066?
      if
       IV
U
          i
          Oi
        Regulatory Impact Analysts

                     of

The National  Ambient Air Quality Standards

                    for

             Particulate Matter
                                        Second Addendum
                                         December 1986
                Prepared  by:
   Strategies and Air Standards Division
    U.S. Environmental  Protection Agency
        Research Triangle Park, N.C.

                 HEADQUARTERS LIBRARY
                 ENVIRONMENTAL PROTECTION AGENCV
                 WASHINGTON, D.C. 20460  . :•

-------

-------
                                    1-1                        -

 I.   Introduction
      This  addendum to the Regulatory  Impact Analysis  (RIA)  of the National
 Ambient  Air  Quality Standards  (NAAOS)  for Particulate Matter (PM) reports
 on  new analyses "that have been conducted  since the March 1984 proposal
 of  those standards.  The  original  PM  RIA  (dated February 21, 1984)  was
 prepared to  fulfill the requirements  of Executive Order 12291 (E.G. 12291).
 It  attempted to quantify  and  inform the public of the costs, benefits,
 and economic impacts of various NAAQSs for PM.
      During  the Agency's  own  internal  review cycle and during subsequent
 review by  other Agencies  and  the public a number of questions were  raised
 regarding  the underlying  data  bases and analyses discussed  in the RIA.
 Many of.the  questions raised  centered  on  the cost and environmental
 impact analyses.   To address  a number  of  these questions, the Agency
 modified the cost model used  and made  other more limited changes  to the
 benefits analysis.  The number and extent of the changes were constrained
 by  the underlying model structure  and  data limitations.  This addendum
 summarizes the changes made and presents  the results  of new analyses.
      It  should be noted that,  in proposing the revised NAAOSs for
 particulate  matter, the agency solicited  comment on the role, if  any, of
 economic analyses in setting  secondary NAAOS.   Public commenters  were
 divided  on the general question of whether economic studies should  be
 used. Even  among those who favored the use of such studies there was a  .
 significant  body  of opinion which  held that the PM RIA was  not  of sufficient
 quality  to be used in standard setting.   The Agency has evaluated the
 revised  analyses  and believes  that despite the very significant improvements
.made, there  remain fundamental  questions  regarding certain  aspects  of the

-------
                                   1-2
methodology used.  As a result, the Agency has not considered this RIA in
the review of the PM NAAQS.
     This addendum is organized similar to the RIA itself.  Section II
below addresses changes made in the Cost and Environmental analyses.
Section III reports on changes in the Benefits Analysis, while Section IV
presents the Cost Benefit analysis.  Inasmuch as no changes were made to
the Economic analysis this topic is not addressed in the Addendum.

-------
                                II-l
II.  COST AND ENVIRONMENTAL IMPACTS
     A.  INTRODUCTION
         The original RIA released in 1984 was based on a county by county
air quality and cost analysis.  The first steps in the process involved a
county level projection of air quality, a county wide roll-back analysis
of the emission reductions needed to meet the current or alternative PM
standards, and an estimation of the cost of those emission reductions.  A
simplified outline of this process is presented below.  A more detailed
description is contained in the 1984 particulate matter RIA (EPA, 1984).
     1) An annual and 24-hour average design value for all counties with
measured ambient air quality data was obtained.  Design values were
always in terms of TSP.  A single national ratio was used to convert
ambient TSP values to ambient PMiQ.  The design values generally reflected
1977-78 conditions.
     2) In each study county an emissions inventory of point and area
sources was assembled.  The inventory is also nominally representative of
1978 conditions.  A PM^g inventory was derived from the TSP emissions data
using source category specific ratios.
     3) Economic growth estimates were used to project emissions growth
and a linear roll back model used to estimate the impact on future air
quality.  The use of roll  back presumes that each source in a county
contributes to the design value proportional to its emissions.  Clearly
this represents an extreme simplification but was dictated by the necessity
of completing a national analysis.
     4) For each county projected to be in non-attainment in Step 3 above
a source specific control  strategy was derived.  A list of control  options
and associated costs was developed for each source.  The expected air

-------
                                11-2

 quality improvement for each source/control  option combination was then
 estimated using linear roll  back.  Then on a county wide basis, control
 options were selected until  the standard under consideration was met.
 Source specific options were selected on a least cost basis (i.e., lowest
 dollar per microgram (ug) of air quality improvement).
      5) In as many as half the counties analyzed the application of all
 controls did not result in full attainment.   The major reasons for this
 "residual" non-attainment include an incomplete baseline inventory, a
 predominance of sources for  which no control  options were available in
 the options file, and difficult air quality  problems.  On a county basis
 the additional  cost of full  attainment was estimated by multiplying the
 cost of the strategy by the  ratio of the remaining air quality improvement
 needed to the air quality improvement already achieved (this procedure was
.known as Scenario B).
      This entire process contained many analytic assumptions and used data
 bases with known limitations.   The resulting  estimates of non-attainment
 and cost were thought to capture  the relative differences among  various
 standards.  However, the original  RIA contained a  thorough discussion of
 the limitations and possible biases in the analysis (EPA, 1984).  By
 and large outside reviewers  of the RIA raised the  same limitations  and
 biases in their comments. The major concerns can  be summarized  as  follows:
      1) New Source Control:  In the previous  work  all  new sources  regardless
 of type were subjected to the  same control efficiency.
      2) Least Cost Control Strategy:   The  algorithm used  in  the  1984  analysis
 was a relatively crude approximation of a  true least  cost solution.   Simple
 inspection of the control  strategies in some  counties  revealed lower  cost
 solutions.

-------
                                 II-3
      3)  Accomodative SIP:  In the 1984 analysis, air quality projections
 were  made  to  1995,  but it was assumed that all controls were put in place in
 1989. The controls were made sufficient to  "accomodate" growth.  In many
 instances  this  resulted in "over control" in the first years of the strategy.
      4)  Cost  of Residual Non-Attainment:  The use of national, average
 cost  per microgram  reduction in the  residual non-attainment calculation
 was a source  of concern.
      5}  Age of  Emissions and Air Quality Data:  In the 1984 RIA the air
 quality  and emissions data were already at least six years old and they are
 now at least  eight  years old.
      6)  Economic Growth Factors:  The economic growth factors used in the
 earlier  work  were based on mid-1970's data and indicated positive growth in
 many "basic industries including iron and steel.
      A number of other technical issues were also discussed in the RIA and
 raised by  outside commenters.  This  section of the addendum discusses the
 changes  which were  made to the analytic system and presents revised results.
      B.  CHANGES IN THE COST ANALYSIS
         Due  to limitations imposed  by the structure of the model as well
 as resource limitations not all the  concerns raised during the comment
'period could  be addressed through changes in the analytic system.  However,
 many  of  the major, concerns were addressed directly, and still others were
 addressed  through sensitivity analyses.  This section addresses the changes
 made  to  the RIA model.  Sensitivity  analyses are discussed in Section II.0.
 below.
      1}  New Source  Control:  In the  previous analysis it was assumed that
 all new  sources would be controlled  to 99% except in the case of the iron
 and steel  industry  where a new source control efficiency of 98% was assumed.

-------
                                 II-4
While this assumption greatly simplified the original analysis, it failed
to recognize that new source control efficiencies vary from locality to
locality and can become quite stringent depending on local air quality.
     In this revised analysis an effort was made to model this variation.
In essence, four levels of new source control were specified.  These were:
     *  Level 0:  The base year controls found in the RIA inventory
        constituted Level 0.
     *  Level 1:  The controls required by an applicable New Source
        Performance Standard {NSPS) were used at this level.  Only sources
        controlled by NSPS were controlled at this level.  The control
        efficiencies were taken from EPA's NSPS Cost Effectiveness File and
        from the Background Information Documents (81Ds).
     *  Level 2:  The control required here was either the most stringent
        control option used in the RIA for existing sources or an estimate
        of Reasonably Available Control Technology.  This level ensures
        that growth and replacement sources will be controlled at least as
        stringently as existing'sources subject to a control strategy.
     *  Level 3:  The third level  of control was set equal to the estimated
        Lowest Achievable Emission Rate (LAER).  The LAER estimates were
        drawn from the 8ACT/LAER Clearinghouse.  LAER is required by the
        Clean Air Act of all new or modified sources in non-attainment
        areas.
     The model is set up to run with any of the four levels specified as
the most stringent level of new source control.  In practice all  of the
runs reported here had Level 3 (i.e., the most stringent level) specified.
At this level, the model produces  a report on the number of non-attainment
counties under each of the lower control levels.  Table 1 below summarizes

-------
                             II-5
this data for the promulgated standard (PM^o 50/150).  More detailed data
concerning this procedure are available in the supporting documentation
(Argonne 1986).  As might be expected the data show that increasingly
stringent new source controls bring down the number of counties estimated
to be in non-attainment.
                                Table II.B.I
                                PMlQ 50/150
       NonrAttainment Counties Under Alternative New Source Controls
                              I	Year
Control Level
     0
     1
     2
     3

     2) Least Cost Control and Multiyear Strategies:  In the original
analysis, the model chose control options on a lowest cost per yg/m^ of air
quality improvement basis.  In other words the model always selected the
option which resulted in the lowest dollar cost per Mg/m3 of air quality
improvement first.  Inspection of the intermediate output files revealed
that this process did not always result in the lowest overall  strategy
cost.  Most often this occurred when the model selected a number of cost
efficient controls at sources which resulted in small total air quality
improvements and then choose a less efficient option which produced a  large
improvement more than sufficient to meet the NAAQS.  In such situations the
standard might have been met at lower total  cost by selecting  the option
with the large air quality improvement first even though it was
less efficient in a cost per pg/m-* sense.
1989
329
288
231
188
1991
370
319
241
190
1993
421
365
261
191
1995
463
402
275
. 197

-------
                              .  II-6
     To remedy this problem, a new "least cost" algorithm called the Maximum
Cost Reduction (MCR) algorithm was developed.  The MCR algorithm is superior
to the previous algorithm in the sense that county wide control costs are
always less than or equal to those produced by the previous algorithm.  In
broad outline the MCR algorithm involves the following steps:
     1) The most stringent control option available is assigned to each
source in a given county.  If this does not result in attainment, the
county is considered to be in residual non-attainment and the model moves
to the next county.
     2) If Step 1 results in attainment, the model considers the next less
stringent control option at each source.  [N.B. Many sources have only one
control option and such sources are therefore not considered at this step.]
The model then determines which of these options could replace the most
stringent option without causing a violation of the NAAQS.  Each possible
replacement represents an allowable relaxation.  If no allowable relaxation
can be found then the set of most stringent options is the solution.
     3) If more than one allowable relaxation is found the model  determines
the one which if implemented would result in the greatest overall  cost
reduction (the MCR relaxation).  This relaxation is implemented and the
model loops back to Step 2.
     Figure 11.B.I presents a schematic flow.of the algorithm for the multiyear
strategy considered. -It was noted above that  in the original  RIA work an
"accomodative SIP" approach was used.  This approach often resulted in
overcontrol  in the early years  of a strategy in order to  accomodate growth
in the out years.  To overcome  this problem, a year-by-year approach to
strategy development was employed.  If the  first and final, years  are the

-------
                         II-7
                |   Year-*-First Year
                          £
                       Implement
                   Maximum Controls
                          ±
                  Compute Atr Quality
              C
                          1
        Any Violations Of
         AQ Standards?
                              Yes
                            No
                    Determine The
                    MCR Relaxation
Implement
   The
Relaxation
Yes
               1
     Was A Relaxation Found?)
                            No
                   Current Controls
                   Are The Solution -
                 Process The Solution
Year-*-Year* 1
                          I
                   Year-Final Year?
                           J>
                    Final Processing
                  Qo To Next  Courty
                                                    County Is
                                                    Intractable
                                                     Maximum
                                                    Control is
                                                   The Solution
                      Figure  II.B.I

-------
                                I1-8  .                         .

same, then the MCR algorithm operates as described in the three steps outlined
above.  Where a multi-year strategy is required, the problem is somewhat
more complicated.  In such situations, it had to be decided how to handle
sources controlled in an earlier year.  It was assumed that once  a control
option was applied to a source in a given year it could not be removed in
some later year.  An exception was made in the case of sources where more
than one control option was available.  For such sources, it was  assumed
that a control option could be removed in a later year to be replaced by a
more stringent option.         .
     3)  Cost Computation:   The use of the MCR algorithm in a multi-year
mode requires that costs be handled in an appropriate manner.  Outlined
below are the steps followed in computing costs:
     0  If the candidate relaxation would lead back to base-year
        controls, then the associated cost reduction.was computed
        from:
         A C(t) = * (t)-COM + CAP-CRF(T)]
        where t is the year for which the strategy is being developed;
         + (t) = 1 - K(t)  + bL(t)  + aM(t)
        The quantities K,  L, M depend on  the projection year t  and are
        computed as follows:
             Define  G = (1  + g)   t-1
             and     R =  l-(l-r)    t
        where g is the growth rate (fraction per  year),  r the replacement
        rate (fraction per year),  and A t is the  difference between the
        projection year t  and the  year.   Then K,  L,  M can be defined  as:

-------
                             II-9
      K  - K  if G _>  -R,  and  K  =(G)  if  -R  >  G
      L .  • K  if G _>  0, and L = K  -  (G)  if U >  G
      fl  * G  if G _>  U, and M = 0  if 0  > G.  T  is the equipment
lifetime, taKen to be  15 years.   OM  and CAP  are the operating/
maintenance and capital costs for the control option, and CRF(N)
is the  capital recovery factor  for N  years:
              R(1+R)N
      CRF(N) =
R denotes the real interest rate.
If the candidate relaxation would lead to an option not previously
implemented, then the associated cost reduction was calculated from:
 A C(t) - * (t)'  (0"k + CAPk*CRF(T))-
                  {OMk_1+CAPk-1«CFR(T))
where the relaxation is assumed to be from option k to k-1.

If the option to which the candidate relaxation would lead has
already been imposed in a previous year, then the associated
cost reduction was calculated from:
   C(t) = * (t) •   (OMk+CAPk-CRF(T))
                    + CAPk ,-CRF(T) -      CRF(T)
                                        CRF(T + A t) -fc
where A t = (current year)- (year in which option k-1 was imposed),
and fc denotes the fraction of the capital cost of option k-1 that
is assumed to have been spent on equipment, such as ductwork, that
can be used as part of the option k system.  In this analyses,
fc was taken to be 0.5.

-------
                                11-10
     In addition, In order to facilitate proper comparison of control  costs
with the dollar value of benefits computed from air quality improvements
over the analysis period, it was necessary to truncate costs at the end of
the analysis period, even though the control  equipment lifetime was assumed
to be such that controls would remain in place beyond that point.   The
following procedure was used to compute yearly annual 1 zed costs:
     1.  If a source is controlled only once  during the analysis period
     (in year tj), then the associated annual i zed  cost is given by:
          AC = OM + CAP*CRF(T)
     The effective bef ore-tax annual i zed cost in year t is:
                    0        if t < ti
          BTAC(t) =
                      (t)*AC if t!  £ t
     2. .If a source is controlled twice during  the  analysis  period  (in
     years tj_ and t2,  with A t = tg - t]. >  0), then  define AC^ and AC2 by
          ACj - Ofy +  CAP^CRFt  A t)
                    +-f'CAP1'[CRF(T+  A t) - CRF( A t)]
          AC2 = OM2 +  CAP2*CRF(T)
                    +  f'CAP1*[CKF(T+  At)- CRF(T)]
BTAC(t) is then given by:
                   AC  if
BTAC(T) =  * (t) .
                                       <  t  <
                             AC2  if  t2 _< t
     Given the BTAC(t)  values for each year t  in the analysis period,
     the discounted present  value associated with each is given by:

                  8TAC(t

-------
                                  11-11
     and the total cumulative present value by:
        pvTOTrt
     Two control options at most were available for any given source
     in the current study, and the expressions given above cover all
     possible cases.
     4)  Cost of Residual Non-Attainment:  In the original 1984 RIA as well
as in this analysis, there were a number of counties which remained in
non-attainment even after the imposition of all available control options.
These counties were defined as being in residual non-attainment.  To estimate
the cost of full attainment, a procedure known as "Scenario B" was developed.
In essence, "Scenario B" assumed that the unit cost of the remaining air   .
quality improvement needed in any county would be equal to the national
mean unit cost incurred in the strategy.
     A number of. reviewers took exception to the use of the mean unit cost
measure and suggested that the marginal  unit cost was a more appropriate
measure.  Moreover, it was suggested that the cost measure (whether mean
or marginal) should be defined at a county level.
                                                   {
     There are a number of reasons to believe that, despite its conceptual
appeal, the marginal cost measure would  be inappropriate.   The use of a
marginal multiplier rests on the notion  that the last increments of
control obtained from a source are normally the most expensive.   In situations
where the reduction of residual  non-attainment depends on  more stringent
control of already controlled sources, the marginal  multiplier would be
appropriate.  However, there are several more common reasons  for residual
non-attainment to occur such as:
     a) the predominance of large uncontrollable area sources especially
        area sources other than  paved roads.  These sources may  be real

-------
                                  11-12
     (e.g., rural fugitive dust) or may be an artifact of the emission
     inventory..
     b)  the predominance of point sources not listed in the control
     options file and which the model recognizes as "uncontrollable."
     In fact, controls do exist for many of these sources and a sensitivity
     analysis was conducted addressing this issue (see Section II. D).
     c)  the utilization of rollback in place of dispersion modeling or
     receptor modeling.
In situations such as those outlined above, the reduction of residual
non-attainment would involve control of sources not addressed by the control
strategy.  That being the case the use of the marginal multiplier would most
likely be inappropriate.
    .Nevertheless, a limited sensitivity analysis was undertaken to determine
what impact the use of a marginal  multiplier would have.   The marginal
multiplier (y(n)) was defined as:
              tn
     u(n)= MAX  A BTAC.
       m         4 XW

     where
           A BTAC » incremental  Before Tax Annual i zed Cost
           & x(n) = incremental  decrease in the ntn measure of air quality
Essentially, pn) represents the cost per unit air quality for the most
               m
cost-ineffective option available in an area.   As a further refinement,
the marginal multiplier was weighted by the ratio of emissions from
sources already controlled to total  emissions  from all  sources.  A mean
multiplier (p(n))  was weighted by the ratio of emissions  from sources
              a
never addressed to emissions from all  sources.  This approach  was then
tested for a sing-le state, Alabama.   As can be seen in  Table II. B. 2,

-------
                                   11-13

even this refined treatment of the marginal multiplier results in an increase
"Scenario B" costs by approximately a factor of 1400.  These results are
quite unreasonable.  In view of these results as well as those presented
in Section II.O."3, and because the underlying logic of the model favors the
use of the mean multiplier, all the results presented in Section II.C.
below use the mean multiplier.  It should be noted in conclusion that
multipliers are now county specific.  This represents a change from earlier
RIA where national averages were used.

                                Table II.B.2
                 Comparison of B-Scenario Annualized Costs*
                                  Alabama

1989.
1991
1993
1995
Marginal u
37,000
53,000
72,000
92,000
He an u
26
39
52
67
1)  106 1984 dollars

-------
                                  IV-14

C.  RESULTS •_
     This section of the Second Addendum presents the revised cost and other
Impact estimates produced by the model.  Non-attainment estimates from
the model are presented along with non-attainment estimates derived from
a probabilistic treatment of a more recent air quality data base.  In
addition national cost and emission reduction estimates are summarized.
Finally industry specific and regional level results are discussed.
     1.  NON-ATTAINMENT
         Table II.C.l summarizes the number of non-attainment counties by
year for each of the strategies analyzed.  It is readily apparent that
the current suite of NAAOS (i.e., TSP (75,150)) results in the greatest
number of non-attainment areas.  The current primary TSP is next most
stringent.  Compared to the 1984 analysis using the old model these
numbers are'some 5-10% lower.  The only exception is the TSP (75,150)
case, where the estimate presented here represents less than a 1% change.
     It was noted in Section 11,A that the model used in this analysis
projects non-attainment (as well as costs) from a 1978 data base.  PMjn,
levels are estimated from TSP levels using a single national ratio of 0,46.
This ratio  is the average found at a number of sites with co-located PHjQ
and TSP instruments.  For comparative purposes, Table II.C.2 presents
projected non-attainment based on the methodology contained in EPA's
probability guideline (Pace and Frank, 1984).  This methodology is designed
to estimate the probability of exceeding a given PM^o level from TSP
data.  It makes use of a distribution of PM10/TSP ratios.  The results in
Table II.C.2 are based on an analysis of 1983-1985 data.

-------
                                  11-15
                               Table II.C.I
                Number of Non-Attainment  Counties  by  Year*
Scenario
TSP (75,150)
TSP (75,260)
PM10 (50,150)
PM10 (65,250)
PM10 (50,150) TSP (90,-)
PMiQ (65,250) TSP (90,-)
1989
480
275
188
55
205
138
1992
500
281
188
58
210
145 .
1995
524
285
197
66
223
158
^Based on a projection of 1978 air quality  levels  forward  to 1989 through
 1995.  Data set included 1231 counties.
                               Table II.C.2
               Estimated Number of  Non-Attainment Counties
                           By Probability Level1   .
Scenario	  p >  0.95	p > 0.5
TSP (75,150)
TSP (75,260)
PM10 (50/150)
PM10 (65/250)
3602
1362
76
9


191
35
1Based on 1983-1985 TSP air quality  data in 661 counties.
^TSP non-attainment estimates  are deterministic and  have no estimated
 probability.

-------
                                  11-16

Any straightforward comparison of the results from this probabilistic
procedure to those obtained from the RIA model is impossible.  In the
first place the probabilistic results represent 1984 conditions while the
RIA represents a projection to 1989-1995.  Moreover, the RIA results are
deterministic and based on a single national ratio while the others are
probabilistic and based on a full distribution of ratios,  with these
substantial caveats in mind, it can be seen that the estimated TSP non-
attainment levels are some 25-50% lower using the 1983-85 data and comparing
to the 1989 RIA estimates.  This comparison of TSP nets out the two
different PMjg/TSP conversion procedures.  Therefore the differences are
due to 1) the projection to 1989 made in the RIA and 2) differences in
the underlying data.  For PMjo a 2% increase in non-attainment is seen when
one compares the p >_ 0.5 estimates to the RIA estimates for PMjQ 50/150.
On the other hand at PMio 65/250.a 34% decrease in non-attainment is seen.
The p ^0.5 level was selected for comparison because the RIA ratio of
0.46 represents the mean of a distribution of ratios.
     Recent trends in TSP air quality suggest that levels have been
coming down with a leveling off in the mid-?1980's.  Emissions have shown
similar trends.  With this as background it does not seem likely that
actual growth between 1984 and 1989 will make the RIA estimates of non-
attainment correspond to actual.  Thus, it appears that the growth pro-
cedures used in the RIA are resulting in an upward bias in the non-attainment
and therefore cost estimates.  Other data pertinent to this conclusion
are presented in the Sensitivity Analyses (Section II.D).
     A more troubling comparison results from contrasting the relative
stringency of the current primary TSP NAAQS with the PM^Q (50/150) case.
Using the RIA data one would conclude that the TSP standard is some 45%
more stringent.  On the other.hand, the probabilistic assessment indicates

-------
                                   n-17

that the TSP primary NAAQS is about 30% less stringent (p >_0.5).  EPA
1s undertaking an assessment of the limited number of sites with co-located
instruments to determine the reliability of the probabilistic method compared
to actual data.
2.   National Costs
     Table II.C.3 presents a summary of the estimated national Before Tax
Annualized Costs (8TAC) and for the Discounted Present Value (DPV).  Since
the model now simulates a multi-year strategy, with controls being put in
place only when needed, the results are calculated for each year.  Table
II.C.3 presents national results for three years.  More detailed results are
available in the supporting documentation (Argonne 1986).  In all cases, it
was assumed that the standards would be implemented in 1989 and maintained
through 1995.  Before turning to a discussion of these results, several
caveats are in order.  In the first place, these cos.ts represent the cost
of going from 1989 projected air quality levels to attainment and maintenance
through 1995.  They do not include capital investments made in earlier
periods.  To place the costs in perspective, the 1984 RIA estimated that
the pre-1978 capital costs were on the order of $11 billion.  [1978 was
selected as a benchmark because the baseline air quality data used in this
analysis were generally representative of 1978 conditions.]  In addition,
it does not include the cost of new source control.  Even though new source
controls improve air quality and make the job of attaining a standard
easier, the requirements for these controls are generally independent of
the NAAQS {e.g., NSPS, PSD).
     The costs in Table II.C.3 include both the cost of the control strategy
and the Scenario B costs (i.e., the cost of reducing residual  non-attainment).
As can be seen, the-suite of TSP standards would be the costliest to attain.

-------
                                 11-18

                                Table II.C.3
               Estimated Nationwide Costs Including Reduction
                   of Residual  Nonattainment (Scenario B)
Nationwide Cost Including Reduction of
Residual Nonattainmenta»btC
BTAC(106$/yr)
Scenario
TSP(75,150)
TSP{75,260)
PM10(50,150)
PM10(65,250)
PM10(50,150) TSP(90,-)
PM10(65,250) TSP(90,-)
1989
1375
761
572
151
690
568
1992
1385
824
622
182
694
567
1995
1430
851
640
193
736
606
19b9
776
430
323
85
390
321
DPV(10°$)
1992
587
349
264
77
294
240
1995
456
271
204
61
234
193
Total0
4193
2418
1859
528
2110
1733
aBTAC:  Before tax annualized cost.
 DPV:  1983 discounted present value.

bAll  costs are in first-quarter 1984 dollars.

cValues were computed for each year from 1989  to  1995.'
 Only three yearly values are tabulated.

dTotal  for seven year period, 1989-1995.

-------
                                  11-19

The inclusion of the secondary component (150 ug/nr*, 24-hour TSP) results
in a 73% increase over the cost of the current primary alone (TSP (75,260)}.
In comparison to the current standards, all of the PM^o alternatives result
in a significant reductions in costs.  However, based on the non-attainment
analysis presented in the preceding section, there is reason to believe that
the cost comparison between the current TSP primary standards and the PM^Q
(50,150) case may not be valid.  As discussed above, the use of a constant
0.46 ratio to relate TSP to PMio is not as reliable as the probabilistic
methods now in use.  Moreover, recent measured PM^g data indicate that the
probabilistic method may also result in an underestimate.  It can also be
seen that adding a secondary TSP standard of 90 Annual Arithmetic Mean (AAM)
would increase the DPV of the PMio (50/150) case by only about 13%.   On the
other hand, the inclusion of such a secondary standard with a primary PMjQ
(65/250) would result in an approximate 230% increase in DPV.
     Table II.C.4 provides a more detailed breakdown of the estimated costs
and also provides the number of initial and residual non-attainment  counties.
As can be seen as the stringency of the standard increases, both the absolute
number and percent of counties defined in residual non-attainment increases.
In addition, the Scenario B costs as a percent of total costs increases.
For example, under the PMiu (65/250) case, the Scenario B costs represented
only 20% of the total; while under the current TSP-Standards (TSP (75,150))
they represent some 55% of the total.  It was pointed out in Section II.8
above that the mean multiplier (ua) was used in deriving the Scenario B
costs.  Since the mean was weighted rather heavily by the marginal costs,
it was felt to be a reasonable estimator.  Nevertheless, it sho.uld be
understood .that the Scenario B costs are considerably more uncertain than
the strategy costs.  Once again, in Table II.C.4, it can be seen that the

-------



























{/)
3
3

O>
s
c
T"
«3- iQ
* +»
O -«->
•«*
him*
— -a
c

0
CJ)

qj
•r"
•gp
£
0
4J
ra
z




































"
4J
I/I
O
O

•5*
0.

02
n
O O>
•r- 4J
L. 10
10 e
Ci-
01 +4
O I/I
(/I UI

VJ-
0

t_
a>
^3
€
3
Z











HM»
O

•u
O
O












<4_
O o—
n
C- i-
9} 4J
•§•£
3 —.
^2





























VI-
O

C
o

4-J
0
3
•o
01
ec

t_
o
<^M

,_
3
•o

(/I
0)
o£

vn
cn
f-i




•o
• *

4->
10
C
o








a
o
«c
»—















f*
u
IO
tfl
OJ

c
3
O
U


























1*-^^
M^

^5
^^
^_^

4->
C
QJ


>—

«

4J
C
3
O
O



m
en
3^
^••4


CM
W
3^
f4


en
00
CT^
pMl|


0)
a.
o




CT>
3ft
'^


CM
ey>
cn
^

^^
so
2

















0
•r-

•0
C
03
U
t/1





,



• O ^
p-l 1— <
CO CM
CM <->










OO •— '
CO «*
CM ^






CM CO
00 00
• vn co



f*. i— «
r* CM
yjj  VI
^_ |_









^ cn
^f *^
00 ^











^^ PO
CO CM







ro vo
(••4 ^"^
fl *-^



|MM| sO
^f* *sf
C*ir| ^^



ro 3
CM ?M
ro ^^



iO '30
•-4 cH
3 ^





cn vo
«™«


00 CO
oo vn



00 '^
00 -^
^













<^^^"^
3 3
uft trt
PH CM
* *
O UI
LO *
CM
41
'.n


,i>
^
iO
^
^k


^
?u


•
I^J
en
^^

o
**
cn
oo
en

g
Q
U
tQp-


3
a.
s
0

0)
c.
0)
Tp


irt

^—
O
L-
c
o
u

"4-
o
c
o

IO
u

'a.
S"

0)
t_ • »
o • ui in
*-  o o cn
D o -a co
T- +j cn
•*-> c -a TT »H
C 4J D 00
3 «/> N en i/i
o a> •«- *-H L.
U . &. i— iO
CL iO t- 01
4-» 3 0) >,
^« ^3 ^3 ^J
fll 4) C ^ C
E 4-> IQ 
t- 3 x er ai

*5 S i t/> t«

90 ai e a.

*
^n
o^
^^

c

W)
ot
^
c
3
O
u

1 1
c
i
•^
IO

ie
c
o


O"*
IO
3
•o

(/I
a>
ce


-------
                                 11-21
addition of a secondary TSP standard of 90AAM would not add significantly
to the cost of a PMig (50/150) but would do so to a PMm (65,250).
     Table II.C.5 provides information on the emission reduction and associated
discounted present value costs by source type (i.e., stack emissions, non-
traditional fugitives, and municipal paved roads).  To provide a basis of
comparison for the emission reductions achieved, Table II.C.6 summarizes the
projected inventories.  It can be seen that controls on municipal  paved
                                Table II.C.6
               Nationwide Projected Inventory by Source Type
                                   tons/year)
Source
Stack
Non-Traditional
1978
TSP PMio
5041
433
3668
226 .
1989
TSP PM10
4331
526
3197
275 .
1992
TSP PM10
4230
558
3137
293
1995
TSP PM10
4165
597
3112
306
  Fugitive
Effective Area
230    60
234    63
235
64
237    6b
roads account for the largest single fraction of emission reductions,
but add relatively little to total  costs.  Emission reductions from stacks
are next largest in all  cases.  However,  these stack controls are significantly
more costly and account'for over 90% of total costs.
     Since the control of municipal  paved roads.is comparatively in-
expensive these sources  will always  be controlled first by the model in  the
early years of a scenario.  It should be  understood that only a fraction of
paved road emissions are assumed to  contribute to the county design value.
These are called the effective fraction.   Varying the effective fraction
has a large impact on estimated non-attainment and cost.  A sensitivity
analysis is presented in Section II.D.

-------











01
Q.

>~
Ol
U
t_
3



^^
J3

" 0
•^
+J
U
3
(V
-ft Of
:_J C
• 0
— . V .
— • w»
trt
(U t-
^— £
e ^U

a.
Q


1
S
C
0
•-
£
Z


.



1
1

1
1
1
1
1
1
1
1
1
1
1
1

l_








u
|i
2^
V» 4J
10
§o

^•^
4)
3 "O
p— 01
IB >
^* O
*F"
+J ^»
C U
3} *
4! I/I
(- C
0. O
•o *->
•W 3
c -o
3 O)
Q t*£
 ifl
00 •*-
ON £
«-i LU








































•••I















oi
Q.

^*

41
U
U
3
(/>
























































•Hi




O r—
f- 
<1> -r- -C
> e *
<0 3 O
o. Z as









..!• 01
"S *

0_
C,


c.

a.

Q.
to
1-



^^
Q.
^d




~
Z
a.


a.

h-

-

a.
a



 CO CO VO 00 UO <^ CO ^9* fO *C^ OC r**»   *«i" cr* w> t^* ^ oo
ir> tr> u~i co^^ CM CM CM ^-» ^-» ^* CM CM co CM CM CM


U^ C^ CO ^^ O CM, CO CAJ **^ C3 CO ^i CM O^ ^D ^f t^ ^f
^C CO C\J **Q O^ ^^ ^* C^ CM O% C^ CO OO ^^ UO ^^ CO ^*
^^ ^^ oo ^n u^ ys co co ^t* ^^ CM c^ co ^f <^ o^ co co



VO^^T ^TCOCSJ OJCMCM p^^4^H COCVICM CMCM^H





ioovc courts ir> co ^r OOC^H ^aoo O\I«^M
f-H ^H

oc cj^ o CM c\j ^ oo CT\ o ^o co ^f o ^ ^^ r** oo G\
*— 4 i— 1 CM i-H <— 1 r^t . . ^-H ^H^H^-t.



oo ^^ ^"^ ^^ CT* r*»- oo (fi to co co CM c^ r>» 10 f**» ^o ^^
^^ ^^ ^^ i-*





^O -CO ^^ C\J l^ ^^ tiO O OO ^^ ^f OO ^* ^> CO CO ^O ^O
^O ^^ ^^ ^M ^M ^^ ^^ OO ^"^ CO ^^ ^^ ^O ^^ C^J CO CM ^^



l^ i^f ^^ ^f ^H OC CNJ ^O C^ OO CO CO 1^ CM VO ^^ *^3 OO
CO i^ co r*»« cO ^^ oc cy* ^^ ^f ^* oo co ^^ c^ c^ co r^*
^o LO to co co co <^ ^^ ^ CM cvi ^^ ^ «-* ^^



S5J^ n»S2 ^fJoSSS? i£i?o SSS
CO CMJ *"^ CM ^* ^"* v'H *~H CM *"^ ^^ ^"^ ^^



OtCMUO CTiCXJU^ O>CM>X> O> CM l^ ^ CM Ul O*v CNJ :^")
oo c^ cn oo -^ o% oo ON cj* co crt o^ co c^ c^ cO o^ CT*
ON ON CT> O^ CT> ON O> O^ ON ON O> ON  o -n
*iO ' UO LO "O LO MO
f***. f**» •+*•••* i...._.r >«*' »w^
*— «— • o o o o
CL a. <— i <— » ^^ ^<
VO t/) 	 S 	 3E g JE 	 	









i
4^
(O
3
•8
*->

Oi
^


^»^
^~-

00 3
t- •— >)
n3 ^9 ^^
•— > c
^ o
0 *-
•a o
«T CT LO
00 C ON
2 •52
c
t_ 3 O
a> o •*•>
*-> t_
o ON
>
rtt t~ ^J
QJ ^ •**
3 U f
P—  wi T3 ai
•MO >» u
c •— ^ ^

U 3
•o -c a.
01 tn -a E
•*-» c * o
CO U
3 ••- 4J
O tf> O 41
U U) C C-
- S >t 3
"5 LU *^
E v>

-------
                                  11-23
     3.  Industry and Regional Results
     This section provides cost and emission reduction information
disaggregated by industry type {4 digit SIC) and by EPA Region.  The industry
costs are reported in Table II.C.7 with associated emission reductions in
Tables II.C.8 and 9,  The costs in Table II.C.7 represent strategy costs
only and do not include the Scenario 8 estimates.  [Scenario 8 was
calculated on the basis of air quality and can not be assigned to specific
sources or industries.]  The SIC's reported are listed either because they
contributed >^ 3% of national OPV or because of national policy interest
(e.g., Primary copper).  Table II.C.7 indicates that regardless of scenario,
the two industries which are likely to incur.the highest costs are utility
power plants (SIC 4911) and Iron and Steel  (SIC 3312).  This should not be
too surprising, given that these two make up over 5U% of tne base inventory
and given the use of rollback.  Nevertheless,  this result may simply be
an artifact of the analysis.  In the case of utilities, the elevated release
of the bulk of their emissions undoubtedly  diminishes their real  world  impact.
If dispersion modeling*were used in place of rollback, the utilities would
most likely not be as important a part of the  control  strategy.  In the case
of Iron and Steel, several  factors are suspected.  In the first place, the
growth factors used in projected emissions  indicated net positive growth
for Iron and Steel throughout the 1978-1995 period.   For the first 5-6
years of that period, the industry has declined.   In the second place,
the industry's capacity has been shifting out  of  the older,  integrated
facilities into newer and cleaner "mini-mills".  Both  of these factors  suggest
that the Iron and Steel  results may represent  an  overestimate. A limited
sensitivity analysis of the growth issue  is presented  in Section  II.D.

-------
                               11-24
     The emission reductions shown in Tables II.e.8'and .9 9enerally
parallel the cost breakout.  An important exception relates to municipal
paved roads. 'Hence, although the costs are relatively low, the emission
reductions are the largest of any single "industry."
     Table II.C.10 provides a regional  summary  of costs.

-------
                     11-25
I I












M
«T
I
i
\


4





S
i



























"1
i
Til
:
















i
s
c
8
E

1
8
c
i
s
s
t

9
i




I
s
2
s



i
S

1
X





i

wi

8
£
j

|

i

i
i
i
fw
i

•
j.
i


i

|
i
i
!

i
i
1
•3


vt
-»2 = 	

-52 s , , , , , 	 . „_,,.„.

— *• i —



"f- - ' ' ' ' ' ~ 	 ' "" ' ' '"'

~sa s • • ' ' - "••- • .--••'--

" S3 S ~ ~ ~ ' ™ ~ «• - '. « « i i • * ~
S ** 2 —
• O* « 1 1 ( 1 1 «w» t Mllllll
.^•jj 1 1 1 1 1 -_. < --l«l(l
«•••> W 1 1 1 » 4 «««•» 1 Mvlllll


""
~'S2 t ....•---.-.--•.•--

"**. * .""".' ""
** -


"52 " - ' ' 	 ' 	
••sa 2 *• - - ™ " •»•""»'' «•»-!(«•>
s2s 8 a - - = a s * s - ««-•-««
*SS I " - ' - " " * - ~ -•»-••--
=- *
Sf 1 : , ] 1 J.I S ? '
«:9i::..s;. s. x : * ie
Ijsl.i ,|« |l'.|s > | j Jlif-s
•5*^— "j^ *is2«S—i?fcC— S •• *"5^l S"5*? —
flsilllilllilllllli-lllslll:^
iSifi a S 5 £ * 2 .3.3 5 S£C£S|
'isi I 1 I I s I ill' § iillll
»

*

2

fi

I

s

ff
•*
9
3
3


=
8

3
8

I
5
a
i
'
2
s
j

1
i
^
:
i
•3
i  ^
5  I
S  I
                                                 I
                                                 ^   M
                                                 3   £
                                                 1   :
                                                 a   s
                                                    22
                                                 s
                                                  =. 5 :
                                                 lii!

-------
       o
       i
 &
 — ^
nl
•MM U
4) tt
3°*
si
 3!'
 I—
 S

       z
       g,
             » !•» *»  -»
             n^p^i
               **
              S^rm
              «. «•
             — •* M
IS*
                                           «"• — I — ft*
             *** *^ 2
                    2S(
             O irt «  <•* P^
            sag  a
               - S  s   « «i  « ^ • a. «  e     «  eu   c  -  -
               • •««  -o e   fu  UK  1  -o  v  a «   —  «  ^
               i-SS^gJjSS  SI.-^-S  5  1  i:  IliSSiS  s
               •o o «i  wucg^ea*«*«x**4  euo^o     3 --  —•

             •s=gii1-3li1jg5s.ils^8-iIltiJll-3  ^
             Isiu  5  S  3  £  £  £  £33  5  5 £ £ £ £ I  £
                                                 -^  ^
                                           £
                                           §  -=
                                           S  3
I  1

I  1
•g  «
   X
I  «
a  «
   w


11 j
•9 w
e  •
                                                        f - I
                                                        Si*
                                                        III
                                                        3 1-S
                                                        slf
                                                        •j> a o
                                                        J i^
                                           '1 s
                                            * I

-------
    I
    «
*.   §
              52


              I


              g
              s
             31
                 3
I
                                           11-27
SSS
                                                         *«0:9   M   ^hi*^e«*   u   ««y^e__  -__
           i*o a *   ^^su^BOi^^XB'^   euo«o        3»«   -*
         XtSOib^autiaO*   X -C   ^ -"   03    l.  u  X X  >, U1  «    *
       -.  « «   3««^k>   —   — M O 3   UTJ^JI—  Ul-t^a    e





       I  S - •''
        <5   o


= 2=   a   a
                                                                        »«yhJM
                                                                        a. a- ^ !fl
                                                                        PI <•* <•*> -* ^
                                                                                          "   I
                                                                                          IM   M
                                                                                          0   *

                                                                                          M   3
                                                                                          p->   JS
                                                                   |   e
                                                                    tl   *




                                                                    -   !



                                                                    OT •-- X


                                                                    ||1


                                                                    59*
                                                                    a a ^
                                                                    M o ^
                                                                   £- 5
                                                                    "e 1

                                                                    is I
                                                                    4J   9
                                                                    *• TJ U
                                                                    OE 3 V

                                                                    - 2 »

-------
11-28















3

J
« •;
— *•»
^ ^
i =
5 *•
M
CJ
I

'S
I
I




















•
,


*
1
4
*
a
4*


5



;,



^
M




»



5














|


i
u


>
a
I
1
i
^

i

^

i
u
^


i


g

i
„
i


i
u '
3

|




0
••I
|
1

n A
If i- 5-

SS 52 SS

3S 2^ =2
S2 5- RS
• O « « ^» ^
01 9 — » ffi 10 m
MA » -, M 4
C C *? « C ^
33 - a-

32 32 «.*


^ ^ ^ 5» ^ ??
3 S * """ "*
si 2" -S
^ _ _ e _ 0
2^ «5 «i


as =° =c
c« «5 --
C3 Sr« s«
CN
ft« M. *«

TT
5*5*
a^ o^

H H
C C * ® 00
O O m *A in «n
«• M * * • -
* • o «rt o in
•n m tn « m «
f* p*» «^ ^  >«•
w e e oo
§a if ii











i
41

at
«.
b>
|
41
>
• • |

«

5
I
*

2
S f I
^ £ J
^ » u
2 I S
2 S ~
!• ^
2 83
u 4 — <
3 >. -2
r = s
- 1 5
2 •
2 * "S •
-S ^ S
.. -s
553
5 • s
M « i
(S M »•<
• W <•
1 S £
s « *
•o >

2e S 1
s s 3 s
J 1 » S
11 Jl 1

-------
                                     11-29
0.   SENSITIVITY ANALYSES
     As with any analysis, the PM RIA cost results presented above depend
on the data bases used and the assumptions made.  Care was taken to ensure
that the data were as good as possible and that the analytic assumptions
were realistic and did not unduly bias the results.  Nevertheless, the
use of other data sets and other assumptions would change the final results.
A number of sensitivity analyses were undertaken to determine how much the
results might change.  This section reports on those sensitivity analyses.
     1)  Growth and Effective Fraction Parameters
     The impact which growth will have on future air quality is heavily
dependent on the fraction of new and replacement sources which are subject
to 8ACT requirements.  In the model, the quantities a and b were defined as
the fraction of new growth sources and replacement sources, respectively,
controlled at current rather than "BACT" levels.  In the results, reported
above in Section II.C., a » 0 and b - 0.5.
     A second major determining factor in the growth computations was  the
treatment of area source emissions.  Area sources were divided into two
categories  1)  Paved .municipal roads for which a control option was
available, and 2) "other" area sources for which no controls were available
in the model.  The following parameters were then defined:        /
      o *the fraction of county-wide paved road emissions assumed to
         affect the design value.
      8 =the fraction of "other" area source emissions assumed to affect
         the design value.
      Y =the fraction of "other" area source emissions assumed to grow in
         proportion to the population.  [N.B. Paved road emissions are
         completely "grown" at the population growth rate.]

-------
                                  11-30

Changes In any of these parameters can have a large impact on the final
results.  Table II.0.1 presents the results of a sensitivity analysis of
             :                  Table II.0.1
     Sensitivity Analysis of Growth and Effective Fraction Parameters
                              PM10 (50/150)
Point
a
0
0
0
0
Sources
b
0
0.5
1.0
0.5
Area Sources Number Non-Attainment Counties
a 8 Y 1989-Initial 1995 Residual
0.01
0.01
0.01
1.0
0.01
0.01
0.01
1.0
0
0
0
1
160
138
232
223
65
87
117
171
Total
OPV l
832
1015
1168
3936
these parameters using the PMiQ (50,150) case.  The case shown in the second
row of the table represents tne base case conditions.  The results show that
subjecting all replacement sources to "BACT" (i.e., b * 0) reduces the cost
of the attainment.-strategy.  Likewise exempting replacement .from "BACT'r
(b = 1) or increasing the effective fraction of all area sources and growiny
them all ('«»!, Bal, y*l) can greatly increase the cost of attainment.
     2)  Assumed Growth Rates
     The industrial  growth rates used in this analysis are the same as those
used in the 1984 analysis and were originally derived in 1979 from Bureau
of Economic Analysis (BEA) data.  Economic forecasts from that period
predicted positive economic growth for the entire 1978 to 1995 period over
which growth was being simulated.  It was jnfeasible to recalculate and
replace all growth rates.  However, a limited analysis was made for the
utilities and iron and steel which constitute nearly 50% of the inventory.
Table II.D.2 compares the growth rates (N.B.  In the base analysis rates
were state specific whereas a national average rate was used in the sensitivity
analysis.)

-------
                                    11-31
                               Table II.0.2
                   Comparison of Growth Rates (% yr"1)
             .                 Base Analysis          Sensitivity
                             Max.       Min.          Analysis
Iron and Steel  (SIC 33)      10.3       -O.ll          -2.92
Utility (SIC 49)              6.2        2.4            1.64

     Wyoming only.All other states had positive growth.

Table II.D.3 compares the national non-attainment and cost results when
the new growth rates are substituted for the old.  As can  be seen, 5 more
counties came into attainment and the national  costs (DPV) are reduced by.
almost 25%.  A more detailed review indicated that Iron and steel  costs
(DPV) fell by almost 50% and Utility costs were reduced by 33%.

                               Table II.0.3
                      Effect of Revised Growth  Rate
                                   (50, 150)
Number of Non-Attainment Counties
1989 Initial 1995 Residual
Revised
Base
176
188
70
87 '
Total
. DPV 1
763
1015
1)  106 1984 dollars, the national  total  for 1989-1995.
     In addition to help resolve the question of growth  rates  a comparison
was made between the projected RIA  inventory for 1982 with  the actual
NEDS inventory for the same year.  The projected RIA inventory used  all
the base case assumptions.  The 1982 NEDS inventory was  modified to
parallel the RIA (i.e., only the "effective fraction" of area  sources
were considered).  For total national  emissions  in 1978, the ratio of
NEDS to the RIA'was 0.88.  In other words NEDS was 88% of the  RIA inventory,

-------
                                  11-32

This was not unanticipated since the RIA  inventory deleted  all sources
£5 TPY and added estimates for industrial fugitive emissions, Moreover,
there have been updates and changes to NEDS for 1978 since  the RIA file
was compiled.  However, by 1982 the ratio of NEDS to the RIA falls to
0.48.  This suggests that, at least for the first five years of the
projection period, the RIA growth procedure overestimates growth.  For
the 1978-82 period, the RIA model predicted about a 3% decline in emissions
while NEDS showed a 47% decline.
     In summary, these comparisons of growth rates and growth suggest a
net positive bias in the model.  This implies that the cost and non-attainment
estimates may be too high.  It is theoretically possible that over some
longer period (1978-1995) the growth procedure may prove to he more accurate.
However, given the magnitude of the discrepancy in the early years this is
unlikely.
     3.  Generic Control of Point Sources
     In the discussion of residual non-attainment in Section II.B., it was
pointed out that the model does not possess control options for a number
of point source categories.  On a national basis, these categories do
not contribute very much to total emissions.  However, in some counties
they may be important.  A review was made of the categories involved
and generic controls were assigned to each based on engineering judgment
and a comparison to similar source categories listed in the control
options file.  This procedure and the control efficiencies used are
discussed in more detail in the supporting documentation (Argonne, 1986).
Associating costs with these options would have been a more major undertaking
and was not feasible within the resource constraints of this RIA.  A run
was made of the model with these generic controls in place  and analyzing

-------
                                  11-33

the PMjQ (50,150) case.  The number of Initial non-attainment counties was
the same as in the base run, but the availability of the generic controls
cut the residual  non-attainment by nearly 40% (87 counties down to 54).
This result suggests that, in fact, the predominance of sources not
listed in the control options file is a major reason for residual  non-
attainment.  It,  therefore, lends support to the decision to use the mean
             (m)
multiplier (u a ) and not the marginal.

-------
                                  11-34
                                References

1)   Argorine National  Laboratory (1986),  Revised Costs  and Air Quality
     Impacts of Alternative National  Ambient  Air Quality  Standards for
     Particulate Matter,  Technical  Support Document, Argonne National
     Laboratory, Argonne, 111.,  1986.

2)   T.G. Pace and N.H.  Frank,  "Procedures for Estimating Probability
     of Non-Attainment of a PMiQ NAAQS  Using  Total Suspended Particulate
     or Inhalable Particulate Data" Monitoring and Data Analysis Division,
     U.S. Environmental  Protection  Agency, Research Triangle Park, N.C.,
     Draft, February 1984.

3)   U.S. Environmental  Protection  Agency (1984), Regulatory Impact Analysis
     of The National  Ambient Air Quality  Standards for Particulate Matter,
     Strategies and Air  Standards Division, Research Triangle Park, N.C.,
     February 21,  1984.

-------
                                  III-l   .

III. BENEFITS ANALYSIS ESTIMATES   .    '
     A.   INTRODUCTION                       .
          The benefits analysis quantifies in dollar terras some of the
improvements in health and welfare associated with particulate matter
(PM) reductions.  The selection of concent ration-response functions to
determine the risk associated with various levels of PM,  the methods of
determining exposure, the approach used to value health and  welfare
effects, and the aggregation of benefits within and across effects cate-
gories are described in detail in Manuel et al. (1983).  A general  dis-
cussion of the benefits analysis and the strengths and weaknesses  of the
analysis can be found in the Executive Summary and Chapter VI  of this
RIA.  As stated in the Executive Summary, the reader should  be aware that
neither the benefits methodology used nor the results obtained have been
subjected to formal  peer review.
     B.   BENEFIT ANALYSIS CHANGES
          A number of changes were incorporated into the  original  Manuel et  al
(1983) analysis.  The changes can be grouped into the following categories:
     0    changes to the underlying concent ration-response functions
          used to estimate effects (Mathtech (1986), "Revisions to the PM
          NAAQS Benefit Analysis")
     9    changes to the underlying data used to calculate benefits
     0    changes to the alternative aggregation procedures  used to
          present the benefit estimates.
     Changes within each of these categories are discussed briefly below.
     1)   Changes to Concentration-Response Functions
          The model  used in aggregation procedures C,D, and  E  to estimate
household soiling benefits (Manuel et al. (1983)) is an econometric model
of household demand with environmental  variables included as shift parameters.

-------
                                .  III-2

In the original  statistical  estimation of the model, the PM measure
selected for inclusion was the second high reading of 24 hour total
suspended particulate (TSP)  averages measured over the calendar year.  In
subsequent reviews of the household model, the plausibility of using an
extreme measure of the distribution to characterize materials soiling was
questioned.  In view of these comments, the household model  was revised
to include the annual mean TSP measure instead of the second high measure.
The net effect of this change is to increase the benefits associated with
alternative PM reductions.  Under some of the alternative standards
examined in this addendum, an approximate doubling of household benefits
can be expected from the substitution of the annual  mean TSP for the
second high measure.
     the morbidity studies by Ostro (I983a, b) form the basis of the acute
morbidity benefit estimates  used in several of the aggregation procedures.
Previously, the benefits reported for this study relied upon unpublished  work
by Ostro.  Since that time,  Ostro's work has been peer reviewed and  published
in two journals, and the functions are more applicable to the general
population than the ones used originally.  The net effect of this change
is to increase benefits by approximately 15 percent for the most stringent
alternative examined.
     2}   Changes to Data
          Soon after completion of the 1983 Manuel et al. benefits analysis,
1980 census data became available on computer tape.  These data were used
to estimate the benefits reported in this addendum.  The change has  a
relatively small impact on estimated benefits.
     To value changes in mortality risk, the current analysis uses dollar
values that are approved by the Environmental Protection Agency.  In 1984

-------
                                  III-3

dollars, these values range from a minimum of $0.43 for a unit reduction  of
1.0 x 10'6 in annual mortality risk to a maximum of $7.46 for the same unit
reduction.  A point estimate for the value of mortality risk reductions is not
currently available.  Compared to the values used in the 1983 benefits analy-
sis, the minimum value is approximately 5 percent lower while the maximum
value is approximately 100 percent higher.
     Previously, the morbidity benefit calculations for Samet et al.  (1981)
were based on data for respiratory disease admissions through the emergency
room.  The Samet et al. study, however, is actually based on respiratory
disease emergency room visits.  Therefore, data on visits were substituted.
The acute morbidity benefits for Samet et al. decline by approximately 85
percent as a direct result of this change.
     A number of changes were made to the air quality data that are used  to
calculate benefits.  The changes are documented elsewhere in this  addendum.
Benefit categories that rely on concent ration-response functions which use  the
design value annual arithmetic mean (DV-AAM)  TSP as the index of exposure are
particularly sensitive to the changes in air quality data.  .Under the  most
stringent alternative examined, the change in air quality data  leads to a 65
percent increase in benefits for concent ration-response functions  that use the
DV-AAM.  Table III.B.I provides a summary of the extent of the  change  in the
air quality data.  As the table indicates, the mean baseline DV-AAM in
the current analysis is approximately 25 percent higher than the mean
baseline used in Manuel  et al. (1983).  The standard deviation  is  also
significantly higher in the current analysis.  In addition,  the change in
the mean OV-AAM between baseline and attainment is about 35  percent
larger in the current air quality data set.   Because the air quality

-------
                                  III-4
                              Table III.B.l

  COMPARISONS BETWEEN MEAN  1989 DESIGN  VALUE  TSP  ANNUAL  ARITHMETIC  MEANS
                                (in yg/m3)
                    Manuel  et  al.  (1983)

                  Baseline     Attainment1
                                  Current Analysis

                              Baseline    Attainment*
    Mean
110.1
90.9
137.5
111.7
    Standard
    Deviation
 32.3
22.8
 41.8
 25.2
    Number of
    Counties
        279
                         296
1)  Represents technologically feasible  attainment  levels.

-------
                                  III-5

levels in some populated counties may be significantly higher than the
means reported in Table 111.B.I, it is not surprising that the change in
the air quality data has a significant impact on the benefit estimates.
     3}   Changes to Alternative Aggregation Procedures
          Six different procedures (A through F) were developed  in Manuel
et al.  (1983) to aggregate benefits within and across benefits  categories.
The procedures give differing weights to the issues  of double counting of
benefits, coverage of effects, and the strength of the underlying  studies
upon which the benefits are based.  Aggregation procedure  A is most
restrictive, avoids double counting, and probably underestimates benefits,
while aggregation procedure F seeks to provide more  complete coverage  of
benefits but risks double counting.  Procedures B through  E are  gradually
less restrictive than A but more restrictive than F.   A complete discussion
of the rationale used to select studies for the alternative aggregation
procedures can be found in the Manuel  et al. report  and Chapter  VI  of  the
1984 PM RIA.
     In the current analysis, the acute morbidity study by Ostro has been
moved from aggregation procedure C to aggregation procedure D at the request
of the Office of Air Quality Planning and Standards  (OAQPS).  In place of
Ostro, the acute morbidity study by Samet et al. previously used in procedure
B is used in procedure C also.  Benefits reported for aggregation  procedure C
decrease significantly as a result of the change. The decrease  is  mitigated
somewhat by the increase in the soiling benefits estimated  from  the Manuel  et
al. household soiling model.
     C.   BENEFIT ANALYSIS ESTIMATES
          Tables III.C.I and III.C.2 report  the incremental  benefits for
the alternative PM^o (particulate matter of  10 micrometers  and smaller)

-------
                                                           III-6
         «*    o

         CO    VO
         ro    en
         CM    ~+
                              •    vo
                            in      •
                            in    m
                            f-4    CO
                                un
                                00
       oo
       i
co
o.
00
            CM
                    ie
                    CT
                    at

                    en
                    01
u
LU Z  l/»
•—    z <
      ec
at    LU _j
00-
u.

crt
N-

U.
      LU

      09
        10
          *


        O
        f-4








        en
                          CM    r-»
                          o    vo
                          CM    —I
                            m
                              *


                            vo
                                                    vo
                                                     •
                                                    <—I
                                                    CO
                                                                m
                                                                vo
                                                                vo
                                                                             vo
                                                                             CO
                                                                              •

                                                                             o
                                                                             vO
                                                   CM
                                                   CO    CO
                                                   «-«    r*.
                                              r-.     —i
                                                *      •
                                              o     tn
                                              *     CM
                         CO
        CM



        VO






        CO









        in







         •

        m
o
i—i

CO
       00

       en
               CO
               f-4


               oo"



               CO
                               


O    vO
  •      •

-.    o
                                            o

                                            CO

                                            VO

                                            o
                                                               VO

                                                               00

                                                               o
                                                                     oc
                                                              CM
                                                                •

                                                              CO
CO

 1
 <0


 10



 at
                                                              O
                                                              en
                                                              a.
                                                              in
                                                                     in
                                                                     vo
                                                                      *

                                                                     o
                                                                          o
                                                                          en
                                                                    oo
       o    o
       in    vo
       ~*    CM
                        in
                              in
      O-     Q.
i—    oo     oo
                                           o
                                           in
                                                  o
                                                  — <

                                                 OL.
                                                       O
                                                       in
                                                       CM

                                                       m
                                                       vo
                                                  o
                                                  i— i

                                                 Q.
                                            o     o
                                            in     LO
                                            »-     CM
                                                              in
                                                  in
                                                  vo
                                                                                    o
                                                                                    N
                                                           O 00
                                                          f t-

                                                           io


                                                          5 IP
                                                           at C
                                                          f at
                                                          p- *j

                                                             e

                                                           at"""
                                                                                   o   9*
                                                                                    oo  at
                                                                                 «-» sn _
                                                                                  C F-*  O
                                                                                  at    —
                                                    CO


                                                    o


                                                     o
                                                                              wt
                                                                                 at
                                                                                        c.
                                                                                        10
                                                                                  at  «-  o
                                                                                 *j  »  .
                                                                                  o     >>
                                                                                  o  c  7
                                                                                  c/i  at  c
                                                                                 »•  >  at
                                                                                 •o  at  en
                                                                                     (/)  C
                                                                                 CO    f-
                                                                                 oo  at  c.
                                                                                 en f *j
                                                                                                 1 -a

                                                                                                  O
                                                    O  (A
                                                    Q.
                                                    at  at


                                                      4-1

                                                    C.
                                                    10 r-


                                                     t-
                                                                                                     O <0
                                                                                                    ••- +J
                                                                                                    4-1 «-
                                                                                                     (B O
                                                                                                     c E
                                                                                               O
                                                                                               U
                                                                     ^ *•"


                                                                     5v
                                                                      in O


                                                                     |x


                                                                     1°.
                                                                     O _4
                                                                     O

                                                                     a. o

                                                                     e e
                                                                     o o

                                                                     *-> +->
                                                                      t_ ^-    o
                                                                                          L.


                                                                                          O

                                                                                          J=
                                                                                           I

                                                                                          *sr
                                                                                          CM
                                                                                          II
                                                                                                                 •o

                                                                                                                 TJ
                                                                                                                 4->
                                                                                                                 in
                                                                                                                 n
                                                                                                                 x
                                                                                              
                                                                                              at  
-------
III-7




































CM
*
O
•
*— 4
4}
s
ITS
1—





















































•— 4
CO
O CM
§ £
§3
^3
1— 0
(/I O
O. Q.
(— C
o
C2 •*""
Z 4J
^C 'O
91
O 4)
£ cn
0.1— cn
Z <
UJ UJ — - •
3* X <«
*•"* JgJ C
SSz
0£ h--^
tjj *SC "^
Sd"
3 •— '
O£ U_
0


I—
|B*4
U.
L&J
z




u_





VO
*
CO
o
m
-•»
o
" •
r*»
CM
. CM
T






UJ





CO
•
^
CO
J*
VO
•
r-4
f*.
• r-t






O



CM
5
PO
uo
*
us
CO






(J
cn
CM
*s^
9>
*
O
t-H





OS



CM
VO
•^
CM
•
CM

UJ
CO

_^

1—
Jjg
UJ
uj
ee




<


O

m
^^
CO
•-4
z




CO
"E
fO
^2
*
4-*
$
«
^

4)
4**
^




I
in"
r^


CL
^

VO
VO
cn
CO
*^s
00
•
VO
p*lk
1— 1



CO
•
«I"J»
4*
. CO
CO
•


00
VO

^

1-4
CM

m

<— i



m

<— i

00
o


^— Ifc
1

o
cn

a.
to
I—
o
m
o"
m
**^

0
f
a.

^
^
VO
CM
^(^
cn
9
ro
^^
0p^



o

^•s
f*^
CVJ
00

rn
CO




CO
00
1— •
^
•
VO




vo
*
12
*^
i-4
10



cn
O
•->
CO

_4



co

o
f*4

O


J^™*
1

o"
cn

0.
{C
o
vn
CM
tO
10

^^ j^J

e
  cn
OO 4)
. 4-> cn f—
C 1-4 U
 g IM
Q-<»- 0.

-a co 4J
01 t- 0
*j  4) C
.^. > 4>
•a 4> cn
— • f> e
! co •>-
O CO 41 *-
^*
a.

cn js 4^
««•«
•— «
e
3
CO 4)

* iT3
O «/>
^^j>
41

1 «
C
o t-
o
(/< 10
10 ^^
^Q •
i^^
 **
C
4J O
in

*^ flu
•*— in
t
C I—-
CD *B
*J C-



§fd
3
U C
 1
wt O
4)
(- x
1°.
U ^-i

(^
o. o
e c
o o
4-> 4->
- C
•C C O
t \ 3 «^»
41 *O O

l_ t- -CJ
O O 4»
^« 4^ ^
^«^
CM





































•g
>
X
Ck.
co

o
o
IM-4
a.


^
4)
_— ^
CO

-------
                                  III-8
and TSP standards under consideration.   The benefits are incremental
since they are calculated for incremental  improvements in air quality from a
baseline scenario that factors in existing controls  and new source  regula-
tions.  Table III. C.I lists the benefits associated  with partial  attainment,
or attainment based on technological  feasibility,  whereas the benefits
reported in Table 1 1 I.e. 2 are based on  the assumption that all  counties  are
forced to comply with the standard under consideration.  Obviously, the
incremental benefits in Table III.C.2  are higher than those reported in
Table II I. C.I.  As expected, benefits increase as  one reads from  left to
right.  The probable exclusion of benefits categories in aggregation
procedures A and B and the likely double counting  of benefits in  Procedures
E and F suggest that benefits are "best" represented by categories  C  and 0.
     The tables indicate that the incremental  benefits are lowest for the
(65,250) standard and highest for the TSP (75,150) standard.  It  should  be
stated that the studies upon which the  benefit estimates are based  all use TSP
as the index of exposure.  The PMio standards  are  examined in terms of the
                                                              ,       »
equivalent level of TSP.  Consequently, the benefits of these standards  repre-
sent the effects of TSP stringency, not the effects  of particle size.
     Tables III.C.3 and III.C.4 report  the expected  incremental reduction
in adverse health effects for the six standards considered.

-------
m-g





I

.


to
o
UJ
u.
UB
UJ
I—
•f
UJ
3;
CO
uj O
co ec
uj in o
CO «< -4 UJ *—
• _C CO
o z a z
• i— i z •— > uj
fc? o  00
A »
CM CM




_i H- a.
«C LU
H- CQ
UJ
Q£
^^
z



o
VO
C\4
*
JC
^ .,
0
uj a.
lm__ * A
^^
0
w *

O O
0 . 0
o in
co co



UJ
a.
X
0








LO
^•4
A
r-.
*••*•

a.
H-

O 0
0 0
CO CM
co *r




;



•—
 CO W»
. -M •-- 00
UJ «/l _J t~t _J


0
M
VO


O
O
M
f*.




O
o

•
o






o
o
CM
«.
I-M^
^..j







o
o

o
CM






o
o
°-
CO
CM





I/I
^^
^3 *O
^ CZ)
U
3 >>
tJ *-
OJ •!-
oe >

CO •«•»
O v^
--« «c


o
o
o
A
rH


O
o
o
A
1— 1




o
in
r-.








o
o
o-

^__j








o
o
0^
__^







o
o
°A
f— 1



1
Q.
trt U1 d)
0) 0) V)
u a: «
C (U
Q> O V>
•a t- .f
•i- e o
u o

""* (j O
CO *J

,_ O «-


o
o
M
CO


o
o
t— 1
A
*




o
o
1—4
A
CM






o
o •
01

co








o
o

^r







o
o

 OL 
u ui  CK w>
•a •»-
«-- 4> a
0 +J
C 3 >>
_!<_)<_
< O
fO -M
O •»- "O
--• O U


















*





















4
.*•*
0)
t.
3
"S
U
O
u
*•
c
o

*
.4)
fc.
O^
o>


c
o
•a
0)

^OT
CD
i.

-------
111-10







trt
UJ
LL.
LL.
UJ

z
h-
UJ

UJ
CO
ac

• LU
O Z O £
1-4 *"' «« *
« 02 t
M 00 H-
2 y '^
10 O Z _l
h- UJ UJ ^3
OS UJ U.
1 ^^
^J ^™
< UJ
H- CD
. LU
JS
LU
oc
o
1— •
J»^
UJ
u
UJ
CL
X
UJ
















1







i
co
0

O
H-

LU
1—
ee.
UJ
JT






















•••*
r> i
\J
* M
/) O
iO O^

ooc,
E t—
a.
^H«
m i
pH
0*0*
10 OS
«v«>^
oa.
a.
O
O
"I
c*>


o
o
as
CO



^— *
o
CM
VO

C
a.


§
00
*
<— 1




^•^
o
*
o
un
""^
c
^B
2


o
o
c-
co





o
VO
CM
*"^
^^
to
^^


0
o
ur>
^T


^^

^^
O
CO
*

>^^

o.
H-

o
o

^
CD








U  t-
 ^ J

go o
0 2
•— i ir» o
4% • «•

CM

O O O
00 0
^o o o
^f (JO ^
CM




00 0
o o o •
£3 u9 • CT*
* M
CM CM
^^




,
o o o
00 0
CM O^ O^
«^- ^- •— <
CM






0 O 0
o o o
urT CM -<
CO




00 0
00 0
rl
a.
 41
W) 

•0 «* = * S
So a> o 
u) u -o ••- i-
^ >» 3 >» "^50
o o a> -^ c *• >>
3 «» ^-2 *^oo
o o o w o "*• "
S _i "See Sot.

o

U »
— 0 «-
<: o
o «- *
S 0 C.
































jj
„
t_
1
u
o

a.
c
o
*J
0>
V
t_
01
o>
<
o
•a
10
CO

-------
                                   IV-1

IV.  BENEFIT-COST ANALYSIS
     The benefit-cost analysis evaluates the alternative standards in
terms of economic efficiency.  Incremental  costs are subtracted  from
incremental benefits for each alternative standard.   The standard  producing
the highest positive incremental  net benefits is preferred  in terms  of
economic efficiency.  Negative incremental  net benefits suggest  that, on
the basis of efficiency, the baseline scenario is preferred to the standard
examined.
     Tables IV.1 and IV.2 tabulate the incremental net benefits  for  partial  and
full attainment, respectively.  With the exception of the incremental  net
benefits of aggregation procedures A and B, the results of  both  the  partial
and full attainment scenarios indicate that the most stringent TSP standard
(75, 150) is preferred on the grounds of economic efficiency. Under aggre-
gation procedures A and B, the least stringent standard [PMjQ (65, 250)]
is generally preferred when the lower value of mortality is used.  On the
other hand, the most stringent standard [TSP (75, 150)] is  preferred under
the higher value of mortality.
     With the exception of the lower bound  estimates for aggregation proce-
dures B and C, the incremental net benefits are higher than those  reported in
the 1984 Executive Summary and First Addendum.  The  reduction in the incremen-
tal net benefits for these two procedures is due to  the reduction  in the
incremental benefits calculated for Samet et al. and moving the  Ostro study
from Procedure C to 0.  The increase in incremental  net benefits for the
remaining aggregation procedures can be attributed to two factors.  First, a
different algorithm is used to calculate costs in the current analysis.  The
algorithm predicts costs that are substantially lower than  the costs estimated

-------
IV-2


















~lr*
o
5
o
*X
t— CM
CO UJ
Q£
Q. •• =3
CO O
h- UJ
^2 G3
z at
«C O-
0 Z
•J4 O
^C H*»*
O- H"
UJ IE CD
«S* UJ ' UJ
«- « P zln" tfl
> Z < O «£
M Q£ (— •»-
UJ t— —
flO !••• ^£ I""""'
2 «X _j S
£ ecSt*
o £•*—
U- Q£
co a.
H-
u.
UJ
z
UJ
03
t—
UJ
z
^J
5
i -
UJ
O£





u.




in
VO
CO
CM
**»»^
•r-.
CO
. O







UJ



" 0
•
«— «
o
CM
CM
pj
p**»







a




CO
*
in
T— 1
in*
•
>— 4
**





0



VO

£J
in
*
.CO




. 'CD





•
CO
VO

O
1




^




en
•
J^
CM
•— i
1

CJ
^£
^•4



en
"2
(O
•o
c
•O
(/)
«
•*••
4->
ia
c
t
CU
if
«*






S
^^
»
m
p^
«M«^

CL
>—

00
*

Qf>
^•4
*^%.
**!
CO
00



CM
*
*&
t^
r- 1
cn
CM
Ij0




cn
•
* p«^
CO
r- >
VO

^»
CO



in
•
H3
^
*
CO




ao
^

o
i


VO
•
^
O
1













o"
CM

r>
^ 	 ^

• a.
«—

VO
TT
in
^•4
^N^
**\
CO
^o



^
•
(•^
CO

cn
r*.
^^




CM
•
^^
f-4
t— «
in

f^»
CM



CM

O
»-4
00
•
CM




"".
O

o



0
•
f.
CD
1












^-^
O
m
»
o
in
^•^

o
^
a.

CM
in
00
^^
CM
^
CO





o

CO

CO
CO
CM





<—i
•
in
VO
^
•
\ft
^



CM

VO
°i
t-4





in
•*•
CM

O



co
•
^
o













^m^
O
in
CM
«i
in
^•^

o
£
a.

w
CO
yD
«pH
"*X>
C3
00
VO



VO

iy\
CO
r- 1
CO
i-H
in




CO

i^t.
Z3
^j
' •
cn
CM



m
•
o
en"
•
CM




•
^J

O



i— i
•
l*»
in
o
i



^**
i

4k
a'
cn

a.
bO
t—
^^^
O
m
A
o
m
%•*

o
3E
Q.

^J
in

plMt
**^»
CO
cn
m
VI

e
o
N
•|tt»
o to


I 0


^
•
^^
fVJ
t— 1
crT
^T
^^




^.^
•
^^
CM
r~.

m
CM




in
cn
•*•.
VO

CM
f—
+J (A
• E
JS OJ
t— 4-1

c
V "^
4J 4)

C Qm
0)  TS
a. c
to
o
^ 0

^— ^



05
VO
0

0



in
•
 o (-
C fc-  *- o
C flU
S^x
O CE O
tn ai c
in j •»• > ai
vo T) co cn
^—^

a
3E
a.
i/> e
CO *i-
QQ 4U (•>
O> JS 4J
—


CO 4>
«*• E

O •/>
vV

«J
O C
o
T3 ^—
 O
^ "n
*. CU
1*- (/I
CU ^
J—
1— in .


* *o
"S §
t_ OJ
o 
a.
g, ^
|-*'

d* •v
^
O
2 >>
4-> 4->
e -^
C) ^O
"c 4^
||


O 3
O C
e
•£ TJ
 I
(U
«- X

"O t^2
CU *
U ^H
o
C. M_
a. o

e c
O O
4^ 4^
 2. .2
0) 4J
•P- 3

vJ 3 CU
(0 «-
ai a

<_{_.,_
o o e
1^1^ (j^t ••!
J*L I




































»
1
TS

ts
+-J
I/I
I

1
CM
II

*
^•B
C
1
fO
in

C
c
(O
II
X

«^-^
^^
^
X
a.

^^
^
o
o
4^^
a.

• *
>H
ai
^••k

-------
IV-3
1
f


















CM
•
;>
Ol
2
a
t—











































,-H
CO
o
or CM
4
O UJ
i §
1— O
oo UJ
«J
0. O
co as
f- a.
o st
5 S
0 <
x* S
^ s
Q- OS,
1 1 1 )-* ^jj

l-H IU«—
1- Z «»
•iT" JP C t J
i


i

T

^ CM
. •
en «*
en en
J. ^f CO
•>•» "N.
CO ^f
* «
CM *»•
CM P-
CM '<-t



U



<-i en
• •
O •— i
CO *
j *t m
^**» *•«»
* *
• •
i** ^
VO CO
^ t»"<
a
1
00
f«. •
• CO
O P*.
CM CM
CO CO
• •
CM VO
00 VO





zS o i
5E3M-
U4 1— -—
L^ 1^_ JBW
5si
^j
^V 1 M&-
Q£ .J *«•
O" ^ "i J
^ ^—
!_•_ i i
Un .U*
co
1--

LL.
UJ
UJ
ml



r-» vo
* •
O 00
CM •-!
*•«•. *s» .
p«. co
• •
VO • VO






00


O r».
• *
CM •—<
^4 ^H

O VO
* •
CM O
1 1
I
i
,_ j
C
z 1
£
SE

7
A.
LU
dU
^C



«c




00 CO
* •
O 0
1*4 1*4
•++. ~**,
CM in
• •
CO i-H
1 1
0
z
»-*


CO
•p
w
fQ
"P5
s
10
* *
^* ,

•^
^J
». "x.
""1 °»
--4 O
1—4 *&
w—r ^%l
1-4


CO
• 0
f** •
CM CM
CM cn
*»• *»^
o o
• •
VO O
SO 00

cn
• VO
in •
00 O
<-4 00
•-W *>x.
vo en
• •
«S- 00
^ -*




co
. • CM
^ •
t— • r~-
CM CO
• •
^- CM




m --<

cn vn
"•^ -»v.
m CM
• 9
0 O
1 1




cn cn
• •
«^
^*
. .
^-^ ^^
*-* O
1 1









^_ma^ ^ r L
o o
m vn
*-< CM
«t • »
o m
in vo
0 O
x £"
a. a.
CM O ' "1
• •
^t CO
en vo
CM CM
lf> CM
• *
VO CM
CM *-4
r*4 r-1


VO CO
^•* co"
in CM
CM CM
in r*4
0 •
r** r*»
cn co

I** VO
• •
VO '-O
o co
CM •-!
*-. •».
r-. *r
• •
cn ^»-
«r ^f
•f»
o
M
.fm
C- Q.
O CO
£ •—

m «»-
i o
•*—
^j vi
« c
f CU
t— *J
c
fit
Ql
W  co
^^ -—i
^^ ^^^
^^» ^^s
r* ^*
. •
^ ^*




O
• CM
O •
•— « cn
vo <*
• •
O 0




«9- VO
t • '
o
•^ o
^04

,— vv
r— -O
O
•0 vn
c
^- o
00 <"
CTl -^
s *-

*- Q.
O E
0
(A U
c
o
•^ •
»— in
•— en
•-- cn
^3 --<
cn co c J=
*•••. *«v
CO O
• •
t-4 ^->
1 1


^* T"

.
M * |
0 O
cn en
**** I
A. fH
W» UH
CO CO
l_ 1—

o" c?
in in
•-4 CM
o m
m vo
~li-* ^*^ . Wl yz.
i- cn
a
vi O >
41 I. ft)
3 £ M
.— *J -f-

> cn
03 Of
+J Cf» »—
C 1-4 0
At •«•»
U) £ ^
£S^
Q.H- Q.
•O VI 4J
0) C_ O
*j 
3 >> *
O X
u e u
in vv c
•»- > cu
•O 0» cn
*4« .^
CO -r-
O O CO  (/I
1— 1


*J
CO t-
^t- e
• 3
. ^ <1i
^FT W
vn
e
o ~
v» **
•r- C
vw a
reported. The
econd is based <
i/i
O!
<- eg
 J
O j*
2 
•U •!—

C
0 >>
•i- 4->
•^J *^*
*Tl ^^
•w ^^
C IQ
••- 4-»
.a t-
§o
E
o

•p 
cn u
O> 3
ia -o •
w c
.c * o
U -r-
•O 4-> 4->
4) -t- O
«- §«3
o ai
u. * t.
f«J.




















•
•^
C

vn

(&
«•
«)
C
>

*
X
GL
cn
^_
^^
1_
o
o
(••4
£
a.
^^
•^

-------
                                   IV-4

in previous analyses.  Second,  the changes  to  the  air quality  data  result  in a
significant increase in benefits.
     Finally, it should be noted that the incremental  net  benefits  analysis
should be viewed in light of the caveats  and limitations associated with the
underlying benefit and cost analyses.  Of particular importance  is  the  fact
that the incremental net benefits  analysis  still provides  no information on
the preferred particle size.  Incremental  net  benefits are expressed  only  in
terms of TSP stringency.  Although the TSP  (75,150)  standard appears  to be the
preferred standard based on economic efficiency, it  may not be preferred in
terms of particle size.

-------
REFERENCES

Argonne National Laboratory (1986-).  Revised Costs and Air Quality Impacts, of
     Alternative? National Ambient Air Quality Standards for Parti oil ate
     Matter.  Draft technical report prepared for the U.S. Environmental
     Protection .Agency, February.

Manuel et al. (1983).  Benefit and Net Benefit Analysis of Alternative
     National Ambient Air Quality Standards for Particulate Matter.   Final
     report prepared for the U.S. Environmental  Protection Agency,  March.

Mathtech (1986).  Revisions to the PM NAAQS Benefit Analysis.   Report
     prepared for the U.S. Environmental  Protection Agency, July.

Ostro, B.O. (1983a).  The Effects of Air Pollution on Work Loss and Morbidity.
     Journal of Environmental  Economics and Management,  10:371-382.

Ostro, B.D. (1983b).  Urban Air Pollution and  Morbidity:   A Retrospective
     Approach.  Urban Studies.

Samet,.J.M. et al. (1981).  The Relationship Between Air Pollution  and
     Emergency Room Visits in an Industrial  Community.   Journal of the Air
     Pollution Control  Association, 31:  236-240.

U.S. Environmental  Protection Agency (1984).   Regulatory Impact Analysis on
     the National  Ambient Air Quality Standard for Particulate Matter.  Pre
     pared  by the  Strategies and Air Standards Division, Office of Air, Noise,
     and Radiation, February 21.

-------

-------
                                    V-l

V.   Rationale for the Promulgated Action
     [Text to be added.  Text will  summarize Part 50 Federal  Register Notice
rationale.]

-------

-------