, but DeSerpa suggests that it may be more useful to consider a distinction made by
Tipping between leisure and intermediate goods.36 Leisure involves free choice in the consumption
of goods and services (such as recreation), while intermediate goods are activities (such as travel
to a recreational site) that make the consumption of this leisure possible. Variations on this
conceptual model underlie several of the approaches applied in the recreation literature examined
earlier in this chapter, as well as the related debate over the appropriate valuation of travel and on-
site time.37
This model also suggests that it is important to consider the choices of activities considered
by individuals in determining their opportunity costs of time. For example, as Cesario notes:
The value of time for an individual in a given situation is conditioned by what activities are
being traded off If the individual is trading off travel time for work time and there is no
marginal utility or disutility associated with work or travel, then there is some basis for
valuing travel time at the wage rate. However, it seems farfetched to assume that the
recreational tripmaker is trading off time for travel with time for work. It seems much more
likely that the trade off is between time for travel and time for leisure activities... The value
of travel time in a recreational tripmaking context thus reflects the value placed on
alternative uses of leisure time by the individual, for this is the relevant opportunity cost.
If we posit that travel per se carries with it a marginal utility or disutility, then it can be
shown that the value of saving travel time will diverge from the value of leisure time3*
This theoretical framework, as well as the results of empirical studies (such as the 1983
Smith, Desvousges and McGiverny study discussed earlier) suggests that the value of time is likely
to vary across individuals and activities, and that researchers need to carefully consider the context
for valuation. For example, illness may force an individual to involuntarily participate in activities
(such as bed rest or doctor visits) in lieu of his or her preferred normal activities. The implied trade
off is different than in the case of a individual choosing among of variety of recreational
opportunities.
2.3.2 The Labor-Leisure Trade Off
As noted earlier, the basic neoclassical economic model assumes that individuals will
allocate time between paid work and other activities up to the point where, at the margin, the value
of compensation received is equal to the value of uncompensated activities. The assumptions behind
36Tipping, David G., "Time Savings in Transport Studies," Economic Journal, December 1968, pp.
843-854, as cited in DeSerpa (1971).
37See, for example, Wilman (1980) for an early theoretical analysis of the implications of this
approach for recreation studies.
38Cesario (1976).
2-15
-------
this approach are summarized in the Gold et al. study of "best practices" for cost-effectiveness
analysis:
... Thefundamental assumption ofthis literature is that people will take their opportunity cost
into account when allocating their time, choosing to devote it to the activities that produce
the greatest utility. They will work an extra hour, for example, if the compensation they
receive exceeds the value they place on their time in other activities...
... The labor-leisure trade off, which is at the heart of the theory of labor supply, illustrates
the method used to value time which is not spent at work: if there is perfect competition; if
workers and employers are perfectly well informed; if the worker has declining marginal
utility of leisure time (i.e., the more time spent away from work, the lower the value of each
incremental increase in leisure time) and diminishing marginal utility of income; and if the
quantity of labor supplied in the market is continuously variable, then the worker
"consumes" leisure time up to the point at which the value of an additional hour of leisure
equals the (hourly) wage that he or she can receive by working.39
As indicated by this quote, the use of wage data to estimate opportunity costs is based on a number
of simplifying assumptions regarding the operations of labor markets and the process by which
individuals choose among different activities, many of which have been investigated in the
theoretical and empirical literature.
One key assumption that affects this trade-off is the ability to work as many, or as few, hours
as desired. The wage rate may over- or understate the value of leisure time when individuals do not
have complete flexibility in work hours. If this inflexibility causes them to work more hours than
desired (i.e., they would prefer to spend time, at the margin, in leisure rather than work), the
marginal value of leisure is likely to be greater than the marginal wage. If they are working less than
desired, the opposite is likely to be true. This assumption has been investigated in empirical work,
such as in the Bockstael, Strand, and Hanemann 1987 study and the Feather and Shaw 1999 study
discussed earlier. These studies find that the shadow price of time may be higher or lower than the
wage rate in cases where individuals do not have flexible work hours, but that time is valued close
to the wage rate when hours are flexible.
Building on the work of Bockstael, Strand, and Hanemann and others, Larson notes that the
degree of flexibility in work hours also depends on whether the perspective is short term or long
term.40 He describes the two stage budgeting model, where in the first stage the individual decides
how to allocate his or her time between work and nonwork activities, and in the second stage the
individual decides how to allocate the resulting nonwork time across different activities. As noted
39Gold, M.R., L.B. Russell, J.E. Siegel, and M. C. Weinstein (eds.), Cost-Effectiveness in Health and
Medicine, New York: Oxford University Press, 1996, p. 40.
40Larson, D.M., "Separability and the Shadow Value of Leisure Time," American Journal of
Agricultural Economics, Vol. 75, August 1993, pp. 572-577.
2-16
-------
in the prior section (in the discussion of Cesario's approach), recreation demand modeling generally
focuses on decisions in this second stage. However, over the long run, individuals may have more
flexibility in choosing the hours worked, if they can choose among different jobs offering varying
wages or work schedules. From this long run perspective, at the margin, the wage rate than may be
the appropriate measure of the value of leisure time.
These theoretical constructs and related empirical work suggest that the wage rate may
represent the opportunity costs of leisure time under short-term or long-term conditions where there
is flexibility in work schedules and income. It also provides some indication of the extent to which
such opportunity costs may diverge from the wage rate in cases where labor market choices are more
constrained than in the simple neoclassical model; for example, suggesting that wage rates will
understate the value of leisure time when work hours are inflexible and greater than desired.
The utility maximizing behavior that underlies the neoclassical model also suggests that
individuals will allocate their time across all nonwork activities so as to equalize the marginal value
of time spent in each activity, and that this marginal value will equal the marginal wage rate; i.e. the
amount that one would earn from working an additional hour. However, as Winston has noted, the
use of time is constrained by the time period considered — limited options exist for working at night
or during the weekend, golfing may be possible only during the day and in good weather, etcetera.41
This limited availability may lead to variations in the value of time over different periods. Valuing
time spent in discrete individual activities presents a number of practical challenges, however, and
has been infrequently attempted in the empirical literature. Many activities are bundled (e.g., one
may listen to the radio while vacuuming, or one must travel to go fishing), making it difficult to
value them separately.
In addition, it is often hard to separate marginal from average values. The standard
assumption of decreasing marginal utility suggests that the last, or marginal, unit will be valued less
than the preceding units. (In other words, an individual would value the second hour of leisure time
received less than the first hour, and the average value of an hour spent in that activity will exceed
the value of the final hour.) However, separating and defining activities in a manner that allows the
switch point (or marginal unit) to be identified can be quite complex. The advantage of focusing
on the overall labor-leisure trade off is that the presence or absence of wages can be used to define
the marginal unit.
The focus on marginal trade offs in the labor-leisure decision is similar to the focus of the
human capital model on marginal decisions by the employer, as discussed earlier in this chapter. The
human capital approach assumes that employers hire workers up to the point where, at the margin,
the value of the worker's output is equal to the value of his or her cost to the employer. In this case,
total compensation (pre-tax wages plus benefits) can be used to value marginal changes in output.
41Winston, G. C., "Activity Choice: A New Approach to Economic Behavior," Journal of Economic
Behavior and Organization, Vol 8, 1987, pp. 567-585.
2-17
-------
2.3.3 Conclusions
The discussion in this chapter suggests that there are no "off-the-shelf readily available
estimates of the value of illness-related time losses. It does suggest, however, that the neoclassical
economic model provides a framework for valuing the opportunity costs of time in a social welfare
context, arguing that compensation or wages can be used to value time at the margin. However,
theoretical and empirical investigation of the assumptions underlying this model suggests that there
are cases where the value of time will diverge from these rates, and argues for careful consideration
of related uncertainties. The following chapter discusses the use of compensation data to measure
the opportunity cost of time in more detail — considering paid work, unpaid work, leisure, and sleep
separately — as well as the limitations of this approach.
The preceding discussion also indicates the need to take into account the differences between
the context for regulatory analysis and the context underlying much of the available empirical
research. When valuing time losses due to illness, the analyst is concerned with a condition that may
force an individual to spend time in activities (such as bed rest or doctor's visits) other than their
preferred, normal activities. This differs from the situation generally modeled in the recreation or
transportation literature, where the analyst is concerned with the choices an individual makes in
determining how to "normally" spend his or her time; e.g., in selecting modes of transportation or
recreational opportunities.
In addition, in the context of regulatory analysis, analysts are not able to identify the specific
individuals who might be affected by the risk reductions nor are they likely to have detailed
information on the types of time use affected. The risk assessment may provide limited information
on the characteristics of affected individuals; e.g., the age range of individuals most likely to incur
the risks (e.g., the very young or very old). The analysis of contaminant occurrence may, in some
cases, suggest that certain geographic regions will be disproportionately affected. Whether the
analyst has detailed information on the extent to which the health effects of concern curtail different
types of activities or on the duration of these restrictions is likely to vary from illness to illness. The
analyst generally will need to estimate values based on the general characteristics of the potentially
affected population, rather than on detailed characterization of each potentially affected individual.
This contrasts sharply with the context for most of the transportation and recreation studies
described above, which survey identifiable individuals and value more discrete activities. Thus
while the recreation literature supports the notion that different activities will have different values
and that these values will vary across individuals, the regulatory analyst is often looking for values
that provide averages or expected values that cover a range of individuals and activities. The
following chapter discusses approaches for developing these values in more detail, and includes an
example of the calculation of the value of time losses in the context of regulatory benefit-cost
analysis.
2-18
-------
3.0 CALCULATING THE VALUE OF LOST TIME
The available empirical literature reviewed in the prior chapter provides little information
on the specific dollar values individuals place on avoiding the effects of illness on their ability to
pursue normal activities. It does, however, provide a framework for determining the value of time
in a social welfare context, based on the concept of opportunity costs. Neoclassical economic theory
includes two key related tenets: (1) employers hire workers up to the point where, at the margin,
the value of the worker's output is equal to the value of his or her cost to the employer, and (2)
individuals allocate time between paid work and other activities up to the point where, at the margin,
the value of compensation received is equal to the value of the uncompensated activities.
While these tenets are based on a number of simplifying assumptions regarding the
functioning of labor markets and the factors influencing individual choices, they provide a starting
point for valuing time losses. They suggest that compensation data can be used to estimate the value
of time (i.e., its opportunity cost) in its preferred, or normal, use. This value of time in its normal
use can then be compared to illness-related time use to estimate the value of the losses that occur
when an individual's activities are restricted by illness. Uncertainty analysis then can be applied to
assess the effects of simplifying assumptions on the resulting values.
This chapter extends the discussion of the value of time by breaking time use into four
categories: paid work, unpaid work, leisure, and sleep. The first section discusses different
approaches for valuing each type of time use based on the available theoretical and empirical
literature, suggests an approach for valuation in the context of regulatory analysis, and notes some
of the key limitations of the approach. The following section discusses the valuation of time for
individuals not currently employed (including children, the elderly, those seeking employment, and
those not in the labor market).
The next section then discusses how the extent of loss can be assessed, considering both
complete losses (e.g., where the individual is unable to participate in normal activities due to illness)
and partial losses (e.g., where the individual continues to participate in normal activities but finds
them less productive or enj oyable than usual due to illness). In other words, it considers the impact
of illness on the quantity of time spent on normal activities, as well as its effect on the quality of the
experience. In addition, it discusses the impacts of illness on caregivers. The chapter concludes
with an example of the calculation of the value of time lost due to illness, and describes the
limitations of the approach that can be explored through uncertainty analysis.
3.1 VALUING TYPES OF TIME USE
From a social welfare perspective, the value of time depends in part on the type of activities
in which an individual is engaged. The available valuation literature varies in the ways in which it
categories these activities. For example, much of the human capital research focuses on the
productivity of paid (market) and unpaid (nonmarket) work time, while recreational studies often
consider only the difference between the value of travel time and other nonwork or leisure time —
3-1
-------
defining leisure broadly to include any uncompensated activity. In reality, individuals may value
each specific activity differently. This variation may result from the need to engage in a less
desirable activity (e.g., travel) to participate in a highly desirable activity (e.g., outdoor recreation)
or because the choice of activities may be constrained at a particular point in time (e.g., by time of
day or weather conditions). However, a comprehensive system for estimating the variation in the
value of time across the full range of work and nonwork activities is not available. Hence this
chapter focuses on four maj or time use categories discussed in the available literature: market work,
nonmarket work, leisure, and sleep.
Market labor involves an explicit transaction between the buyers (employers) and sellers
(employees) of the labor to compensate the worker. Nonmarket labor includes all forms of
uncompensated activities that produce output for internal household consumption rather than for
external buyers. Housekeeping, childcare, auto repair, lawn care, and similar activities are examples
of these nonmarket labor activities. In addition, volunteer work and other productive, unpaid
activities undertaken outside the home constitute nonmarket activities. Nonmarket labor may be
pursued by students, retirees, homemakers, and others, as well as by individuals who spend part of
their time in the paid workforce. This chapter defines leisure as all time not spent working or
sleeping, hence all activities (other than sleep) not classified as market or nonmarket labor are
assigned to this category.1 These categories are summarized in Exhibit 3-1 below.
Exhibit 3-1
DEFINITIONS OF TIME USE CATEGORIES
Category
Definition
Market work time
Time that an individual spends engaged in productive activity in exchange for a salary
or wage at a price set in the labor market.
Nonmarket work time
Time spent on productive work without monetary compensation in the home (e.g.,
housework) or outside the home (e.g., volunteer work).
Leisure time
Time spent engaged in activities other than market work, nonmarket work, or sleep.
Sleep time
Time spent sleeping.
The rates (i.e., "dollar per hour" values) for these four categories are discussed below. For
each category, the discussion first describes the major alternatives most frequently used in the
available empirical literature. It then notes the approach that is likely to provide the best estimate
of the opportunity costs associated with each category of time use, based on currently available data.
Exhibit 3-2 below summarizes the base values suggested for each activity category; analysts will
often wish to explore the effects of alternative assumptions in sensitivity or probabilistic analysis
of uncertainty.
'This definition ignores the distinction between "pure" leisure (activities undertaken by free choice,
such as time spent on recreational activities) and intermediate activities (such as time spent on travel to
recreational areas) noted in the prior chapter because the data available on time use generally does not allow
the analyst to distinguish clearly between these two categories of leisure activities.
3-2
-------
Exhibit 3-2
OVERVIEW OF VALUATION APPROACH
Category
Valuation Approach
Market work time
Gross (pre-tax) wage plus benefits, reflecting both the opportunity costs to both the
individual (i.e., lost wages) and to the employer and society (i.e., lost product)
Nonmarket work time
Net (post-tax) wage, based on the opportunity costs to the individual
Leisure time
Net (post-tax) wage, based on the opportunity costs to the individual
Sleep time
Zero, due to lack of support for a specific dollar value and difficulties inherent in
estimating the net impact of illness on sleep time
The rationale for selecting these values is discussed in more detail in the following sections.
3.1.1 Market Work Time
In the case of market labor or paid work, the selection of a rate for valuing time losses is
relatively straightforward. The available data (discussed in Appendix B) support three choices for
estimating the value of lost work time.
1.	Total compensation: This approach includes all compensation, such as taxes, wages, and
benefits, and reflects the total costs of lost time from the perspective of the employer. The
national data on benefits are less extensive and detailed than the data on wages or earnings,
sometimes requiring the use of simplifying assumptions to implement this approach.
2.	Pre-tax wage rate: While this approach is an incomplete measure of compensation,
extensive earnings data are available, both in the aggregate and broken out by age, gender,
and occupation. Hence this approach is often applied in the empirical literature.
3.	Post-tax wage rate: This approach reflects the value of lost time from the perspective of the
individual. The national data on income taxes paid (including all Federal, state, and local
taxes) are less detailed than the data on pre-tax earnings, hence simplifying assumptions may
be needed to apply this approach.
Neoclassical economic theory suggests that total compensation is the preferred measure of
value of lost market work time. This approach is consistent with the human capital method for
valuing lost productivity, and is based on the assumption that the compensation provided by an
3-3
-------
employer equals the value of each worker's marginal output.2 In other words, an employer would
not pay an employee more, in salary plus benefits, than that employee is worth to the company (i.e.,
the value of the employee's marginal product) and hence to society.
In a social welfare context, the value of marginal changes in market work time is comprised
of two components: (1) the value of the time loss to that individual, and (2) any additional value to
the rest of society. Total compensation is the most representative measure of the full social welfare
impact of lost work time because it incorporates both the loss to the individual (in terms of lost
income) and the loss to society (in terms of reduced tax revenue and decreased production of goods
and services).
The total compensation approach recognizes that, when an individual misses work or is less
productive due to illness, he or she loses the utility (or sense of satisfaction or well-being) associated
with working. Under standard neoclassical economic assumptions (where the marginal utility of
paid work is equal to the marginal value of wages received), this loss is measured by the income
which the individual can trade for goods and services. The total compensation approach also
recognizes that the employer (and society) loses the value of the individual's productivity, and (again
based on standard assumptions) this value is equal to total compensation (pre-tax wages plus
benefits). From this perspective, the value of productivity (pre-tax wages and benefits) exceeds the
value of the income (post-tax wages) received by the individual. Some of the total value of work-
related time losses is thus reflected in the employee's take home pay and benefits and the remainder
accrues in terms of taxes paid — reflecting the value of product created above and beyond what is
reflected in the employee's post-tax wages.
In this context, when illness forces an individual to miss work, the reduction in his or her
compensation, discounted over the period of illness, can be used to estimate the value of lost work
time. In some cases, an illness only causes an individual to perform at a reduced level of efficiency,
rather than miss work altogether. In such cases, the wage rate can be fractionally adjusted to
account for the reduced output associated with the illness in question.
Embedded in this approach are a number of simplifying assumptions regarding the operations
of the labor market and the factors that influence individual choice, several of which were
introduced in Chapter 2 of this report. Uncertainties related to these assumptions and their
implications are discussed in detail in Section 3.5 below. In addition, the data sources used to
implement this approach have some shortcomings that are discussed in Section 3.5 and Appendix
B of this report.
2 A number of cost of illness studies use lost earnings to estimate the indirect cost of illness. For
example, the total compensation approach is used in Rice, D.P. and W. Max, The Cost of Smoking in
California, 1989, Sacramento, California: California State Department of Health Services, 1992; Buzby, J.
C., T. Roberts, C.T.J. Lin, and J.M. MacDonald, Bacterial Foodborne Illness: Medical Costs and
Productivity Losses, Economic Research Service, U.S. Department of Agriculture, Agricultural Economic
Report No. 741, August 1996; and Waitzman, N.J., R.M. Scheffler, and P.S. Romano, The Cost of Birth
Defects: Estimates of the Value of Prevention, Lanham: University Press of America, Incorporated, 1996.
3-4
-------
3.1.2 Nonmarket Work Time
The valuation of nonmarket work time varies depending on the goals of the analysis. Under
the human capital approach, time spent engaged in nonmarket labor activities is considered
productive due to the fact that activities such as childcare, cooking, and general home maintenance —
if not performed by a member of the household — could be performed by a professional in return for
compensation (i.e., as market labor). However, unlike individuals employed in the labor market,
those engaged in nonmarket labor activities are not compensated for their work. As a result, the
rationale for selecting a rate for valuing the time spent performing such activities is less
straightforward than for market labor.
Researchers applying the human capital method generally use one of two approaches for
deriving a rate to use in valuing the productivity of nonmarket labor.3 The first approach uses the
wage rate of domestic servants or housekeepers to estimate the value of household production.4
This approach may undervalue the true productivity of nonmarket labor due to the wide range of
activities undertaken in addition to housekeeping. A second approach uses a composite of the wage
rates paid for the diverse range of activities associated with nonmarket work, such as the rates paid
for cooks, childcare providers, gardeners and others.5 Under this latter approach, it is possible to
include the value of volunteer activities outside the home as well as the value of home-related
nonmarket work.6 Time use studies are applied to allocate nonmarket activities across different job
categories to develop this composite. The composite approach, although more complex to
implement, is likely to more accurately reflect the true productivity of this labor than the
3For a survey of studies valuing nonmarket labor, see: Goldschmidt-Clermont, L, Unpaid Work in
the Household, prepared for the International Labor Office, United Nations, 1982.
4A pioneering example of this approach is: Rice, D.P., "Estimating the Cost of Illness," Health
Economic Series No. 6, PHS Publication No. 947-6, U.S. Government Printing Office: Washington, DC, May
1966. A more recent example that uses data on the value of housekeeping services based on Current
Population Survey data from the Bureau of Labor statistics, is: Ray, N.F., M. Thamer, et al., "Economic
Consequences of Diabetes Mellitus in the U.S. in 1997," Diabetes Care, Vol. 21, No. 2, 1998, pp. 296-309.
5One early demonstration of this composite market value approach is: Cooper, B.S. and D.P. Rice,
"The Economic Cost of Illness Revisited," Social Security Bulletin, Vol. 39, No. 2, February 1976, pp. 21-36.
A more recent approach is provided in Hoffman, C., D. Rice, and H. Sung, "Persons with Chronic Conditions:
Their Prevalence and Costs," Journal of the American Medical Association, Vol. 276, No. 18, November
1996, pp. 1473-1479. This later study uses values based on research reported in Douglas, J., G. Kenny, and
T.R. Miller, "Which Estimates of Household Production Are Best?," Journal of Forensic Economics, Vol.
4, 1990, pp.25-45.
6See, for example, Trewin, D., Unpaid Work and the Australian Economy: 1997, Australian Bureau
of Statistics, October 2000; and Robb, R., M. Denton, A. Gafni, A. Joshi, J. Lian, C. Rosenthal and D,
Willison, "Valuation of Unpaid Help by Seniors in Canada: An Empirical Analysis," Canadian Journal on
Aging, Vol 18, No. 4, 1999, pp. 430 - 446.
3-5
-------
housekeeper approach.7
However, this focus on productivity ignores the utility that the individual gains from
choosing to engage in nonmarket, rather than market, work. Under standard neoclassical
assumptions, if (at the margin) the market wage exceeded the utility gained from nonmarket work,
an individual would join the paid work force. In electing to engage in nonmarket labor, an
individual is not necessarily forgoing a job involving similar activities. For example, a practicing
physician may elect to engage in nonmarket labor on a full-time basis following the birth of his or
her child. At a later date, this individual may choose to return to the labor market at a wage rate
comparable to that received before the decision to leave the labor market. For such an individual,
the opportunity cost of choosing not to participate in the labor market is clearly much higher than
estimates generated using the average wage rate for various types of domestic occupations.8 In
addition, an employed physician will choose to spend some of his or her time on nonmarket, rather
than market, work activities.
From the social welfare perspective, the post-tax wage an individual would receive for
market work represents the value, at the margin, of nonwork time. Under this framework,
individuals presumably choose to engage in nonmarket work themselves rather than hire a
replacement worker because (1) the utility they gain from performing nonmarket tasks is greater
than the utility they would gain from spending the time on paid work, and/or (2) the amount they
would earn for paid work is less than the cost of hiring a worker to perform their nonmarket tasks
(i.e., of hiring a replacement worker). Therefore, the individual must value a marginal hour of his
or her nonmarket work time at a rate at least equal to the marginal post-tax wage per hour he or she
could have earned in the job market.
In much of the economics literature that supports this opportunity cost perspective,
nonmarket work time is not clearly distinguished from other uses of uncompensated time. However,
Gronau and others have used a household production framework to explore the valuation of
nonmarket work in more detail.9 While Gronau notes that the line dividing nonmarket work and
leisure may be difficult to define precisely, he distinguishes between household production (i.e.,
nonmarket work, which could be performed by someone else for pay) and household consumption
7The replacement cost approach will understate the cost of using market labor if it ignores the
relatively high transaction costs related to hiring domestic workers, including the need to pay taxes and
maintain and submit related records.
8In addition, an individual who chooses to engage in nonmarket work because the cost of hiring a
replacement worker exceeds his or her market wage presumably values nonmarket work time at minimum
at his or her market wage rate.
9Gronau, R., "Home Production -A Survey," Handbook of Labor Economics, Vol. 1, (O. Ashenfelter
and R. Layard, eds.), New York: North-Holland, 1986.
3-6
-------
(i.e., leisure, which is almost impossible to enjoy vicariously).10 He reviews the arguments related
to using market (replacement) costs or opportunity costs (foregone wages) to value home
production, and concludes that market costs are likely to be an inappropriate measure because: (1)
the utility gained from home production may exceed the costs of a replacement worker, and (2) the
quality or quantity of the output of the nonmarket worker may be superior to that of a market
substitute.11 Thus the opportunity cost approach is more consistent with the social welfare model
of individual utility maximization, since it takes into account both the productivity and utility
associated with nonmarket labor.
This report suggests the use of post-tax wages for valuing nonmarket work time, consistent
with standard neoclassical assumptions, since this approach takes into account the opportunity costs
of decisions to engage in nonmarket work and the effects of this choice on individual utility.
(Approaches for estimating this value for individuals who do not engage in paid work are discussed
later in this chapter.) This value accounts for the wage rate an individual could command if he or
she elected to spend more time in the labor market (either by taking a j ob or by increasing paid work
hours) rather than engaging in nonmarket labor.
This approach is based on many of the same simplifying assumptions as the approach for
valuing lost market work time, several of which were introduced in Chapter 2 of this report. These
and other uncertainties related to the approach for valuing lost nonmarket work time are discussed
in Section 3.5 below. Concerns related to available data sources are also explored in Appendix B.
3.1.3 Leisure Time
The valuation of leisure time is the area where there is the greatest divergence between the
human capital approach and approaches that provide more complete consideration of impacts on
social welfare. Analysts who value lost time using the human capital approach rarely, if ever,
include a value for lost leisure time due to the emphasis of these studies on measuring productivity.
While leisure time may affect an individual's productivity, some researchers believe this impact is
at least partially captured in the valuation of market and nonmarket work.12 Hence from the human
10Gardening is an example of the problems related to defining particular activities as nonmarket work
or leisure; it is an important leisure activity for many individuals but one could hire gardener at a market
wage.
nGronau suggests that analysts use both rates to value output. However, this conclusion appears to
be based on the fact that (1) he is interested primarily in the value of output, not its utility, and (2) he notes
that job satisfaction is rarely considered in valuing paid work. Thus his arguments for using a replacement
cost approach are similar to those used in the human capital literature, and do not take into account the full
social welfare costs of related decisions.
12For example, in the extreme case, an individual who has no leisure time is likely to be less
productive than an individual with some leisure, due to the stress and exhaustion that could result.
3-7
-------
capital perspective, including a separate value for leisure runs the risk of some degree of double-
counting. Researchers using the human capital approach, however, note that the exclusion of the
value of leisure time leads them to understate the full social welfare impact of illness.13
In contrast, from the opportunity cost perspective, the theory behind estimating the value of
lost leisure time is the same as that used to value lost nonmarket work time — in both cases the
individual presumably chooses to engage in activities other than market work because, at the margin,
the utility gained from these other activities is greater than the utility gained from market work.
Hence under the opportunity cost approach, the value of leisure time can be estimated based on the
after-tax wage rate, assuming that this rate represents the income an individual forgoes in choosing
to engage in leisure rather than market work.14 This approach is consistent with the approach
applied in the recreational literature which suggests that desirable leisure time (i.e., on-site time)
should be valued at the after-tax wage rate.15
In addition to its consistency with standard neoclassical assumptions regarding the labor-
leisure trade-off and the concept of opportunity costs, this approach has a practical advantage. Using
the same rate for both nonmarket work and leisure circumvents difficulties related to clearly defining
different uncompensated activities as either nonmarket work or leisure, and then determining the
extent to which illness affects the time spent in each type of activity. A number of activities (such
as gardening) could be considered both nonmarket work and leisure.
Using the post-tax wage rate to value leisure time is subject to uncertainties related to the
operation of the labor market and its effects on wage rates, such as those discussed in the section on
valuing paid work time. In addition, as discussed in the section on valuing nonmarket time,
individuals may consider factors other than wage rates (such as non-reimbursed work-related costs
13For example, Ungar et al., in their analysis of asthma, use daily wages to calculate the value of lost
work time but do not address other categories of time losses. They note that their approach may mislead
policymakers regarding the real social costs of asthma by excluding consideration of the opportunity costs
of nonmarket work time and leisure, as well as the impact on caregivers. Ungar, W. J., P. C. Coytem and the
Pharmacy Medication Monitoring Board, "Measuring Productivity Loss Days in Asthma Patients," Health
Economics, Vol. 9, 2000, pp. 37-49.
14Examples of the valuation of lost nonwork time are provided in two studies that are summarized
in Appendix A: Harrington, W., A. J. Krupnick, and W. O. Spofford, Economics and Episodic Disease: The
Benefits of Preventing a Giardiasis Outbreak, Washington, DC: Resources forthe Future, 1991; and Kocagil,
P., N. Demarteau, A. Fisher, and J.S. Shortle, "The Value of Preventing Cryptosporidium Contamination,"
Risk: Health, Safety and Environment, Vol. 9, pp. 175-196, 1998.
15As discussed in Chapter 2, under a two-stage model of time allocation, the recreation valuation
literature subdivides nonwork time into necessary intermediate activities that may involve disutility (such as
travel) and hence may be valued at less than the wage rate, and "pure" leisure (such as recreational activities)
which are generally valued at post-tax wages. However, researchers assume that the total amount of nonwork
time (including both intermediate and pure leisure activities) is chosen so that its marginal value is equal to
post-tax wages.
3-8
-------
and foregone benefits) in determining whether to engage in paid work. The implications of these
limitations are discussed in more detail later in this chapter.
3.1.4 Sleep Time
Perhaps the most difficult type of time loss to value is the effect of illness on sleep. Similar
to lost leisure time, sleep time is normally not valued under the human capital approach because of
its lack of direct association with a "product." Because individuals generally perform work and
leisure activities at a decreased level of efficiency when receiving inadequate amounts of sleep,
under the human capital approach analysts often assume that the value of sleep manifests itself at
least in part in the productivity associated with other activities.
As in other applications of the human capital method, this approach ignores any utility that
the individual receives from sleep that may be above and beyond its productive value. The
neoclassical approach, which assumes that an individual will trade-off time in different activities so
as to equalize their value at the margin, implies that the value of changes in sleep time would be
equal to post-tax wages, consistent with the valuation of nonmarket and leisure time. However, the
fact that some minimum amount of sleep is a biological necessity limits the extent to which an
individual can in fact exchange sleep time for time spent in other activities.
Biddle and Hamermesh explore this and related issues, noting that relatively little attention
is paid to sleep in the economic literature that explores time allocation.16 This literature tends to
implicitly include sleep with other leisure activities or to focus solely on the allocation of waking
time. Biddle and Hamermesh review several studies of sleep time and note that, while some
minimum amount of sleep is needed, individuals also exercise choice in determining the amount of
sleep they achieve beyond this minimum. This amount may vary due to economic incentives and
other factors. For example, they examine the effects of wages on sleep time and conclude that
changes in the time spent in market work affect the time spent in sleep as well as in nonmarket work
and leisure, suggesting that sleep involves a choice between resting and engaging in other work or
nonwork activities.
Illness may involve counterbalancing effects on sleep — it can disrupt normal sleep patterns
but can also lead one to nap rather than engage in normal waking time activities. Assuming that
some minimal level of sleep is necessary, naps may compensate for any lost night sleep time
associated with illness. Greater than normal nap or night sleep time may also be necessary simply
to cope with the impacts of the illness. This need for sleep may in turn displace some of the time
the individual would otherwise spend engaged in leisure, nonmarket work, or market work activities.
Given this complexity, analysts are likely to find it difficult to determine the amount of sleep
16Biddle, J.E., and D.S. Hamermesh, "Sleep and the Allocation of Time," Journal of Political
Economy, Vol. 98, No. 5, 1990, pp. 922-943.
3-9
-------
time lost due to illness and its relationship to time spent in other types of activities. Empirical
estimates of the net effects of illness on sleep time are not likely to be available in many cases.
Hence this report suggests that analysts conservatively estimate the base value of sleep time as
"zero," until more research is completed. Sensitivity or probabilistic analysis of uncertainty may be
used to test the impacts of assigning a higher value.
3.2 VALUING TIME FOR INDIVIDUALS NOT ENGAGED IN PAID WORK
Several sub-groups of the population present specific valuation challenges because they
often do not engage in paid work and hence wage rate data may not be available to estimate of the
value of their time losses due to illness. These groups include the following.
•	Children: Children do not generally engage in paid work, and wage rate data is not
generally collected for those who do. The definition of "child" can vary, but the age cut-off
is usually at 16 or 18 years of age. The Bureau of Labor Statistics does not collect
employment data for individuals under the age of 16.
•	Elderly Individuals: As individuals age, they generally leave the work force. However,
retired persons may value their time differently than younger workers who are out of the
labor force, due to the availability of pensions and Social Security. These individuals may
in fact lose benefits if they increase their job-related income.
•	Unemployed Individuals: Individuals not currently engaged in market work, but actively
seeking employment, would prefer to earn a wage but are not currently doing so.
•	Individuals Not in the Labor Force: Individuals not currently engaged in market work, and
not seeking employment, presumably value their time at a rate higher than the earnings they
could gain from paid employment. However, at times they may be engaged in necessary
nonmarket work (such as child care) where the cost of hiring a replacement worker exceeds
the amount they could earn as a member of the labor force.
The concerns related to the appropriate valuation of time losses are discussed for each of these
groups below.
3.2.1 Children
Valuation of the effects of illness on children presents a number of difficult challenges
related to the fact that children are not "economic actors" who can answer valuation questions or
engage in market activities that reflect their own preferences. Under the human capital approach,
the primary issue concerning the valuation of children's time losses is that of double-counting.
Analysts assume that childhood time spent in school or other productive activities will be reflected
in future adult earnings. Hence rather than directly value time losses during childhood, they
3-10
-------
consider the effects of childhood illness, if any, on future earnings.17
This approach contrasts to the social welfare perspective, which ideally would include the
utility losses that accrue to children as well as adults. Children can be profoundly affected by illness
in ways that have little relationship to the effects captured under the human capital approach. These
effects may be direct, i.e., children may be unable to engage in normal activities because they are
ill; or they may be indirect, i.e., children may receive a lessor quantity or quality of care due to the
illness of their caregivers. The dollar values of these impacts are highly uncertain.
Freeman notes that risks to children may be valued from three different ethical or normative
perspectives.18 One option is to use children's own values (i.e., assume children's sovereignty) but
these values are not likely to be well-informed. Children also do not control the financial resources
necessary to make the trade-offs implicit in related decision-making. A second option is to use
parental values. While parents are often presumed to have an altruistic interest in the well-being of
their children, they may not always exercise good judgement in determining how to best ensure this
welfare. The third option is to consider the child as an adult. While consistent with basic economic
principles, this approach may be difficult to implement in practice since it involves predicting future
values for individuals who are now children.
Both the OMB and EPA guidance on regulatory analysis suggest that, in the absence of
better data, adult values may be applied to children. For example, the OMB guidance states: "[f]or
rules where health gains are expected among both children and adults and you decide to perform a
benefit-cost analysis, the monetary values for children should be at least as large as the values for
adults (for the same probabilities and outcomes) unless there is specific and compelling evidence
to suggest otherwise."19 EPA's Children's Health Valuation Handbook notes that both the parental
perspective and the child-as-adult perspective may be useful in valuing health risk reductions but,
for practical reasons, analysts are more likely to transfer adult values to children.20
In assessing time losses due to illness, analysts have several choices for implementing this
guidance. For example, the analyst could first calculate the value of time losses for the adult
population potentially affected by a regulation, then assess the values for children based on the
average adult value per statistical case of illness. For some health effects, it may be possible to
17An example of this approach is summarized in Appendix A; see: Waitzman, N. J., R.M. Scheffler,
and P.S. Romano, The Costs of Birth Defects, Lanham: University Press of America, Incorporated, 1996.
18Freeman, A.M. Ill, The Measurement of Environmental and Resource Values: Theory and Methods,
Second Edition, Washington, D.C.: Resources forthe Future, 2003, pp. 340-341.
19U.S. Office of Management and Budget, Regulatory Analysis (Circular A-4), September 2003, p.
30.
20U.S. Environmental Protection Agency, Children's Health Valuation Handbook, October 2003, p.
2-11.
3-11
-------
compare these adult values to values calculated using the human capital approach (i.e., that consider
effect of childhood health problems on adult earnings). However, as noted throughout this report,
the human capital method is likely to significantly understate the value of these time losses from a
social welfare perspective. As always, the limitations of the approach need to be carefully discussed
in presenting the results and addressed by uncertainty analysis as appropriate.21
3.2.2 Elderly, Unemployed, or Out of the Labor Force
The opportunity cost of time for individuals who do not engage in paid work is, conceptually,
the post-tax wage they would earn if they were employed — sometimes referred to as the shadow
wage. This approach is based on the assumption that these individuals do not participate in market
work because they value nonwork time at a rate that is higher than the wage they would earn if
employed. However, due to the lack of employment, this wage rate is not observable for these
individuals.
Researchers generally impute a wage in this case by comparing these individuals to similar
employed individuals. One such approach is the hedonic wage model, which is discussed in the
context of the recreation valuation literature in the prior chapter. This model predicts wages based
on the reported characteristics of each individual in a sample. In the context of regulatory analysis,
the average or median post-tax wage for the employed members of the affected population may be
used to estimate the shadow wage for those members of the population not engaged in market work.
These approaches are based on the assumption that wage earners and nonwage earners do not differ
in respects that would affect their wage rates. This simplifying assumption may under- or overstate
the actual shadow wage depending on the characteristics of the groups of concern.
For example, many elderly individuals are not employed in the labor force for reasons not
directly related to the wage rate. Several factors, such as forced retirement and Social Security,
distort their employment decisions so that an imputed marginal wage rate (or the actual wage earned
in any post-retirement employment) may not reflect their opportunity cost of not working.22 It is not
clear whether elderly individuals not engaged in market work are more likely to be similar to those
who are technically unemployed (i.e., would still like to work) or to those not in the labor force (i.e.,
not working by choice).
By definition, the unemployed (i.e., individuals not in the paid work force, but actively
seeking employment) would prefer to spend at least part of each day employed, rather than in
seeking employment or in nonmarket work and leisure activities. For these individuals, the offered
21EPA's Children's Health Valuation Handbook provides a detailed discussion of the uncertainties
introduced when adult values are transferred to children.
22For a brief discussion of the impacts of social security on the employment decisions of elderly
individuals, see: Gold, M.R., J.E. Siegel, L.B. Russell, and M.C. Weinstein (eds.), Cost-Effectiveness in
Health and Medicine, Oxford: Oxford University Press, 1996, p. 202.
3-12
-------
wage is presumably below the wage they would like to receive (i.e., their reservation wage)
suggesting that they value nonwork time more than the wage they could currently command in the
labor force. This assumption, however, ignores the complex set of factors that contribute to
unemployment. In general, wage rates for the employed population may provide an upper bound
estimate of the wage rates unemployed workers could command, given the potential differences in
the characteristics of each group.
In contrast, individuals not in the labor force include homemakers and other individuals
(including many elderly) who are not actively seeking employment. As discussed earlier, these
individuals presumably value the time spent in uncompensated activities at a rate greater than their
potential take-home (i.e., post-tax) market wage, and therefore have chosen not to engage in paid
labor. However, in some cases this decision may reflect constraints on individual choice. For
example, individuals may engage in necessary nonmarket work (such as childcare) because the full
cost of hiring someone else is greater than the wage they could earn in the job market. Another
example is a severely mentally or physically handicapped person who may be unable to engage in
market labor.
For individuals not engaged in paid work, the framework applied in this report suggests that
the shadow price (or opportunity cost) of time should be estimated based on potential post-tax
wages, and that this value should be used to estimate the social welfare effects of time losses due
to illness. The limitations of this approach for each of the groups addressed above can be discussed
qualitatively along with the results of any related quantitative analysis of uncertainty.
3.3 DETERMINING THE EXTENT OF TIME LOSSES
In addition to determining the value of time, analysts must estimate the amount of time lost
due to illness. Some of the illness-related loss may be relatively complete; i.e., individuals may be
forced to spend time in unpleasant activities that have little or no utility, rather than in their preferred
normal activities. Other illness-related losses may be partial; e.g, individuals may find that the time
spent in normal activities is less productive or pleasurable than in the absence of illness, or may need
to substitute less desirable activities for those that they prefer. These effects may be long term as
well as immediate. For example, illness may affect decisions to enter or leave the workforce, to
select particular types of jobs, or to work full time or part time — as well as whether to spend a
particular day resting rather than engaged in normal activities. Time losses are not limited to the ill
individual. For example, others may take time off from compensated or uncompensated work, or
lose leisure time, to care for the ill individual. Children who are dependent on the ill individual for
care may be affected as well. Approaches for estimating these impacts are discussed below.
3-13
-------
3.3.1 Types of Time Losses
Ideally, estimates of time losses should be developed by comparing time usage when well
to time usage when ill. These estimates of the amount of the loss (e.g., in terms of hours affected)
can then be multiplied by the dollar values (e.g., post-tax wages) to determine the total value of time
losses. This section provides a general description of the options for developing estimates of the
amount of time lost, while Appendix A describes examples of some of the specific methods applied
in the empirical literature. Appendix B discusses a number of data sources that can be used to
estimate the normal allocation of time across market work, nonmarket work, leisure, and sleep, as
well as sources of information on the effects of illness on this normal time use.
The approach used to estimate the amount of time losses will vary across different regulatory
benefit-cost analyses depending on the characteristics of the illness and the affected population, the
available data, the importance of the resulting estimates to the overall analytic conclusions, and the
time and resources available. Some of the options an analyst might consider include the following.
1.	First, he or she may use an existing empirical study of time losses that directly addresses the
types of health risks and the populations affected by the regulatory options of concern.
Because the match between the available research and the context of a particularly
regulatory analysis is almost always imperfect, in most cases the analyst will need to choose
an alternative approach.
2.	Second, he or she may transfer estimates from a research study that addresses the same or
similar health effects in a comparable, but not identical, population. For example, the
available literature may include a study of time losses due to myocardial infarction for
patients at a particular health maintenance organization. This study may be used as an
indicator of the value of time losses due to ischemic heart disease (for which myocardial
infarction is one of several symptoms) for a rulemaking with nationwide impacts. Factors
related to assessing the quality and suitability of studies for transfer, as well as issues related
to the limitations of the benefit transfer approach, are discussed in a number of other
documents and hence are not explored in detail in this report.23
3.	Third, the analyst may construct his or her own estimates based on survey data or other
information sources. In this case, analysts may need to consult a variety of data sources —
e.g., on the frequency of hospitalization, doctors visits, and other illness-related activities —
and compile the results to determine the total time losses. In the case of nonfatal chronic
illnesses that last for several years, the likelihood that an individual will survive the course
of the illness (i.e., not die of other causes) must also be considered.
23For example, information on conducting benefit transfers is provided in: U.S. Environmental
Protection Agency, Handbook for Non-cancer Health Effects Valuation, December 2000.
3-14
-------
In pursuing these options, the analyst may decide to rely on a single approach to estimate time
losses, or to use more than one approach and compare the results.
Under each of these options, the amount and types of time losses may be determined from
individual research studies or from large scale, national surveys. For example, the health economics
literature includes several studies of specific populations (e.g., at a particular health maintenance
organization or hospital) that may provide information on time allocation.24 Alternatively, many
studies rely on national self-reported data from the National Center for Health Statistics' National
Health Interview Survey.25 This survey provides data on lost time including the number of disability
days, doctor's visits, and hospitalizations. Estimates for individuals with the illness of concern can
be compared to those for individuals without the illness, to separate out the effects of other factors
that affect time use.
An alternative approach involves compiling data from different sources to develop composite
estimates. For example, one source can be used to estimate the number of doctor visits associated
with the illness, another can be used for hospitalization rates, etcetera. Several national databases
maintained by the National Center for Health Statistics and other agencies provide these types of
data (see Appendix B). Because these data sources generally focus on specific treatment-related
activities, they often lack information on time spent at home (i.e., bed rest days or days during which
activities are restricted) and may need to be supplemented by other sources.
In some cases, the data sources noted above will identify the activities from which time is
diverted; e.g., the number of hours lost from market work, nonmarket work, leisure, and (possibly)
sleep activities. In other cases, the data sources may only indicate the duration of the illness-related
activities. For example, information may be available on the number of days spent ill or on the
number of hours spent on doctor visits, but not on what the individual would have been doing during
these times in the absence of illness. In these cases, the analyst may need to compare this illness-
related time allocation to the allocation found in studies of typical time use (see Appendix B). For
example, if a study of typical time usage indicates that individuals spend three hours per day in paid
work (averaged across employed and unemployed individuals and across seven days per week), then
the analyst could assume that individuals too ill to engage in paid work lose an average of three
hours of paid work time per day of illness.
24For example, the Hartunian study summarized in Appendix A relies on a study by Weinblatt et al.
to example the functional status of heart attack survivors (Weinblatt, E., S. Shapiro, C.W. Frank, and R.V.
Sager, "Return to Work and Work Status following First Myocardial Infarction," American Journal of
Public Heath, Vol. 65, No. 2, 1966, pp. 169-185). See: Hartunian, N.S., C.N. Smart, and M.S. Thompson,
The Incidence and Economic Costs of Major Health Impairments: A Comparative Analysis of Cancer, Motor
Vehicle Injuries, Coronary Heart Disease, and Stroke, Lexington, Massachusetts: D.C. Heath and Company,
1981.
25There are numerous examples of the use of the National Health Interview Survey to estimate time
losses, including Ray et al. (1998) and Waitzman et al. (1996), which are described in Appendix A.
3-15
-------
These comparisons raise difficult questions regarding whether the time loss is complete or
partial. In some cases, illness may force an individual to substitute a less desirable activity (e.g.,
reading in bed) for a preferred activity (e.g., paid work or outdoor recreation). In this case, the
substitute activity presumably has some utility that is less than the utility of the preferred activity.
This partial loss could be reflected by assigning a value to the substitute activity that is less than the
wage rate (i.e., the value of a complete loss) but greater than zero (the value of continued
participation in normal activities). Another example of a partial loss is when an individual continues
to engage in normal activities when ill, but is less productive or finds them less enjoyable than usual.
These losses can be represented by prorating the value of time; for example, the loss associated with
a day reported as 30 percent less productive would be 30 percent of the wage rate. In other cases,
the loss may be complete. For example, an individual is likely to assign little, if any, utility to time
spent suffering from a high fever, vomiting, or coping with diarrhea. In such cases, the time spent
ill may have zero value compared to the normal use of time, and the loss would be represented by
100 percent of the wage rate.
3.3.2 Acute vs. Chronic Illness
The approach used to assess time losses depends in part on the duration of the illness. In the
case of acute illnesses, such as nonfatal cases of giardiasis or cryptosporidiosis, the effects generally
do not linger beyond the year of incidence and often may cover only a few days or weeks.
Individuals suffering from such illnesses (as well as family or friends acting as unpaid caregivers)
generally are able to return to their full schedule of normal work and leisure activities within the
year of incidence and experience no permanent or long-term effects. In these cases, there is no need
to adjust the calculations for year-to-year changes, e.g., in employment status or survival rates, nor
is there a need for discounting to reflect the time value of money.
The time lost to these types of acute illnesses is sometimes characterized by defining day
scenarios.26 Such scenarios reflect the effect of a given illness on the daily allocation of time to
labor and leisure activities. These days can be defined by the degree and type of time loss, including
bed-rest days, work-loss days, and restricted activity days. For example, bed days may be defined
as days an individual is unable to perform any type of normal activity, resulting in a relatively
complete loss of the value of labor and leisure time. Low productivity days could be used to
characterize days when an individual can participate in work at a reduced rate or for a reduced
period of time, reducing the value of output for a given individual by a marginal amount, depending
on the symptoms and severity of illness. During these time periods, individuals may also experience
a loss in utility when engaged in leisure activities, as the illness reduces their ability to enjoy these
activities or pursue them with the same degree of efficiency as normal. In some cases, the effects
of the illness will vary by day; e.g., a few days of severe symptoms may be followed by several days
of milder impacts. Time losses can then be computed for each day scenario then summed across the
days.
26See, for example, Hartunian (1981).
3-16
-------
More detailed survey data may be available to aid in defining these scenarios for some
illnesses. For example, Harrington et al. collected information on time losses associated with
giardiasis, including data on work loss days for the ill individual and his or her caregiver.27 They
also collected information on the extent to which the ill individual's productivity decreased while
at work during the course of the illness. These examples suggest that the availability and nature of
the data characterizing the particular health effect of concern will determine the degree of detail that
can be applied using the day scenario approach.
For chronic illnesses, that may last for many years, estimating time losses is more complex.
The ill individual may experience time losses associated with both labor and leisure activities
throughout a significant portion of the remainder of his or her life. In some cases, the effects may
be permanent and prevent return to a normal schedule of work and leisure activities. Even in cases
where the individual is able to return to his or her former schedule of activities, he or she may not
be capable of performing them at the same level of efficiency or of deriving the same level of utility
from them.
For long term illnesses, estimating the number of years an individual lives with the illness
generally requires consideration of life expectancy. For nonfatal cases, this life expectancy is
defined based on when the individual is likely to die from causes other than the illness of concern.
This life expectancy may be shorter for ill individuals than for the general population because the
illness itself may contribute to premature mortality from other causes. For example, diabetics have
a higher rate of heart and renal disease than the general population, and hence may die prematurely
from these associated conditions rather than directly from diabetes.28
The life span of ill individuals should be compared to the life span of similar healthy
individuals in calculating the value of time losses.29 A simple approach would involve considering
the average age at death for a person with the illness who dies of other causes, and comparing it to
the average age at death for the general population. Times losses would then be calculated over the
duration of the illness (if not life-long) or over the remaining life expectancy (if "incurable"). Data
related to developing a more complex approach, which address survival rates for each year of age
with and without the illness, are provided in Appendix B.
27Harrington, W., A. J. Krupnick, and W.O. Spofford, "The Economic Losses of a Waterborne
Disease Outbreak," Journal of Urban Economics, Vol. 25, No. 1, 1989, pp. 116-137.
28Addressing these issues requires clear communication between economists and risk assessors to
ensure that both parts of the analyses apply the same definition of a fatal case. A case of diabetes that leads
to death from an associated cause (such as renal failure) may be counted as either a fatal or nonfatal case of
diabetes, depending on whether the risk assessment considers these types of associated conditions or focuses
solely on deaths attributable directly to diabetes.
29 An early example of this approach, which could be refined based on the valuation methods and data
sources discussed in this report, is provided in Hartunian (1981).
3-17
-------
3.3.3 Caregivers
Caregiver time losses may fall into the same four categories as discussed earlier for ill
individuals.30 A caregiver may take time from paid work, reduce the time spent on nonmarket work,
or lose leisure or sleep time to care for sick family members or friends.31 While caregiver losses are
only occasionally quantified in human capital studies (due to data limitations and other factors), the
importance of including these losses is often noted.32 In its Children's Health Valuation Handbook,
EPA states that the inclusion of caregiver losses is of particular importance when children are
affected.33
Ideally, analysts would rely on data that specifically describes the duration and types of time
losses for caregivers. In the absence of empirical data, analysts will need to develop a method to
estimate these losses on a case-by-case basis. For some health effects, it may be possible to estimate
caregiver losses in proportion to the time lost by the ill individual. For example, a parent who takes
time off from work to care for an ill child at home may lose an amount of time equal to the time loss
of the child. In contrast, a woman who takes time off to transport her sick husband to the doctor,
but not to care for him while he is resting at home, will accrue only a portion of the time loss of the
ill individual. In this case, the losses could be estimated from data on the number and length of
doctor visits (including transport time) for the illness of concern.
As when evaluating the lost time of ill individuals, the use of wage data to estimate the value
of lost caregiver time may under- or overstate the actual value of the losses. This is particularly true
if caregivers are able to perform some of their regular activities while caring for the ill individual,
such as working from home, completing household chores, or engaging in leisure activities. Again,
it will be important to discuss the limitations of the chosen valuation approach and its implications
30This section considers only unpaid caregivers such as friends and family. Paid caregivers, such as
doctors and nurses, are covered in the medical cost component of the analysis and hence are not considered
in the analysis of time losses.
31Caregivers may also become ill as a result of their caregiving activities. Communicable illnesses
can be transmitted from the patient to the caregiver. In addition, caregivers may suffer from emotional
distress or depression as a result of their responsibilities. The extent to which these effects are addressed
(quantitatively or quality) will depend on the availability of related data as well as the likelihood of such
effects given the illnesses addressed.
32An example of the inclusion of caregiver losses is provided in Buzby et. al (1996) which is
summarized in Appendix A. Other studies discuss the need to include these losses (especially in cases where
children are affected) such as: Rice, T.D., A. K. Duggan, and C. DeAngelis, "Cost-Effectiveness of
Erythromycin versus Mupirocin for the Treatment of Impetigo in Children," Pediatrics, Vol. 89, No. 2,1992,
pp. 210-214. Inclusion of caregiver losses is also suggested in health economics texts, such as Dravove, D.,
"Measuring Costs," Valuing Health Care, (F.A. Sloan, ed.), Cambridge, England: Cambridge University
Press, 1995, p. 74.
33U.S. Environmental Protection Agency (October 2003), p. 2-16.
3-18
-------
when presenting the results of the analysis.
3.4 EXAMPLE OF CALCULATIONS
The previous sections discuss the process for estimating a dollar value per unit of time and
for estimating the amount of time lost or degraded by illness. This section provides an example of
an approach for deriving specific numerical values. While the actual approach will depend on the
data and context for a particular analysis, this illustration provides some general information how
the available data sources can be used in regulatory benefit-cost analysis.
Ideally, estimates of wages and total compensation and of the quantity of time lost would be
developed for the specific population affected by the rulemaking.34 However, it is generally not
possible to identify this population in much detail. The risk analysis will provide an estimate of the
number of statistical cases averted over time under each regulatory option, and may provide some
information on distribution of these risk reductions by age and/or gender. In addition, if the
regulation addresses contaminants that are prevalent only in a few geographic regions, certain parts
of the country may be disproportionately affected. In cases where these types of information are
available, the analyst will have some ability to tailor the valuation of time losses to fit the
characteristics of the affected population. In other cases, even if some identifying features are
known (e.g., that individuals with suppressed immune systems are more likely to contract the
disease), data on wage rates and time losses may not be available for the particular group affected.
Most often, the effects of regulations are dispersed across the nation and across different
demographic groups to such an extent that national data will provide the most appropriate estimates.
If data are available on the specific population affected by the rulemaking, average values
for this population will provide the best estimates of the central tendency or expected value.
However, when using national data to estimate wages and compensation for the small proportion
of this population that is affected by a rulemaking, analysts may wish to use median rather than
average values if available.35 The distribution of income in the U.S. is highly skewed due to the
small number of people who are extremely highly compensated, hence the average is significantly
higher than the median.36 Using the median is consistent with the notion that the small fraction of
34For consistency, the approach for valuing time use should reflect the same population as the
approach for valuing the medical cost of illness. For example, if regional values are used, both compensation
and medical costs will reflect regional economic conditions ~ such as whether wage rates and the cost of
living (including medical costs) are lower than national averages.
35See, for example: Abt Associates, Final Heavy Duty Engine/Diesel Fuel Rule: Air Quality
Estimation, Selected Health and Welfare Effects Methods, and Benefits Results, prepared for the U.S.
Environmental Protection Agency, December 2001, p. 3-15.
36In some cases, a regulation may disproportionately affect a group whose median earnings are less
than the U.S. average (e.g., the elderly, minorities, women). Using the earnings rate specific to this group
3-19
-------
the U.S. population affected by most rulemakings are likely to be better reflected by the median
(which is in the center of the income distribution) than by the mean (which is closer to the upper tail
of the distribution).
Often, these calculations will involve developing a profile of a "typical" individual affected
by the regulatory option. In other words, if the regulation averts 100 nonfatal statistical cases of the
disease throughout the United States, then the goal is to develop a value that can be multiplied by
100 to estimate the total national time losses averted throughout this population. Because the
specific individuals affected by the illness are not known, the value would be a weighted average
of the characteristics of the affected population. For example, if the population is 45 percent male
and 55 percent female, this profile would weight the values for males and females accordingly to
develop a composite (or weighted average) value. As described above, nationwide per capita values
can be used to represent this composite individual in most cases, since risk reductions are often
spread across a large population of unidentified individuals for whom the expected value of wages
and compensation is best approximated by the national median.
For an individual in either well or ill status, the total value of time reflects the value of time
spent in different activities.
Value of time (total) =
Value of market work time + Value of nonmarket work time
+ Value of leisure time + Value of sleep time
The total value of time is equivalent to the dollar value per unit of time (i.e., the measures
of compensation or wages discussed earlier) times the amount of time spent in each activity, which
then can be summed across the different types of time valued using the equation above. For
example:
Value of market work time = Time spent on market work * Market work rate
Value of nonmarket time = Time spent on nonmarket work * Nonmarket work rate
Value of leisure time = Time spent on leisure * Leisure rate
Value of sleep time = Time spent on sleep * Sleep rate
To determine the value of time losses, the value of time use without illness is compared to
time use with the illness for a typical individual. In other words:
Value of time losses =
Value of time in the absence of illness - Value of time with illness
as a benefit measure is consistent with the focus of welfare economics on individual preferences and
economic efficiency. However, this approach may raise concerns about equity, since it implies that society
values benefits to these groups at a lower rate than benefits to other groups. In these cases, analysts may wish
to explicitly recognize this concern in presenting the analysis, and possibly use the U.S. median (rather than
rates specific to the age, sex, or minority group of concern) to address equity concerns.
3-20
-------
An example of these calculations, for a representative individual experiencing a single day
of illness, is provided below. This calculation conservatively assumes a zero value for lost sleep
time, since (as discussed earlier) it is generally difficult to determine the amount of sleep time lost
and to assign an appropriate dollar value. In addition, this example combines nonmarket work and
leisure time in the calculations, since (as also discussed earlier) the same dollar per hour value is
used to value losses in both categories. The basic equation for calculating time losses in this
example is presented below, and compares the value of work and nonwork time (excluding sleep)
when well to the value when ill.
Value of lost time per day = ([Wwdl - Wm]- C) + ([Nwell - Nm]- E)
(Equation 3-1)
where...
WweU = work time when well; the number of hours that an individual without the
illness will be employed in the labor market
Wm = work time when ill; the number of hours that an individual with the illness
will be employed in the labor market
C = compensation (pre-tax wages and benefits); the hourly value of market work
time
Nwell = nonwork time when well; the number of hours that an individual without the
illness will be engaged in nonwork (uncompensated) activities, including
nonmarket labor and leisure but excluding sleep
Nin = nonwork time when ill; the number of hours that an individual with the
illness will be engaged in nonwork (uncompensated) activities, including
nonmarket labor and leisure but excluding sleep
E = earnings (post-tax wages); the hourly value of nonwork (uncompensated)
time including nonmarket work and leisure
The numerical values for these variables could be derived from a number of sources. For
this example, the values for total compensation and post-tax wages are estimated from national data,
discussed in detail in Appendix B.37 The starting point for the development of these estimates is
median weekly earnings for the year 2001 for full time workers ($597 per week) as reported by BLS
(replicated in Appendix B, Exhibit B-3).38 This value is derived from the Current Population Survey
37The example focuses on a typical or composite individual, but these values could be calculated
separately for each gender and/or age group if data on the amount and type of time losses are available at this
disaggregate level. In the latter case, the results for each gender or age group would need to be weighted to
reflect the characteristics of the affected population (i.e., the overall proportion of the population of each age
and sex) in determining the national impacts.
38This approach assumes that the hourly earnings for part-time workers will be the same as for full-
time workers; analysts may wish to explore the extent to which this assumption introduces uncertainty into
3-21
-------
and includes wages and salaries but not other costs (e.g., benefits) paid by the employer. The next
step is conversion of this value to earnings per hour. Individuals usually working full time averaged
42.9 hours per week at work in 2001, again according to Current Population Survey data reported
by BLS.39 This means that median earnings per hour averaged $13.92 ($597/42.9).
For market work time, the measure of opportunity costs used in this example is total pre-tax
compensation from the perspective of the employer. The earnings number reported above does not
reflect employer paid benefits. To adjust this estimate upwards to reflect total compensation, this
example uses the average ratio of wages and salaries to total compensation as reported by the Bureau
of Labor Statistics for private industry workers for 2001. These data show that total compensation
per hour averages 1.393 times wages and salaries for full time workers ($23.55/$ 16.91 per hour).40'41
Using this factor to adjust median hourly earnings (as reported above) leads to an estimate of $19.39
per hour for total compensation.
For nonwork time (excluding sleep), the measure of opportunity costs used is post-tax
earnings; i.e., the "take home" pay of the median individual. This example relies on Current
Population Survey data on household income before and after taxes to determine the percent of
earnings paid as taxes.42 In 2000, the median before tax income was $42,151 and median after tax
income was $34,667.43 Thus after tax income was 82.2 percent of the pre-tax amount. Applying
this factor to median hourly earnings leads to estimated after tax earnings of $11.44 per hour.
The results of these calculations are reported in Exhibit 3-3 below.
the results.
39 U.S. Bureau of Labor Statistics, "Table 19: Persons at work in agricultural and nonagricultural
industries by hours of work," Employment and Earnings, undated (http://stats.bls.gov/cps/cpsaatl9.pdf, as
viewed September 2002).
40U.S. Census Bureau, "Table No. 626: Employer Costs for Employee Compensation Per Hour
Worked: 2001," Statistical Abstract of the United States: 2001, Washington, D.C.: U.S. Government Printing
Office, November 2001, p. 406.
41The earnings per hour estimates vary across data sources depending on sample characteristics,
whether mean or median values are reported, and other factors.
42U.S. Census Bureau, "Table No. RDI-1, Household Income Before and After Taxes : 1979-2000",
March 2002 (http:www.census.gov/hhes/income/histinc/rdi01.html as viewed September 2002).
43This median income estimate differs from the earnings estimates cited earlier because it reflects
household income rather than individual earnings and relies on a different data source. It also represents data
from the year 2000, because year 2001 data on pre- and post-tax earnings are not reported.
3-22
-------


Exhibit 3-3


DOLLAR PER HOUR VALUES USED IN EXAMPLE CALCULATION
Variable
Time Loss Category
Basis for Estimate of Value
Dollar Value
C
Market work time
Median gross (pre-tax) wage plus benefits.
$19.39 per hour
E
Nonwork (uncompensated) time
(nonmarket work and leisure)
Median post-tax wage
$11.44 per hour
Notes:
See text for derivations of values.
Sleep time losses are conservatively valued at zero.
As discussed earlier, the source of estimates for the value of lost time will vary depending
on the illness of concern and other considerations. For simplicity, this example assumes that the ill
individual loses a full day of normal activities and is unable to engage in any sort of desirable leisure
activities. Based on the data on normal time use provided in Exhibit B-l of Appendix B, this loss
will include 3.3 hours of work time and 12.6 hours of nonworktime (excluding sleep) for the typical
individual (averaged across employed and unemployed individuals and seven days per week). In
other words, in this example, zero time will be spent in work or nonwork activities while ill.
Equation 3-1 can be rewritten as follows using the above data
Value of lost time per day = ([Wwell - Wm]- C) + ([Nwell - Nm]- E)
(Equation 3-1)
= ([3.3-0] • 19.39)+ ([12.6-0] • 11.44) = $208.13
If the illness lasts for more than one day and the time loss is constant across days, then this
value can be multiplied by the number of days of illness to determine the value of the time losses
per case. (Adjustments will be needed if the time losses vary from day to day during the course of
the illness.) This value then can be multiplied by the number of statistical cases averted by the
regulation to determine the total value of the avoided time losses throughout the affected population.
This calculation will become more complicated in cases where the losses vary over the time period
of the illness, or where data limitations require the development of additional assumptions to
complete the analysis.
The above example assumes that the loss is complete; i.e., that the affected individuals do
not engage in any normal activities. This approach can be adjusted to estimate the value of partial
losses; e.g., in cases where individuals continue to engage in normal activities while ill, but are less
productive or find the activity less enjoyable. For example, if illness reduces productivity or
enj oyment by 3 0 percent, then 3 0 percent of the hourly rate (represented by C and E in the example)
can be used to estimate the value of the losses. This approach can also be applied when an ill
individual is forced to substitute a less preferred, but somewhat pleasant, activity (e.g., reading) for
3-23
-------
a preferred activity (e.g., outdoor recreation).
Calculation of the impacts of nonfatal chronic illnesses require consideration of a number
of other factors. For example, chronic illness may require consideration of changes in status (e.g.,
employment and survival rates) from year-to-year, as well as of the time value of money (i.e., using
discounting to estimate the present value of future losses). As in the example above, the calculation
of time losses for nonfatal chronic illness compares an ill individual to a similar well individual to
determine the net loss over the duration of the illness. In this case, however, the duration is
determined by the interaction of two factors: (1) the likely duration of the health effect (i.e., the
number of days or years with the illness), and (2) the likelihood that an individual will survive (i.e.,
will not die from other causes) throughout the period of the illness.44 In cases where the time loss
is expected to accrue over a multi-year period, the analyst will need to discount the results so as to
estimate the present value of the time losses in the base year, using the discount rates applied
elsewhere in the analysis.45 These factors will need to be considered in assessing time losses for
unpaid caregivers as well as for the ill individual.
The assessment of chronic illness requires more data than the assessment of acute illness,
examples of which are provided in Appendix B and elsewhere. For valuation, in addition to data
on total compensation and post-tax wages (C and E in Equation 3-1 above), data are needed on the
expected change in these earnings over time, the variation in the fraction of time spent in market
work and other activities, and the discount rate. To estimate the amount of time lost, data are needed
on age at onset, the duration of the illness, and survival probabilities, as well as on the effect of the
illness on normal work and nonwork activities over time.46
Application of this approach is relatively straightforward for illnesses that strike in
adulthood, but may be more difficult for illnesses that begin in childhood.47 As discussed earlier in
44As noted earlier, nonfatal cases can contribute to premature mortality if the definition of fatal cases
(based on the risk assessment) includes only deaths directly attributable to the illness. For example, fatal
cases may include only deaths directly attributable to diabetes, whereas nonfatal cases may include deaths
from associated conditions (e.g., premature mortality from diabetes-related ischemic heart disease).
45A detailed explanation of the concept of discounting, as well as the rates recommended for use in
EPA analyses of social costs, is provided in: U.S. Environmental Protection Agency, Guidelines for
Preparing Economic Analyses, September 2000, EPA 240-00-003, Chapter 6.
46 Because EPA expects to use this approach only in the valuation of nonfatal illnesses, the "with
illness" survival probabilities should be based on data for individuals who survive the disease (i.e., die of
other causes).
47All values should be real values (net of inflation) indexed to a common base year that is used
throughout the regulatory analysis. For example, for an analysis conducted in 2004, all cost and benefit
estimates may be based on year 2003 dollars since that would be the most recent year for which inflation data
are available. In cases where the age at onset is after this base year, an annual productivity factor would be
used to adjust the compensation and wage rate values from the base year to the year at onset.
3-24
-------
this chapter, adult values may be used when calculating time losses during childhood years in the
absence of better data.48 In general, calculations will end at age 85, because most life tables assign
a value of zero to the probability of a member of the general population surviving beyond that age.
The market work time proportions can be adjusted to reflect employment rates at each age; the
amount of time spent in paid work drops in the years after age 65 as many members of the
population retire. In some cases, there may be a need to adjust the rates used for total compensation
and post-tax wages to reflect the fact that individuals with the illness, while continuing to engage
in paid work activities, are no longer able to be employed in occupations with earnings equivalent
to those they received when well.
As noted previously, these approaches for valuing time losses for acute and chronic illnesses
can be applied to caregivers as well as to ill individuals, and limitations related to applying this
approach to children, the unemployed, and those out of the labor market should be discussed in
presenting the results. For simplicity, the example does not include assessment of uncertainty or
variability, which can be addressed quantitatively through sensitivity analysis of the effects of
changes in key assumptions or through probabilistic analysis.
Finally, the example uses a combination of hypothetical assumptions and real data for
demonstration purposes. EPA regulatory analysts should use the most current sources of data rather
than the information in the examples. Many of these data are periodically updated and will need to
be tailored for the health effects and populations considered in each individual analysis.
3.5 LIMITATIONS AND ASSESSMENT OF UNCERTAINTY
Analysts may be interested in characterizing three types of uncertainty related to this
approach: uncertainty in the relationship of the resulting estimates to the preferred measure of
willingness to pay for risk reductions, uncertainty in the estimates of the amount and types of time
losses due to illness, and uncertainty in the unit dollar values assigned to these losses. The effects
of these limitations should be clearly stated in the analysis, and can be explored through sensitivity
analysis or probabilistic analysis using lower or higher rates, applying methods that are discussed
in detail elsewhere.49
Relationship to willingness to pay: As discussed in Chapter 1, estimates that add the direct
medical costs of illness to the indirect costs associated with lost work time are generally expected
to understate willingness to pay in most cases for a variety of reasons. A major contributing factor
is the lack of consideration of the value of averting the pain and suffering associated with illness,
which is manifested at least in part in restrictions on nonwork activities. The approach discussed
in this report addresses this deficiency by proposing an approach for valuing losses in nonwork time.
48Data on earnings and employment rates are generally not available for individuals 16 years and
younger.
49See, for example, U.S. Environmental Protection Agency (September 2000).
3-25
-------
It is conceivable that this enhanced approach will still understate true willingness to pay; e.g.,
because it does not address the value of avoiding pain and suffering beyond its impact on normal
activities. However, the extent of understatement is uncertain and difficult to quantify.
Estimates of time use with and without illness: The uncertainty in the estimates of the
amounts and types of time losses associated with illness will depend on the approaches and data
sources used in the particular analysis. Because these approaches are likely to vary greatly, it is
difficult to generalize about the potential magnitude or direction of any bias or to develop any
specific suggestions for addressing it quantitatively.
In many cases, data on time use with illness will be more uncertain that estimates of time use
in the absence of illness. As discussed in Appendix B, a number of surveys collect data on normal
time use. While different sources do not yield identical estimates, the estimates are relatively
similar. For example, estimates of average sleep time appear to be in the range of 8 hours per night,
and estimates of paid work time suggest that full time employees in the U.S. tend to work slightly
more than 40 hours per week.
Information on the effect of illness on these normal activities is less prevalent and more
uncertain, in part because of the vast variety of possible health effects and the diverse ways in which
individuals respond to their symptoms. While data may be available on the likely duration of illness
(e.g., the average number of days that elapse between when symptoms appear and disappear, or less
ideally, between diagnosis and "cure"), information on the change in activities over this time period
may be relatively general and incomplete. In particular, it may be difficult to differentiate between
periods when the time loss is complete (i.e., the illness prevents any participation in enjoyable
activities) and when the time loss is partial. Partial losses may occur when a less preferable activity
(e.g., watching TV in bed) is substituted for a preferred activity (e.g., paid work or outdoor
recreation), or when an individual continues to engage in normal activities but is less productive or
finds them less enjoyable than usual. Hence in many cases the analyst may wish to test the
assumptions regarding the degree of loss using sensitivity analysis or other approaches. Regardless,
it is important that analysts clearly discuss the limitations of the data and assumptions.
Estimates of dollar values per unit time: Uncertainty in the dollar values per unit of time
may stem from two sources: (1) uncertainty in the data used to estimate the value of wages and
compensation, and (2) uncertainty that underlies the assumptions related to the use of these data to
estimate the opportunity costs of time. The first source of uncertainty is probably less significant
than the second. Data on U.S. wages and compensation are plentiful and (as in the case of normal
time use) different sources provide reasonably similar values. Hence the estimates of post-tax wages
and total compensation are likely to be subj ect to less uncertainty than other aspects of the analysis.
The uncertainty related to the assumptions that underlie the use of these data to value time
losses requires consideration of more complex issues. The use of total compensation (wages and
benefits) to estimate the value of paid work time, of post-tax wages for the value of uncompensated
work and leisure time, and of a "zero" value for sleep time are based on a number of simplifying
assumptions regarding the operations of the labor market and the factors that influence individual
3-26
-------
choice. The key uncertainties are summarized in Exhibit 3-4 and discussed below.
Exhibit 3-4
SUMMARY OF IMPACTS OF KEY ASSUMPTIONS
RELATED TO DOLLAR VALUES FOR TIME USE
Time use
category
Valuation Approach
Key Limitations
Comments
Market
work time
Gross (pre-tax) wage plus
benefits, reflecting both the
opportunity costs to both the
individual (i.e., lost wages)
and to the employer and
society (i.e., lost product)
• Based on a number of
simplifying assumptions
regarding the relationship
between wages and
productivity.
Effect is indeterminate,
however, approach is
widely accepted for
valuing lost work time.
Nonmarket
work and
leisure time
Net (post-tax) wage, based on
the opportunity costs to the
individual
•	May be reasonably accurate for
individuals with flexible work
schedules, but may understate
or overstate values for
individuals with inflexible work
schedules.
•	May exclude some costs and
benefits that affect employment
decisions.
•	May understate or overstate
values for individuals who do
not engage in paid work.
Effect is indeterminate,
but impacts appear to be
counterbalancing to some
(unknown) extent.
Sleep time
Zero, due to lack of support
for a specific dollar value and
difficulties inherent in
estimating the net impact of
illness on sleep time
• Value may be similar to that of
other nonwork activities.
Likely to underestimate
value by a potentially
significant amount.
All
Average values may differ
from marginal values
Excludes consideration of
involuntary nature of losses
•	Theory suggests average values
may exceed marginal values,
but constraints on activity
choice leads to variation in this
relationship
•	Available research suggests
avoidance of involuntary
adverse effects may be more
highly valued.
Effect is indeterminate,
but may underestimate
value.
The approach for valuing paid work time is based on two assumptions: (1) that the marginal
value of an employee's output is likely to be equal to the marginal value of his or her total
compensation; and (2) that the value of this output is likely to be greater than the utility an individual
gains from working. The first assumption is based on the basic neoclassical model of profit-
maximizing behavior on the part of firms. Underlying this model are a number of simplifying
assumptions such as perfectly competitive markets and complete information, and the extent to
3-27
-------
which total compensation will under- or overstates the actual value of the worker's output to society
is uncertain. For example, the friction cost approach (discussed in the previous chapter) suggests
that this approach may overstate net national productivity losses if illness-related absenteeism results
in the employment of previously unemployed workers.50
In addition, as discussed in Appendix B, the data sources used to estimate total compensation
are likely to provide median or average rather than marginal values; i.e., to report total values for
a given time period rather than values for small changes in the number of hours worked within that
time period. Under standard neoclassical assumptions, marginal productivity is expected to decrease
with each unit of labor (in simple terms, the first hour spent in production is expected to provide
greater output than the second hour), implying that average values may exceed marginal values as
long as this assumption holds. Under this assumption, use of median values may overstate the value
of relatively small changes in time use, but may be accurate for larger changes.
Under the second assumption, the difference between the value of work to the individual and
the value of work to the employer and society is equal to the costs of benefits and taxes. If an
individual gains utility from work that exceeds post-tax wages but is less than the value of total
compensation, using total compensation to measure the societal value of illness-related losses should
not introduce additional bias into the analysis. If, however, individuals gain utility from paid work
that is greater than total compensation, then this approach may understate the value of related social
welfare losses. Despite these limitations, the use of total compensation as an estimate of the value
of lost work time is widely accepted in the cost of illness literature and is likely to be subject to less
debate than the approaches used to value uncompensated time.
The approach for valuing nonmarket work and leisure is based on the assumption that, at the
margin, individuals allocate their time between work and leisure so that the value of leisure time is
equal to the wage rate. Work by Bockstael, Strand, and Hanemann, and by Feather and Shaw
(summarized in Chapter 2) suggests that the wage rate provides a reasonably accurate estimate of
the value of uncompensated time when an individual can choose the number of hours he or she
spends in paid work.51 Where hours are inflexible, the marginal value of uncompensated time may
be more than the wage rate (if a worker is forced to work more hours than desired), or less than the
wage rate (if a worker is forced to work fewer hours than desired). While Feather and Shaw found
that their sample of employed individuals was split roughly into thirds (one-third with flexible hours,
50While the friction cost approach suggests that this approach may overstate the value of lost output
in cases where unemployed workers replace ill workers, more research is needed to determine the extent to
which such replacement actually occurs. In addition, the friction cost approach does not take into account
other factors that contribute to the uncertainty inherent in the approach.
51Bockstael, N.E., I.E. Strand, and W.M. Hanemann, "Time and the Recreational Demand Model,"
American Journal of Agricultural Economics, Vol. 69, pp. 293-202, 1987; Feather, P. and W.D. Shaw,
"Estimating the Cost of Leisure Time for Recreation Demand Models," Journal of Environmental Economics
and Management, Vol. 38, pp. 49-65, 1999; and, Feather, P. and W.D. Shaw, "The Demand for Leisure Time
in the Presence of Constrained Work Hours," Economic Inquiry, Vol. 38, No. 4, pp. 651-661, October 2000.
3-28
-------
one-third overemployed, and one-third underemployed), Bockstael, Strand, and Hanemann found
that individuals with inflexible hours generally valued leisure higher than the wage rate. These
results suggest that it is possible that overemployment is more prevalent than underemployment and
hence post-tax wages may understate the value of nonwork time.
The focus on marginal values introduces uncertainty into the valuation of uncompensated
time, as well as compensated time (as discussed earlier). Neoclassical economic theory suggests that
average values will exceed marginal values. The standard assumption is that each hour spent in an
activity has decreasing utility, and hence the marginal hour may have a lower value than the average.
In reality, as discussed in Chapter 2, the value of activities is likely to rise and fall depending on a
number of factors — such as constraints on an individual's ability to engage in certain activities at
certain times (e.g., to golf at night, or to enjoy the beach in the winter) — and the relationship
between average and marginal values is uncertain. In addition, the time losses associated with many
illnesses may lead to greater than marginal changes in time use, and use of the average or median
value per hour may provide more accurate estimates than relying on the marginal value for the last
hour.52 As noted above, the available data sources generally provide median or average rather than
marginal estimates of both total compensation and post-tax wages.
Other factors also influence the accuracy of post-tax wages as a measure of the value of
nonmarket work and leisure time. In particular, this approach ignores some costs that may a
necessary part of the decision to engage in paid work, such as those associated with commuting or
childcare, as well as the potential utility (or disutility) associated with these activities. For example,
commuting imposes dollar costs that decrease an individual's disposable income, and may be a
source of discomfort or inconvenience. Childcare also imposes costs, however, an individual may
gain substantial utility from taking care of his or her own child. Presumably, these costs would be
netted out from post-tax wages in an individual's consideration of the impact of work on income, and
the individual would also consider any utility gains or losses associated with these activities.
This approach also ignores the impact of employer paid benefits on individual utility —
working fewer hours will decrease the receipt of any benefits that are tied to hours worked, and
deciding to not work at all will eliminate any benefits. In addition, illness-related absence may not
reduce the disposable income of the employee due to the availability of sick pay and disability
benefits (as long as the absence does not exceed the time period covered by these benefits), and
small changes in hours (or short absences from work) may not affect total compensation particularly
for salaried employees.53 The net direction and magnitude of these effects is difficult to determine,
52 This factor may be one explanation for Gronau's findings, as discussed earlier, In his work on the effects of
the wage rate on the hours of home work for employed wives, Gronau found that the value of output was almost twice
as high as would be predicted under the opportunity cost approach. Gronau (1986), p. 299.
53If the individual has access to paid sick leave, a marginal loss of work time (within certain limits)
will not result in an immediate loss of income. However, a loss will accrue to the employer, who pays wages
without the benefit of the worker's productivity. The individual also loses the ability to save this sick leave
for another time.
3-29
-------
especially since many of these concerns have counterbalancing effects.
The approach of using estimates of post-tax wages for individuals not engaged in paid work
raises a number of other concerns. This approach assumes that individuals who do not engage in
paid labor are similar to those who do. For children, there are clear differences, and applying adult
values introduces significant uncertainty into the calculations. Adults who are involuntarily
unemployed may differ from the working population in ways that suggest that wage rates will
overstate the value of their time. Those not seeking employment may be forced to do so by
retirement policies (in the case of the elderly) or by the cost of childcare (for homemakers), rather
than by choice, and factors such as the availability of Social Security or welfare benefits complicate
these decisions. The net effects of using wage rates to estimate the value of time for these
individuals is unclear.
The approach used to value sleep is conservative and likely to undervalue related losses.
Available literature suggests that sleep may be valued at approximately the same marginal rate as
nonmarket work and leisure, but there a number of difficulties related to estimating the net effect
of illness on sleep time. Hence analysts may wish to test the effects of assigning a larger value to
sleep-related losses if it is possible to estimate the net change in sleep time due to illness.
Finally, two sources of uncertainty have cross-cutting effects. As described in the previous
paragraphs, the use of marginal vs. average values has implications for the valuation of all of the
time loss categories. In addition, the approach for valuing lost time is based on the theory of
consumer choice, where individuals are able to make decisions regarding the allocation of time
across different activities. However, illness limits this choice, and the activity restrictions involved
are generally involuntary. Available research on risk perception suggests that individuals may place
a higher value on avoiding involuntary or uncontrollable adverse effects than on avoiding such
effects voluntarily. For example, voluntariness and controllability are among the nine major risk
dimensions that Slovic, Fischhoff, and Lichtenstein identify as influencing individuals' perceptions
and rankings of risks.54 In addition, Cropper and Subramanian find that individuals tend to choose
programs that address risks that are more difficult to avoid (i.e., less controllable) when selecting
among public health and environmental programs with different risk characteristics.55 In addition,
two recent reviews of the risk perception literature interpret it as suggesting that the combined
effects of dread, lack of voluntariness, and lack of controllability in environmental risk reduction
settings compared to a typical accidental death risk scenario could increase willingness to pay
54Summarized in: P. Slovic, B. Fischhoff, and S. Lichtenstein, "Perceived Risk: Psychological Factors
and Social Implications," Proceedings of the Royal Society of London. Series A: Mathematical and Physical
Sciences, Vol. 430, No. 1878, 1981, pp. 17-34. Also evaluated in P. Slovic, "Perception of Risk," Science,
Vol. 236, April 1987, pp. 280-285.
55M.L. Cropper and U. Subramanian, "Public Choices Between Lifesaving Programs: How Important
Are Lives Saved?" Policy Research Working Paper 1497, The World Bank, 1995.
3-30
-------
estimates derived from accidental risk scenarios by a factor of two or more.56 These studies suggest
that the value of involuntary time losses attributable to illness may exceed the value of time reflected
in the normal allocation of time between labor and leisure activities.
56R.L. Revesz, "Environmental Regulation, Cost-Benefit Analysis, and the Discounting of Human
Lives," Columbia Law Review, Vol. 99, No. 4, 1999, pp. 941-1017; and P. Rowlatt et al., Valuation of
Deaths from Air Pollution, prepared for the (British) Department of Environment, Transport and the Regions
and the Department of Trade and Industry, 1998.
3-31
-------
REFERENCES
Advisory Council on Clean Air Compliance Analysis. May 2004. Review of the Revised Analytical
Planfor EPA's Second Prospective Analysis - Benefits and Costs of the Clean Air Act 1990-
2020. Prepared for the U.S. Environmental Protection Agency. EPA-SAB-COUNCIL-
ADV-04-004.
Abt Associates. December 2001. Final Heavy Duty Engine/Diesel Fuel Rule: Air Quality
Estimation, Selected Health and Welfare Effects Methods, and Benefits Results. Prepared
for the U.S. Environmental Protection Agency.
Berger, M., Blomquist, G., Kenkel, D., and Tolley, G. 1987. "Valuing Changes in Health Risks:
A Comparison of Alternative Measures." The Southern Economic Journal. Vol. 53. Pages
977-984.
Biddle, J.E. and Hamermesh, D.S. 1990. "Sleep and the Allocation of Time." Journal of Political
Economy. Vol. 98, No. 5. Pages 922-943.
Bockstael, N.E., Strand, I.E., and Hanemann, W.M. 1987. "Time and the Recreational Demand
Model." American Journal of Agricultural Economics. Vol. 69. Pages 293-202.
Brody, W.H. August 1975. Economic Value of a Housewife. Research and Statistics Note 9.
DHEW Publication No. SSA75-11701. Washington: Social Security Administration, Office
of Research and Statistics.
Buzby, J.C., Roberts, T., Lin, C.T.J., and MacDonald, J.M. 1996. BacterialFoodborne Disease:
Medical Costs and Productivity Losses. Food and Consumer Economics Division, Economic
Research Service, U. S. Department of Agriculture. Agricultural Economic Report No. 741.
Cesario, F.J. February 1976. "Value of Time in Recreation Benefit Studies." Land Economics. Vol.
52, No. 1. Pages 32-41.
Chestnut, L.G., Colome, S.D., Keller, L.R., Lambert, W.E., et al. October 1988. Heart Disease
Patients Averting Behavior, Costs of Illness, and Willingness to Pay to Avoid Angina
Episodes. EPA Report 230-10-88-042.
Cooper, B.S. and Rice, D.P. February 1976. "The Economic Cost of Illness Revisited." Social
Security Bulletin. Vol. 39, No. 2. Pages 21-36.
Cropper, M.L. and Krupnick, A. 1999. "The Social Costs of Chronic Heart and Lung Disease."
Valuing Environmental Benefits: Selected Essays of Maureen Cropper. (M.L. Cropper,
ed.). Cheltenham, UK: Edward Elgar Publishing.
Cropper, M.L. and Krupnick, A. June 1990. The Social Costs of Chronic Heart and Lung Disease.
Ref-1
-------
Resources for the Future Discussion Paper QE89-16-REV. Washington, DC.
Cropper, M.L. and Subramanian, U. 1995. "Public Choices Between Lifesaving Programs: How
Important Are Lives Saved?" Policy Research Working Paper 1497. The World Bank.
DeSerpa, A. C. December 1971. "A Theory of the Economics of Time." The Economic Journal.
Vol. 81, No. 324. Pages 828-846.
Douglas, J., Kenny, G., and Miller, T.R. 1990. "Which Estimates of Household Production Are
Best?" Journal of Forensic Economics. Vol.4. Pages 25-45.
Dravove, D. 1995. "Measuring Costs." Valuing Health Care. (F.A. Sloan, ed.). Cambridge,
England: Cambridge University Press.
Englin, J. and Shonkwiler, J.S. 1995. "Modeling Recreation Demand in the Presence of
Unobservable Travel Costs: Toward a Travel Price Model." Journal of Environmental
Economics and Management. Vol. 29, No. 3. Pages 368-377.
Feather, P. and Shaw, W.D. 1999. "Estimating the Cost of Leisure Time for Recreation Demand
Models." Journal of Environmental Economics and Management. Vol. 38. Pages 49-65.
Feather, P. and Shaw, W.D. October 2000. "The Demand for Leisure Time in the Presence of
Constrained Work Hours." Economic Inquiry. Vol. 38, No. 4. Pages 651-661.
Freeman, A. M., III. 2003. The Measurement of Environmental and Resource Values: Theory and
Methods. Second Edition. Washington, DC: Resources for the Future.
Gold, M.R., Siegel, J.E., Russell, L.B., and Weinstein, M.C. (eds.). 1996. Cost-Effectiveness in
Health andMedicine. Oxford: Oxford University Press.
Goldschmidt-Clermont, L. 1982. Unpaid Work in the Household. Prepared for the International
Labor Office, United Nations.
Gronau, R. 1986. "Home Production - A Survey." Handbook of Labor Economics, Vol. 1. (O.
Ashenfelter and R. Layard, eds.) New York: North-Holland.
Harrington, W., Krupnick, A. J., and Spofford, W.O., Jr. 1991. Economics and Episodic Disease:
The Benefits of Preventing a Giardiasis Outbreak. Washington, DC: Resources for the
Future.
Harrington, W., Krupnick, A.J., and Spofford, W.O., Jr. 1989. "The Economic Losses of a
Waterborne Disease Outbreak." Journal of Urban Economics. Vol. 25, No. 1. Pages 116-
137.
Ref-2
-------
Harrington, W. and Portney, P. 1987. "Valuing the Benefits of Health and Safety Regulations."
Journal of Urban Economics. Vol. 22. Pages 101 - 112.
Hartunian, N.S., Smart, C.N., and Thompson, M.S. 1981. The Incidence and Economic Costs of
Major Health Impairments. A Comparative Analysis of Cancer, Motor Vehicle Injuries,
Coronary Heart Disease, and Stroke. Lexington, Massachusetts: D.C. Heath and Company.
Heckman, J. July 1975. "Shadow Prices, Market Wages, and Labor Supply." Econometrica. Vol.
42, No. 4. Pages 679-694.
Hoffman, C., Rice, D., and Sung, H.Y. November 1996. "Persons With Chronic Conditions: Their
Prevalence and Costs." Journal of the American Medical Association. Vol. 276, No. 18.
Pages 1473-1479.
Hutubessy, R.C.W., van Tulder, M.W., Vondeling, H., and Bouter, L.M. 1999. " Indirect Costs of
Back Pain in the Netherlands: A Comparison of the Human Capital Method with the Friction
Cost Method." Pain. Vol.80. Pages 201-207.
Johannesson, M. and Karlsson, G. 1997. "The Friction Cost Method: A Comment." Journal of
Health Economics. Vol. 16. Pages 249 - 255.
Kocagil, P., Demarteau, N., Fisher, A., and Shortle, J.S. 1998. "The Value of Preventing
Cryptosporidium Contamination." Risk: Health, Safety and Environment. Vol. 9. Pages
175-196.
Koopmanschap, M.A. and Rutten, F.F.H. 1993. "Indirect Costs in Economic Studies."
PharmacoEconomics. Vol. 4, No. 6. Pages 446-454.
Koopmanschap, M.A., and F.F.H. Rutten. 1996. "A Practical Guide for Calculating Indirect Costs
of Disease." PharmacoEconomics. Vol. 10, No. 5. Pages 460-456.
Koopmanschap, M.A., Rutten, F.F.H., van Ineveld, B.M., and van Roijen, L. 1995. "The Friction
CostMethod for Measuring Indirect Costs of Disease." Journal of Health Economics. Vol.
4. Pages 171-189.
Landefeld, J.S. and Seskin, E.P. 1982. "The Economic Value of Life: Linking Theory to Practice."
American Journal of Public Health. Vol.6. Pages 555-566.
Larson, D.M.. August 1993. "Separability and the Shadow Value of Leisure Time." American
Journal ofAgricultural Economics. Vol. 75. Pages 572-577.
Magat, W.A., Viscusi, W. K., andHuber, J. 1992. The Death Risk Lottery Metric for Valuing Health
Risks: Applications to Cancer and Nerve Disease. Prepared for the U.S. Environmental
Protection Agency.
Ref-3
-------
McConnell, K.E., and Strand, I. 1981. "Measuring the Cost of Time in Recreation Demand
Analysis: An Application to Sportfishing." American Journal of Agricultural Economics.
Vol. 63. Pages 153-156.
MVA Consultancy et al. 1994. "Research Into the Value of Time." Cost-Benefit Analysis (R.
Layard, and S. Glaister, eds.). Second Edition. Cambridge: Cambridge University Press.
Pages 235 -272.
National Research Council. 2000. Time Use Measurement and Research: Report of a Workshop.
Commission on Behavioral and Social Sciences and Education, Committee on National
Statistics (M. Ver Ploeg, J. Altonji, N. Bradburn, J. DaVanzo, W. Nordhaus, and F.
Samaniego (eds.)). Washington, D.C: National Academy Press.
Revesz, R.L.. 1999. "Environmental Regulation, Cost-Benefit Analysis, and the Discounting of
Human Lives." Columbia Law Review. Vol. 99, No. 4. Pages 941-1017.
Ray, N.F., Thamer, M., Gardner, E., Chan, J.K., and American Diabetes Association. 1998.
"Economic Consequences of Diabetes Mellitus in the U.S. in 1997." Diabetes Care. Vol.
21, No. 2. Pages 296-309.
Rice, D.P. May 1966. "Estimating the Cost of Illness." Health Economic Series No. 6, PHS
Publication No. 947-6. Washington, DC: U.S. Government Printing Office.
Rice, T.D., Duggan, A.K. and DeAngelis, C. 1992. "Cost-Effectiveness of Erythromycin versus
Mupirocin for the Treatment of Impetigo in Children." Pediatrics. Vol. 89, No. 2. Pages
210-214.
Rice, D.P. and Max, W. 1992. The Cost of Smoking in California, 1989. Sacramento: California
State Department of Health Services.
Ries, L.A.G., et al. (eds). 2001. SEER Cancer Statistics Review, 1973-1998. Bethesda, MD:
National Cancer Institute.
Robb, R., Denton, M., Gafni, A., Joshi, A., Lian, J., Rosenthal, C., and Willison, D.. 1999.
"Valuation of Unpaid Help by Seniors in Canada: An Empirical Analysis." Canadian
Journal on Aging. Vol 18, No. 4. Pages 430 - 446.
Rowe, R.D. and Chestnut, L.D. undated. Valuing Changes in Morbidity: WTP v.s. COIMeasures.
Prepared for the U.S. Environmental Protection Agency and the California Air Resources
Board.
Rowlatt. P. et al. 1998. Valuation of Deaths from Air Pollution. Prepared for the (British)
Department of Environment, Transport and the Regions and the Department of Trade and
Industry.
Ref-4
-------
Shaw, W. D. 1992. "Searching for the Opportunity Cost of an Individual's Time." Land Economics.
Vol. 68, No. 1. Pages 107-115.
Shaw, W. D. and Feather, P.. 1999. "Possibilities for Including the Opportunity Cost of Time in
Recreation Demand Systems." Land Economics. Vol. 75, No. 4. Pages 592-602.
Slovic, P. April 1987. "Perception of Risk." Science. Vol. 236. Pages 280-285.
Slovic, P., Fischhoff, B., and Lichtenstein, S.. 1981. "Perceived Risk: Psychological Factors and
Social Implications." Proceedings of the Royal Society of London. Series A: Mathematical
and Physical Sciences. Vol. 430, No. 1878. Pages 17-34.
Small, K.. 1992. Urban Transportation Economics. Luxembourg: Harwood Academic Publishers.
Smith, V. K., Desvousges, W.H. andMcGivney, M.P. 1983. "The Opportunity Cost of Travel Time
in Recreation Demand Models." Land Economics. Vol. 59, No. 3. Pages 259-278.
Statistics Canada. 1999. "Average time spent on activities, total population, and participants, by
sex." General Social Survey, 1998.
Tipping, D. G. December 1968. "Time Savings in Transport Studies." Economic Journal. Pages
843-854.
Trewin, D. October 2000. Unpaid Work and the Australian Economy: 1997. Australian Bureau
of Statistics.
Ungar, W. J., Coytem, P.C., and the Pharmacy Medication Monitoring Board. 2000. "Measuring
Productivity Loss Days in Asthma Patients." Health Economics. Vol. 9. Pages 37-49.
U.S. Census Bureau. November 2001. Statistical Abstract of the United States: 2001 (121st
edition). Washington, DC: U.S. Government Printing Office.
U. S. Department of Transportation. April 1997. Departmental Guidance for the Valuation of Travel
Time in Economic Analysis (Memorandum from F. E. Kruesi).
U.S. Department of Transportation. February 2003. Revised Departmental Guidance, Valuation of
Travel Time in Economic Analysis (Memorandum from E.H. Frankel).
U.S. Environmental Protection Agency. October 2003. Children's Health Valuation Handbook.
U.S. Environmental Protection Agency. May 2003. Benefits and Costs of the Clean Air Act,
1990 - 2020: Revised Analytical Plan for EPA 's Second Prospective Analysis. Prepared
by Industrial Economics, Incorporated.
Ref-5
-------
U.S. Environmental Protection Agency. February 2001. Cost of Illness Handbook.
U.S. Environmental Protection Agency.	December 2000. Handbook for Non-cancer Health
Effects Valuation.
U.S. Environmental Protection Agency.	September 2000. Guidelines for Preparing Economic
Analyses. EPA 240-R-00-003.
U.S. Environmental Protection Agency. August 1997. Exposure Factors Handbook (Final
Report), Washington, D.C.: Office of Research and Development.
U.S. Office of Management and Budget. September 2003. Regulatory Analysis (Circular A-4).
Waitzman, N.J., Scheffler, R.M., Romano, P.S. 1996. The Cost of Birth Defects: Estimates of
the Value of Prevention. Lanham: University Press of America, Incorporated.
Weinblatt, E., Shapiro, S., Frank, C.W. and Sager, R.V. 1966. "Return to Work and Work
Status Following First Myocardial Infarction." American Journal of Public Heath. Vol.
65, No. 2. Pages 169-185.
Weinstein, M.C., Siegel, J.E., et. al. 1997. "Productivity Costs, Time Costs and Health-Related
Quality of Life: A Response to the Erasmus Group." Health Economics. Vol.6. Pages
505-510.
Wilman, E.A.. 1980. "The Value of Time in Recreation Benefit Studies." Journal of
Environmental Economics and Management. Vol. 7. Pages 272- 286.
Winston, G.C.. 1987. "Activity Choice: A New Approach to Economic Behavior." Journal of
Economic Behavior and Organization. Vol 8. Pages 567-585.
Ref-6
-------
APPENDIX A: EXAMPLES FROM THE EMPIRICAL LITERATURE
The preceding chapters of this report provide detailed information on calculating the
value of lost time, and include footnoted references to existing studies that illustrate particular
aspects of the approach. This appendix provides more information on some of the studies
referenced previously. It is not intended to be a comprehensive review of the literature; rather it
provides selected examples of alternative approaches to the valuation of time losses due to
illness. As indicated by these examples, much of the existing literature applies the human capital
approach and focuses on lost work time. Few researchers have considered the social welfare
losses associated with the effects of illness on other types of time use (e.g., leisure) using the
approach described in this report.
The studies selected as examples are listed in Exhibit A-l, which provides information on
the types of time losses costs assessed, the health effects studied, and the type of estimates
developed. These studies are summarized to reflect the variety of methods and results within the
body of relevant literature. The first two studies provide examples of the human capital
approach, while the remaining studies provide examples of alternative approaches to estimating
indirect costs. Descriptions of each study, as well as a summary of the results, are provided in
the following pages.
A-l
-------
Exhibit A-l
EXAMPLES FROM THE LITERATURE
Authors and
Publication Date
Time Losses Assessed
Health Effects
Assessed
Type of Estimates
Hartunian et al.,
1981
Lost earnings and
housekeeping time for ill
individuals
Cancer, ischemic heart
disease, stroke, and
motor vehicle injuries
Lifetime costs of U.S. cases
identified in 1975
Ray et al., 1998
Lost earnings and
housekeeping time for ill
individuals
Diabetes mellitus
Annual costs for U.S. cases in
1997
Cropper and
Krupnick, 1999
Changes in earnings and
labor force participation rates
for ill individuals
Various chronic heart
and lung diseases
Predictive model based on
1977 national sample of males
Buzby et al., 1996
Lost earnings for ill
individuals and caregivers
Various bacterial
diseases
Lifetime costs per case using a
1993 base.
Waitzman et al.,
1996
Lost earnings and
housekeeping time for ill
individuals
Various birth defects
Lifetime costs of California
cases identified in 1988
Harrington et al.,
1989
Lost earnings, household
work, and leisure time for ill
individuals and caregivers
Giardiasis
Per case losses due to 1983-
1984 contamination incident
Kocagil et al.,
1998
Lost earnings and nonwork
time for ill individuals
Cryptosporidiosis
Per case losses due to 1996
contamination incident
Sources:
See footnotes and reference list at the end of this report for full citations.
Hartunian et al. (1981): This study uses an incidence approach to estimate direct and
indirect costs of cancer, motor vehicle injuries, coronary heart disease, and stroke.1 The authors
separate each condition into a number of sub-conditions to distinguish between those with
different cost impacts. The direct costs considered vary somewhat by health effect, but generally
include: emergency assistance; initial inpatient hospital care; physical and surgeon services;
vocational and physical rehabilitation; nursing home and home attendant care; drugs, medical
supplies, and appliances; outpatient medical and surgical care; re-hospitalization (due to
recurrences); home modifications; paramedical and miscellaneous expenses; insurance
administration; and legal and court expenses.
For nonfatal cases, the authors estimate the indirect costs of lost market work, reduced
'Hartunian, N.S., C.N. Smart, and M.S. Thompson, The Incidence and Economic Costs of Major
Health Impairments: A Comparative Analysis of Cancer, Motor Vehicle Injuries, Coronary Heart Disease,
and Stroke, Lexington, Massachusetts: D.C. Heath and Company, 1981.
A-2
-------
productivity at work, and lost nonmarket work due to morbidity in surviving patients. For fatal
cases, they estimate morbidity prior to premature mortality (the morbidity increment) as well as
lost earnings due to death. Data on employment and housekeeping rates are from the
Department of Labor. For employed individuals, lost time is valued at market wage rates,
however, it is unclear whether these rates include taxes (or benefits). The value of household
labor is estimated in two ways: (1) based on Brody (1975) which uses a composite market-value
approach, and (2) based on the opportunity cost of household labor, valued at the average
earnings of same-age and same-sex peers in the labor force.2
This study is an early and precedent-setting example of an incidence-based approach to
valuing indirect costs; much of the available literature focuses on annual estimates rather than on
lifetime costs per case. It provides the foundations for several aspects of the valuation methods
discussed in this report, including the valuation of reduced productivity as well as more complete
work-related losses, the valuation of nonmarket work based on earnings (i.e., opportunity costs)
compared to replacement costs, and the calculations used to estimate the value of time losses
associated with both acute and chronic illnesses. While the authors do not estimate the value of
caregiver losses, they note that including such losses would improve their estimates.
Ray et al. (1998): This study provides a more recent example of a straightforward
application of the human capital method, and illustrates how the data sources discussed in
Appendix B can be used to estimate the amounts and types of illness-related time losses. It
assesses the direct medical expenditures and indirect costs attributable to the roughly 7.5 million
estimated cases of diabetes in 1997.3 The direct cost estimates include nursing home stays,
hospital stays, physician visits, emergency room care, medication, and related professional
services. The indirect costs estimated include lost earnings and housekeeping time due to
disability and premature mortality.
Ray et al. estimate three types of time losses. First, they consider short-term disability,
using self-reported data on bed days, restricted activity days, and work loss days from the
National Health Interview Survey. They compare estimates of lost days for persons with and
without diabetes, to estimate the net impact of diabetes. For employed individuals, lost work
days were valued at median gross (pre-tax) earnings for full time workers from the Bureau of
Labor Statistics; bed days and disability days were valued based on median gross earnings for
housekeepers from the same source. Second, Ray et al. also consider long term permanent
disability time losses. These losses are estimated based on lost earnings data (exclusive of
disability payments) from the Social Security Administration. Third, the analysis also considers
time losses due to premature mortality.
2Brody, W. H. Economic Value of a Housewife. Research and Statistics Note 9, DHEW Publication
No. SSA 75-11701. Washington: Social Security Administration, Office of Research and Statistics, August
1975.
3Ray, N.F., M. Thamer, et al., "Economic Consequences of Diabetes Mellitus in the U.S. in 1997,"
Diabetes Care, Vol. 21, No. 2, 1998, pp. 296-309.
A-3
-------
Cropper and Krupnick (1999): This study applies a different and somewhat unusual
approach to valuing time losses, using an econometric model to estimate the effects of illness on
earnings and labor force participation.4 Cropper and Krupnick characterize the effects of chronic
respiratory and circulatory disease on labor force participation and earnings for men aged 18 to
65. Data used in the model are from the Social Security Survey of Disability and Work, and
include individuals with and without illness. Annual losses in expected earnings are estimated at
different ages, by age at onset, for different health conditions. The independent variables include
data on earnings, household characteristics, age, level of education, location, and other
characteristics as well as health status. To capture an individual's decision to remain in the work
force, additional variables are added to the model, including whether the individual is aware of
Social Security benefits, whether the individual is a veteran, and the size of the individual's debt.
The study also estimates direct medical costs from the National Medical Care Expenditure
Survey for some conditions, including the costs of medical visits, hospitalization, and drugs.
The researchers note that the theoretically correct measure of social benefits from
reducing the incidence of a disease is the sum of ill individuals' willingness to pay to avoid the
disease plus the costs of the disease borne by others. This study focuses on only the second half
of this equation; estimating costs (e.g., of hospital stays and lost productivity) that are often
borne by others (such as insurance companies or employers) rather than directly by the ill
individual. Therefore, the authors suggest that the results of this study could be added to an
individual willingness to pay measure to better estimate the social welfare benefits of risk
reductions.
Buzby et al. (1996): This study assesses medical costs and productivity losses for a
number of bacterial diseases, and provides an example of an approach for assessing caregiver
losses. For four bacterial diseases (salmonellosis, campylobacteriosis, staphylococcus aureus,
and Clostridium perfringens), the authors rely on previous research that does not separate out
medical costs from the costs of lost time. For the remaining two health effects, E. coli disease
and listeriosis, the authors separate medical costs from the value of lost productivity. In addition
to assessing lost time due to morbidity, they assess mortality costs by relying on adjusted value
of statistical life estimates based on a study by Landefeld and Seskin.5 The discussion below
focuses on the approach for valuing lost time for acute cases of these latter two illnesses, because
chronic cases are valued in part based on adjusted value of statistical life estimates.
4Cropper, M.L. and A.J. Krupnick, "The Social Costs of Chronic Heart and Lung Disease," in Valuing
Environmental Benefits: Selected Essays of Maureen Cropper, M.L. Cropper (ed.), Cheltenham: Edward
Elgar, 1999. (A previous version of this study was published as: Cropper, M.L. and A.J. Krupnick, The
Social Costs of Chronic Heart and Lung Disease. Resources forthe Future Discussion Paper QE89-16-REV,
Washington, DC. June 1990.)
5Buzby, J.C., T. Roberts, etal., Bacterial Foodborne Disease: Medical Costs and Productivity Losses,
Food and Consumer Economics Division, Economic Research Service, U.S. Department of Agriculture,
Agricultural Economic ReportNo. 741,1996; Landefeld, J.S. and E. P. Seskin, "The Economic Value of Life:
Linking Theory to Practice," American Journal of Public Health, Vol. 6, 1982, pp. 555-566.
A-4
-------
For acute E. coli disease, the authors estimate time lost from paid work both for
caregivers of ill children and for ill adults. The number of days a child is home from school is
estimated based on the severity of the condition. For ill adults, days lost due to illness, physician
visits, and hospitalization and subsequent recuperation are considered. For both caregivers and
ill adults, time losses are valued using average weekly gross (pre-tax) earnings, including fringe
benefits. These estimates are based on data for production of non-supervisory workers in private
nonagricultural jobs, collected by the Bureau of Labor Statistics. A similar approach is used for
acute cases of listeriosis. However, in this case, only time losses due to hospitalization and
subsequent recuperation are considered. Costs are estimated separately for pregnant women and
other adults, and for cases of differing severity. The amount of lost time is based on estimates of
the length of the hospital stay and of the subsequent recuperation time at home. Lost work time
is valued using the same approach as described above for E. coli disease.
The researchers acknowledge limitations of their approach to valuing these illnesses. In
particular they note that in the case of the salmonellosis study, if the value of lost leisure were
included in their estimate and valued at the prevailing wage rate, the cost estimate of
salmonellosis would more than double their current estimate. The authors state, "[w]hen using
the COI estimates presented here, one should bear in mind that they underestimate the true
economic value of bacterial foodborne illnesses to society because they exclude such costs as:
(1) pain, suffering, and lost leisure time of the victim and her/his family, (2) lost business and
costs and liabilities of lawsuits affecting agriculture and the food industry, (3) the value of self-
protective behaviors undertaken by industry and consumers, and (4) resources spent by Federal,
State and local governments to investigate the source and epidemiology of the outbreak."
Waitzman et al. (1996): This study provides an example of the valuation of the impact
of childhood illnesses on future market and nonmarket work; it estimates the lifetime costs of
several types of birth defects for children in California.6 Indirect costs of morbidity are
separated into two categories based on the severity of the lifelong effects: (1) the affected
individual is totally unable to work, or (2) the affected individual is limited in the type and
amount of work he or she can perform. The percent of individuals with work limitations of each
type was estimated largely from the National Health Interview Survey, supplemented by data
from the Survey of Income and Program Participation. The authors consider both market work
and household production, and assume that time losses for household work will be concomitant
with the reductions in market work. The authors combine these data with information on
survival rates to determine the time periods over which losses occur.
Waitzman et al. use information from Rice and Max (1992) to value time losses.7 They
value lost market work time at gross wage rates that include earnings and benefits, assuming a
6Waitzman, N.J., R.M. Scheffler, and P.S. Romano, The Costs of Birth Defects, Lanham: University
Press of America, Incorporated, 1996.
7Rice, D.P. and W. Max, The Cost of Smoking in California, 1989, Sacramento: California State
Department of Health Services, 1992.
A-5
-------
one percent increase in market productivity per year. The Rice and Max study is also used to
value lost household production, applying a composite rate that reflects the market costs for
different household activities.8 Waitzman et al. provide estimates discounted at rates of two,
five, and ten percent. In addition, the researchers estimate lost earnings from premature
mortality, medical costs (such as inpatient and outpatient care, pharmaceuticals, laboratory tests,
x-rays, appliances, and long-term care), and the costs of developmental and special education
services. Waitzman et al. acknowledge that excluding caregiver costs, as well as the impacts of
pain and suffering, results in a conservative estimate of the social costs of birth defects.
Harrington et al. (1989): These researchers estimate the losses associated with
giardiasis due to a drinking water contamination episode in Luzerne County, Pennsylvania.9
Their study is somewhat unusual in that it includes the valuation of lost leisure time and provides
a more complete estimate of the effects of illness (on both caregivers and ill individuals) than the
previously cited studies. The medical cost component of the analysis considers visits to doctors,
the hospital, and the emergency room, as well as laboratory tests and medication. Time costs
include time spent on these activities and associated travel. In addition, the indirect cost
component of the analysis considers work time losses, reduced productivity while at work, and
lost leisure and housework time from the perspective of affected individuals and their caregivers.
The researchers also assess losses due to averting actions.
To determine the direct and indirect costs of illness, the authors surveyed individuals
with reported cases of giardiasis. They collected information on work days lost for the employed
and for homemakers, including both the ill individual and his or her caregiver. In addition, they
collected data on the self-reported reduction in productivity due to illness while at work for
employed individuals and homemakers. Leisure time was calculated as time not spent working
or sleeping.
For the employed, the researchers value lost time at the before-tax hourly wage rate for
survey respondents. Leisure time losses are valued at the after-tax wage, calculated from gross
wages by netting out an amount equal to the U.S. average income tax rate (including federal,
state and local taxes and FICA). Three different methods are used to value the lost nonmarket
time of homemakers, retirees, and unemployed persons: (a) the after-tax wage rate for employed
respondents, (b) the after-tax minimum wage rate, and (c) zero.
8This approach in turn is based on a methodology developed in: Douglas, J., G. Kenny and T.R.
Miller, "Which Estimates of Household Production Are Best?," Journal of Forensic Economics, Vol. 4, No.
1, pp. 25-45.
9Harrington, W., A.J. Krupnick, and W.O. Spofford, Jr., "The Economic Losses of a Waterborne
Disease Outbreak," Journal of Urban Economics, Vol. 25, No. 1, 1989, pp. 116-137. (More detailed
information on this study is available in: Harrington, W., A.J. Krupnick, and W.O. Spofford, Jr., Economics
and Episodic Disease: The Benefits of Preventing a Giardiasis Outbreak, Washington, DC: Resources for
the Future, 1991.)
A-6
-------
Kocagil et al. (1998): The Kocagil study is similar to the Harrington study in that it
considers the value of averting actions, medical costs, and time losses associated with a
contamination event.10 However, this team of researchers uses a slightly different approach for
valuing time losses. Kocagil et al. assess the impacts of cryptosporidiosis, using a 1996 outbreak
in Lancaster County, Pennsylvania as a case study. The medical cost component comprises
over-the-counter medications, physician or emergency room visits, and hospitalization (in severe
cases). Time losses for ill individuals include both work and leisure time. (Leisure is defined as
including both nonmarket work and other nonwork activities.) The researchers also include time
losses in their analysis of averting behavior, including boiling water (electricity and time),
hauling water from outside the contaminated region (travel time), and purchasing bottled water.
For ill individuals, Kocagil et al. present two estimates that vary in the approach used to
estimate the value of time losses (medical costs are the same under both approaches). First, they
value all time losses at the after-tax wage rate to reflect the cost of these losses from the
individual perspective. Second, to reflect the societal perspective, they value lost work time at
the average before-tax wage rate, and lost leisure time at the after-tax wage rate. They assume
that individuals sleep for eight hours per day, and that 2/7 of the remaining 16 hours are spent in
leisure and 5/7 in work. Alternative assumptions are explored through uncertainty analysis. The
researchers also estimate the value of cryptosporidiosis-related mortality based on different
estimates of the value of statistical life. The researchers describe their results as providing a
lower bound estimate of the value of preventing these contamination events, since they do not
include the potential costs associated with pain and suffering.
Exhibit A-2 below provides the estimates of the value of morbidity time losses from each
study, both in dollars and as a percentage of medical costs.11 Many of the studies in the exhibit
address time losses associated with mortality as well as morbidity, however the table excludes
post-mortality losses unless otherwise indicated. For all illnesses, the estimates of morbidity
losses include both fatal and nonfatal cases. In other words, the exhibit includes the costs
associated with morbidity prior to death for those who die of the disease, as well as morbidity
costs for those who die of other causes. In cases where the study reports total (e.g., national)
estimates, the estimates are converted to per case values based on information reported on the
number of cases. Note that data from Hartunian et al. (1981) are not provided in this exhibit
because the researchers do not report time losses for morbidity separate from time losses from
mortality.
10Kocagil, Patricia, Nadia Demarteau, Ann Fisher, and James S. Shortle, "The Value of Preventing
Cryptosporidium Contamination," Risk: Health, Safety and Environment, Vol. 9, 1998, pp. 175-196.
nThe studies do not provide adequate information to convert the cost estimates to a comparable
metric (e.g., average value of time losses per day of illness in year 2001 dollars).
A-7
-------
Exhibit A-2
MORBIDITY-RELATED COST ESTIMATES
Authors and
Date of Study
Value of Time Losses
Time Losses as a Percent of Medical Costs
(medical costs in parentheses)
Ray et al., 1998
(1997 dollars,
annual costs per case)
Diabetes: $4,947
Diabetes: 67% ($7,402)
Cropper and Krupnick,
1999
(1977 dollars,
annual losses per case)
Allergies: $742 - $1,300
Chronic bronchitis: $1,770 - $3,083
Emphysema: $4,212 - $5,336
Other lung disease: $961 - $1,647
Arteriosclerosis: $2,921 - $7,226
Heart attack: $607 - $5,744
Hypertension: $556 - $983
Stroke: $5,369 - $6,635
Other heart disease: $1,149 - $2,358
Chronic bronchitis: 1,830%-3,187% ($97)
Emphysema: 666% -843%) ($633)
Hypertension: 258%-455% ($215)
Other health effects:
comparable medical costs not reported
Buzby et al., 1996
(1993 dollars, total costs per
case, 3 % discount rate)
Acute E. coli: $175 - $2,815
Acute listeriosis: $1,166 - $1,469
Acute E. coli: 8%-219% ($160-$36,185)
Acute listeriosis: 10%o-12%o ($12,117)
Waitzman et al., 1996
(1988 dollars, total costs per
case, 5 percent discount rate)
Spina bifida: $53,000
Truncus arteriosus: $16,000
Transposition/DORV: $26,000
Tetralogy of fallot: $32,000
Single ventricle: $20,000
Cleft lip or palate: $30,000
Upper limb reduction: $24,000
Lower limb reduction: $88,000
Down syndrome: $171,000
Cerebral palsy: $92,000
Spina bifida: 53% ($99,000)
Truncus arteriosus: 8%o ($210,000)
Transposition/DORV: 37%) ($69,000)
Tetralogy of fallot: 30% ($108,000)
Single ventricle: 20 % ($99,000)
Cleft lip or palate: 265% ($11,000)
Upper limb reduction: 456% ($5,000)
Lower limb reduction: 551%) ($16,000)
Down syndrome: 313%o ($55,000)
Cerebral palsy: 125% ($74,000)
Harrington et al., 1989
(1984 dollars,
total costs per case)
giardiasis: $604 - $1,001
giardiasis: 238 - 394% ( $254)
Kocagil et al., 1998
(1998 dollars,
total costs per case)
cryptosporidiosis: $460 - $547
crypto sporidio sis: 78%o - 81 % ($129)
Notes:
See Exhibit A-l and preceding text for information on methods used and types of costs assessed.
Kocagil et al. estimates exclude time losses related to averting actions.
Excludes Hartunian (1981) due to lack of per case data.
Sources:
See footnotes and reference list for full citations.
As can be seen from the exhibit, the estimates of the value of time losses vary greatly due
to the types of time losses considered (e.g., work vs. nonwork time), the approach used to value
these losses, and the differences in the effects of each illness on time use. While the studies
exhibit some agreement in how they account for direct medical costs, they diverge in their
analyses of indirect costs. All incorporate estimates of lost work time for ill individuals, and
A-8
-------
several include valuation of losses in nonmarket work time. Two of the seven studies include
the valuation of lost leisure time, and two include caregiver losses. Some of the studies account
for productivity losses for ill individuals who continue to engage in normal activities in addition
to accounting for more complete losses of normal time use. In general, the researchers note that
their approaches are likely to understate the full value of related social welfare losses, due to the
exclusion of some types of time losses as well as the incomplete consideration of the value of
avoiding pain and suffering.
The precise estimates derived in these studies are not entirely comparable because of the
differences in approaches as well as in the overall goals of each study. Exhibit A-2 suggests,
however, that the value of time related-losses can be substantial. Hence ignoring these effects
could lead an analyst to significantly understate the value of risk reductions.
Analysts interested in valuing time losses for the health effects listed in the table should
consult the original studies rather than relying solely on the materials in this appendix. The brief
summaries provided do not include the sort of detailed information needed to determine the
suitability of individual studies for use in a particular regulatory context. In addition, as noted
earlier, these studies were selected to represent different methodological approaches and hence
do not represent a complete review of the approaches applied in the literature. More recent
and/or more complete estimates of time losses may be available for many of these illnesses.
A-9
-------
APPENDIX B: DATA SOURCES
This appendix provides additional information on some of the data sources that can be
used when valuing time losses. It first discusses data sources that provide information on daily
time usage both with and without illness. It then presents U.S. data on: (1) the probability of
employment and labor force participation; (2) earnings and benefits, including tax impacts; and
(3) growth in productivity over time. It concludes by describing an approach for assessing
survival probabilities for illnesses that last for more than one year. Most of the data sources
reported in this appendix are frequently updated, and analysts may wish to consult the sources
cited for more recent information.
As discussed in the main text of the report, regulatory analysts may transfer data from an
existing study of time losses rather than construct estimates from the types of national data
sources referenced in this appendix. In this case, the national data sources may provide a means
of verifying the values in particular research studies, or may allow the analyst to adjust the
values in the studies to reflect estimates for the population of concern. For example, for a
rulemaking with national impacts, the analyst may wish to use the national earnings data from
the sources cited in this appendix rather than the local values applied in a research study of a
particular population.
The national databases referenced in this appendix are typically administered by
government agencies and involve large sample surveys. Data from such sources usually can be
disaggregated in a number of different ways (e.g., by age, race, sex, or region). The agencies
that maintain these databases are staffed by individuals with detailed knowledge of the data sets,
who are available to answer questions or provide more detailed data than are reported publicly.
This appendix includes website addresses for most of the data sources; information on people to
contact for further information is generally posted on these sites.
B.l TIME LOSS DATA
The Center for Disease Control's National Center for Health Statistics (NCHS) publishes
several national databases containing information on work loss days, doctors visits, and hospital
stays for a variety of illnesses (for a list of all NCHS surveys, see: http://www.cdc.gov/nchs
/nhcs.htm). Published reports may provide summary statistics that are useful for the analysis of
time losses; however, in many cases analysts will need to acquire and analyze the raw data (often
for a fee) to derive the information needed. Key databases include the following.
• The National Health Interview Survey (NHIS) is a national cross-sectional household
interview survey of illnesses, injuries, impairments, and chronic conditions (see
http://www.cdc.gov/nchs/nhis.htm). It contains information on a variety of medical
costs, as well as data on activity limitations caused by chronic conditions, utilization of
health services, and other health topics. It is designed to collect data that can be used to
understand disability, to develop public health policy, to produce simple prevalence
B-l
-------
estimates of selected health conditions, and to provide descriptive baseline statistics on
the effects of disabilities. The NHIS is conducted on an annual basis and covers hundreds
of diseases.
The National Hospital Discharge Survey (NHDS) provides information annually on the
inpatient use of hospitals in the United States (see: http://www.cdc.gov/nchs/about/
major/hdasd/listpubs.htm). Data are collected on diagnoses, surgical and nonsurgical
procedures, and patient characteristics from a national sample of approximately 500 non-
Federal, short-stay hospitals. The information is abstracted from a sample of medical
records from each sample hospital for a total of about 270,000 records each year.
The National Nursing Home Survey (NNHS) is a continuing series of national sample
surveys of nursing homes, their residents, and their staff which provides information on
nursing homes from two perspectives — that of the provider of services and that of the
recipient (see: http://www.cdc.gov/nchs/about/major/nnhsd/nnhsd.htm). Data about the
facilities include characteristics such as size, ownership, Medicare/Medicaid certification,
occupancy rate, number of days of care provided, and expenses. For recipients, data are
obtained on demographic characteristics, health status, and services received. Data for
the survey have been obtained through personal interviews with administrators and staff
and occasionally through self-administered questionnaires in a sample of about 1,500
facilities.
The National Hospital Ambulatory Medical Care Survey (NHAMCS) is designed to
collect data on the utilization and provision of ambulatory care services in hospital
emergency and outpatient departments (see: http://www.cdc.gov/nchs/about/major/
ahcd/ahcdl.htm). Findings are based on a national sample of visits to the emergency
departments and outpatient departments of noninstitutional general and short-stay
hospitals, exclusive of Federal, military, and Veterans Administration hospitals.
National Survey of Ambulatory Surgery (NSAS) is designed to meet the need for
information about the use of ambulatory surgery services in the United States (see:
http://www.cdc.gov/nchs/about/major/hdasd/nhds.htm). Data are available on patient
characteristics including age and sex; administrative information including patient
disposition, expected sources of payment, and region of the country where procedure was
performed; and medical information including diagnoses and procedures performed
coded using the International Classification of Diseases, 9th Revision, Clinical
Modification (ICD-9-CM).
National Home and Hospice Care Survey (NHHCS) is a continuing series of surveys
of home and hospice care agencies in the United States (http://www.cdc.gov/nchs/
about/major/nhhcsd/nhhcsd.htm). Information is collected on agencies that provide home
and hospice care and about their current patients and discharges. Data are collected on
referral and length of service, diagnoses, number of visits, patient charges, health status,
reason for discharge, and types of services provided.
B-2
-------
The health economics literature includes a number of examples of how these data can be
used to estimate time losses, including the examples cited in Appendix A of this report. For
example, analysts can use the NHIS data to compare time losses for individuals with the illness
of concern to other individuals, to determine the time losses attributable to the particular illness.
Alternatively, data from these sources can be used to compile information on the number of
doctor visits and hospital stays associated with the illness to be used in estimating time losses
attributable to these activities.
In some cases, analysts may have data on the quantity of time lost (e.g., on the total
duration of the illness, or on the days of associated bed-rest), but not on the types of activities
affected (e.g., work vs. nonwork time). In these cases, data on normal time use allocation can be
used to infer the allocation of losses across each usage category. For instance, if data are
available on the number of days lost to illness, analysts can use normal time use allocation data
to estimate the allocation of these losses across market work, nonmarket work, leisure, and sleep
activities.
The allocation of normal time across different types of activities is an active area of
research and a number of time use studies have been completed internationally. Time use
researchers have developed detailed schemes for categorizing different uses of time. For
example, the categorization scheme for created by the National Research Council for a proposed
U.S. survey includes nine major groupings (personal care, employment activities, education
activities, domestic activities, care for dependent household members, purchasing activities,
voluntary work and care, social and community interaction, and recreation and leisure) which
are further subdivided into 99 subgroups, each of which is then subdivided into a number of
discrete categories.1
The National Research Council study lists over 50 major time-use surveys that have been
completed internationally.2 However, the majority of the studies completed in recent years
address countries other than the U.S., including Australia, the European Community, Japan, New
Zealand, and Canada. The most recent U.S. studies were completed by the University of
Michigan in 1981-1982 and by the University of Maryland in 1985; the U.S. Bureau of Labor
Statistics (BLS) is in the process of developing a new time use study and expects the results to be
available late in 2004.3
1 For an overview of available time use research as well as information on the proposed U.S. survey,
see: National Research Council, Time Use Measurement and Research: Report of a Workshop, Committee
on National Statistics, M. Ver Ploeg, J. Altonji, N. Bradburn, J. DaVanzo, W. Nordhaus, and F. Samaniego,
(eds.), Commission on Behavioral and Social Sciences and Education, Washington, DC: National Academy
Press, 2000.
2National Research Council (2000).
3See www.bls.gov/tus/ for more information.
B-3
-------
Ideally, time allocation would be determined based on a recent U.S. study because the
allocation of time can vary significantly across cultures and over time. However, the most recent
comprehensive U.S. study was conducted over 15 years ago and is not likely to reflect recent
trends.4 In contrast, Statistics Canada has completed a recent study that addresses a population
which may have time use patterns that are similar to the U.S.5 It seems reasonable to expect that
Canadian time use patterns will be similar to U.S. patterns due to the proximity of the two
countries and the extent of interaction between their populations. Thus analysts may wish to use
the Canadian results until more recent U.S. data are available.
Exhibit B. 1 summarizes the results of the Canadian study, which provides national
estimates for all individuals ages 15 and older (regardless of employment status) in 1998. The
hours per day estimates are based on a seven day week for all time categories. The study
allocates time usage across numerous categories, which are aggregated in the exhibit into the
three major categories (market work, nonmarket work and leisure, and sleep) used for the
calculations in this report.
4 EPA's National Human Activity Pattern Survey, conducted in 1992-1993, also provides data on time
use (see http://www.epa.gov/heasd/herb/hap.htm for more information). However, the easily accessible data
from this survey focus on time spent in selected activities and micro-environments for the purpose of
exposure assessment, and do not provide the comprehensive summary data necessary for the analysis of time
losses. For more information on this and related exposure studies, see: U.S. Environmental Protection
Agency, Exposure Factors Handbook (Final Report), Washington, D.C.: Office of Research and
Development, August 1997, Chapter 15 - Activity Factors.
Statistics Canada, "Statistics Canada, "Canadian Time Allocation: Average time spent on activities,
total population and participants, by sex," General Social Survey, 1989, 1999.
B-4
-------

Exhibit B-l


AVERAGE DAILY TIME ALLOCATION IN CANADA
(1998)
Time Category
Women
Men
Total Population
Market Work
2.8 hours
(11.67 percent)
4.1 hours
(17.08 percent)
3.3 hours
(13.75 percent)
Nonwork (nonmarket
work and leisure)
13.0 hours
(54.17 percent)
11.9 hours
(49.58 percent)
12.6 hours
(52.50 percent)
Sleep
8.2 hours
(34.17 percent)
8.0 hours
(33.33 percent)
8.1 hours
(33.75 percent)
Total (all activities)
24.0 hours
(100.00 percent)
24.0 hours
(100.00 percent)
24.0 hours
(100.00 percent)
Notes:
Detail may not add to totals due to rounding.
Market work includes paid work time only. Nonmarket work and leisure includes all other activities except sleep,
including unpaid work-related activities such as commuting time. Sleep includes night sleep only.
Includes all individuals aged 15 and older (regardless of employment status).
Source:
Statistics Canada, "Canadian Time Allocation: Average time spent on activities, total population and participants,
by sex," General Social Survey, 1989, 1999.
B.2 EMPLOYMENT AND WAGE DATA
The U.S. Bureau of Labor Statistics (BLS) provides well-established, detailed, and
widely used data on national employment and compensation. Because BLS collects data
throughout the year, these data can be used to calculate long-term averages or describe seasonal
employment patterns. BLS collects data on a number of different characteristics of the labor
market, including wage rates, employment rates, and employee benefits, by age, sex, race, and
occupation (see: http://www.bls.gov).
The Current Population Survey (CPS), a joint project of BLS and the Census Bureau, is a
monthly survey of approximately 50,000 households that provides data characterizing the labor
force (see: http://www.bls.census.gov/cps/cpsmain.htm). CPS includes information on the
number of employed and unemployed individuals in the civilian labor force, the number of
individuals not in the labor force, and average earnings.
The level of demographic detail available in on-line data sources and in print publications
tends to vary over time. CPS data are typically provided by age, sex, race, marital status, and
level of educational attainment. However, data broken out by age and sex are generally
published online only for the current year. Similar tables are available in written form in
Employment and Earnings, a periodical published by BLS. Additional data are collected and
B-5
-------
compiled internally by BLS and can be obtained by calling the Division of Labor Force
Statistics.
In interpreting these data, it is important to consider the definitions of the following terms
used in the CPS, which is the source of most of the data presented in this section.6 The CPS is
conducted monthly, and addresses the status of respondents in the week prior to the week during
which the survey interviews were conducted.
•	Employed: all persons who, during the reference week...(a) did any work at all (at least
1 hour) as paid employees, worked in their own business, profession, or on their farm, or
who worked 15 hours or more as unpaid workers in an enterprise operated by a member
of the family, and (b) all those who were not working but... were temporarily absent,
whether or not they had paid time off.
•	Unemployed: all persons who had no employment during the reference week, were
available for work, except for temporary illness, and had made specific efforts to find
employment in the 4-week-period ending with the reference week.
•	Labor force: all persons classified as employed or unemployed in accordance with the
criteria described above in the definitions of "employed" and "unemployed."
•	Not in the labor force: all persons...who are neither employed nor unemployed; includes
the retired, students, discouraged workers, people keeping house, the ill, and the disabled.
Exhibit B-2 presents BLS data on employment rates and rates for nonparticipation in the
labor force. These rates do not include the unemployed; however, the rate of unemployment can
be estimated by subtracting the sum of the rates in the table from 100 percent [i.e., for males
ages 16 to 19, 42.6 percent (employed) + 49.3 percent (not in labor force) = 91.9 percent; 100
percent - 91.9 percent =8.1 percent (unemployed) for males in this age group].
6 Detailed information on these and other definitions used by BLS is provided at http://www
.bls.gov/bls/glossary.htm.
B-6
-------
Exhibit B-2
U.S. EMPLOYMENT AND LABOR FORCE PARTICIPATION RATES, 2001
(as a proportion of the civilian noninstitutional population > 16 years of age)
Age Interval
Proportion Employed
Proportion Not in the Labor Force
Male
Female
Combined
Male
Female
Combined
All ages (16 +)
0.708
0.573
0.638
0.256
0.399
0.331
16 to 19
0.426
0.427
0.427
0.493
0.506
0.500
20 to 24
0.742
0.674
0.708
0.185
0.271
0.229
25 to 29
0.871
0.722
0.795
0.084
0.239
0.163
30 to 34
0.900
0.718
0.807
0.064
0.245
0.156
35 to 39
0.895
0.731
0.811
0.071
0.239
0.157
40 to 44
0.889
0.753
0.820
0.079
0.220
0.150
45 to 49
0.874
0.759
0.816
0.097
0.215
0.157
50 to 54
0.837
0.721
0.778
0.135
0.260
0.199
55 to 59
0.748
0.599
0.670
0.227
0.384
0.309
60 to 64
0.545
0.413
0.475
0.435
0.576
0.509
65 to 69
0.293
0.194
0.240
0.697
0.800
0.753
70 to 74
0.176
0.105
0.137
0.819
0.892
0.859
75+
0.082
0.033
0.052
0.916
0.966
0.946
Notes:
Annual averages for the year 2001.
Source:
Derived from data from the Current Population Survey, conducted by the Bureau of Census for the Bureau of Labor Statistics,
http://www.bls.gov/cps/cpsaat3.pdf as viewed September 2002.
Exhibit B-3 provides data on compensation rates. The starting point is BLS data on
median pre-tax weekly earnings in 2001 for full time employees from the Current Population
Survey. These data are adjusted to reflect employee benefits (the second set of columns) using
the average ratio of total compensation to wages and salaries in 2001 from the BLS. These data
are reported by BLS for private industry workers based on a sample of establishments. In the
third set of columns, pre-tax weekly earnings are adjusted downwards to net out taxes, using an
estimate of the year 2000 average tax rate based on data on household income before and after
taxes from the Current Population Survey (at the time this exhibit was prepared, year 2001 data
were not yet available for this variable).
B-7
-------
Exhibit B-3
MEDIAN WEEKLY EARNINGS BY AGE AND SEX, 2001
(Full-Time Wage and Salary Workers)
Age
Interval
Median Pre-Tax
Weekly Earnings
Median Pre-TaxWeekly
Earnings Plus Benefits
Median After-Tax
Weekly Earnings
Male
Female
All
Male
Female
All
Male
Female
All
All ages
(16+)
$672
$511
$597
$936
$712
$832
$553
$420
$491
16 to 19
319
287
304
444
400
423
262
236
250
20 to 24
410
375
395
571
522
550
337
308
325
25 to 29
589
505
546
820
703
761
484
415
449
30 to 34
657
523
601
915
728
837
540
430
494
35 to 39
735
539
647
1,024
751
901
604
443
532
40 to 44
773
551
669
1,077
768
932
636
453
550
45 to 49
788
585
683
1,098
815
951
648
481
562
50 to 54
811
592
705
1,130
825
982
667
487
580
55 to 59
793
551
659
1,105
768
918
652
453
542
60 to 64
714
518
610
995
722
850
587
426
502
65 to 69
575
372
478
801
518
666
473
306
393
70+
512
372
459
713
518
639
421
306
378
Notes:
Pre-tax wages plus benefits are derived using an average ratio of wages to wages plus benefits of 1.393.
After-tax wages are derived using an average tax rate of approximately 17.8 percent.
Source:
Pre-tax weekly wages: Personal communication between Michel Woodard Ohly, Industrial Economics, Incorporated and Mary
Bowler, Bureau of Labor Statistics, unpublished tabulation from the Current Population Survey, September 2002.
Benefits adjustment: U.S. Census Bureau, "Table 626. Employer Costs for Employee Compensation Per Hour Worked: 2001,"
Statistical Abstract of the United States: 2001, November 2001.
Average tax rate: U.S. Census Bureau, "Table RDI-1. Household Income Before and After Taxes: 1979 to 2000," March 2002
(http://www.census.gov/hhes/income/histinc/rdi01 .html as viewed September 2002).
The exhibit reports median rather than average earnings (although both are available
from BLS) because in most cases analysts may wish to use the median values in regulatory
analysis. The distribution of income in the U.S. is highly skewed, due to the small number of
people who are extremely highly compensated, hence the average is significantly above the
median. Using the median is consistent with the notion that the small fraction of the U.S.
population affected by most rulemakings are likely to be better reflected by the median (which is
in the center of the distribution) than by the mean value (which is closer to the upper tail of the
distribution).
B-8
-------
Because of differences in the level of aggregation in the reported data, the employment
rates in Exhibit B-2 include part time workers whereas the compensation data in Exhibit B-3 are
for full time workers only.7 Combining these data in the calculations will overstate earnings as a
result. In contrast, the data on hours worked in Exhibit B-l include all individuals regardless of
whether they are employed, unemployed, or out of the labor force.8 Analysts should take care to
consider these differences when combining data from different sources and determine the
appropriate level of aggregation on a case-by-case basis, depending on the level of detail desired
and the time and resources available for the particular analysis. Any potential bias introduced by
the data limitations would be discussed in presenting the results of the analysis.
Analysts should also note that the result will be median value per hour across all hours
worked during the reporting period (e.g., a week or a year). The theory of decreasing marginal
utility (discussed in the main text) suggests that, in a perfectly functioning competitive labor
market, marginal values will be less than median or average values. However, illness often leads
to greater than marginal changes in activities. Hence the impact of using the median to value
time losses is unclear and may vary depending on the magnitude of the losses.
BLS also collects information that characterizes national trends in the prices of various
goods and services and in the productivity of labor. Data on the annual change in labor
productivity is available on the BLS website. Exhibit B-4 contains the BLS data on the
historical rate of growth in output per hour, which can be used to predict the increase in real
wages over time if needed for long term illness or for regulatory analyses that predict impacts far
into the future.
Exhibit B-4
ANNUAL PERCENT CHANGE IN OUTPUT PER HOUR
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Ten-Year
Average
3.7
0.5
1.3
0.9
2.5
2.0
2.6
2.4
2.9
1.1
2.0
Notes:
10 year average is for 1992-2001; the 40 year average (1962 - 2001) is also 2.0 percent.
Data reflect annualized rates for output per hour for nonfarm business.
Source:
Derived from data from the Major Sector Productivity and Costs Index, maintained by the U.S. Bureau of Labor Statistics,
http://www.bls.gov/lpc/ as viewed September 2002.	
7 For Exhibit B-3, the ratio of total compensation to wages is for full-time workers, consistent with
the data on median earnings. However, the same data source also includes information on part-time workers
and for all workers.
8Data on hours worked for employed individuals only are available in: U.S. Census Bureau, "Table
582. Persons At Work by Hours Worked: 2000," Statistical Abstract of the United States: 2001, November
2001.
B-9
-------
B.3 SURVIVAL STATISTICS
The period over which time losses accrue will depend on the life span of affected
individuals as well as the duration of the illness, particularly in cases where the health effects are
experienced over a multi-year period. A simple approach to determining this time span could
involve estimating the time that elapses between incidence and death on average for members of
the affected population (e.g., based on national U.S. data).
A more complex approach involves considering the probability that affected individuals
are likely to survive for different time periods after the onset of illness. This approach requires
considering both the likelihood of death at a particular age and the likelihood of surviving to that
age. For example, assume an illness starts at age 30 and continues for the remainder of an
individual's life. Whether the individual accrues related time losses at age 70 depends on (a) the
likelihood that he or she survives to age 70; and (b) the likelihood that he or she dies at age 70.
Regardless of whether a simple or complex approach is used, the analysis should
compare the life expectancy of a healthy individual to the life expectancy of an ill individual to
estimate the net time loss. If losses were to be assessed for a fatal case of illness, the life
expectancy for the ill individual should be that for individuals who die of the disease. For
nonfatal cases, this life expectancy should be based on individuals with the disease who die of
other causes. Even for these nonfatal cases, life expectancy differ from that of an individual
without the illness, if the illness is associated with other conditions that can result in premature
mortality.9
EPA has developed an approach for calculating these survival probabilities which is
described in detail in its Cost of Illness Handbook. These calculations are quite complex, and
analysts interested in using them should consult the Cost of Illness Handbook for detailed
information. An example of these calculations for stomach cancer is provided below.10
For cancers, the survival rate calculations rely largely on data from the National Cancer
Institute's Surveillance, Epidemiology, and End Results (SEER) program.11 Data on other
9Care must be taken to ensure that the definition of a fatal case used in the analysis of lost time is the
same as that used in the risk assessment. In some cases, fatal cases may be estimated based only on the
disease of concern; in other cases, fatalities from associated diseases may also be considered.
10U.S. Environmental Protection Agency, Cost of Illness Handbook, February 2001, pp. II.2-7 to II.2-
16.
nSee, for example, Ries, L.A.G., et al. (eds.), SEER Cancer Statistics Review, National Cancer
Institute, 1973-1998, Bethesda, MD, 2001. More information on this data source is available at
http://seer.cancer.gov/.
B-10
-------
illnesses are available in the Cost of Illness Handbook, the general health science literature, and
data sources maintained by non-profit health research institutions (such as the American
Diabetes Association).
The starting point for these calculations is the relative survival rate for the illness of
concern. This rate is the number of observed survivors among patients, divided by the number
of "expected" survivors among persons with the same age and gender in the general population
(observed divided by expected). The relative survival rate takes into account competing causes
of death, which increase with age. For example, the relative survival rate for stomach cancer
patients during the first year post-diagnosis is 46 percent. This value indicates that a person with
stomach cancer would have, on average, a one-year survival probability that is 46 percent of
someone of the same age and gender in the general population. Because these survival rates are
comparative (i.e., the analyst needs to know the survival rate for the general population to
estimate the likelihood that a stomach cancer patient will die in a particular year), a number of
calculations are needed to translate these rates into estimates that can be used in a particular
analysis of the costs of illness or time losses.
The results of such calculations are presented below for stomach cancer. For this illness,
the average age at diagnosis is approximately 70 years, and survival rates are estimated for the
first 10 years post diagnosis; i.e., for affected individuals aged 71 to 80. The data indicate that
most stomach cancer patients who die of the disease will do so by the tenth year, whereas
stomach cancer patients who die of other causes have slightly lower survival rates than the U.S.
population. Stomach cancer is fatal in over 80 percent of all cases.
These survival rates are provided in Exhibit B-5. Column (1) provides the survival rates
for members of the general population. Columns (2) and (3) provide the probabilities that should
be used for ill individuals; the rates in Column (2) apply to fatal cases (persons with the disease
who die of the disease) and the rates in Column (3) apply to nonfatal cases (persons with the
disease who die of other causes). Each column provides two estimates, one for the likelihood of
surviving that year and one for the likelihood of dying in that year.
B-ll
-------
Exhibit B-5
CONDITIONAL SURVIVAL AND MORTALITY PROBABILITIES
Years Post-
Diagnosis
(1)
General Population
Conditional probability of:
(2)
Stomach Cancer Non-
Survivors (fatal cases)
Conditional probability of:
(3)
Stomach Cancer Survivors
(nonfatal cases)
Conditional probability of:
Surviving
through the
«th year
Dying
during the
«th year
Surviving
through the
«th year
Dying of
stomach
cancer
during the
«th year
Surviving
through the
«th year
Dying of
some other
cause during
the «th year
1
0.973
0.027
0.343
0.657
0.894
0.106
2
0.945
0.028
0.188
0.155
0.834
0.060
3
0.915
0.030
0.121
0.067
0.787
0.048
4
0.884
0.031
0.082
0.039
0.745
0.042
5
0.852
0.032
0.057
0.026
0.706
0.039
6
0.817
0.035
0.039
0.018
0.669
0.037
7
0.782
0.035
0.025
0.014
0.634
0.035
8
0.745
0.037
0.015
0.010
0.600
0.034
9
0.707
0.038
0.007
0.008
0.567
0.033
10
0.667
0.040
0.000
0.007
0.535
0.032
Source:
U.S. Environmental Protection Agency, Cost of Illness Handbook, February 2001. General population probabilities are
derived from data presented in Table II.2-3. Probabilities for stomach cancer patients are taken directly from Table II.2-7.
B-12
-------
-------