-------
                            Figure  4-1
  SCHEMATIC REPRESENTATION OF THE SHORT-RUN AND LONG-RUN  IMPACTS
OK GASOLINE CONSUMPTION OF AN INCREASE IN THE GASOLINE  EXCISE  TAX
 P+At

    P
                q2
                                95

-------
     Suppose however that we desire to know the impact of
the tax increase, say one year hence, and take DSR as the
demand schedule appropriate to a short run of this duration.
Then such a policy will cause gasoline consumption to be
reduced to the level qr  The total reduction of desired
gasoline consumption in this case is  [q - q-,} which, of
course, is less than before.  That is, in the short run,
when all the economic effects of the policy have not had
time to work themselves out, a given tax increase will lead
to a smaller reduction in consumption than in the long run.
This expresses the basic economic result that demand is
less price-elastic  (i.e., price-sensitive) in the short
run than in the long run.
     Under the assumptions implicit in the above discussion,
then, an increase in the federal excise tax on gasoline will
result in reduced gasoline consumption and thus improved
ambient air quality.  The same argument also shows that the
long-run effects of a  "once and for all" tax increase are
greater than the short-run effects.  Both of these results
are partial equilibrium results, in that they do not take
account of all of the other factors influencing gasoline
consumption.  Because  these other factors are constantly
changing, it is often difficult in practice to observe directly
and precisely the effects of a policy change.  For example,
in prosperous times more and larger cars tend to be sold,
leading to an increase in gasoline consumption.  If these
times happen   to coincide with an increase in the excise
tax on gasoline, it might appear as though the tax had no
effect.  The difficulty of interpreting  the data is compounded
over longer time periods, as more extraneous influences are
felt.  The deliberately simplified analysis of the policy
changes ought not to be regarded as directly applicable to
                              96

-------
the prices and quantities observed in the market.   In the
empirical analysis, we use an econometric model to disen-
tangle the different influences on gasoline consumption.

     The Effects on Alternative Modes of Transportation
     The discussion thus far has focused on the direct effects
of these policies; that is, what happens to gasoline consumption
and emissions by automobiles when the price of gasoline is
increased.  The reduction in gasoline consumption may arise
from several different sources — a decrease in trip length,
an increase in auto occupancy, an increase in fuel economy
of the fleet, and a decrease in the number of trips.1
     Some of the reduction in auto trips comes about, how-
ever, because drivers and passengers shift to alternative
modes of transportation — buses, subways  (in a few cities),
trains, and (for intercity trips) airplanes.  These other
modes of transportation also consume energy and emit pollu-
tants.  As this study focuses on urban air quality, we
will be concerned primarily with fuel consumption and
emissions from buses.
     Economic theory does not furnish any necessary condi-
tions for the response of total  (all modes) fuel consumption
and emissions to an increase in the excise tax on gasoline
(especially since over 80 percent of municipal bus system
fuel consumption is diesel oil).  In principle, that is,
total fuel consumption might increase.   This perverse

   1These responses are those predicted by economic theory
as the underlying reasons for the responsiveness of gasoline
consumption to increases in the price of gasoline.  We do not,
however, have quantitative estimates of their relative impor-
tance.
                             97

-------
response might come about if certain very implausible con-
ditions held; for example, if total fuel consumption per
passenger mile was greater for buses than for autos and if
all the reduction in passenger miles by auto was diverted
onto buses.
     For a number of reasons, however, such conditions are
quite farfetched.  Several of the consumer responses men-
tioned above — a decrease in average trip length, an
increase in average auto occupancy, and an increase in fleet
fuel efficiency — do not imply an increase in passenger
miles by alternative modes.  Moreover, some of the reduction
in total trips probably comes about because people schedule
their trips more carefully  (deferring some trips to combine
them with others, for example) or simply do not make some
trips that they did at lower gasoline prices.
     Finally, fuel consumption and emissions per passenger
mile are, at present, lower for alternative modes than for
autos, given current load factors  (average proportion of
vehicle capacity being used).l  With an increase in the
demand for public transportation, these load factors prob-
ably increase somewhat in the short run, before additional
capacity can be added to the  system.  Higher load factors
imply still  lower gasoline  consumption  and emissions per
passenger mile.
     In summary, then, it seems most unlikely that, in the
short run, the increase in  fuel consumption and emissions

    1As discussed later in this chapter, however, emission
factors for automobiles are projected to fall much faster
than for buses.  Consequently, current and expected exhaust
emission standards imply that some pollutant emissions per
passenger mile are likely to be higher for buses than for
autos between 1981 and 1987, if the relative occupancy
rates of buses and autos do not change.
                             98

-------
from alternative modes will completely offset the reduc-
tion in fuel consumption and emissions from automobiles.
These increases do, however, need to be taken into account,
and they are considered later in this chapter.

The Imposition of Gasoline Rationing
     We consider a policy of gasoline rationing with a "white"
market in coupons.  That is, the coupons necessary to pur-
chase gasoline may be legally bought and sold.  As will be
shown below, this rationing scheme is quite similar to
increasing the excise tax to allocate gasoline among users,
except that the income from the effective increase in prices
is retained by consumers of gasoline instead of the increased
tax receipts accruing to the Treasury (and, assuming no
increase in total tax collections, to the general taxpayer).
     The impact of gasoline rationing on gasoline consump-
tion and on ambient air quality follows directly if
rationing is effective, that is, if the number of gasoline
coupons (the size of the stock to be allocated) is less
than the market-clearing quantity of gasoline which would
be consumed in the absence of rationing.  In this case,
rationing leads to less gasoline consumption and, through
reduced automotive emissions, to improved air quality.
We show below that, for any quantity of gasoline to be
rationed,  there exists an excise tax increase which leads
to an equivalent reduction in gasoline consumption.  Fur-
thermore,  the white market price of coupons under the given
rationing scheme will be equal to this excise tax increase.1

   JThis equivalence assumes that the costs of buying and
selling coupons are so small that these costs do not
appreciably reduce the market value of the coupons.
                             99

-------
     In order to see this, consider the diagrammatic repre-
sentation of a rationing scheme given in Figure 4-2.  Here the
market is initially in equilibrium at point E3  with the con-
sumption of qQ gallons of gasoline per year at the price P„.
Now suppose that a policy of gasoline rationing with a -white
market in coupons is imposed, and that the aggregate allot-
ment is q  gallons per year.  As may be easily seen in the
diagram, if DSR represents the short-run demand for gasoline,
then in the short run consumers would desire to purchase
exactly the allotment qp, if the price of gasoline were PQ +t 2'
Thus, increasing the excise tax on gasoline by the amount
T  will induce a reduction in consumption in the short run
 £j
equivalent to that achieved by the rationing policy.
     Notice however that while the tax increase of T« is
sufficient to achieve the desired reduction in consumption
in the short run, it will continue to reduce consumption below
the desired rate if maintained indefinitely.  This is because,
as noted above, the long-run demand schedule is more price-
elastic than the short-run demand.  Thus, if one desires to
maintain consumption at the reduced rate of qy for a prolonged
period of time, the excise tax increase necessary to accomplish
this must be reduced as time goes by.  Let VLR in Figure 4-2 repre-
sent the demand schedule corresponding to the indefinite long
run.  Then examination of Figure  4-2 reveals that the tax increase
which sustains consumption at the rate q  approaches the level
T7 as time recedes indefinitely.
     Let us now consider the determination of the white market
price of the ration coupons.  This price will in general be
exactly equal to the excise tax necessary to effect a reduction
in consumption equivalent to that achieved by the rationing
policy.  This means that the price of coupons will vary over
                                100

-------
                          Figure  4-2

   SCHEMATIC  REPRESENTATION OF THE SUPPLY OF AND DEMAND
             FOR COUPONS ON THE WHITE MARKET
 T2
Tl
                'SR
                              101

-------
time, starting at the level T^ in the short run, but eventually
approaching the long-run equilibrium level T^.   We may substan-
tiate this claim for the long run by examination of Figure 4-3.
Here we depict the long-run supply and demand schedules for
gas coupons, deduced from Figure 4-2.  Supply is of course fixed
at the level a , since the quantity of coupons available will
              r
equal the quantity of gasoline to be rationed.   Demand for cou-
pons at a given price, on the other hand, will equal the demand
for gasoline at the effective price of gas implied by the coupon
price.  That is, if the price of gasoline at the pump remains
constant at PO (as in Figure 4-2) after the rationing scheme is
imposed, and if the price of a coupon is at some level T^, then
the effective price of gasoline is p0+^2'  Tnis is so because
when an individual buys a gallon of gasoline, he must pay $P Q in
cash, and present a coupon worth $7^.   Had he not purchased the
gasoline, he could have sold the coupon on the open market for
$T-.  Symbolically,
where D  represents the demand for coupons (in gallon-entitle-
       G
ments) on the white market, and D  is the demand for gasoline
(in gallons).  The white market will clear at price T^, if and
only if the gasoline market clears at price P 0+i .   However,
this also means that an excise tax increase of i  will reduce
gas consumption to the rate q .
     This conclusion holds in principle,  but in practice it is
impossible to know with certainty the demand curve for gasoline.
Consequently, an excise tax allows us to  estimate with reason-
able precision the increase in price, but the quantity of con-
sumption is subject to considerable uncertainty.  Conversely,
under rationing the amount of consumption is known with
                              102

-------
                          Figure 4-3
SCHEMATIC REPRESENTATION OF  THE EQUIVALENCE OF  A  GIVEN  RATIONING
          SCHEME TO THE APPROPRIATE EXCISE  TAX  INCREASE
                                                          0
                              103

-------
reasonable certainty, but the value of a coupon (and, hence,
the effective price of gasoline) is subject to considerable
uncertainty.1  In addition to uncertainties about the shape
OT the demand curve, there are other determinants of demand
besides price (such as the stock of cars and their average
fuel economy), and variations in these determinants cause
consumption to vary.  Thus, the relative certainty about the
quantity consumed under rationing may be an advantage, if
uncertainty about effective price is less of a concern than
uncertainty about quantity.   (Administrative costs of ration-
ing are discussed later in this chapter.)
     Finally, one important difference between the rationing
and equivalent excise tax schemes should be noted.  While they
have the same effect on gas consumption, the income distribution
effects are, at least in principle, different.  This difference
is discussed in the section of this chapter entitled "Secondary
Impacts."
     In the qualitative analysis, we observed that an
increase in the federal excise tax on gasoline would increase
the market price of this fuel, leading to a reduction in
the consumption of gasoline by automobiles.  It was further
observed that reduced consumption of fuel by autos implies
reduced automotive emissions  of various pollutants and, as
a consequence, improved air quality.  Now we wish to quan-
tify these effects.  The linkage may be symbolically repre-
sented as

     [At]  •> [Ap]  f [A0]  + [Ae] -> [Aa]
          1231*

    1Total consumption under rationing  is not known precisely
because the  number of licensed  drivers  is  not  known with  certainty,
Indeed, rationing is likely to  create  incentives  for  eligible
citizens without drivers' licenses to apply for them.
                               104

-------
(where t = excise tax on gasoline, p = tax-inclusive price
of gasoline, g = gallons of gas consumed by autos, e =
quantity of automotive emissions into the atmosphere, cc =
some measure of ambient air quality, and A = change in).
     Links 3 and 4 in this chain  (the effect of a change
in gasoline consumption on emissions, and the effect of a
change in emissions on ambient air quality) were discussed
in Chapter 3.  The rest of this section deals with the
first two links.

Effects of a Tax Increase on the Price of Gasoline
     Link 1 embodies the effect of increasing the excise tax
on gasoline on the market price of the fuel.  In general, the
relationship between a change in the tax and the change in
the equilibrium price depends on both the demand and supply
schedules for gasoline.  We simplify the analysis, however,
by assuming that the supply of gasoline is perfectly elastic.
In this instance, an excise tax increase of any given size
will lead to the same increase in the market price of gasoline.1
     The effect of this assumption may be seen with the
aid of Figures 4-4tand 4-5.  Figure 4-5 depicts the case of
perfectly elastic supply while Figure 4-4 shows the situation
when this assumption does not hold.  In each instance we
assume the market initially in equilibrium at the intersection
of the demand schedule D and the  pre-tax schedule S  .  The

   1We have made this assumption because, to our knowledge,
there are no reliable estimates of the elasticity of sup-
ply of gasoline.  Moreover, as shown in Section A of Appen-
dix C, under reasonable assumptions about the elasticity of
supply, almost the full amount of the tax is translated
into the price.  To the extent that supply is less elastic
than assumed, however, the equilibrium price rises by less
than the full amount of the tax increase, and our estimates
of the change in gasoline consumption overstate the change
that will occur.
                               i05

-------
                              Figure 4 -4

     SCHEMATIC REPRESENTATION  OF THE  EFFECT OF A  CHANGE IN THE GASOLINE
        TAX  ON THE EQUILIBRIUM PRICE  OF GASOLINE, WHEN SUPPLY IS NOT
                         PERFECTLY ELASTIC
"prVAT
                                                               •* q
                                 106

-------
                                 Figure  4-5
      SCHEMATIC  REPRESENTATION OF THE EFFECT OF  A CHANGE IN THE GASOLINE
         TAX ON  THE  EQUILIBRIUM PRICE OF  GASOLINE, WHEN SUPPLY  IS,
                           PERFECTLY ELASTIC
PfP0+AT
                                  107

-------
market clearing price is p  and Q0 is the quantity consumed.
In Figure 4-4 the pre-tax supply schedule S  is less than
perfectly elastic, having a positive slope.  This reflects
the assumption that an increase in the price of gasoline
will increase the amount of fuel which producers desire to
supply to the market.  Figure 4-5, on the other hand, shows
a horizontal pre-tax supply schedule which is perfectly
elastic.  In this case  (corresponding to constant average
production costs) producers are willing to supply at the
price P  whatever amount is demanded.
     In either case, we may think of an increase in the excise
tax of At as shifting the supply schedule vertically by
that amount at each point.  This is so because producers
will be willing to supply the same quantity after the tax
increase as before if and only if the net price they
receive for their product remains unchanged.  But this is
true only if the market  (gross) price is increased fay the
amount of the tax increase.  Thus at each quantity of the
post-tax supply schedule, S  lies above the pre-tax supply
schedule by the amount of the tax increase.  The post-tax
market clearing price and quantity under the respective
assumptions about supply elasticity may be compared by
observing the intersection of the post-tax supply schedules
with the unchanged demand schedule in the two figures.
     It is clear in Figure 4-5 that, when supply is per-
fectly elastic, a given increase in the excise tax leads
to the same increase in the market clearing price, as
asserted above.  Similarly, from Figure 4-4 we can see that
if the assumption of perfectly elastic supply fails, then
an increase in the excise tax causes the market clearing
price of gasoline to increase by less than the price amount
                              108

-------
of the tax (p  < P ).  It is also clear in this instance
that the reduction in consumption caused by the tax
increase is less than predicted under the assumption of
perfect elasticity adopted for the purpose of our analysis.
However, as is shown in Section A of Appendix C, as long as
the elasticity of supply is "large" relative to the elas-
ticity of demand, the size of these effects will be negli-
gible.  Since the demand for gasoline is notably inelastic
in the short run,1 and the supply is probably very elastic
in the long run,2 the assumption that price increases by
the full amount of the tax should not give misleading
results.

Effect of Price Increase on Gasoline Consumption
     For the purpose of analyzing increases in the excise
tax on gasoline, link 2, the effect of a price change on
the quantity of gasoline demanded, is critical.  This
effect is embodied in the price elasticity of demand for
gasoline.
     Knowledge of this aggregate elasticity, however, is
insufficient by itself to permit the further determination
of the effects contained in links 3 and 4.  This is because
the effect which a given reduction in gasoline consumption

   :See the discussion of the price elasticity in Chapter 3.
   2 If, in the long run, refiners respond to permanent
shifts in demand by increasing or shutting down refinery
capacity (as the case may be), and if new and old refin-
ing capacities have  similar costs, then the long-run supply
schedule will be very elastic.  Verification of these
hypotheses is beyond the scope of this report, although
the gradual decline in real gasoline prices during the
1960's suggests that the assumptions are reasonably correct.
                             109

-------
has on the quantity of automotive emissions of various
pollutants (Ikg] -*• l&e])r depends on the extent to which
the initial drop in consumption results from a reduction in
the number of automotive trips made as opposed to a reduction
in the average  length of such trips.  Furthermore, the effect
of reduced emissions on ambient air quality ([&&]  •*• l&a] )
depends on where (urban vs. rural emissions) and when
(peak hour vs. offpeak hour emissions) such a reduction in
emissions takes place.
     Consequently, analysis of the effect on air quality of
a given excise tax increase requires disaggregation along
three dimensions of the overall effect on consumption embodied
in link 2.  We must know the effect of [Ap] on [A#] by time
and place, and for each such subaggregate we must decompose
the change in gasoline consumption into changes in the number
of auto trips and changes in the average length of these trips,
     For this purpose we assume two mutually exclusive and
collectively exhaustive temporal and spatial categories of
gasoline consumption:  peak hour vs. offpeak hour consump-
tion and urban vs.  rural consumption, respectively.
     This leaves eight parameters to be determined (a trip
frequency and trip length elasticity for each time-of-day
and place-of-consumption category).  Unfortunately, not all
of these disaggregated elasticities have been estimated.
Various assumptions and implicit constraints must be used to
make point estimates and define plausible ranges for these
parameters.  The procedures are described in detail in Sec-
tion A of Appendix C.  The net effect of the assumptions,
however, is to increase the uncertainty about the urban
trip-making elasticities that are central to the air
quality problem.  That is, these assumptions do not affect
                             110

-------
the direction of the results, but the uncertainty about them
increases the uncertainty about the exact size of the para-
meters of interest.  This increased uncertainty is  (partially)
reflected in the size of the "confidence intervals" for the
estimated effects given in Section B of Appendix C.
     In general, urban peak gasoline demand is less  price-
elastic than urban offpeak gasoline demand, because  peak
travel contains relatively more work trips than offpeak
travel, and work trips are less responsive to increased
travel costs than are trips for other purposes.  The esti-
mated urban gasoline demand elasticities vary by state,
depending on the price of gasoline in the state as well as
the degree of urbanization.  States in which the rural
share of gasoline consumption is small generally have less
elastic urban peak and offpeak gasoline demands than
states with larger relative rural auto use.

Short-Run vs. Lpng-Run Gasoline Demand
     We noted in the qualitative analysis that the demand
for gasoline is likely to be more responsive to a given
tax increase in the long run than in the short run,
because changes in consumption habits and in the stock of
cars in response to a change in the price of gasoline
take time to occur.  Thus, for a given increase in the
excise tax on gasoline, annual consumption is reduced
somewhat in the year following the tax increase as con-
sumption partially adjusts to the price change.  If  the
new, high price persists, in subsequent years gasoline
consumption falls still further, reaching a floor as
                             111

-------
complete adjustment is made.1  This process is schematically
shown in Figure 4-6.  The initial  (equilibrium) consumption
of gasoline per unit of time is g  .  Imposition of a tax increase
results in a new, lower long-run equilibrium quantity demanded/
given by g~.  But since adjustment takes time, g(t), the
actual flow demand for gasoline at any time t after the tax
increase, lies between 9$ and g^ •   Gasoline consumption
approaches g^ as time goes by.
     Under certain reasonable assumptions  (described in
Section C of Appendix C) it is possible to measure how
quickly demand responds to a change in price.  For example,
if one assumes that each year demand adjusts from its current
level toward the long-run equilibrium level by some constant
fraction (s) of the difference,2 then it is possible to
estimate from time series data the size of s.  Knowledge
of this parameter, in conjunction with an estimate of the
short-run price elasticity of demand, permits determination of
the change in consumption for a given change in price over
any duration.  This parameter was estimated to be 0.788.
It was used in the analysis of the effects of proposed
policies in the years 1981 and 1987.  This analysis pro-
ceeds  (using disaggregation techniques described above)
exactly as with the short-run analysis for 1975, replacing

   *For simplicity, we assume in this discussion that all other
factors — such as population, the number  of automobiles, tech-
nology, fuel economy, and so forth — are  held constant.  In the
quantitative analysis, of course, base gasoline consumption is
increasing over time, and the long-run elasticity is reflected
in a reduced growth rate.
 2In Figure 4-4, this implies the differential equation
[g(t)-g^], with exponentially decaying consumption.  The
eventual solution is the asymptote, g (°°) = g~.
                               112

-------
                                      Figure 4-6

          SCHEMATIC REPRESENTATION OF THE TIME PATH OF GASOLINE CONSUMPTION

          IN  RESPONSE TO A SUSTAINED  INCREASE IN THE EXCISE TAX ON GASOLINE

                              (ALL OTHER THINGS EQUAL)
Gasoline
consumption
per unit of
time
                                                                   Time
                                      113

-------
the short-run demand elasticity with a modified elasticity
appropriate to the increased duration of the period of
analysis.
     The estimate of the long-run adjustment coefficient
implicitly takes into account changes in the stock of
cars in response to changes in gasoline prices.  We also
assume, however, that average emission factors of the stock
of cars are not affected by these policies.  In a strict
sense, this assumption is probably not correct.  Higher
gasoline prices may cause the stock of cars to be replaced
more rapidly, as older fuel-inefficient cars become less
attractive relative to new, fuel-efficient models.  Since
new cars have lower emission factors than cars sold before
the exhaust emission standards were implemented, the assump-
tion made here probably overstates the average emission
factors for the fleet.  We expect this effect to be very
small.  Since the difference in gasoline costs between new
and old cars is such a small fraction of the cost of car
ownership and operation, it is unlikely that an increase
in gasoline prices will cause the substantial change in
scrappage rates that would be required to affect average
emission factors significantly.  In any case, the direction
of the error is to overstate the emissions associated with
gasoline taxes and rationing, and these policies are shown
below to lead to a substantial reduction in emissions.

-------
     Table 4-1 shows the central elasticity estimates used
in this study for 1975, 1981 and 1987.  These estimates
assume that the price change occurs at the beginning of
1975 and continues in effect to 1987.  The elasticities
for 1981 and 1987, therefore, reflect the long-run adjust-
ment parameters.  These elasticities are disaggregated
into peak and offpeak components, which, in turn, are
weighted averages of work and nonwork trip elasticities.
The weights are the relative shares of each kind of trip
during peak and offpeak hours.
     We also disaggregate these elasticities by number of
trips and average trip length, because the number of
trips affects cold start and hot soak emissions.  As can
be seen in Table 4-2, most of the reduction in gasoline
consumption can be attributed to a reduction in the num-
ber of trips.  Because there are no separate long-run
estimates of trip and average trip length elasticities,
we assumed that the ratio of trip to average trip length
elasticities was constant over time  (although peak and
offpeak ratios differ).
     These elasticities are used in the next section to
derive the impact of the different policies on gasoline
consumption.*

   JFor brevity, only national average elasticities are
shown in Tables 4-1 and 4-2.  The elasticities vary from
state to state, however, and the actual computations used
the state-by-state elasticities to derive gasoline con-
sumption by city size and by the representative air qual-
ity cities.

-------
                             Table 4-1

                  NATIONAL AVERAGE DISAGGREGATED
               URBAN GASOLINE DEMAND ELASTICITIES1
Year                             Peak        Offpeak        Total
1975                            -0.149        -0.173        -0.164
19812                           -0.539        -0.626        -0.592
I9872                           -0.600        -0.697         -0.659
   Elasticities assume gasoline prices at their post-embargo levels.

   2AII  elasticities assume that the price change occurs at the
beginning of 1975 and remains in effect throughout the period of
ana Iysis.
                                 116

-------
                             Table 4-2
                 NATIONAL AVERAGE ELASTICITIES OF
              NUMBER OF TRIPS AND AVERAGE TRIP LENGTH
                          Number of Trips
Year

1975

1981

1987
Peak
-0.134
-0.484
-0.538
Offpeak Total
-0.167 -0.153
-0.604 -0.553
-0.672 -0.616
                       Average Trip Length
Year
1975
1981
1987
Peak
-0.016
-0.056
-0.062
Offpeak
-0.006
-0.023
-0.025
Total
-0.010
-0.036
-0.041
                               117

-------
                        Results

     In this section, we discuss the results of the different
policies affecting gasoline consumption directly.  We dis-
cuss first the change in gasoline consumption, as measured
from the base line forecast gasoline consumption, for the
different policies.  We next discuss the change in emissions,
on a national level, for the different pollutants.  We also
discuss changes in concentrations of the different pollutants,
by the appropriate time of day, as measured by an index of
concentration of these pollutants in 13 different cities
spread across the United States.

Gasoline Consumption
     Table 4-3 presents a summary of our best estimates of
gasoline consumption in 1975, 1981, and 1987 for each of
the four policies.  These forecasts assumed the policy goes
into effect in 1975.  The forecasts shown in Table 4-3 use
the point estimates from the gasoline demand equation and
the long-run adjustment equation.  These estimates are based
on high gasoline prices.
     These estimates reflect the extreme insensitivity of
gasoline consumption to changes in price in the very
short run.  For example, a tax of $0.10 per gallon (which
is double the amount suggested in one recent policy)
leads to only a 3 percent reduction in consumption in the
base line forecast.  A $0.50 per gallon tax on gasoline,
which implies almost a doubling of the current price, leads
only to a 15 percent reduction in consumption.   Rationing,
on the other hand, is much more effective in reducing con-
sumption.  The cost of this reduction of almost 40 percent
                            118

-------
                                 Table 4-3
                  MEDIUM ESTIMATE OF GASOLINE CONSUMPTION
             UNDER POLICIES AFFECTING GASOLINE DEMAND DIRECTLY
                          (Billions of Gallons)
                                                               As Percentage
Year
1975



1981



1987



Policy
$0. 10/gal.
$0.25/gal.
$0.50/gal.
Ration! ng
($l.27/gal .)
$0. 10/gal.
$0.25/gal.
$0.50/gal.
Ration i ng
($0.39/gal .)
$0. 10/gal .
$0.25/gal .
$0.50/gal.
Ration! ng
($0.37/gal .)
10 Km
Cities
17.54
16.74
15.39
1 1.26
20.24
16.64
10.62
13.27
24.55
19.848
1 1.30
15.60
35 Km
Cities
36.97
35.27
32.44
23.73
42.42
34.86
22.26
27.80
51.17
40.820
23.56
32.53
Rural
47.18
45.00
41.36
30.18
54.43
44.68
28.41
35.55
66.00
52.57
30.17
41 .79
Total
101.69
97.01
89.19
65. 17
1 17.09
96.18
61 .29
76.62
141.72
1 13.24
65.03
89.92
of Baseline
Forecast,
97.0
92.6
85.1
62.2
89.4
73.4
46.8
58.5
88.1
70.4
40.4
55.9
Assumes (I) high gasoline prices; and (2) medium sensitivity assumptions.
                                    119

-------
is, however, a fairly high price per coupon.   The estimated
price of a coupon under a rationing scheme limiting
licensed drivers to 10 gallons per week is almost $1.30,
more than double the current price of gasoline.
     In the longer run, however, as the drivers have time
to adjust their patterns of consumption and as demand
adjusts to the higher prices, demand becomes much more
elastic.  By 1981, for example, the same $0.10 per gallon
increase in the excise tax has led to greater than a 10
percent reduction in gasoline consumption.  The estimated
reduction in consumption from a sustained increase in the
excise tax of $0.50 per gallon is over 50 percent of base
line consumption.  The increasing elasticity of demand
implies, however, that the price of a coupon falls over
time.  By 1981, the price of a coupon implied by policy
restricting gasoline consumption to 10 gallons per licensed
driver per week is about $0.39 per gallon.   (It should be
noted, of course, that the policy as we have defined it
implies a growth in gasoline consumption concomitant with
the growth of the number of licensed drivers.)
     Tables 4-4, 4-5, and 4-6 show a more detailed disag-
gregation of gasoline consumption in 1975, 1981, and 1987,
respectively, for the best estimates of these  policies.
These tables show that, for the point estimates used here,
there is not much difference in the percentage reduction
during peak hours as opposed to offpeak hours.  For example,
in 1975, for the most  severe reduction in gasoline consump-
tion, peak consumption falls by about 35 percent, while
offpeak consumption falls by about 40 percent.  This close-
ness arises from two factors.  First, the less the overall
percentage reduction in demand, the more  narrow will be  the
spread between  the peak and  the offpeak reductions.  For
                             120

-------
                         Table 4-4

     URBAN GASOLINE CONSUMPTION IN 1975, MEDIUM SENSITIVITY
                    (Billions of Gallons)
                   0 KM
         35 KM
         TOTAL
           FRACTION
           OF BASE
   TAX: $0.10

        PEAK:
        OFF-P:
        TOTAL

   TAX:$0.25

        PEAK:
        OFF-?:
        TOTAL

   TAX. '$0.50
        OFF-P:
        TOTAL
7.54442
9.99632
17.5407
7.22787
9.507
16.7349
6.70028
8.69147
15.3918
15.9016
21.0695
36.971 I
15.2344
20.0382
35.2725
14.1224
13.3193
32.4416
23.446
31.0658
54.5118
22.4622
29.54b2
52.0074
20.8227
27.0107
47.8334
0.972789
0.968398
0.970282
0.931972
0.920995
0.925704
0.863945
0.841991
0.851409
   RATIONING

        PEAK:
        OFF-P:
        TOTAL
5.07532
6.17965
11.255
10.6974
13.025
23.7224
15.7727
19.2047
34.9774
0.65442
0.598656
0.622579
SENSITIVITY:  MEDIUM

PRICE ASSUMPTION:  HIGH
                            121

-------
                          Table 4-5

       URBAN GASOLINE CONSUMPTION IN 1981, MEDIUM SENSITIVITY

                     (Billions of Gallons)
                   10 KM
         3D KM
         TOTAL
          FRACTION
           OF BASE
   TAX:$0.!0

        PEAK:
        OFF-P:
        TOTAL

   TAX:$0.25

        PEAK:
        OFF-P:
        TOTAL

   TAX:$0.50

        PEAK:
        OFF-PS
        TOTAL

   RATIONING

        PEAK:
        OFF-P:
        TOTAL
8.77114
11.4715
20.2427
7.35415
9.28117
16.6353
4.99249
5.63058
10.6231
6.03162
7.23684
13.2685
18.3796
24.0382
42.4176
15.4104
19.4464
34.8588
10.4616
11.7967
22.2603
12.6391
15.1646
27.8036
27.1503
35.5097
62.6605
22.7645
28.7296
51.4941
i 5.4541
17.4293
32.8834
18.6707
22.4014
4!.0721
0.90277
0.887082
0.893312
0.756926
0.717704
0.73453
0.513852
0.435407
0.46906
0.620805
0.559618
0.585867
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
                              122

-------
                         Table 4-6

      URBAN GASOLINE CONSUMPTION IN 1987, MEDIUM SENSITIVITY

                    (Billions of Gallons)
                    0 KM
         35  KM
         TOTAL
           FRACTION
           OF BASE
   TAX:$0.!0

         PEAK:
         OFF-P:
         TOTAL

   TAX:$0.25

         PEAK:
10.6494
13.895
24.5444
8.69834
22.2043
28.9715
51.1758
32.8537
42.8665
75.7202
0.89]156
0.873593
0.881127
 8.1363   26.8346    0.72789
        OFF-P:
        TOTAL

   TAX:$0.50

        PEAK:
        Oi-r-P:
        TOTAL

   RATIONING

        PEAK:
        OFf-P:
        TOTAL
10.8791
19.5775
5.4466
5.85267
11.2993
7.13751
8.46643
15.6039
22.6833
40.b196
11.3563
12.203
23.5593
14.8819
17.6527
32.5346
33.5624
60.3971
i 6.6029
lb.0557
34.8566
22.0194
26. I 192
46. 1 386
0.683981
0.702818
0.455779
0.367963
0.405636
0.597277
0.532293
0.560171
SENSITIVITY:  MEDIUM

PRICE ASSUMPTION:  HIGH
                              123

-------
example, under a $0.10 per gallon tax, peak demand falls
by 2.7 percent, while offpeak demand falls by 3.2 percent.
Since, according to the way the peak and offpeak elasticity
estimates have been constructed, the ratio of the percentage
decrease in peak consumption to the percentage decrease in
offpeak consumption will be the same fcr a given set of
point estimates and price assumptions, the spread between the
decreases will naturally increase with the decrease in total
gasoline consumption.  Second, although work trip demand is
considerably less elastic than the demand for other kinds
of trips (assumed to be similar to shopping trips in this
analysis),  not all work trips occur during peak hours, nor
do all shopping trips occur during offpeak hours.  This
mixing of trip types at different times of the day tends to
cause peak and offpeak elasticities to be closer to each
other than if different trip types were rigidly made at
different times of the day.  Even for the most extreme
percentage decrease in consumption — that associated with
a $0.50 per gallon tax in 1987 — the reduction in peak
demand is about 55 percent, while the reduction in offpeak
demand is less than 65 percent.
     This similarity in the elasticities of peak and off-
peak demand suggests that, as a first approximation, it is
not too misleading to assume that peak and offpeak demands
have similar elasticities, from the point of view of deter-
mining the impact of changes  in fuel consumption on air quality,
That is, applying the overall elasticity of demand will give
approximately right results for both peak and offpeak gaso-
line consumption.  Certainly, the difference between the
peak and the offpeak demand redaction is less than the un-
certainty associated with the estimate of overall gasoline
demand itself.
                             124

-------
     The central elasticities used here were estimated
from historical data.  Some of the policies considered
(such as a $0.50 increase in the excise tax or rationing)
imply price increases well outside the range of the sam-
ple.  Consequently, these estimates may be especially
subject to error for large price increases.  Figure 4-7
illustrates the difficulties of extrapolating beyond the
range of observed prices.  The method used here assumes
that the demand curve, estimated over the price range PT
to Pn, continues with the same slope over the range P_
    C                                              '  0
to p.. (segment AB) .  In fact, however, it might very well
become either more elastic or less elastic in this range,
as represented by segments AC and AD, respectively.
     There is no econometric method of estimating demand
elasticities outside of the observed historical range;
the economy has not performed the requisite experiments.
One way to evaluate the uncertainty, however, is to use
different elasticity estimates and determine how the
results change.  For this purpose, we assume that the true
coefficients lie within one standard error on either side
of the point estimates of the short-run price elasticity
and of the long-run adjustment coefficient.
     The 1975 results are not especially sensitive to
which elasticity estimate is used; a $0.50 per gallon
increase in price leads to a decrease in consumption
between 8 and 22 percent of the base case level.  The
1987 estimates vary markedly according to the elasticity
assumed; a $0.50 per gallon tax surcharge leads to
reductions in gasoline consumption between 90 and 24 per-
cent of the base case.  This wide range largely reflects
relatively small differences in year-to-year adjustment
                            125

-------
                      Figure 4-7
       SCHEMATIC REPRESENTATION OF DEMAND CURVES
         OUTSIDE THE OBSERVED RANGE OF PRICES
Pri ce
                                                Quantity
                         126

-------
of gasoline consumption over a long period of time.  The
results of using different elasticities are presented in
full in Appendix E.

Carbon Monoxide
     The percentage reduction of emissions of carbon mon-
oxide are quite similar to the percentage reductions in
gasoline consumption, with some interesting exceptions.
Our best estimates of urban carbon monoxide emissions,
disaggregated into peak and offpeak for the two different
size cities, are shown in Tables 4-7 to 4-9.  The percent-
age reductions in 1975, shown in Table 4-7, are virtually
identical to the percentage reductions shown in Table 4-4.
By 1987, however, the percentage reductions in carbon
monoxide (shown in Table 4-9)  are not as great as the
percentage reductions in gasoline consumption.  The dis-
crepancy, however, is quite small — on the order of 2
percent.  This difference arises because, as the price of
gasoline goes up, the length of the average auto work
trip tends to fall.1  In the model from which this elas-
ticity estimate was derived, average vehicle trip length
falls for two reasons:  (1) at higher gasoline prices,
longer auto trips are more expensive than short trips,
relative to the transit alternative; consequently, rela-
tively more long auto work trips are diverted to transit;
(2) those who live farthest away from work have the

   1In the terminology of the transportation analyst, the
average auto-driver trip length falls, while the average
of auto-driver plus auto-passenger trips remains constant.
Some additional effects will be introduced by the tendency
of auto-owners to substitute fuel-economic cars in the
longer trips.
                            127

-------
                        Table 4-7

   URBAN CARBON MONOXIDE EMISSIONS  IN 1975, MEDIUM SENSITIVITY
                       (Millions  of Kilograms)
                   10 KM
         35 KM
         TOTAL
           FRACTION
           OF BASE
   TAX:$0.10

        PEAK:
        OFF-P:
        TOTAL

   TAX:$0.25

        PEAK:
        OFF-P:
        TOTAL

   TAX:$0.50

        PEAK:
        OFF-P:
        TOTAL
4286.44
5118.26
9404.7
411 1 .79
4869.52
6981.31
3820.71
4454.95
8275.66
9163.02
10896.7
20059.7
8790.08
10367.3
19157.4
6168.52
9484.94
17653.5
13449.5
16015.
29464.4
12901.9
15236.8
28138.7
11989.2
13939.9
25929.1
0.973574
0.968626
0.970879
0.933935
0.921563
0.927195
0.867872
0.843121
0.854387
   RATIONING
         OFF-P:
         TOTAL
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
2924.16
31 78.09
6102.25
6254.12
6767.34
13021.5
9178.28
9945.43
19123.7
0.664394
0.601526
0.630143

-------
                        Table 4-8

   URBAN CARBON MONOXIDE EMISSIONS IN  1981, MEDIUM SENSITIVITY
                        (Millions of Kilograms)
                    10 KM
         35 KM
         TOTAL
           FRACTION
            OF BASE
    TAX:$0.10

         PEAK:
         OFF-P:
         TOTAL

    TAX:$0.25

         PEAK:
         OhV-P*
         TOTAL

    TAX:$0.50

         PEAK:
         Of F-P:
         TOTAL

    RATIONING

         PEAK:
         OFF-P»
         TOTAL
1210.92
!337.06
2547.98
1024.74
1085.07
2109.01
714.442
665.093
1379.54
850.973
849.884
1700.86
2593.78
2848.67
5442.45
2195.58
2312.05
4507.63
1531.9
1417.67
2949.57
1823.92
1811.2
3635.12
3804.7
4185.73
7990.43
322C.32
3397.12
6617.44
2246.34
2O82.76
432-?. 1 1
2674.89
2661.08
5335.98
0.907114
0.8*8413
0.89722
0.761767
0.721032
0.743052
0.535571
0.442062
0.486102
0.637746
0.5648!
0.5991 6
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
                              129

-------
                       Table 4-9

   URBAN CARBON MONOXIDE EMISSIONS IN 1987, MEDIUM SENSITIVITY
                      (r.llllons  of Kilograms)
                  10  KM
         35 KM
         TOTAL
           FRACTION
           OF BASE
  TAX*$0.10

       PEAK:
       OFF-p:
       TOTAL

  TAX:$0.25

       PEAK:
657.513
615.436
1272.95
547.715
1426.09
1328.98
2755.07
1188.21
2083.6
1944.42
4028.02
0.699892
0.876735
0.888563
1735.93   0.749733
       OFF-p:
       TOTAL

  TAX:$0.50

       PEAK:
       OFF-P:
       TOTAL

  RATIONING

       PEAK:
       OFF-P:
       TOTAL
485.601
1033.32
364.717
269.208
633.925
459.876
381.732
841.608
1048.74
2236.95
791.738
531.67
1373.41
997.902
824.545
1822.45
1534.34
3270.27
1156.46
850.87d
2007.33
1457.78
1206.28
2664.06
0.691832
0.721406
0.499464
0.38366
0.442309
0.629603
0.543909
0.587678
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
                             130

-------
greatest incentive to form carpools; in addition, longer
trips are associated with greater freedom of route choice;
this freedom tends to offset the greater density of poten-
tial ridership associated with shorter work trips to the
central business district.1  The fall in average auto trip
length means, however, that average emissions per mile
increase, because cold start emissions do not change with
gasoline consumption.  Therefore, the carbon monoxide emis-
sions do not fall at the same rate as gasoline consumption.
This argument, while of theoretical interest, is seen to
be rather unimportant, at least for these estimates,
because the difference between the fall in gasoline con-
sumption and the fall in carbon monoxide emissions is
very small.

Hydrocarbons
     Tables 4-10 to  4-12 show the forecasted emissions of
hydrocarbons for the medium estimate for 1975, 1981, and
1987 for the four policies.  These emissions follow the
same pattern relative to gasoline consumption as do the
carbon monoxide emissions, except that, since cold start
emissions are a smaller fraction of total emissions than
in the case of carbon monoxide, the percentage reduction
in hydrocarbons is closer to the percentage reduction  in
gasoline consumption.

   !The model and its results are described in Charles
River Associates, Economic Analysis of Policies for Controlling
Automotive Air Pollution in  the Los Angeles Region (draft report
to the Environmental Protection Agency, March 1975),
Chapter 4 and Appendix A.
                            131

-------
                        Table 4-10

     URBAN HYDROCARBON EMISSIONS IN  1975,  MEDIUM SENSITIVITY
                        (Millions of Kilograms)
                    0 KM
          35  KM
         TOTAL
                                FRACTION
                                OF BASE
   TAX:$0.10

         PEAK:
         OFF-P:
         TOTAL

   TAX2$0.25

         PEAK-'
         OFF-P:
         TOTAL

   TAX:$0.50
570.603
742.836
1313.44
547.521
706.929
1254.45
1223.77
1592.2
2615.97
           M 74 . 33
           1515.26
           2689.59
                     1794.37
                     2335.04
                     4129.41
           1721 .65
           2222. 19
           3944.04
0.973763
0.968789
0.970944
0.934407
0.92197
0.927359
PEAK:
OFF-P:
TOTAL
509.051
647.085
1 156.14
1091 .92
1387.04
2478.96
1600.97
2034. |3
3635. 1
0.868309
0.843943
0.854717
   RATIONING
        PEAK:
        OFF-P:
        TOTAL
390.565   838.121    1228.69   0.666778
462.764   992.12     1454.88   0.60362
853.329   1S30.24    26d5.^7   0.630985
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:   HIGH
                              132

-------
                                 Table 4-11

        URBAN HYDROCARBON EMISSIONS IN 1931, MEDIUM SENSITIVITY
                          (Millions of Kilograms)
TAX: $0.10

    PEAK:
    OFF-PEAK:
    TOTAL:
                        10 km
258.019
326.229
584.248
               35 km
 551.141
 696.307
1247.448
               Total
 809.16
1022.536
1831.696
             Fraction of
             Base
0.906720
0.888886
0.896677
TAX: $0.25

    PEAK:
    OFF-PEAK:
    TOTAL:
218.102
264.897
464.999
 466.006
 565.389
1031.395
 684.103
 812.286
1496.394
0.766590
0.706117
0.732535
TAX: $0.50

    PEAK:
    OFF-PEAK:
    TOTAL:
151.589
162.558
314.1472
 324.128
 347.I 59
 671.287
 475.7172
 509.717
 985.4342
0.533074
0.443095
0.482403
RATIONING:

    PEAK:
    OFF-PEAK:
    TOTAL:
180.834
207.622
388.456
 386.523
 443.138
 829.661
 567.357
 650.76
 1218.117
0.635763
0.565703
0.596309
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
                                133

-------
                        Table 4-12

     URBAN HYDROCARBON EMISSIONS IN 1987, MEDIUM SENSITIVITY
                       (Millions of Kilograms)
                    0 KM
         35  KM
          TOTAL
            FRACTION
            OF BASE
   TAX:$0.fO
         OFf-P
         TOTAL

   TAX: SO. 25

         PEAK?
154.225
184.401
335.626
327.595
390.534
716.129
127.13^    270.14
481.82
574.935
1056.76
          397.2o
0.8^5277
0.875017
0.864139
          0.73e>lV
        OFF-P:
        TOTAL

   TAX:$0.50

        PEAK:
        OFF-P»
        TOTAL

   RATIONING

        PEAK:
        OFF-P:
        TOTAL
144.862
272.021
81 .9952
79.0156
161.011
105.47
113.266
218.736
306.873
577.014
174.364
167.439
341.823
224.177
239.945
464.122
451.755
849.035
256.319
246.45o
502.834
329.647
353.211
682.858
0.687544
0.710349
0.476382
0.375089
0.420699
0.612522
0.537566
0.571316
SENSITIVITY:  MEDIUM

PRICE ASSUMPTION:   HIGH

-------
Nitrogen Oxides
     Tables 4-13 to 4-15 show the best estimate of nitrogen
oxide emissions for 1975, 1981, and 1987 under the four
policies affecting gasoline demand directly.  These emissions^,
not surprisingly, follow gasoline consumption quine closely.
Cold start emissions account for only a small fraction of
total nitrogen oxide emissions.  Indeed, for recent model
years, the estimated breakout between cold start and running
emissions shows that running emissions per mile increase with
average trip length.1 That is, a decrease in average trip
length leads to a decrease in nitrogen oxide emissions per
mile.  It is this relationship that, paradoxical as it may
seem, leads to nitrogen oxide emissions falling more rapidly
than gasoline consumption for each of the policies and years,
although the difference is most pronounced for the $0.50
per gallon tax in 1987.  This phenomenon is, of course, just
the obverse of the phenomenon with cold start and hydrocarbon
emissions, where, because of the large positive contribution
of cold start emissions, total average emissions per mile
fall with trip length.

Concentrations of Pollutants
     Tables 4-16 to 4-19 show the national index of con-
centration for these pollutants.  Because of the simple
diffusion model used, the percentage change in the index
of each nonreactive pollutant follows very closely the
percentage change in the corresponding emissions.   For

   !J. R.  Martinez,  R. A Nordsieck and A.  Q. Eschenroeder,
"Morning Vehicle Start Effects on Photochemical Smog,"
Environmental Science and Technology, Volume 7, Number 10
(October 1973), pp.  917-923.
                            135

-------
                        Table 4-13

   URBAN NITROGEN OXIDE EMISSIONS IN 1975, MEDIUM SENSITIVITY
                        (Millions of Kilograms)
                    0 KM
         35 KM
         TOTAL
           FRACTION
           OF BASE
    TAX:$0.10

         PEAK:
         OFF-P:
         TOTAL

    TAX:$0.25

         PEAK:
         ()FF-P:
         TOTAL

    TAX:50.50

         PEAK:
         OFF-P:
         TO i AL
353.011
454.993
808.004
338.317
432.762
771.079
313.828
3pf;. 7 1
709.538
746.965
961.471
1708.44
715.886
914.497
163O.38
66-4.086
836.209
1500.3
1099.98
1416.46
2516.44
1054.2
1347.26
2401.46
977.914
1231.92
2209.83
0.973006
0.963457
0.97044
0.932517
0.92114
0.9261
0.865034
0.842261
0.8522
    RATIONING

         PEAK:
         OFF-P:
         TOTAL
238.4
281.591
519.991
504.544
595.08
10P9.62
742.944
876.671
1619.62
0.6571 86
0.599393
0.624589
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:   HIGH
                             136

-------
                        Table 4-14

   URBAN NITROGEN OXIDE EMISSIONS IN 1981, MEDIUM SENSITIVITY
    TAX* $0.10
    TAX:$0.25

         PEAK:
         OFF-P:
         TOTAL

    T4X:$0.50
    RATIONING
                      (Millions of Kilograms)
                     0 KM
          35 KM
          TOTAL
149.041
184.985
334.026
313.032
388.202
701.234
462.073
573.187
1035.26
SENSITIVITY:   MEDIUM
PRICE ASSUMPTION:  HIGH
           FRACTION
           OF BASE
PEAK:
OFF-P :
TOTAL
177.59
228.58
406. 17
372.976
479.683
852.659
550.566
708.263
1258.83
0.90321 7
0.887199
0.894134
0.756042
0.717997
0.735335
PEAK:
OFF-P:
TOTAL
101.46
1 12.326
213.786
213. 125
235.734
448.859
314.585
343.06
662.645
0.51 6085
0.43599-
0.47067
PEAK:
OFF-P:
TOTAL
122.396
144.296
266.692
257.084
302.82
559.904
379.48
447. I 1 6
826. bV6
0.622546
0,560075
0.587123
                             137

-------
                        Table 4-15

   URBAN NITROGEN OXIDE EMISSIONS IN 1987, MEDIUM SENSITIVITY
                        (Millions of Kilograms)
                    10 KM
         35 KM
         TOTAL
                     FRACTION
                     OF BASE
    TAX:$0.10

         PEAK:
         OFF-P:
         TOTAL

    TAX:$0.25

         PEAK:
90.466)
121 .787
212.253
73.7272
187.771
253.222
440.993
          278.237
          375.009
          653.246
          0.890154
          0.873341
          0.880424
153.009   226.736    0.7253c9
         OFF-P:
         TOTAL

    TAX:$0.50

         PEAK:
         OFF-P:
         TOTAL

    RATIONING

         PEAKs
         OFF-P:
         TOTAL
95.2959
169.023
45.8289
51.1433
96.9722
60.336
74.1026
134.439
198
351
    134
    143
95.0706
106.32
201.391
125.198
154.063
279.261
293.43
520.166
          140.9
          157.463
          298.363
           185.534
           228.166
           413.7
0.683355
0.701063
          0.450775
          0.366709
          0.402124
          0.593572
          0.531364
          0.557571
SENSITIVITY:  MEDIUM
PRICE ASSUMPTION:  HIGH
                               138

-------
                             Table 4-16

RELATIVE ONE HOUR CARBON MONOXIDE CONCENTRATIONS FOR 13 CITY AVERAGES
           (As a Percentage of Baseline Concentrations)
  Year          Policy          Light-Duty  Vehicle          Total
1975 $O.IO/gal.
$0.25/gal .
$0.50/gal.
Ration! ng
($1 .27/gal )
1981 SO.IO/gal.
$0.25/gal.
$0.50/gal.
Rat i on i ng
($0.39/ga!)
1987 SO.IO/gal.
$0.25/gal .
$0.50/gal.
Ration ing
($0.37/gal)
97.4
93.5
87.0
67.0
90.9
77.2
54.3
64.4
90.2
75.5
51.0
63.7
97.7
94.2
88.4
70.6
93.8
84.4
68.8
75.6
95.1
87.7
75.4
81.8
  The  cities for which the concentrations are averaged  are:

            Portland,  ME               Miami, FL
            New  York,  NY               Spokane,  WA
            NashviIle, TN              Denver,  CO
            Pittsburgh,  PA             Seattle,  WA
            Little  Rock, AR            San Francisco, CA
            Oklahoma City, OK          Los Angeles,  CA
            Tampa,  FL

  Note:   Based on high gasoline prices  and medium sensitivity  assumptions,
                                   139

-------
                              Table  4-17

RELATIVE EIGHT HOUR CARBON MONOXIDE  CONCENTRATIONS FOR 13 CITY  AVERAGES
           (As a Percentage of Baseline Concentrations)
  Year         Policy         Light-Duty Vehicle         Total

   197?      SO.IO/gal.               97.1                 97.6

            $0.25/gal.               92.8                 94.0

            $0.50/gal.               85.6                 88.1

            Ration!ng
            ($l.27/gal)               63.6                 69.8


   1981      SO.IO/gal.               89.9                 94.4

            $0.25/gal.               74.7                 85.9

            $0.50/gal.               49.2                 71.5

            Rat ion i ng
            ($0.39/gal)               60.5                 77.9


   1987      SO.IO/gal.               89.1                  96.0

            $0.25/gal.               72.6                 89.9

            $0.50/gal.               45.3                 79.8

            Rat ion i ng
            ($0.37/gal)               59.5                 85.0


   The cities  for which  the concentrations  are averaged.are:

            Portland, ME               Miami, FL
            New  York, NY               Spokane, WA
            NashviI le,  TN              Denver,  CO
            Pittsburgh,  PA             Seattle, WA
            Little  Rock, AR            San Francisco,  CA
            Oklahoma City, OK          Los Angeles,  CA
            Tampa,  FL

   Note:   Based on high  gasoline prices and medium sensitivity assumptions.
                                    140

-------
                           Table 4-18
 RELATIVE ANNUAL  NITROGEN OXIDE CONCENTRATIONS FOR 13 CITY AVERAGES
           (As a  Percentage of Baseline Concentrations)
Year         Policy         Light-Duty Vehicle         Total
\f)Tj !f.O. lO/g.'jl .
$0.25/gal .
$0.50/gal.
Ration! ng
($l.27/gal )
1981 SO.IO/gal.
$0.25/gal .
$0.50/gal .
Rationing
($0.39/ga!)
1987 SO.IO/gal.
$0.25/gal .
$0.50/gal.
Ration i ng
($0.37/gal)
97.1
92.8
85.5
63.3
89.7
74.2
48.4
59.7
88.4
71.0
41 .9
57.0
99. 1
97.8
95.7
89.0
98.4
96.0
92.0
93.8
99. 1
97.7
95.4
96.6
The cities for which the concentrations are averaged are:
          Portland, ME               Miami, FL
          New York, NY               Spokane, V.'A
          NashviI le, TN              Denver, CO
          Pittsburgh, PA             Seattle, WA
          Little Rock, AR            San Francisco, CA
          Oklahoma City, OK          Los Angeles, CA
          Tampa, FL

Note:  Based on high gasoline prices and medium sensitivity assumptions.

-------
                           Table 4-19

   RELATIVE  ONE HOUR OXIDANT CONCENTRATIONS FOR 13 CITY AVERAGES
           (As a Percentage of Baseline Concentrations)
Year         Policy         Light-Duty Vehicle         Total
1975 SO.IO/gal.
$0.25/gal .
10.^0/gul.
Rat ioni ng
($l.27/gal )
1981 SO.IO/gal.
$0.25/gal.
$0.50/gal.
Ration i ng
($0.39/gal)
1987 SO.IO/gal.
$0.25/gal.
$0.50/gal.
Rat ion i ng
($0.37/gal)
79.7
79.4
84. 1
41.7
68.7
40. 1
38.9
49.6
1 1 .2
0.0*
0.0*
0.0*
93.7
88.2
90.6
75.3
83.0
69.0
68. 1
73.0
87.5
75.0
74.2
81.7
The cities  for which the concentrations are averaged are:

          Portland, ME               Miami, FL
          New York, NY               Spokane, WA
          Nashville, TN              Denver, CO
          Pittsburgh, PA             Seattle, WA
          Little Rock, AR            San  Francisco, CA
          Oklahoma City, OK          Los  Angeles, CA
          Tampa, FL

Note:   Based on high gasoline  prices and  medium sensitivity  assumptions.
        *Va!ues of zero denote mode! limitations.

-------
example, the percentage change in the index of one-hour
concentrations of carbon monoxide under the different
policies is almost exactly that of the percentage reduc-
tion in peak emissions of carbon monoxide.  Similarly, the
percentage in the eight-hour concentration levels of carbon
monoxide is almost exactly that of the percentage reduction
in total urban carbon monoxide emissions.  Concentrations of
nonreactive nitrogen oxides and hydrocarbons tend to follow the
emissions of these pollutants equally closely.

Changes in Fuel Consumptionand Emissions
by Alternative Modes
     So far in this study we have analyzed the impact of
various policies on the consumption of gasoline by auto-
mobiles and automotive emissions.  Little has been said
about the cross-effects which these policies will have on
fuel consumption by other transit modes, however.  In this
section we present analysis intended to suggest the
magnitude of these effects.  This discussion will proceed
in three parts.  First we will give a general statement of
the problem, and review some attempts to measure these
cross effects.  Next, we will examine the impact of
rationing and gasoline excise tax policies on fuel consumption
by public transportation, indicating the assumptions and
methodology used to infer our results.  Third, we will
consider increases in emissions from alternative urban modes.

     General Discussion of This Problem
     All of the policies analyzed in this study are aimed
a I reducing the automotive consumption of gasoline.  Pur-
suing these methods to reduce automotive fuel consumption

-------
will certainly affect the frequency with which individuals
choose to use the public transportation services available
in all urban areas.  Consider an increase in the federal
excise tax on gasoline, causing people to consume less
gasoline.  One way to economize on gasoline is to take
the bus instead of a car to work.
     Thus, analysis of the fuel conservation impact of
these policies must take account of the effects which
they have on non-automotive fuel consumption.  Because
a policy-induced reduction in auto use will lead to an
increase in the use of public transportation, we have
overstated the reduction in fuel consumption that will
result from the policies in question.  It is more diffi-
cult, however, to determine the size of this overstate-
ment.  This depends on how individuals choose among alter-
native transport modes for various kinds of trips, and
how their choices are affected by changes in the relative
costs and convenience of travel of alternative modes.
     Estimation of the intra-city demand for travel on
alternative modes as a function of the line-haul costs
and travel times associated with these modes has been
attempted in a previous CRA study.1  Unfortunately, prob-
lems encountered in the course of the study prevented
determination of estimates of the cross-effect which an
increase in the cost of travel by auto has on demand for
travel by public transit.  These cross-elasticities were
ultimately constrained to be zero.  To our knowledge,
there exist no direct estimates of these cross-effects
   1 Thomas Domencich, Gerald Kraft, and Jean-Paul Valette,
"Estimation of Urban Travel Behavior:  An Economic Demand
Model," Highway Pesearch Record  (No.  238, 1968).

-------
 for  intra-city  travel.   While  much  work  has  been  done  on
                                      \       ,
 the  modal  shift effect  in  inter-city  travel,   it  is  not
 directly applicable  to  this  problem.   Inter-city  travel
 has  quite  different  attributes from intra-city travel,
 including  the relative  costs in time  and money of alter-
 native  modes, trip and  occupancy patterns and, perhaps
 most important,  availability of different modes.   All  of
 these considerations suggest that cross-elasticities for
 inter-city travel would be quite different from those
 for  intra-city  travel.   The  assumptions  we make below  to
 put  bounds on the cross-elasticities  for urban travel
 are  also appropriate to inter-city  travel.   To the extent
 that alternative inter-city  modes use more fuel or emit
 more pollutants per  passenger  mile  than  intra-city alter-
 natives, the  calculations  below will  understate somewhat
 the  increase  in fuel use and emissions from  drivers
 changing modes.  The error is  likely  to  be small, however.
      Another  approach to this  problem, which has  been
 used by CRA in  other work, is  to infer cross-elasticity
 values  from direct estimates of the own-elasticity and
 assumed invar iance of overall  travel  demand.   The basic
 idea is best  illustrated by  the case  of  work trips.  It
 is reasonable to assume that no one quits a  job or changes
 residence  in  response to an  increase  in  the  price of gaso-
 line.   Thus,  the total  passenger miles of travel  on  trips
 to work do not  change with the price  of  gasoline. We  may
 assume  then that the reduction in passenger  miles traveled  by
 automobile is offset by a  one-for-one increase in passenger
 miles traveled  by public transit.   Since we  know  the price
         for example, Richard Quandt, ed., The Demand for
Travel:  Theory and Measurement   (Lexington, Mass.:  D. C. Heath
and Company, 1970) .

-------
elasticity of demand for gasoline and the average fuel
economy of the auto stock, we can translate a gasoline
price increase into a decrease in vehicle miles traveled
by auto.  If the auto occupancy rate were constant, we
could then determine the decrease in auto passenger miles
traveled and the consequent increase in transit passenger
miles traveled.1  Assuming that the fuel efficiency of the
transit system  (measured in gallons per passenger mile)
remains constant, we have a complete link from the
initial gasoline price change to the resultant increase in
transit fuel consumption.  While some of these assumptions
are relaxed below, this is the basic procedure we follow
in analyzing the effect which the modal shift will have
on gasoline consumption.

     The Effect of Rationing and Gasoline Excise Tax
     Increases on Fuel Consumption by Public Transit
     To a large extent the quantities of fuel consumed by
automobiles and public transit systems are not directly
comparable.  Autos run almost exclusively on gasoline,
whereas over 80 percent of urban buses use diesel fuel.
Thus the modal  shift in intra-urban travel resulting from
these policies will cause a decline in gasoline consump-
tion and an increase in diesel consumption.
     In this section we give estimates which bound the
size of the actual modal  shift effect on transit fuel
consumption.  The bounds  will be generated by making
alternative assumptions about the response of the overall
    JAuto occupancy  rate will,  in general,  increase  in
 response to  these policies,  and we  use  an  estimate  of
 this  increase  in the  calculations below.

-------
demand  for  travel  to  an  increase  in  gasoline  price.   The
actual  response will  be  seen  to lie  between these  two
extremes.   In  this way we  can determine  the maximum  amount
by which our previously  reported  estimates of the  policy-
induced reduction  in  fuel  consumption overstate  the  actual
reduction that will occur.
     We restrict our  modal  shift  analysis to  intra-urban
travel, as  discussed  in  the preceding section.   Further,
we assume that the shift of travel from  auto  to  mass
transit will cause an equivalent  increase in  the number
of passenger miles of travel  by buses alone.   This assump-
tion may be justified on the  basis of two observations.
First,  fuel consumption  per passenger mile by electric
mass transit is only  10  percent higher than that by  bus,
while buses account for  1.8 times as many passenger
miles of urban travel as does electric transit.1   Conse-
quently, this  assumption understates energy use  per  pas-
senger  mile by, at most, less than 4 percent.  Second,
the increased  ridership  of  public transit will require
expansion of capacity.   Although  subway  systems  may  add
trains, no  new subway systems will be constructed  by 1987
in response to these  policies.  Thus,  the bulk of
increased passenger miles of  travel  will be absorbed by
increasing  the size of the  intra-urban commercial  bus
fleets.
     If the price  of  gasoline increases, we can  put  two
bounds  on the  response of total travel.  First,  the  total
demand  for  travel  cannot increase.   That is,  an  increase
   *Eric Hirst, Energy Intensiveness of Passenger1 and Freight
Transport Modes:  2950-1970  (Oak Ridge,  Tennessee:   Oak  Ridge
National Laboratory, April  1973).

-------
in the price of gasoline could not lead to people desiring
to make more frequent or longer trips than they made before.
This follows both from common sense and basic economic
theory.  Secondly, the reduction in total travel will not
exceed the reduction in auto travel resulting from the tax
increase.  This is plausible because the cost of other travel
modes has not changed, so there is no reason that they should
lose ridership by virtue of auto travel becoming more expensive,
     Therefore, the reduction in total passenger miles of
travel must be between zero and that quantity implied by
the fall in auto fuel consumption.  From this conclusion
we may infer bounds on the size of the increase in
transit fuel consumption resulting from the modal shift
in the following way:  If the reduction in total travel
is equal to the reduction in auto travel, then there is
no modal shift and consequently no increase in transit
fuel consumption.  Conversely, if the reduction in total
travel is zero, then the increase in transit travel is
equal to the reduction in auto travel.
     From the previous analysis, we estimated the reduc-
tion in VMT's due to these policies.  The increased
price of gasoline will encourage car pooling, increasing
the average auto occupancy rate.  However, the elasti-
city of auto occupancy  (by trip purpose) with respect
to increased auto travel costs has been estimated else-
where by CRA.l  By adopting this estimate, we can determine
the reduction in passenger miles of auto travel resulting

   *The overall elasticity of auto occupancy is -0.11.  See
Charles River Associates Inc., "Study of Alternatives to
Gas Rationing in the Los Angeles Area," in preparation for
the Environmental Protection Agency under Contract #68-01-2235.
                               148

-------
from the policies, and hence the equivalent increase in
passenger miles of transit travel.
      At this point, we can convert the increase in transit
ridership into an increase in fuel consumption by using
the fuel efficiency of the public transit mode.  In 1970
buses consumed about 0.027 gallons of fuel per passenger mile,
using an average load factor of 18 percent.1  It is necessary
to assume that bus fuel efficiency remains constant after
the increase in ridership.  It is likely that the fuel efficiency
will increase however, since increased ridership will probably
mean a greater load factor for intra-urban bus travel.  Nonetheless,
this possible understatement of transit fuel efficiency is
consistent with our desire to give an upper bound on the size
of increase in transit fuel consumption.
     On the basis of the above observations, we can now
calculate an upper bound on the fraction of fuel savings
stemming from reduced auto use which is offset by the modal
shift effect.  Our assumption of no reduction in total
travel demand requires that total passenger miles of travel
be conserved.  Thus the reduction in auto passenger miles leads to
an equivalent increase in transit passenger miles.  Assuming
both fuel efficiency parameters constant (in gallons of fuel
per auto or transit passenger mile of travel), the ratio of
the quantity of fuel conserved in auto travel to the quan-
tity of increased fuel usage in transit travel is simply
the ratio of gallons per passenger mile in transit to gal-
lons per passenger mile in auto travel.  Since transit is
more fuel-efficient than auto travel, fuel is  still conserved,
even though total travel remains constant.   As indicated

         Hirst, op. cit. ,   Table 6,  p. 14.

-------
earlier, 0.027 gallons per passenger mile are consumed in
urban bus transit, while the averate auto fuel efficiency
for urban travel is approximately 0.048 gallons per passen-
ger mile.1  Thus the shift effect could offset fuel conser-
vation from reduced auto use by as much as  56  percent.
Note that this percentage increase is independent of the
size of the reduction in gasoline consumption by auto,
because of the assumption of conservation of passenger
miles.  It should be emphasized that this figure almost
certainly overstates the offsetting effect of the modal
shift, but a tighter estimate is not possible without fur-
ther empirical investigation.
     Using the methods described above, we can also give
upper bounds on the increase in fuel consumption by transit
associated with each of the excise tax policies and the
rationing policy.  The rationing policy is, of course,
equivalent for purposes of this calculation to an excise
tax policy achieving the same reduction in fuel consumption.
Since these calculations depend on the base price of gaso-
line assumed and the value of the elasticity of gasoline
demand utilized, Table 4-20 presents results for all four
policies under the pre-embargo and post-embargo base gaso-

     1 Using an average fuel efficiency for all auto travel of
13.5 vehicle miles per gallon with a relative gallons per
mile in urban vs. rural travel of 1.42, a rural share of VMT
of 0.57, and an average urban occupancy rate of 1.9 passengers
per vehicle, one arrives at the cited figure.  For overall fuel
efficiency, see FHWA, Highway Statistics, 1972, Table VM-1,
p. 52; the relative urban auto fuel efficiency is taken from
1975 Gas Mileage Guide for New Car Buyers, EPA; the rural share
is given in Nationwide Personal Transportation Study, Vol. 8,
1969;  finally, one finds urban and rural vehicle occupancy
rates in Eric Hirst, op.  cit. , p. 32.
                             150

-------
line price assumptions, using the low, medium, and high
sensitivity estimates of the gasoline demand elasticity.
We assume that base transit fuel consumption remains
constant at 548.3 million gallons per year.1  While the
rapid growth of the auto as a means of intra-urban travel
has caused a steady decline in the total passenger miles
of travel by transit, it seems rather unsafe to project
such a continued relative growth of auto transit  for
future years.  For this reason, we assume  constancy of the
absolute level of transit ridership.  Though transit's
relative share of total urban travel is thus assumed -to
continue to decline, it declines at a slower rate than
has historically been the case.
     Because only 7.5 percent of all urban passenger miles
of travel occur on public transit, even a  small relative
decline in auto travel will cause a large  percentage
increase in transit fuel consumption.  Thus, Table 4-20 indi-
cates that a $0.10 excise tax could have the effect of more
than doubling transit fuel consumption by  1981.   with a
$0.50 increase in the price of gasoline, fuel consumption
by public transit could be as much as 10 times greater by
1987 than it is now.  It would seem that more precise esti-
mates of these effects would have a substantial payoff in
terms of more intelligent policy making in the near future.

   JThis number was derived by summing the gallons of fos-
sil fuel and the gallon-equivalents of electricity con-
sumed by transit in 1974.  Electricity used was converted
by multiplying by 10,500 BTU/kwh  (average  heat rate in
1971, from Steam-Electric Plant Construction Cost and Annual Pro-
duction Expenses, 1971, p. XXVIII) and dividing the  result
by 136,000 BTU/gal.  Fuel use was taken from American
Public Transit Association,  '74-'75 Transit Fact BOOK (March
1975), p. 29.
                             151

-------
                            Table 4-20
    MAXIMUM INCREASES IN FUEL CONSUMPTION BY PUBLIC TRANSIT IN
 RESPONSE TO POLICIES REDUCING AUTOMOTIVE CONSUMPTION OF GASOLINE
                      (Millions of Gallons)
Policy on Gasoline
$0.10/gal.  surtax
      1975
      1981
      1987
$0,25/gal.  surtax
      1975
      1981
      1987
$0.50/gal. surtax
      1975
      198!
      1987
Ration!ng
      1975
      1981
      1987
Low Gasoline  Medium Gasoline  High Gasoline
Consumption1   Consumption2    Consumption3
    163.7
   1227.5
   1699.6
  66.1
 757.0
1074.2
410.7
3079.8
4392.3
173.1
1897.9
2697.4
    818.4
   6154.9
   8775.2

   1410.I
   3202.6
   4392.3
 335.2
3789.6
5391.7

 849.8
2958.7
3994.2
  0
  II .0
 199.9

  0
 339.9
 497.3

  0
 678.3
 996.2

  0
1438.4
2108.8
NOTES:
   aAssumes post-embargo gasoline prices and high elasticity estimates;
   2Assumes post-embargo gasoline prices and medium elasticity estimates;
   3Assumes pre-embargo gasoline prices and  low elasticity estimates.
Fuel  consumption by transit  is  in units of diesel fuel; electricity
consumed has been converted  to diesel equivalents as described in
footnote I on page 4-60.
                                  152

-------
     Increase in Emissions from Alternative Sources
     Because so little is known about cross elasticities
between travel by automobile and by other kinds of trans-
portation, there is a wide range in the possible increase
of emissions due to travelers shifting from auto to alter-
native types of transportation.  On the one hand, if these
cross elasticities are zero, there will be no increase in
fuel consumption by alternative transportation modes and,
hence, no increase in emissions from these modes.  At the
other extreme, if the entire reduction in automobile pas-
senger miles is diverted onto public transportation, the
increase in fuel consumption will be as shown in Table 4-3,
while the increase in emissions will depend on the relative
emissions per passenger mile of buses and automobiles.
These emission factors, in grams per passenger mile, are
shown in Table 4-21.  The values for light-duty vehicles
are shown for 1975, 1981 and 1987, based on the base fore-
casts.  The bus emissions per passenger mile assume that
average occupancy  (defined as passenger miles divided by
vehicle miles) was 9.29, its value in 1970.l  The emissions
   JThis figure was derived by dividing total bus passen-
ger miles  (Eric Hirst, op.  oit., p. 14) by total bus vehicle
miles  (American Public Transit Association, '74- '75 Transit
Fact Book, p.  26).   This estimate  agrees  quite  closely with
a  figure derived  for  1969,  8.99,  according to the  follow-
ing procedure.  We subtracted  intercity bus passenger
miles  from total  bus  passenger miles (Automobile Facts and
Figures, 1972-74,  pp.  34-36)  to obtain an  estimate  of  urban
bus passenger miles.  This estimate was then  divided by
total  urban  commercial bus miles,  taken from  the Federal
Highway Administration,  Highway Statistics,  1969 Edition,
p. 73.
    Average occupancy was slowly declining over  the  1960-
1970 period,  so that use of a 1970  figure probably over-
states the base period occupancy rate.  The occupancy
rate would, of course, tend to rise with an increase  in
bus ridership.
                             153

-------
                            Table 4-21

   EMISSIONS PER PASSENGER MILE OF BUS  AND LIGHT-DUTY VEHICLES,
                        1975,  1981 and  1987

                         (Grams Per Mile)


                  Bus (All Years)            Light-Duty Vehicles


                                             1975    1981

CO                     3.63                 23.18    5.29

NO                     3.34                  1.98    0.84
  x
HC1                    0.49                  3.28    1.22


    1 Excludes diurnal evaporative emissions.




SOURCES:

      Heavy-duty gasoline emission factors:  Compilation of Air
Pollutant Emission Factors^ p. 3.1.5-2,
      Heavy-duty diesel emission  factors:  Ibid., p. 3.1.4-3.

      Share of  bus emissions from diesel:  calculated as diesel
fuel as  fraction of  total fuel *  diesel emission factors plus
(I - diesel fuel fraction)  * gasoline emission factors; 1972 fuel
breakdown from  American Transit Association,  '73-'74 Transit Fact
Book, Table No.  16,  p.  19.
      Passenger miles  per vehicle mile:
      Buses:   Average  occupancy  of 9.29 derived  by  dividing  1970
bus  passenger-m'r les, from Hirst,  op. cit.3  p.  14,  by 1970  bus
vehicle miles,  from '74-'75 Transit  Fact  Book,  p.   26.

-------
per  passenger  mile for light-duty vehicles assume an aver-
age  occupancy  of 1.9,  the average for all  urban automobile
trips  in  1969.1
     Average emission  and occupancy rates  are used here
because they reflect what can be expected  from policy
changes that,  in large measure,  affect urban gasoline
consumption proportionately at all times of day.   These
figures would  not necessarily, therefore,  be appropriate
for  evaluating policies that affect,  say,  rush-hour traffic
only,  when  bus occupancy rates are very high.   Diversion
of a single-passenger  trip to transit at peak hours might
have quite  different implications from one during offpeak
hours.
     These  numbers suggest, surprisingly,  that the reduc-
tion in emissions from diverting a passenger from auto to
bus  is not  very  great,  assuming that the passenger travels
the  same  number  of miles by both modes.  For example,  the
emissions per  passenger mile of nitrogen oxides are
greater for buses than for automobiles in  every forecast
year.
     The  short-run effects on carbon monoxide and hydro-
carbon emissions,  however, favor bus over  auto.   By 1981,
carbon monoxide  emissions per passenger mile by bus are
about  70  percent of those by light-duty vehicles,  but  by
1987 buses  will  have higher emission rates of  carbon mon-
oxide  per passenger mile than light-duty vehicles.   Hydro-
carbon emissions per passenger mile are roughly the same
by 1987.  This shift over time occurs because  the emission
standards for  light-duty vehicles are reduced  sharply  over
 	 	T—  *   . ._ 	  _		L-l	  	    -- ._-	L_.		  	-     	-
   JThis  number  was taken from.the Nationwide Personal Trans-
portation Study, Report No.  1,  Table 2,  p. 10.
                             155

-------
the forecast period, while the emission standards for
heavy-duty gasoline and diesel powered vehicles do not
change.
     Even if the 1975 California standards for heavy-duty
vehicles were adopted nationally, the conclusions presented
here would not require substantial revision.   Only the
nitrogen oxide  standard is substantially stricter than
the U.S. standard; even that standard implies that bus
emissions of nitrogen oxides per passenger mile will be
above those of autos throughout the forecast period.
Table 4-22 presents a comparison of the California and
U.S. standards from 1975 on.
     Use of the average auto occupancy rate is based on
the assumption that the reduction in auto trips comes
about from a proportionate reduction in all trips at all
occupancy levels.  On intuitive grounds, it is not clear
whether single-passenger or multi-passenger auto trips
are most likely to be diverted to transit.  On the one
hand, it might be that those who value the convenience
and privacy of driving alone will continue to do so at
higher gasoline prices, while carpoolers will shift to
transit.  On the other hand, a given increase in the
cost of an auto trip represents a greater cost per person
for the single-occupant trip than for the carpool member,
leading to a disproportionate reduction in single-occupant
trips.  We are not aware of any studies that provide evi-
dence on the substitution of transit trips for auto trips
with different numbers of passengers.  It seems reason-
able to assume, however, that the appropriate occupancy
rate to use in calculating the change in emissions lies
somewhere between the average occupancy rate and single
                            156

-------
          J-  S-
           •
 re  co'
 i.  re
O D-
             O
CTi
  •

CN
                             VO
ro   in


CD   —
cr>
t/o
UJ

                      to
                             CM
                             hO
                                  —   in
                                  vo
CM
CVI
^-

 CM




                                       in
CO
co
                  re
                 CD
                             oo
                                       ro
                        —   co

                        •<3-   CM
                      co
                                      CM
                             o   ro
                             8
                                  o  o
ards: Environmental Protection Agency
•o
c
ro
4-
in

CD
c
.„
»
O
in
ro
CO

ro

c
O

•_
• •
CO
z
o

E
•s.

CO
co
^^

Q
Z


CO
UJ
o
o:
ID
o
CO
—
ro
o

TO
n
ro

.
co

ID



jf- s.
—
N^





•
ro
1
•^J-
*
•
ro
QL
•^
8
o
•P
O
G
ft,

SS
O
•^
CO
.eo

g
[j]

"M
§
-P
3

r-i
O


.^
"^C

^4— J
o


5^
_O

-£i
^J
t"-i
*r^
1*
o
Co
CM
in
to
B
Q.

•t
,
*^J
*r^
tCl



• •
in
TD
<
ro
TQ
C
ro

in

_
CD
01
CD
-o

.
CO

i



*—*
CM
^-s





phone conversation with Joe! Horowitz
20, 1975. Horowitz disaggregated the
0
"CD
4-


••
i/i
-o
L.
(D
TO
C
ro
4-
01

—
CD
in
CD

TO

ro

c
t_
o

•—
— .
ro
o



s~^
ro
*^x





JC
O
L.
ro
s

%
>^
o
c
CD
CO
<

C
0

O
CD
4-
O

Q-

—
ro
c
CD
E
C
O

• —
>
c:
LLJ

CD
JZ
+•
*4—
0












•
X

~^

-o
c.
ro

0
X

U
o
H-
T3
!„
ro
-o
c
ro
-H
01

-o
CD
C
» —
JD
0
U
cr.
r^
CO
o
XI
TO
CD
+-
JC
CO
"co
2
in
ro
3:
0
01
CD
TO
CD
CO
ro
L
CD
>
ro
TJ
CD
4-
JZ
CO

CD
3

0
JC
4-

>+-
O

C
Q
4-
(D
-|_
^
CL
E
O
o
0
JZ
4-

C
—



^— *
^rj-
*%^>





t the share of fuel consumption In
c
CD
Ul
CD
1_
10
CV
•Sh
tv
•
"CD
c
O
1/1
ro
CO
JC
4-
5
T3
CD
•o
ri
o
c
CD
JD
O
4-

CLTD
CD
L.

01
4-
-C
CO

CD
5

CD
in
CD
JC
j_


.
—
CM
—
O

>^
XI

CD
C
•—
—
O
01
ro
CO

c
ro
CD
E
•3
01
01
ro

CD
c
ro
Q-
O
Q-
•^^
•
01 CT>
^ CM
JO
*
S- Q.
o
CD A;
CL O
>- o
4- 03

-d -p
U O
ro CJ
CD Ct,

>* -^i
JO -i^
CO
ro S
O\ ?,
— &H
assumed.
in
ro
•5.

a\
CM
.
Q>

«4_
O

CD
-j—
ro
i_

>^
o
c
ro
Q.

o
o
o

CD
O

JZ
CD
>.



*~*
in
%-*•





                                            157

-------
occupancy.  Using both rates will, at least, provide
bounds on the likely range.1 If, therefore, we assume
that all of the reduction in automobile passenger miles
occurs from single-occupant vehicles, the relevant emis-
sion rates for automobiles, for purposes of comparison,
are those per vehicle mile.  These rates for the differ-
ent pollutants for the forecast periods are shown in
Table 4-23.  Under the assumption that all of the diver-
sion in automobile passenger miles is from single-occupant
automobiles, the comparison is much less ambiguous.  The
emissions per passenger mile of buses are about the same
as those of single-occupant autos by 1987 for carbon mon-
oxide and hydrocarbons.  Emissions of nitrogen oxides are
higher for both 1981 and 1987.  There is, however, a dra-
matic contrast between the short-run carbon monoxide emis-
sions of automobiles and those of buses.
     Of course, if the demand for bus transportation increases,
average occupancy will tend to increase, particularly in the
short run.  After the bus system has had time to add addi-
tional vehicles, the average occupancy rate will fall,
although it may still be above the level before the change.
From 1950 to 1970, for example, when bus passenger miles
fell from 29 to 13 billion, the average occupancy rate
fell only from 12.63 to 9.29. 2  If w'e assume that, once
     *If those who switch from autos to transit are carpoolers
with an average occupancy rate higher than 1.9, then bus
emissions per passenger-mile will be even greater relative to
those from the reduced auto trips.  For purposes of illustra-
tion, if the average auto occupancy rate of those switching
to transit is 3, the relevant auto emission rates per passenger-
mile become:
                            1975          1981          1987
Carbon monoxide
Nitrogen oxides
Hydrocarbons
Comparison of these figures with those in Table 4-22 shows
that bus emissions per passenger-mile of carbon monoxide
and nitrogen oxides exceed those of autos by 1981.  However,
this assumption about auto occupancy seems implausible.
     Calculated from Hirst,  loo. cit., and American Public
Transit Association, loo. cit.
                               158

-------
                            Table 4-23
EMISSION RATES PER PASSENGER MILE, SINGLE-OCCUPANT LIGHT-DUTY VEHICLES
                         (Grams Per Mile)

                                 1975            1981              1987

Carbon Monoxide
Nitrogen Oxides
Hydrocarbons1

    Excludes diurnal evaporative emissions.

SOURCE:  Base line forecasts.
44
3.77
6.23
10.05
1 .60
2.31
4.12
0.68
1.10
                                 159

-------
adjustments to bus capacity are made, the maximum occu-
pancy rate is 12.63, the bus emission rates per passenger
mile become:
                              g/passenger mile
          Carbon Monoxide            2.67
          Hydrocarbons               2.45
          Nitrogen Oxides            0.36.
Hence none of the qualitative assertions made above are
affected, although quantitatively the lower rates per
passenger mile reduce bus emissions  relative to autos.
     The increases in emissions of carbon monoxide, hydro-
carbons and nitrogen oxides by buses are shown in Tables
4-24, 4-25, and 4-26, respectively.  These increases are
based on the assumption that total passenger miles are
conserved, and hence represent the maximum increase in
bus emissions.  They do not include  emissions from elec-
tricity generation needed to power subway lines, since
these emissions occur at stationary  sources and include
different pollutant types.  The overall impact of these
policies on air quality depends critically on the cross
elasticity of demand between automobiles and other modes
of transportation.  We were unable,  within the scope of
this study, to do original research  into this question,
but it is important to determine both emissions and fuel
                   >
use from alternative modes of transportation.  Moreover,
the numbers presented suggest that it might be useful to
control exhaust emissions from heavy-duty vehicles as
well, particularly  if any policy is  contemplated that
will divert substantial amounts of travel from automobiles
to public transportation.  Since World War II, public
transportation as a fraction of total passenger miles has
been falling sharply and steadily, so that, in the absence
of policies to reverse this trend, controlling emissions
from these  vehicles more strictly would probably not make
                             160

-------
                           Table 4-24
MAXIMUM INCREASES IN EMISSIONS OF CARBON MONOXIDE BY TRANSIT BUSES IN
 RESPONSE TO POLICIES REDUCING AUTOMOTIVE CONSUMPTION OF GASOLINE
                     (Millions of Kilograms)
Policy on Gasoline

SO.IO/gal. surtax

      1975
      1981
      1987

$0.25/gal. surtax

      1975
      1981
      1987

$0.50/gal. surtax

      1975
      1981
      1987

Rat ion ing

      1975
      198!
      1987
Low Gasoline
Consumption1
    13.09
    98.18
   135.93
    32.85
   246.32
   351.30
    65.46
   492.27
   701.84
   I 12.78
   256.14
   351.30
Medium Gasoline
  Consumption2
       5.29
      60.54
      85.91
      13.84
     151.79
     215.74
      26.81
     303.09
     431.23
      67.97
     236.64
     319.46
High Gasoline
 Consumption3
      0
      0.88
     15.99
      0
     27. 19
     39.77
      0
     54.25
     79.68
      0
    I 15.04
    168.66
NOTES:

   1Assumes post-embargo gasoline prices and high elasticities;

   2Assumes post-embargo gasoline prices and medium elasticities;
   3Assumes pre-embargo gasoline prices and low elasticities.

SOURCES AND ASSUMPTIONS:

      (I)   Fuel consumption is by buses only; bus fuel  consumption
is assumed to be the same proportion of total  transit fuel  consump-
tion as in 1974;

      (2)   Total increase in transit fuel consumption taken from
Table 4-20;

      (3)   Miles per gallon was assumed to be 4.10 (American  Pub-
lic Transit Association, '74-'75 Transit Fact Book, p. 30);  and

      (4)   Emissions per vehicle mile were derived from Table 4-21.
                                161

-------
                            Table 4-25
MAXIMUM INCREASES IN EMISSIONS OF HYDROCARBONS BY TRANSIT BUSES IN
 RESPONSE TO POLICIES REDUCING AUTOMOTIVE CONSUMPTION OF GASOLINE
                     (Millions of Kilograms)
                      Low Gasoline  Medium Gasoline  High Gasoline
Policy on Gasoline    Consumption1    Consumption2    Consumption3
$0.10/gal. surtax
      1975                 1.60            0.65            0
      1981                11.98            7.39            0.11
      1987                16.59           10.48            1.95
$0.25/gal. surtax
      1975                 4.01            1.69            0
      1981                30.06           18.52            3.32
      1987                42.87           26.33            4.85
$0.50/gal. surtax
      1975                 7.99            3.27            0
      1981                60.07           36.99            6.62
      1987                85.65           52.62            9.72
Rationing
      1975                13.76            8.29            0
      1981                31.26           28.88           14.04
      1987                42.87           38.98           20.58
NOTES:
    1Assumes post-embargo gasoline prices and high elasticities;
    2Assumes post-embargo gasoline prices and medium elasticities;
    3Assumes pre-embargo gasoline prices and low elasticities.
SOURCES AND ASSUMPTIONS:
       (I)   Fuel consumption is by buses only; bus fuel consumption
is  assumed to be the same proportion of total  transit fuel consump-
tion  as in 1974;
       (2)   Total  increase  in transit fuel consumption taken from
Table  4-20;
       (3)   Miles  per gallon was assumed to be 4.10 (American Pub-
lic Transit Association,  '74-'75 Transit Fact Book, p. 30); and
       (4)   Emissions per vehicle mile were derived from Table 4-21,
                                162

-------
                            Table 4-26
MAXIMUM INCREASES IN EMISSIONS OF NITROGEN OXIDES BY TRANSIT BUSES IN
  RESPONSE TO POLICIES REDUCING AUTOMOTIVE CONSUMPTION OF GASOLINE
                     (Millions of Kilograms)
Policy on Gasoline

$CK 10/gaI. surtax

      1975
      198!
      1987

$0.25/gal. surtax
      1975
      1981
      1987

S0.50/ga|. surtax

      1975
      1981
      1987

Rationing

      1975
      1981
      1987
Low Gasoline
Consumption1
    12.08
    90.58
   125.41
     30.31
    227.26
    324.I I
     60.39
    454.17
    647.52
    104.05
    236.32
    324.I I
Medium Gasoline
  Consumption2
       4.88
      55.86
      79.27
      12.77
     140.05
     199.04
      24.73
     279.63
     397.85
      62.71
     218.32
     294.73
High Gasoline
Consumption3
      0
      0.81
     14.75
      0
     25.08
     36.70
      0
     50.05
     73.51
      0
    106.14
    155.61
NOTES:

   1Assumes post-embargo gasoline prices and high elasticities;
   2Assunnes post-embargo gasoline prices and medium elasticities;
   3Assumes pre-embargo gasoline prices and low elasticities.

SOURCES AND ASSUMPTIONS:
      (I)   Fuel consumption is by buses only;  bus fuel  consumption
is assumed to be the same proportion of total  transit fuel  consump-
tion as in 1974;
      (2)   Total increase in transit fuel  consumption taken from
Table 4-20;

      (3)   Miles per gallon was assumed to be 4.10 (American Pub-
lic Transit Association, '74-'75 Transit Fact Book, p. 30);  and
      (4)   Emissions per vehicle mile were derived from Table 4-21.
                                163

-------
much difference.  With rising fuel prices, however, the
absolute decline in transit passengers stopped in 1973.
In 1974, transit passengers actually increased; total
passengers rose above the 1971 level.1  If policies are
undertaken that will divert substantial traffic to these
modes, it is important to control their emissions better,
as the emissions per passenger mile are by no means neg-
ligible, even compared with those from automobiles.

                   Secondary Impacts

     For the purposes of this report, the primary impacts
of the policies considered have been the changes in gaso-
line consumption, emissions, and ambient air quality.
There are, however, secondary impacts associated with
these different policies.  In particular, we discuss in
this section differences in the secondary impacts asso-
ciated with increases in the excise tax and gasoline
rationing.  We consider three secondary impacts:  admin-
istrative costs associated with each policy; changes in
the distribution of income due to these policies; and
incentives for efficiency and technical change associated
with each of these policies.  Accurate measurement of the
quantitative importance of each of these impacts is a dif-
ficult and complex job, and it is beyond the scope of this
study.  However, where magnitudes can be roughly quanti-
fied, we present these estimates.

   1American Public Transit Association,  '74-'75 Transit Fact Book,
PD.  16-17.

-------
Administrative Costs
     We are interested here in the marginal administrative cost
of the different policies, that is, what additional adminis-
trative cost will be imposed if these policies are adopted.
The question is posed this way because, to the extent that the
administrative machinery has already been set up for other pur-
poses, it does not seem appropriate to attribute the cost of
the existing structure to the changes in policy.
     There is already a federal excise tax on gasoline, with
an administrative structure to collect the tax and ensure
compliance.  For this reason, it seems probable that an
increase in the level of this tax would lead to, at most,
trivial changes in the cost of administering the increase.
There might be some increase because, at higher tax levels,
there would be greater incentives to evade or to cheat on
tax, but it seems probable that the existing apparatus could
collect the additional federal excise tax revenues.
     There does not exist, at present, an equivalent apparatus
for coupon rationing.  Administrative costs would be incurred
in performing several functions associated with coupon rationing.
First, the coupons would have to be printed and policed.  That
is, the same kinds of safeguards against counterfeiting U.S.
currency would, to a lesser extent, be necessary to ensure
that the coupons (with an estimated market value in 1975 between
$0.86 and $2.64 per gallon) would not be counterfeited.  Second,
there would have to be a distribution network, with provisions
to ensure that people entitled to receive coupons receive
their exact allotment.  Third, with coupons legally transfer-
able, there would need to be a mechanism to ensure that, once
used, they could not be reused.  These and other difficulties
associated with coupon rationing can, in principle, be resolved.
However, they do impose additional administrative costs.
                              165

-------
     When coupon rationing was being considered during the gaso-
line shortage of the winter of 1973-74, it was estimated that
the scheme then considered would have an annual cost of at
least $1.25 billion.  This amount is roughly equivalent to a
surcharge of about $0.015 to $0.02 per gallon,
or, on another basis, an annual cost of between $12 and $13
per licensed driver.  These costs are real costs,  in the
sense that they represent resources used,  rather than a
transfer of income from one group in society to another.
      In addition to the administrative costs, there would
also be transaction costs associated with the buying and selling
of coupons.  Markets for buying and selling coupons would not
arise and function smoothly without some individuals or agencies
undertaking what amounts to a brokerage function, that is,
standing ready to buy or sell coupons.
      It is obviously impossible to predict the costs of organ-
izing and operating a system of coupon exchanges and hence what
the costs of coupon transactions will be.   However, it is per-
haps worth noting in this context the costs of transactions
for exchanges with some similarities to coupon exchanges.
      At one extreme, transaction  costs may be independent
of the size of the transaction.  An example of such a trans-
action is the purchase of postal money orders for which the
current charge is $0.25 per order.  At the other extreme, charges
for transactions are often a simple percentage of the value
of the transaction.  Examples of such charges are the 1 percent
service charge for American Express Travelers Checks and the
2 to 3 percent service charge to buy or sell small quantities
of foreign currencies at commercial banks.  Intermediate between
these extremes are instances where the transaction charges are
a flat fee plus a percentage of the value of the transaction.
                               166

-------
      Although it is difficult to estimate the administrative
and transaction  costs of a coupon rationing scheme, the fore-
going considerations suggest the costs could be substantial
and much greater than the costs of noncoupon schemes.  This in '
turn suggests that if rationing is expected to be required only
for a short period of time, the costs of mounting the necessary
administrative machinery may detract seriously from the benefits
of coupon rationing relative to alternative schemes.  On the
other hand, if it is expected that rationing will be necessary
for an extended period of time, say substantially more than a
year, then the benefits of coupon rationing may well be worth
its administrative burden.

Income Distribution
      Both coupon rationing and an increase in the excise tax on
gasoline result in higher effective prices paid by the consumers
of gasoline.  The increase in price is direct and immediate in
the case of an increase in the excise tax, while, in the case of
coupon rationing, it is implicit in the market value of a coupon.
As with any transaction, someone pays the higher price and some-
one else receives it.  The distribution of these revenues is
different for coupon rationing than for an increase in the excise
tax.
      With an increase in the excise tax on gasoline, it is
clear that the consumers of gasoline are subject to a loss in
income that is equal to the increase in the excise tax times
the quantity of gasoline consumed after equilibrium is reached.
Assuming that this increase in tax revenues is balanced by a

   JThis result is strictly true only if the supply curve is hori-
zontal.  If, however, the supply curve is very close to hori-
zontal, then this result holds approximately.
                               167

-------
decrease in other taxes, such as the personal income tax, the
excise tax redistributes income away from gasoline consumers
to the general taxpayer.  Of course, in many, perhaps most, cases,
the general taxpayer and the consumer of gasoline are the same
             I
person, but the principle of redistribution remains the same-.
      On the other hand, gasoline rationing also involves a
redistribution of income, albeit implicit.  That is, if it
turned out to be the case that each licensed driver consumed
exactly his allotted ration of 10 gallons per week, then no income
redistribution at all would occur.  Because the coupons have
a market value, however, gasoline rationing implicitly gives to
each licensed driver the value of the coupons (in the policy
analyzed here, 520 gallons per year times the coupon price).
In this case, therefore, although there is a redistribution
of income among the consumers of gasoline (from those who use
more than the rationed amount to those who use less),  there is
no redistribution of income between the gasoline consumers in
general and any other category.  That is, through the issuing
of coupons, the entire value of the increased price of gasoline
(due to the mandated reduction in supply) is received by
licensed drivers.
      As discussed above, the amount of income transferred under
increases in the excise tax can be determined by multiplying
the excise tax times the gallons of gasoline consumed after the
increase is put into effect.  This amount will, of course, be
different for different years of the forecast period, both
because of the exogenous growth in gasoline consumption, on the
one hand, and because of the lagged response to price over a
period of years.  Table 4-27 shows the amount of tax revenues
that would be collected from the increase in the excise  tax
for each of the three forecast years — 1975, 1981, and  1987 —
under each of the three levels of increase in excise tax analyzed
                                163

-------
                                Table  4-27
         ESTIMATED ADDITIONAL TAX RECEIPTS FOR EXCISE TAX POLICIES
                           IN 1975, 1981, AND  1987
                           (Billions  of Dollars)
Year
1975
1981
1987
Increase in
Excise Tax
$0.10/galIon
$0.25/galIon
$0.50/ga!Ion

$0.10/galIon
$0.25/galIon
$0.50/galIon

$0.10/galIon
$0.25/galIon
$0.50/galIon
Sensitivity
Low1
$10.02
$23.33
$40.91
$1 1.03
$18.63
$14.98
$13. 17
$21 .97
$ 7.43
Medium2
$10. 17
$24.25
$44.59
$1 1.71
$24.05
$30.65
$14. 17
$28.31
$32.52
High3
$1 1.01
$26.86
$51.50
$13.75
$31.97
$56. 16
$17.07
$39.49
$68.39
Notes:
   'Assumes high gasoline prices, high elasticity  assumptions.
   2Assumes high gasoline prices, medium elasticity assumptions.
   3Assumes high gasoline prices, low  elasticity assumptions.
                                    169

-------
and for the three sensitivity assumptions.  This table shows that
a tax increase of $0.10 per gallon would result in additional
tax receipts of about $10 billion in 1975, $12 billion in 1981^
and $14 billion in 1987.  A tax increase of $0.25 per gallon
roughly doubles these amounts, while an increase of $0.50 per
gallon results in much higher tax receipts in 1975 but,  for the
central estimate, roughly the same receipts in 1981 and  1987.
(It should be noted that estimates of receipts under a $0.50
per gallon increase are quite sensitive to price/sensitivity
assumptions in the later years.  For example, in 1981 the range
is between $15 and $56 billion, while by  1987 the range  spreads
from $7 billion to $68 billion.)
     From the available data, it appears that these increases
in gasoline tax revenues would be paid by all households
roughly in proportion to their incomes.  Table 4-28 shows the
average daily vehicle miles by households in different income
classes.1  If we assume that average fuel economy per vehicle
is roughly the same across income classes, we can estimate
gallons of gasoline bought per year by households in each
class, also shown in Table 4-28.  A $0.10 per gallon surtax
would cost those in the lowest income bracket about $35  per
year, or roughly 1.2 percent of their income.  It would  cost
those in the $4000 to $9999 bracket about $90 per year,  or
about 1.3 percent of their income, while  those in the next
higher bracket would have to pay about $129, or 1.0 percent

   :The figures used in this section are based on the 1969-
1970 Nationwide Personal Transportation Study.  Both the income
ranges and household VMT estimates are understated due to
inflation and the growth in real income and in travel.   For
our purposes,  therefore, these ranges should be interpreted
as relative income classes, rather than as current absolute
income classes.
                               170

-------
                           Table 4-28

   DAILY VEHICLE MILES OF TRAVEL AND YEARLY GASOLINE CONSUMPTION
            BY HOUSEHOLDS IN DIFFERENT INCOME CLASSES
 Income Class
       (1)

 Average Daily
 Vehicle Miles,
Earning a Living
      (2)

Average Daily
Vehicle Miles,
 All  Purposes
     (3)
   Yearly
  Gasoline
Consumption,
 in Gallons
    <$4000

  $4000-9999

$10,000-14,999

   >$I5,000
       4.0

      13.3

      21 .1

      31.7
     12.9

     33.5

     47.9

     66.9
    346.2

    899.1

   1285.5

   1795.5
SOURCES:

      (I) and (2):  Nationwide Personal Transportation Study3
Report No. 7, "Household Travel in the United States," (December
1972), p. 21 .

      (3):  Translation of daily vehicle miles of travel  into yearly
gallons of gasoline consumption assumed (I) 365 days per year, and
(2) 13.6 miles per gallon (Federal Highway Administration,  Highway
Statistics 1969, p. 73).
                                 171

-------
of their income.  Those in the highest bracket would have to
pay about $180 per year, or 1.2 percent.1
     Gasoline taxes higher than $0.10 per gallon would take
a proportionately larger share of income,  but the distribu-
tion across income classes would not be affected.  It does
not appear, therefore, that increases in the gasoline taxes
would be regressive, except slightly as between the two
middle-income classes.
      The implicit amount of income granted to consumers of
gasoline via a coupon rationing scheme can be computed as the
market price of the coupon times the allowed gasoline consump-
tion.  For a rationed amount of 10 gallons per licensed driver
per week, Table 4-29 shows the implicit income for each of the
three forecasted years.  These amounts are different in each
year, both because of the increase in licensed drivers and
because the coupon price changes over time, as gasoline con-
sumers respond to the higher gasoline prices in  a delayed
fashion.
     Note that, because demand is much more elastic in the long
run, the coupon price falls sharply from 1975 to 1981 (and only

    aThis discussion ignores the decrease in consumption
brought about by higher gasoline prices.  We are not aware
of  any data on elasticities by income class but  if, as
seems reasonable, those in the lower brackets cut back
their purchases proportionately more than those  in the
higher brackets, the tax might even exact a larger per-
centage of upper-bracket income than lower-bracket income.
The calculations are also somewhat confused by the large
width of the income classes — it is hard to know, for
example, what the average income of those in the highest
and lowest brackets is.  In calculating these percentages,
we  have assumed the following average household  incomes
for the different classes:  $3000, $7000, $12,500 and
$16,000.
                              172

-------
                                Table  4-29
            ESTIMATED VALUE OF RATIONING COUPONS FOR DIFFERENT
              SENSITIVITY ASSUMPTIONS IN 1975, 1981, AND 1987
Year                                  Low1             Medium2         _ High3
1975          Coupon Price1*          $ 0.86            $ 1.27          $  2.64
              Aggregate value        $56.10            S82.77          $172.15
                of coupons5

1981          Coupon Price"          $ 0.26            $ 0.39          $  1.06
              Aggregate value        $16.42            $29.88          $ 80.49
                of coupons5

1987          Coupon Price"          $ 0.25            $ 0.37          $ 1.06
              Aggregate value        $21.97            $33.27          $94.66
                of coupons5
Notes:
   Assumes high gasoline prices,  high sensitivity.
   2Assumes high gasoline prices,  medium sensitivity.
   3Assumes low gasoline prices,  low sensitivity.
   "in  dollars per gallon, 1975 prices.
   5ln  billions of dollars per year.
                                    173

-------
very slightly from 1981 to 1987), as does the aggregate value
of the coupons.  After 1981, however, the aggregate value of
the coupons increases as the growth in gasoline consumption
outweighs the slight decline in the coupon price.  For the medium
elasticity assumptions, the total value of the coupons distri-
buted to licensed drivers falls from about $83 billion in 1975
to about $30 billion after 1981.  The estimated aggregate value
of the coupons does, however, vary substantially with the
elasticity/price combination assumed, as does the value of the
coupons themselves.
     As shown above, households in higher income classes
drive more.  Households in the higher income classes also
tend to have more licensed drivers.1  The net impact of cou-
pon rationing on income distribution is not immediately
apparent.  Table 4-30 shows desired gasoline consumption and
coupon entitlements for households in different income
classes, while Table 4-31 relates actual consumption under
rationing to entitlements, on the assumption that all house-
holds reduce their desired consumption by the same proportion.2

   *The sharp increase in licensed drivers per household
as income increases is presumably because the expense of
learning to drive is incurred only if a car is likely to
be available, and the number of cars per household
increases sharply with income.  Another source of this
correlation may be that those physically unable to drive
(because of age or infirmities) also tend to have lower
incomes.  If coupon rationing is adopted on the basis of
licensed drivers, the value of the coupons will provide
an incentive for those without licenses but capable of
driving to obtain them.  In this case, coupon rationing
will lead to a greater redistribution of income from the
upper to the lower income classes than estimated below.
   2If those in the lower brackets reduce their consump-
tion more than proportionately, then the income redis-
tribution from upper to lower classes will be greater
than estimated below.

-------
                            Table 4-30

          COMPARISON OF DESIRED GASOLINE CONSUMPTION AND
              COUPON ENTITLEMENTS, BY INCOME CLASS
           (Gallons of Gasoline Per Household Per Year)
 Income Class


    <$4000

  $4000-9999

$10,000-14,999

   >$I5,000
                      (1)
 1969
 Con-
sump-
 tion
 346.2

 899.1

1285.5

1795.5
(2)
1975
Desired
Con-
sump-
tion
443.5
1 151 .7
1646.7
2300.0
(3)
Average
Number of
Licensed
Drivers
per House-
hold
0.75
1.64
2.03
2.36
(4)
Entitle-
ments
per
House-
hold
390.0
852.8
1055.6
1227.2
(2)7(4)


  1.14

  1.35

  1.56

  1.87
SOURCES:

      (I):  Col. (3) of Table 4-28.

      (2):  Estimated as 1969 consumption times the ratio of 1975
forecasted consumption (post-embargo prices) to 1969 consumption.
Forecasted consumption of 104.81 billion gallons is shown in Table
3-10; 1969 consumption of 81.79 bi I I ion gal Ions is shown in Table
3-1.
      (3):  Average number of licensed drivers per household was
derived from Natioraji.de Personal Transportation Study,  Report No.
II, "AutomobiIe Ownership" (December 1974), p. 22.
      (4):  Estimated as Col. (3) times 520 gallons per year.
                                175

-------
                            Table  4-31

                      COMPARISON OF  ESTIMATED
              HOUSEHOLD CONSUMPTION  AND ENTITLEMENTS
                 (Gallons  per  Household per  Year)


                            (1)                 (2)
                     Consumption under        Coupon
 Income Class         Coupon Rationing     Entitlements     (1)/(2)


    <$4000                  306.5               390          0.785

  $4000-9999                795.8               852.8         0.933

$10,000-14,999             1137.9              1055.6         1.077

   >$I5,000                1589.3              1227.2         1.295


SOURCES:
      (I):  Estimated on the assumption that all  households reduce
their desired consumption in the same proportion.  Average licensed
drivers per household are 1.41,  so that average household  entitlements
are 734.9 gallons per year, while  average desired consumption in
1975 is 1062.2 gallons.  The entries in Col. (2)  of Table 4-30 were'
multiplied by (734.9/1062.2), or 0.691, to arrive at  the entries in
Col. (I).
      (2):  Taken from Col. (4) of Table 4-30.
                                  176

-------
     Table 4-32 shows the estimated household payments and
receipts for coupons by income class.  The amounts do not
appear large.  Households in the two lower income classes
would receive, on average, $40 to $100 from the sale of
their unused coupons in 1975, but $20 to $30 by 1981.
Those in the highest bracket would pay about $450 in 1975
for additional coupons, but less than $150 by 1981.  In
summary, then, coupon rationing would redistribute income
from the upper to the lower income brackets, but the
amounts are not significant.

Incentives for Efficiency and Technical Change
      For equivalent levels of gasoline rationing and excise
tax increases — that is, for levels of rationing for which
the market price of the coupon  (disregarding transaction costs)
is the same as the increase in the excise tax — both coupon
rationing and an increase in the excise tax provide similar
incentives for efficiency and technical change.  Under both
sorts of policy, drivers have incentives to use gasoline effi-
ciently and to cut back on their consumption.  Further, under
both sets of policies, automobile manufacturers have an incen-
tive to provide more efficient engines and technologies that
use less fuel per mile, as, at higher prices, drivers naturally
tend to be more concerned about fuel economy of new cars.  This
concern may also lead manufacturers to invest resources in dis-
covering improvements in existing engine designs or, perhaps,
even radical new designs, that will lead to more efficient fuel
consumption.  It is clear that quantification of these incentives
is a very difficult job, beyond the scope of this report.  How-
ever, it also seems clear that these policies provide incentives
for efficiency in technical change that move in the direction
of reduced fuel consumption.  They do not provide incentives
for reducing emissions below the level required by legislated
standards.
                               177

-------
                            Table  4-32

               ESTIMATED TRANSFERS PER  HOUSEHOLD  BY
               INCOME CLASSES UNDER COUPON  RATIONING
 Income Class


    <$4000

  $4000-9999

$10,000-14,999

   >$I5,000
(1)
Net Coupons
Bought1
-83.5
-57.0
82.3
362. 1
(2)
1975 Net
Payments
-$106
-$ 72
$105
$460
(3)
1981 Net
Payments
-$ 33
-$ 22
$ 32
$141
(4)
1987 Net
Payments
-$ 31
-$ 21
$ 30
$134
Price per Coupon
$1.27    $  0.39    $  0.37
NOTES:
   xEach coupon is assumed to represent one gallon of gasoline.
Negative numbers mean a net sale of coupons, while positive numbers
mean a net purchase.  Similarly for dollar receipts and  payments.
SOURCES:
      (I)   Table 4-31, Col. (I) - Col. (2).
      (2) - (4):  Col. (I) times the estimated coupon price, based
on post-embargo gasoline prices and medium elasticity estimates;
shown in Table 4-28.
                                178

-------
          5.  POLICIES AFFECTING THE STOCK OF CARS

                        Introduction

     In  this chapter, we  consider  two sets of policies, both of
which affect fuel consumption, emissions, and air quality through
their direct impact on new car sales and their indirect
impact on the size and composition of the stock of autos.
One set of policies imposes an excise tax on new cars in
inverse proportion to their fuel economy.  That is, for
each mile per gallon less than 20 that an auto achieves in
a certain test, it is subject to an excise tax of so many
dollars.1  The second set of policies requires manufac-
turers to achieve a certain average miles per gallon of
their models, where the average is computed as a weighted
average based on the share of sales in the preceding year.
This latter set of policies directly affects the models
offered by automobile manufacturers, while the former set
has its immediate impact on the relative prices paid by
consumers for different fuel economy classes of automobiles.

   1 For convenience, we frequently refer in this chapter
to these policies as excise taxes based on fuel economy,
even though the relationship is an inverse one.  We also
refer to them as policies based on fuel consumption of
new autos, since fuel consumption is the inverse of
fuel economy, and the taxes may be thought of as being
directly related to fuel consumption.  We use these des-
criptions interchangeably, although we are talking about
the same set of policies in both cases.
                             179

-------
     The rest of this chapter has five main sections. First,
we present a qualitative analysis of the effects of these
policies.  That is, without attempting to assess their
impact quantitatively, we explore what can be unambiguously
determined about the direction of the effect of these poli-
cies.  Second, we discuss some practical difficulties in
determining the impact quantitatively, along with different
approaches that might be used.  Third, we present
the methods used to determine the quantitative impact.
The fourth section presents the results of this analysis.
The last section discusses in qualitative terms the
impact of potential changes in the structure and the
secondary impacts of these policies.
                   Qualitative  Analysis

      In this  section we  present a qualitative  analysis  of
 policies which affect new car  sales.  These policies will
 have  an indirect  effect  on gasoline consumption  and ambient
 air quality because  they will  change  the  fuel  economy and
 emissions characteristics of the stock of automobiles on
 the road,  as  well  as the size  of this stock, after they
 are put into  effect.   This analysis traces through the
 effects which the  above  policies will have over  time, indi-
 cating  where  possible the direction of change  in future
 gasoline consumption and automotive emissions  as a consequence
 of enacting these  policies.
      In this  analysis it will  be necessary to  distinguish
 between the qualitative  effects  which the policies have on
 gasoline consumption  and their  effects on automobile emis-
 sions.   In these policies, unlike  those analyzed in Chapter 4,
                              180

-------
it is not true that a decline in gasoline consumption will
necessarily ensure a decline in emissions.  This ambiguity
arises because the policies discussed in this chapter alter
the vintage distribution of the stock of automobiles.  That
is, the policies lead to a stock of automobiles with relatively
more older cars and relatively fewer new cars than would
have been the case in their absence.  Since older vehicles
emit substantially more pollutants per mile than newer
vehicles, this effect will tend to increase automotive
emissions, in spite of reduced gasoline consumption.1  For
this reason we shall consider the policies' qualitative impact
on gasoline consumption and ambient air quality separately.

Policy Impact on Gasoline Consumption
     The results of the qualitative analysis of the policy
impact on gasoline consumption differ depending upon whether
a short-run or long-run horizon is considered.  In the short
run, it may be the case that taxing poor fuel economy of
new cars will lead to more gasoline being consumed than in
the absence of the tax.  However, this somewhat counter-
intuitive result cannot arise in the long run.  Given enough
time for all cars sold prior to the policy change to be
retired, such policies will lead to decreased gasoline con-
sumption.  For example, a tax on those new cars which do not

     JFor example, the EPA estimates that a car manufactured
after 1975 will emit 4.42 times as much hydrocarbon pollutant
when it is nine years old as when it is new.  See "Compilation
of Air Pollutant Emission Factors," 2nd edition, U.S. Environ-
mental Protection Agency, April  1973,  Table 3.1.2-5.  Moreover,
during the period when the emission standards are in transition,
the older model years have much higher emission factors when
new than more recent models.  Increasing the proportion of
these older cars further worsens the average emission factors
of cars on the road.
                               181

-------
 achieve a certain minimum number of miles per gallon will
 increase the price of these new automobiles relative to
 those which meet the minimum requirement.  This will cause
 a shift in consumers' demand for new cars, leading to the
 purchase of relatively fewer new cars with fuel economy
 below the minimum level.  It follows from this that the
 average fuel economy of the new cars added to the stock of
 automobiles in each year will improve.  Over time the average
 miles per gallon of the stock of automobiles on the road will
 improve, leading to less gasoline being consumed than would
 otherwise be the case.
      In the short run, as mentioned above, the effects may
 be perverse.  An increase in the price of new cars leads to
 reduced scrappage of used cars.  Used cars are driven more
 miles and maintained in the fleet longer.  This leads to
 greater gasoline consumption by these autos than would have
 occurred in the absence of these policies.  The overall impact
 depends, therefore, on the relative importance of a number
 of different factors — the sensitivity of new-car sales
 and average fuel economy to the tax or restriction  and
 the change in used-car longevity and use in response to the
 decrease in new-car sales.  The analysis is formalized in
 Appendix D, but the main results (under a number of simplify-
 ing but not essential assumptions)  may be summarized as fol-
 lows :
     1}  A tax on poor fuel economy in new cars causes an
increase in the relative price of new cars with miles per
gallon below the minimum vis-a-vis those with better than
minimum fuel economy.  As a result, in each subsequent ,year
the fraction of total new car purchases belonging to the
class of autos with miles per gallon above the minimum will
increase.  This causes an increase in tlie average miles per
gallon of oars joining the auto stock in each year after the
policy is initiated.
                              182

-------
     2)  However, the imposition of a tax on some new vehicles
causes the average price of new vehicles as a whole to rise,
leading to a reduction in  the total number of new cars sold.1
That is, increasing the price of low mileage autos causes
people to buy less of them.  Some of these people buy better
mileage cars instead, giving the effect discussed in 1) above.
Others decide not to purchase a new car at all, causing new
car sales to drop.  Thus,  while the policy will lead to an
improvement in the average fuel economy of new autos, fewer
of these autos will be added to the stock of those already
on the road.

     3) At the same time,  this rise in the price of new
cars causes used  car prices to rise  (since new and used cars
are substitutes).  For example, some of the people who decide
not to purchase a new car,after introduction of the tax
causes new cars prices to  rise, will purchase used cars instead.
This shifts the demand for used cars upward leading to a rise
in used car prices.  But this increases the economic life of
older autos.  Used-car dealers will now find it profitable
to repair and market older vehicles which otherwise would
have been scrapped.  Hence the policy will cause fewer older
vehicles to be scrapped than would have been scrapped in its
absence.
     4)  Whether or not in the short run the improvement in
average fuel economy of cars on the road due to the higher
average miles per gallon of new cars is more than offset by
the effect of the increased use and economic life of

   1 This assertion cannot, in general, be proven on a priori grounds
It might be the case, to take a hypothetical example, that a
tax on fuel consumption would cause sales of expensive cars
with poor fuel economy to  fall so sharply that the sales-
weighted average  price after the tax was less than before the
tax.  In the technical appendix to this chapter, we rule
out this kind of behavior  by assumption.  For this kind of
behavior to occur, the cross-elasticities of demand among
different classes of car would have to be very high, but the
quantitative estimates of  the implicit cross-elasticities
(discussed below) suggest  that these cross-elasticities are
quite low.  Available empirical evidence, therefore, supports
the assumption made here.
                               1 °->
                               1o3

-------
the less fuel-economical older cars is an empirical question.1
Under reasonable assumptions, we can state unambiguously
that  the policy under consideration will eventually lead to
an improvement in the average fuel economy of all cars on the
road, and thus to a reduction in gasoline consumption  from
what it would have been in the absence of the policy.   In
the long run, less gas consumption will ensue.
     This result is  intuitively reasonable if one  considers
that from year to year the stock  of autos changes  through
the addition  of newly purchased automobiles and  the  retire-
ment of older vehicles.  For  all  practical purposes,  all
vehicles are  eventually retired.   Since,  as we have  already
seen above, the proposed tax on new autos with low mileage
will lead to  better  average  fuel  economy for  the new autos
added  to the  stock in any year subsequent to  the enactment
of the policy,  it  must eventually lead to a  stock  of auto-
mobiles with  higher  average  miles per gallon  than  would have
occurred without the tax.  This is so because in some
subsequent year, all cars on the  road when the tax was  first
instituted will have been retired. At this  point  the stock
of cars consists of  surviving autos manufactured in  years
after  the tax was  imposed.   But the tax would have improved
the average  fuel economy for autos sold in these years,  and
thus would eventually  lead to a reduction in gasoline con-
sumption below  the level it  otherwise would  have obtained.

      aThis short-run ambiguity  arises because we cannot
ensure a priori  that the combination  of the  improved fuel
economy of the  new additions to the auto stock plus  their
diminished number  will  outweigh the increase in  the number of
older  vehicles  due to reduced scrappage.  However, economic
intuition  suggests  (and estimation would no  doubt  confirm)
that total automobile use could hardly increase  in response
to an  increase  in  new car  prices.  The qualitative analysis
alone  does not  guarantee, therefore,  that in  the short  run
these  policies would reduce  fuel  consumption  below the  level
in the absence  of  these policies.  The empirical results, dis-
cussed below, imply  that these policies reduce gasoline  con-
sumption throughout  the forecast  period.

                              184

-------
     We next analyze the effect of a fuel economy restriction,
imposed' by the federal government on auto manufacturers,
requiring the "average" car manufactured by each auto
producer to achieve some minimum number of miles per gallon.
By "average" we mean the weighted average miles per gallon
across each manufacturer's entire line, where the fuel economy
of each model is weighted by that model's share of the manu-
facturer's sales in the previous year.
     For ease of exposition, we assume that each manufacturer
will respond to such restrictions by increasing the fuel
economy of some or all of his models so that average miles
per gallon rises to the minimum level required by law.1
We further assume that the imposed minimum fuel economy
standard is not currently being met  (or otherwise it would
have no effect) and that auto producers will succeed in just
meeting the standard rather than exceeding it.  This latter
assumption follows from cost-minimizing behavior on the part
of auto manufacturers.
     Since each manufacturer acts so as to just meet the
minimum fuel economy requirement, the actual average fuel
economy of new cars will be equal to the minimum level imposed
     aThe introduction of completely new models creates some
difficulties of interpretation, because the sales-weighted
method for determining average fuel economy cannot be directly
applied.  This difficulty might be taken care of by the following
rule:  any model for which sales data for the previous year do
not exist must meet the minimum miles per gallon standard; the
sales weighted average miles per gallon of models for which
sales data do exist must also meet this standard.  This rule
ensures that the sales-weighted average meets the standard and
also allows for the introduction of new fuel-efficient models
 (although such models contribute to the manufacturer's weighted
average only in subsequent years).
                              135

-------
by the government, if sales of new cars occur in the same
relative proportions as in the preceding year.   Under further
assumptions detailed in Appendix D, this will occur if the
increased cost due to the government restriction is spread
across all models in such a way that their prices rise in the
same proportion.  Thus/ if firms spread their increased costs
in this manner  (and henceforth we assume that they actually
do), the government can legislate the level of the average
fuel economy of the additions to the stock of autos in any
given year.
     These assumptions imply that the average fuel economy
of the new-car additions to the stock of automobiles will
increase and that new car prices will also rise.  Prices, in
this context, should be interpreted as prices per unit of
quality.  That is, a manufacturer might increase fuel economy
without increasing his costs by such methods as decreasing
vehicle weight, lowering horsepower, and so forth.  In this
case, even though money prices stay constant, the amount of
quality supplied has decreased and the price per unit of
quality increased.1
     It can be readily shown, under plausible assumptions,
that the restrictions must lead to an increase in the quality-
corrected prices of new cars.  Improved fuel economy, other
things equal, is viewed by consumers as an attractive
characteristic of a car.  Consequently, if an auto manufacturer
could improve the fuel economy of his models at no cost he
would do so, since he could thereby improve his competitive
     ^Quality, in this context, is used narrowly to refer to
attributes of automobiles valued by consumers and has no
normative significance.
                               186

-------
position.  To improve fuel economy to meet  the restrictions,
therefore, a manufacturer must either use a more costly
known technology or make investments in the research and
development of improved technology. Either  strategy will
result in increased costs and since all U.S. manufacturers
are affected  (by assumption, their model mix prior to the
restrictions does not meet the standards),  these costs will
lead to higher prices for new cars.
     A key assumption here is that the restrictions are
effective , in the sense that manufacturers  would not make
the improvements unless compelled to.  That they haven't
already done so is not necessarily conclusive, because it is
possible that, given sufficient time, higher gasoline prices
will induce them to make the improvements in fuel economy.
If so, then these policies, at the levels of stringency con-
sidered, have no impact either on gasoline  consumption or
automotive emissions.
     From this point on, the analysis follows the same lines
as that of the policies placing excise taxes on new cars in
relation to their fuel economy.  The increase in price of
new cars leads to fewer new cars being sold and fewer old
cars being scrapped.  The long-run effect on fuel consump-
tion is an unambiguous decrease, but the short-run impact on
fuel consumption cannot be unambiguously determined on theo-
retical grounds alone.1  The empirical impacts of these pol-
icies are discussed below.

Policy Impact on Automotive Emissions and Ambient
Air Quality
     We consider the qualitative impact of  the proposed
policies on air quality separately because  its analysis is
complicated by the dependence of a vehicle's emission factor

   JThe remarks of footnote 1, p. 5-6, apply with equal
force here.
                             18?

-------
on its age.l  Since both a tax on poor fuel economy and

federal restrictions on average fuel economy of new vehi-

cles will tend to shift the age distribution of the stock

of vehicles, analysis of these policies should take

explicit account of this dependence.

     In the analysis that follows we make the assumption

that for a given vehicle, automotive emissions per year

are proportional to the annual miles traveled by the

vehicle.2  Furthermore, this emissions factor is assumed

to be the same for all autos manufactured in the same
   *An emission factor is the grams of a given pollutant
emitted per mile.  Thus, the factor times a vehicle's
miles of travel measures the total output of the pollu-
tant from that vehicle.  We also use "emission factor"
to refer to the weighted average emission factor of the
entire stock of vehicles (i.e., that factor which, when
multiplied by total vehicle miles of travel, results in
total automobile emissions of a given pollutant).

   2This assumption differs, at least at first glance,
from the analysis in Chapter 4, where emissions were
taken to depend on the number of trips as well as on the
number of vehicle miles of travel.  The source of the
difference is our assumption that changes in the stock
of cars or in its age distribution do not change average
urban trip length.  In this case, there is a direct cor-
respondence between the two approaches.  We have no way
of verifying this assumption, but it seems plausible.
Although a shift in the age distribution toward older
cars might reduce the number of long trips  (vacations,
for example), this reduction would have only a negligible
impact on average urban trip length.  As these policies
do not substantially affect automobile operating costs,
there is no reason to think that the pattern of urban
auto use should change in one direction or another.
                            188

-------
calendar year. lf 2 New autos of a given year are equipped
with emissions control devices that keep emissions per
mile at  (or below) the EPA standards in effect for the
year model.3  Since these standards become progressively
more strict for vehicles manufactured in future years,
we may presume that the initial emissions factors for
new vehicles added to the fleet in future years will be
less than the emissions factors that were applicable to
currently existing autos when new.1*
     Average emissions per mile by an automobile tend to
increase with the age of the vehicle.5  This results
   *We ignore the slight overlap between a model year
and preceding calendar year.

   2This assumption is not strictly true for automobiles
manufactured before 1968.  Prior to exhaust emission con-
trols, pollutant emission factors varied considerably
among different makes of automobiles, but variations
were not closely correlated with fuel economy.  This
assumption is, moreover, innocuous.  Use of an average
emission factor instead of the distribution of individual
emission factors for pre-1968 cars could make a differ-
ence only if vehicle weight and horsepower were systemat-
ically correlated with emission factors, which does not
appear to be the case.  Moreover, the effect would, in
any case, be negligible, as pre-1968 autos are estimated
to account for less than 30 percent of the stock of cars
in 1975 and about 5 percent by 1981.

   3It is irrelevant, for our purposes, that cars are
designed to meet the standards on average over a driv-
ing span of 50,000 miles.  We assume that, except for
the deterioration factor, the emission factor is con-
stant over the life of the car.

   ''For a further discussion of the variability of ini-
tial emissions factors over time, see Section B of Appen-
dix D.

   5C.f. U.S. Environmental Protection Agency, "Compila-
tion of Air Pollutant Emissions Factors," 2nd Edition,
April 1973.
                           189

-------
principally from the deterioration or malfunction of the
emissions control devices installed on new autos.1  The
extent of deterioration and frequency or incidence of
malfunction will, of course, be greater for older vehi-
cles.  This phenomenon is captured by assuming that the
emissions factor appropriate to a given vehicle increases
by a constant multiplicative factor with each additional
year that the vehicle remains in the auto stock.2  Thus,
to know the initial emissions per mile of an automobile
is to know its lifetime profile of emissions rates.
However, these profiles will be different (perhaps dras-
tically so) for automobiles manufactured in different
years.
     By virtue of the assumed dependence of automotive
emissions on vehicle miles traveled, the policy impact on
the average fuel economy of the fleet is not an issue in
this section of the analysis.  However, we must concern our-
selves with how the policy affects the number of automobiles
of each vintage which will be on the road in future years.
In order to concentrate on this important question we will
neglect the  policy impact on vehicle miles travelled.  We
assume that in each future year every vehicle remaining in
the auto stock in that year is driven the same number of

   *We have not analyzed or considered possible policies
requiring maintenance, inspecting, and testing of emis-
sion control devices on old cars.  These policies are
not directly related to fuel conservation, but their
impact on emissions could be readily analyzed with the
models developed in this study.
   2The assumption of exponential growth of emissions
factors with age at a constant rate is discussed more
fully in Appendix D.
                             190

-------
miles, though this intensity of vehicle usage may vary

from year to year. l  We also assume that instituting the
policies under study will not affect the number of miles

drive by a vehicle in subsequent years.2

     Given the framework described above/ we can identify

three factors crucial in determining the policy impact on

air quality.  These are:  1) the effect of the policy on new

car sales; 2) the policy impact on the rate of retirement

of older vehicles; and 3) the projected emissions factors for

new automobiles manufactured in years subsequent to the

policy initiation.  Since annual emissions of autos manu-

factured in a given year are proportional to vehicle miles

travelled,  (VMT) and since  VMT are in turn proportional to

the number of such autos on the road, the quantity of emissions

will depend on the size of the auto stock.
           assumption of constant mileage for all vehicles
on the road in a given year is, of course, at variance with
the facts.  It is made only to simplify the qualitative analysis
of the policies under study here, but does not affect our
basic results.  As one might suspect, the intensity of vehicle
usage is a decreasing function of vehicle age.  Older cars
are less reliable and more costly to operate, and thus will
be used less frequently for long trips.  For example, in 1969,
the average VMT of a five-year-old car was 56.8 percent that
of a new vehicle, while a ten -year-old auto on average was
driven 37.5 percent as many miles as a new car (Nationwide
Personal Transportation Study, Report No. 2, "Annual Miles
of Automobile Travel," Federal Highway Administration,
April 1972, Table 4) .  Since the policies under study tend
to increase the relative number of older cars on the road,
explicit recognition of the decrease in vehicle usage with
vehicle age would only strengthen the conclusion that the
long run impact of the policies is to decrease automotive
emissions .

     2The reasoning in footnote 2, p. 5-10, can be used
to support this assumption.   Again,  this assumption is
made only for the qualitative analysis.
                              191

-------
     As discussed in the previous section, a tax on poor fuel
economy will cause an increase in the average price of new
cars, leading to a decline in new car sales.  This tends
to diminish the size of the stock of cars on the road in sub-
sequent years, causing a decline in emissions.  On the other
hand, new car prices rising will cause used car prices to
Increase, leading to fewer used cars being scrapped and
increasing the number of used cars on the road in subsequent
years.  Thus the impact of the tax policy on the size of the
auto stock depends on which of these two effects dominates.
     Furthermore, since the emissions rates of older vehicles
exceed those of more recent vintage autos, this tendency to
keep older vehicles in the stock longer will have the effect
of increasing the average emissions rate of the fleet.
Thus, even if the size of the auto stock decreases, absolute
emissions could still increase.  In qualitative analysis
of the short run effects it is not possible to resolve this
ambiguity, and we must rely on the empirical study carried
out  in this report to determine the direction of the policy
impact on automotive emissions.  Observe, however, that if the
emissions factors for older vehicles exceed by a substantial
margin those for new vehicles manufactured shortly after
the  policy is imposed, then this tendency of the tax to cause
older vehicles to be maintained  longer and to diminish the
number of new  (less polluting) vehicles  added to the stock,
may  increase significantly the average emissions per mile of
the  auto stock as a whole.  The  short-run effect thus depends
critically on the extent to which emission factors of new cars
are  lower than those of cars manufactured prior to the pol-
icy.   Ironically, the more rapid is the  decrease in emission
factors of new vehicles, the more likely  it is that a tax
                              192

-------
policy aimed at increasing the fuel efficiency of the auto
fleet will increase emissions and concentrations of pollutants
in the short run, relative to the emissions and concentrations
that would have occurred without the policy.  If, to take a
hypothetical example, emission factors were identical for all
past and future vintages, these policies would, by reducing
VMTs, also lower emissions.  The lower the emission factors of
vintages after the policy  (relative to emission factors of
vintages before the policy), the greater will be the increase
in emissions from a given shift in the age distribution resul-
ting from the policy.
     Once enough time has elapsed for the permanent EPA
emission standards to be achieved, however, we can rule out
the troublesome effects describe above.  Thus the long-run
qualitative impact of the proposed policies depends only on
the first two of the three crucial factors mentioned above.
To the extent that a tax on poor fuel economy diminishes the
size of the long-run auto stock, better air quality will
ensue.  However, since emissions per mile grow with vehicle
age, the policy's effect of increasing the average lifetime
of an automobile will tend to worsen long-run air quality.
If we could assume that in the long run the policy would
decrease not only the overall size of the auto stock, but
also the number of autos of each age that remain on the road,
then we could be sure that the eventual policy effect on air
quality would be a beneficial one.  Whether or not this is
the case depends  (in a rather complex fashion) on the -relative
sensitivities of new car demand and used car scrappage rates
to increases in new car prices.   In Appendix D we state a
condition (A7) which guarantees the long-run compatibility
of the goal of improved air quality with a tax policy aimed
                              193

-------
at increasing the fuel economy of the auto fleet.  Under this

condition (and the other assumptions of Proposition 5) we

may be sure that taxing the poor fuel economy of new vehicles ;

will also have the effect of reducing long-run automotive

emissions and improving air quality.

     Finally, we note that the above analysis applies directly

to the qualitative effect of federal restrictions on the

sales weighted average fuel economy of new cars.  This is

because, as observed earlier in this section, the policy

impact on air quality works entirely through its effect

on new car prices.  In the qualitative analysis of the policy

impact on gasoline consumption, we concluded that the increase

in auto manufacturers' costs necessitated to meet federal

standards for fuel efficiency would cause an increase in new

car prices.   Thus the qualitative analysis of the impact on

air quality of fuel economy restrictions is the same as that
of taxing poor fuel economy}/ 2


     1 The direction of the effects  is identical, but the size
of the effects could differ substantially, depending on which
policy requires the greater increase in new car prices to meet
the same fuel efficiency objective.  We cannot answer this
question a priori.  The size of the price increase due to fuel-
efficiency restrictions depends on technological considerations
external to automobile demand, such as research and
development costs.  With the tax on poor fuel economy, however,
the increase in new car prices is determined by the characteristics
of consumers' demand for autos of different fuel economy classes.
This increase depends on the amount by which the prices of  low
fuel efficiency cars have to be raised in order to cause a
shift in demand to better mileage cars sufficient to  achieve
the desired increase in average fuel economy.

   2This result assumes that the research and development
efforts necessary to meet federal restrictions will not
produce spinoff effects of better emissions control devices.
To the extent that this occurs, fuel economy restrictions,
which encourage research activity by auto manufacturers,
could lead to better air quality even when taxing poor fuel
economy has the opposite effect.
                               13k

-------
       Practical Difficulties in the Quantitative
               Analysis of These Policies

     The policies considered in this chapter are oriented toward
one particular aspect of automobile performance, fuel consumption.
Analysis of these policies, therefore, requires an understanding
of the relationship between this characteristic  of automobiles
and consumer behavior, such as demand for new cars and the
composition of the demand for new cars.  This section of the
chapter explores the question from a theoretical point of view.
It is divided into two parts.  The first part states the problem.
The second part discusses briefly other work related to this
problem.  The next section presents the approach used in this
report.

Statement of the Probler.
     Both sets of policies considered in this section have
two kinds of effects.  First, they change the prices of cars
relative to each other, according to their fuel consumption
characteristics.1   Second, these policies tend to raise the
prices of new cars relative to all other goods, including the
prices of used cars before the imposition of these policies.

     1 This assertion is probably, although not necessarly, true
for the policy restricting the fuel consumption of new cars.  As manu-
facturers are required to increase the fuel economy of their average
car, it seems quite probable that the costs incurred in increasing
the fuel economy of different models will be passed along to the con-
sumer, and that these costs will be different for different models.
It is, in principle, possible that manufacturers would choose to
increase fuel economy of all models by the same percentage, and
that this increase will be associated with the same percentage in-
crease in price of all models.  This outcome seems unlikely, but it
does not substantively affect the analysis in what follows.
                                 195

-------
     As shown in the qualitative analysis section of this
chapter, reasonable assumptions about economic behavior
tell us how consumers will react to these policies.  There
are two kinds of effects.  First, we expect consumers to buy
relatively fewer of the cars whose prices have been increased
more (i.e., the cars with worse gas mileage or higher fuel
consumption per mile) and relatively more of those cars whose
prices have increased less  (i.e., those cars with relatively
good gas mileage or low fuel consumption per mile).  This
substitution will occur because, with the change in relative
prices, some consumers will prefer to buy a smaller car that
gets better gas mileage because of its lower initial price,
when  previously they would have preferred to buy a larger
car in spite of its worse gas mileage.
     Second, fewer new cars will be bought, because some'
consumers with large old cars, which they were considering
trading in, will prefer to maintain these cars and make the
necessary repairs rather than pay the increased price for a
new car.  That is, those customers who, for one reason or
another, need or prefer large cars that get low gas mileage,
faced with the increased prices due to the excise tax will,
on average, buy fewer large new cars.  This kind of behavior
can be expected to affect the sales of every size class of
car whose price has increased due to the excise tax  (after
allowing for the shift to more fuel-economical cars .as a
result in change of relative prices).
     Although the qualitative effects of these policies are rela-
tively straightforward, it is difficult to determine the quanti-
tative importance of these shifts.  However, the second effect
is, in principle, simpler to determine.  If, for example,  we know
the average increase in price implied by the excise tax, we can
use existing estimates of the price elasticity of new car demand
to estimate how many fewer new cars would be bought.
                              196

-------
     The question of the average increase in price, however,
is not simple.  For example, if consumers were indifferent
between all different sizes of cars, then a shift in the
relative prices would mean that no cars with miles per gallon
less than 20 would be bought and all consumer demand would
shift to those cars not subject to the tax.  This kind of
consumer behavior is, clearly, quite unlikely.  At the other
extreme, suppose that each consumer has in mind a particular
size of car to buy.  He plans to buy this car or no car at
all.  If the excise tax increases the price too much, the
consumer will buy no car.  Otherwise, he will buy the same
size of car as he was planning to buy.  This kind of behavior
implies no increase in the sales of smaller cars at the ex-
pense of larger cars.  The percentage increase in price could
be easily calculated from the shares prior to the imposition
of the tax increase, and the shares after the tax could be
readily estimated.  Neither of these extreme assumptions is,
however, appealing.  Consumer behavior almost certainly lies
between these two extremes.  We now turn to a brief review
of different studies that, directly or indirectly, provide
some measure of the way consumers will choose among different
models when the prices of these models have increased differ-
entially in response to excise taxes based on fuel consumption,

Previous Approaches
     We are aware of at least two studies in progress that
bear on this question, but that were not available in time  to
be of use in this study.  First, Charlotte Chamberlain at
the Transportation Systems Center in Cambridge has been work-
ing on a related problem.1   As this work is  still in progress

    aCharlotte Chamberlain, "A Preliminary Model of Auto
Choice by Class of Car:  Aggregate State Data,"  (unpub-
lished paper), January 31, 1974.
                              197

-------
and was not available to us, we do not comment on it further
here.  Second, Charles River Associates, in connection with
a study for the United States Department of Labor, is estimat-
ing a model that describes the distribution,across consumers,
of tastes for different attributes of automobiles.  The mode-1
is designed to predict the share of sales of U.S. autos as either
the prices of different models of cars change or as their at-
tributes change.  This model is, therefore, perfectly matched
to the questions raised in this chapter, as the increase in
the excise tax could readily be translated into an increase
in the price of the different models affected, and these new
prices would then yield predictions of the change in sales
of these models.  Unfortunately, this model has not yet been
sufficiently developed and tested to be used in this study.
     A study performed by Chase Econometrics Associates, Inc.
addressed the question of what determines the share of sales
of different size cars.1   We have seen only a preliminary
draft of their report under this contract, and the estimated
equations may have been changed since this report.  The study
is quite ambitious, as it attempts to explain total new car
sales, as well as the share of new cars in five different size
classes.  For our purposes, however, the estimated equations
are not very satisfactory. 2    According to these equations,
the absolute price of luxury cars affects the share of luxury
car sales; the price of standard size cars relative to all
cars affects the share of standard size car sales; and the
price of intermediate cars relative to standard size cars
affects the share of intermediate size sales.  These are

   1Chase Econometric Associates, Inc., Report No. 2, sub-
mitted to the Council on Environmental Quality under Con-
tract #EQ4AC004, January 1974.
   2We do not evaluate the model here in the context of
the purposes for which it was intended, bur rather for
our own particular uses.
                              198

-------
the only three size classes of cars for which the equations
indicate that the share of sales depends on the price of that
size class.  Consequently, if we attempted to substitute the
increase in price implied by the excise tax into these equa-  '
tions, the results would indicate that the share of new car '
sales in the compact and the subcompact car classes would not
change.  This result, while not formally impossible, strikes
us as unlikely, as it indicates that there is no substitut-
ability between, for example, compact cars and intermediate
size cars.  This result probably also violates the definitional
constraint that the shares of sales must add up to unity.*
It would be possible to tinker with the model by assuming
that the loss in share of the larger size cars was captured
by the combined shares of compact and subcompact cars, but
any division between the latter two sizes would necessarily
be arbitrary.  Finally, because the equations are linear in
untransformed variables, the size of the coefficients depends
on the units of the variables.  Consequently, this model
would be difficult to apply without access to the raw data.
     Instead of the above methods, we decided to use methods
developed by Dewees.2  These methods have several advantages
for our purposes.  First, they bear directly on the willing-
ness of consumers to buy smaller and less powerful cars as
the relative prices of large and more powerful cars increase.
Second, the methods yield results that are internally con-
sistent and plausible.  Third, the equations are analytically

   *We say probably because, in principle, it is possible
that the net change in the share of the three classes of
cars which are affected by price would be zero.
   2Donald N. Dewees, Economics and Public Policy, The Automobile
Pollution Case (Cambridge, Ma.: MIT Press, 1974), pp. 168-178.
                              1S9

-------
tractable.  The equations and the methods used are undoubt-
edly not the last word on this difficult area of analysis,
but, in our judgment, they are the best available at the
•
present time.
     As a full account of his methods can be found in his
book, we only sketch his procedures here.  This analysis
builds on a relatively .long tradition of hedonic analysis
of automobile prices.1  The basic notion is that when a
consumer buys an automobile, he is in fact purchasing a
bundle of attributes or characteristics of that automobile
which can be adequately represented by the physical traits
of that automobile.  Such physical traits might include length,
horsepower, engine  displacement,  gross vehicle  carrying weight,
and  a number of less quantifiable attributes  such  as  handling,
power steering, braking ability,  and  so  forth.  The assumption
is made that the price  paid by  consumers  reflects the  consumers'
evaluation of these different  attributes or characteristics.
When the prices of  automobiles are regressed  on these physical
characteristics, the coefficients of  the characteristics  can
be  interpreted as the  marginal prices of the  characteristics,
          for  example,  Zvi  Griliches,  ed., Price Indexes and
 Quality Change  (Cambridge, Ma.:  Harvard University Press,
 1971), pp.  55-87,  215-239,  and the comprehensive bibliog-
 raphy, pp.  275-281.   The first article was  published by
 Andrew T. Court   in  The Dynamiss of Automobile Demand (New
 York:  General Motors  Corporation, 1939), pp.  99-117.
                               200

-------
such as the price a consumer is willing to pay for one
additional horsepower, for example, or one additional inch
of length.1
     Dewees estimated a number of such equations from cross
sections of data representing the models available in different
periods over the post-war period.  These equations gave him
a number of observations on the implicit price of horsepower,
which he then used as an independent variable to explain the
average horsepower per car over the period.  That is, he
effectively constructed a demand curve, relating the amount
of horsepower purchased to its price.  As explained in the
next sub-section, a tax on miles per gallon can be translated
into a tax on horsepower, which,through the estimated equation,
will lead to a decrease in the average horsepower demanded,
and a corresponding increase in average fuel economy.

          Methods Used to Analyze the Impact
          of Policies  Affecting  New Car Sales

     In this section, we describe in detail the approach
used to analyze the effects of excise taxes on new cars
based on their fuel consumption and restrictions on the

   'This analysis assumes that the measurable dimensions
adequately represent the characteristics of cars which con-
sumers in fact value.  Length and weight, for example, are
frequently used because they represent sturdiness, head and
leg room, carrying capacity, and so forth.  Recent work by
Ohta and Griliches suggests that these physical traits do,
in fact, reasonably represent the quantifiable attributes
valued by consumers.  See, in this connection, M. Ohta and
Zvi Griliches, "Automobile Prices Revisited:  Extension of
the Hedonic Hypothesis," (unpublished Harvard Institute of
Economic Research Discussion Paper #235, October 1973).
                              201

-------
fuel consumption of new cars.1   A number of regression
equations are used, but the discussion is non-technical.
The technical details, along with additional mathematical
tools and results, are presented in Appendix D.
     The procedures involved a number of steps,  so that a brief
overview here may serve as a useful guide to what follows.  One
basic notion is that an auto's fuel economy depends on its weight
and horsepower.  Second, we assume for purposes of analysis that,
of the different attributes of automobiles influencing fuel econ-
omy, consumers attach positive value only to weight and horsepower.
Under these conditions, a tax on fuel economy can, for our purposes,
be viewed as partly a tax on weight and partly a tax on horsepower.
Because it is implicitly a tax on weight and horsepower,  a tax
on fuel consumption leads to a decrease in the average weight
and horsepower of new cars sold, and hence aji increase in
their fuel economy.
     A tax on fuel consumption thus has two effects:  first,
it leads people to shift their purchases of cars to cars
achieving greater fuel economy.  Second, the tax leads to an
increase in the average price of new cars, which
affects the stock of new cars.   The stock is affected in two
ways:  first, the increase in the average price of new cars
   *The term "fuel consumption," as used here, refers to the
number of gallons of gasoline consumed per mile.  It is, there-
fore, the inverse of "fuel economy," which is measured in miles
per gallon.  For our measurement of fuel economy, we use the
results obtained by the Environmental Protection Agency on its
urban driving cycle.   (See, for example, Environmental Protection
Agency/Federal Energy Administration,  1975 Gas Mileage Guide
for New Car Buyers),  This measure may not yield
the same number as other tests of fuel economy  because of
differences in fuel consumption during different driving
modes  (cruising, acceleration, deceleration), the number of
stops and starts, load factors, and other test conditions.
                               202

-------
reduces new car sales; second, because of the increase in
price of new cars and the decrease in new car sales, older
cars will be maintained better and scrapped later in their
service lives.  Thus, the net effect on the stock of cars
is to shift the distribution toward older cars.
     The stock of cars and the average fuel economy of the
stock affect the consumption of gasoline through the short-
run demand equation for gasoline, discussed in Chapter 4.
The emission factors change from their forecasted levels as
well, because older cars have different emissions character-
istics from new cars.  The modified emission factors are
applied to vehicle miles of travel (determined from gaso-
line consumption and fuel economy) to compute the emissions
of the different pollutants  and from this point on, the
analysis is similar to that in Chapter 4.
     The analysis of restrictions on fuel economy for each
manufacturer differs in some respects from the analysis
just sketched.  As the proposed policies discussed here
mandate a certain minimum level of fuel economy, the fuel
economy characteristics of new cars are, in effect, known
by assumption.  It is not so easy to know, however, what
the increase in new-car price implied by these restric-
tions will be.  That is, it will almost certainly cost
the manufacturers more to design cars that will achieve
the new standards of fuel economy, without sacrificing
other performance dimensions, and these costs will almost
                            203

-------
certainly be passed along in the form of higher prices.*' *•
As we have been unable to find any data on the size of
the cost increases implied by meeting these minimum stand-
ards, we make assumptions to bracket the range of these
price increases.  These assumptions will be discussed below.
Once the average fuel economy and the average price increase
are determined, the rest of the analysis follows similarly
to the analysis for the excise taxes on new cars just discussed.
     The res,t of this section is organized as follows:
First, we discuss in detail the methods used to analyze
an excise tax on poor fuel economy.  Then we discuss the changes
in this approach needed to analyze restrictions on the fuel
economy of new cars.
     The methods used to analyze excise taxes on fuel economy
can, for convenience, be broken down into five steps:  First,
we determine the change in fuel economy resulting from the
excise tax.  Second, we determine the change in new car
sales resulting from the tax.  Third, we calculate the
change in scrappage of automobiles resulting from increased new
car prices and the decreased sales.  Fourth, these changes are

   Manufacturers might choose to keep the price of cars
constant, by sacrificing performance dimensions such as
weight, interior space, engine performance, and so forth.
These decreases in automotive quality can, however, be
viewed symmetrically to an increase in price, as consum-
ers are presumably interested in qualities per dollar.
For the purposes of this discussion, therefore, we assume
that performance and other attributes are maintained, and
that the increase in fuel economy comes about because of
design changes made by the manufacturers.
   2As it is the sales-weighted average that is the
standard set by the policy, manufacturers have an ingen-
tive to improve the fuel economy of all models, not just
those below the average set by the standard.
                              204

-------
translated into changes  in gasoline  consumption.   Fifth,
the change in gasoline consumption,  coupled  with  the change
in emission factors, leads to  the change  in  emissions from
the forecast levels.

The Response of Fuel Economy to  an Excise Tax  Based  on
Fuel Economy
     We start by assuming that consumers'  evaluation of
 v
all of the automotive characteristics  that influence fuel
economy can be summarized in their evaluation  of  weight
and horsepower.1  The equation relating miles  per gallon
to weight and horsepower is as follows:2
     Miles per gallon = 25.909 - 0. 02022*Horsepower
                     - 0. 00270*Weight (in pounds)
This equation implies that, other things  equal, increasing
an engine's horsepower by ten  leads  to about a 0.1 mile
per gallon decrease in fuel economy.   Increasing  a car's
weight by 300 pounds, other things equal,  leads to a
reduction of about  0.8 miles per gallon.   At the  sample
means, the elasticity of miles per gallon with respect to
horsepower is -0.12 and with respect to weight is -0.85.
Thus, although this equation implies that vehicle weight
is the more important of the two determinants  of  fuel
economy, differences in horsepower are also  important.  If
both weight and horsepower increase  by 1  percent,  fuel
economy declines by about 1 percent.

   JA discussion of this assumption  can be found  in
Appendix D.
   2This equation is discussed in detail  in  Appendix D.
                            205

-------
     This equation is used in the next step of the analysis,
in which we determine the taxes on horsepower and weight implied
by a tax on fuel consumption.  The implied taxes on horse-
.power and on weight must satisfy two criteria.  First, the
total tax on a given car due to the fuel consumption tax
'must be equal to the sum of the tax on weight and the tax on
horsepower.   (In particular/ the total tax must be zero when
its horsepower and weight are such that they imply a fuel
economy equal to 20 miles per gallon.)  Second, this rela-
tionship must hold for all different combinations of weight
and horsepower of cars.  These criteria imply that a tax on
fuel economy can be split into a unique marginal tax on
weight and a unique marginal tax on horsepower, using the
estimated linear relationship between fuel economy, weight,
and horsepower.  A fuller discussion of the derivation can
be found in Appendix D.  The tax rates per horsepower and
per pound, corresponding to the different levels of tax on
fuel economy, are shown in Table 5-1.
     Taxes on weight and on horsepower imply, of course,
increases in the prices of these characteristics.  Since
these characteristics are not, however, priced separately
but are included implicitly in the price of the new auto-
mobile, we must also determine what these prices are.
These prices were estimated by the technique of hedonic
regression, discussed earlier.  The regression coefficient
of a characteristic (weight, horsepower, and so forth) can
be interpreted as the price of an additional unit of that
characteristic.  Estimation of the implicit prices of weight
and horsepower was beyond the scope of this study, and we
relied on those estimated by Dewees.1

   1Dewees, op. cit-., pp. 168-178.
                             206

-------
                               Table 5-1

                     EQUIVALENT TAXES ON WEIGHT AND
                     HORSEPOWER IMPLIED BY A TAX ON
                    EACH MILE PER GALLON LESS THAN 20
                                  Dollars                    Dollars  Per
Dollars per MPG              Per 100 Pounds of               Horsepower
 Less than 201            Weight Above Threshold2         Above Threshold3

      $50                        $13.50                       $0.51

      $100                        $27                          $1.02

      $200                        $54                          $2.04
NOTES:

      JThe tax is designed so that the rate applies to each mile per
gallon less than 20, tested in a city driving cycle according to the
EPA test procedures.

      2The "threshold" levels of weight and horsepower are such that,
at these levels, miles per gallon (estimated according to the regression
equation in the text) equal 20 (and hence the tax equals 0).

      3See Note 2.
SOURCE:
      For the derivation of these equivalences, see Appendix D.
                                    207

-------
     The last year for which Dewees estimated the price of

weight and horsepower was 1968.  To determine the base prices

of weight and horsepower, therefore, we needed to update these.

prices to 1975 levels.  We assume that these prices have been

rising at the same rate as new car prices.  This assumption,"

besides being the simplest one, is corroborated for the

period 1968 to 1971 by Ohta and Griliches in their recent

study.1  This assumption is also consistent with the

relatively stable average weight of new cars over this
period.2'3

     A central feature of this approach is the estimate of

the willingness of consumers to buy smaller cars as the

price of larger, less fuel-economical cars increases.  These

behavioral relations are embodied in two equations.  One equation


     *M. Ohta and Z. Griliches, "Automobile Prices Revisited:
Extensions of the Hedonic Hypothesis," (unpublished paper,
October 1973), pp. 34-35.

     2Strict corroboration of this assumption would require
not only weight but also all other characteristics to have
remained constant over this period.  In this case, quality,
as measured for the purpose of the regressions, would also
be constant and changes in new car prices would be reflected
directly in the prices of the attributes characterizing cars.
In fact, of course, other attributes have changed somewhat,
but estimation of a relationship for 1975 model-year cars
was beyond the scope of this study.
     3A particular difficulty obscures comparison of 1975
horsepower with 1968 horsepower, as around 1971 there was
a change in the way horsepower is reported.  Whereas it had
previously been measured with most of the engine components
that draw power  (such as alternators) receiving their
power from other sources  (gross horsepower), it has since
been reported on the basis of the power generated net of the power
used by these components  (net horsepower).  Consequently,
a simple comparison of the 1975 horsepower figure with the 1968
ones is misleading, as this comparison seems to imply a
decline in horsepower over this period.
                               208

-------
estimates the demand  for  horsepower as a function of its
price; the other  estimates the demand for weight as a func-
tion of its price.  The prices of weight and horsepower
were derived from the hedonic regressions discussed above.
The equations  (discussed  in detail in Appendix D), are as
follows:
     LOG(Weight)  =  7.350 - 0.0264*LOG(Weighted Average Price of Weight)
                  + 0.103*LOG(Weighted Average Per Capita Income)
and
     LOG(Horsepower) = -3.44 - 0.221*LOG(Weighted Average Price
                    of Horsepower) + 2.258*LOG(Weighted Average
                            Per Capita Income)
In these  equations,  it appears that the demand for horse-
power  is  considerably more elastic than the demand for
weight, although  both are quite inelastic.  That is, the
long-run  elasticity of demand for horsepower is only -0.2,
while  the long-run elasticity of demand for weight is
-0.026, even though both  estimates are significantly dif-
ferent from zero.
     Thus, suppose that the tax on weight  implied by  the
tax on fuel economy doubles the effective  price  of weight.
According to this equation, it will lead to a virtually  im-
perceptible reduction in  the average weight of cars purchased.
Further,  if the tax on fuel economy implies a doubling in
the implicit price of horsepower, the decrease in the average
horsepower of cars purchased will, after three years, be only
about  22  percent.  The inelasticity of these coefficients
implies that consumers will not readily change their  choice
                              209

-------
of vehicle size and power in response to increased prices
associated with these attributes.1
     Given the reduction in horsepower and weight, we then
compute what the average fuel economy of cars subject to
the tax will be.  A weighted average of this fuel economy
and the average fuel economy of cars not subject to the tax
results in the average fuel economy of new cars.  We then use
this average fuel economy, along with the sequence of fuel
economy of cars sold in previous years, to estimate the
average fuel economy of the stock of cars in each forecast
year.
     Given the estimate of average miles per gallon, we can
also calculate the average tax.  For example, if the average
miles per gallon  (after the tax has been imposed) is 18 ,
the average price increase (assuming that the manufacturers
pass along the full tax increase) will be 2  (20-18) times the
tax per mile per gallon.  I.e., if the tax per mile per gallon
is $100, the average price increase will be $200.  This
increase in price plays an important part in determining
how many new cars will be sold, as discussed next.

Change in New-Car  Sales in Response to the Excise Tax
     Given the sales-weighted average price of new cars
before the imposition of the excise tax, we then calculate
the percentage increase in the average price of new cars.
The percentage decrease in sales in new cars in response
to this tax depends on the elasticity of demand for new cars.
          problem with this procedure is that, over, the period
for which we have data, the prices of weight and horsepower
were declining or stable.  There is, therefore, no direct
evidence on consumer reactions to sharply increased prices
of weight and horsepower, as implied by the tax levels in
Table 5-3 .  It is quite possible that, at very high prices
of these characteristics, consumers would be much more willing
to alter their choice of cars, but history has not run the
experiment for us.
                              210

-------
     Estimation of an equation for new car demand was
beyond the scope of this study.  Demand for new cars is
a very complex phenomenon, and the studies which
have addressed this problem have frequently used quite
complicated formulations in attempting to determine the
structure of new car demand.  Among the variables used, for
example, have been variables to reflect credit conditions
in the automobile industry, weighted average stocks of cars
on the road, real disposable income  (in various forms), and
so forth.  All of these studies, however, imply remarkably
similar estimates of the elasticity of demand for new cars.
As discussed in Appendix D, estimates of the elasticity
range between -0.75 and -1.22.   These estimates suggest
that, if the excise tax on new cars causes their average
price to rise by 10 percent, new car sales will fall by
about 10 percent.  As the main estimate in this study,  we
chose a value of -1.0 for the elasticity of new car demand.
This value lies approximately midway between the high and
low estimates.   This elasticity estimate,. along with the
percentage increase in average price implied by the excise
tax, allows us to estimate the percentage reduction in new
car sales below the base case forecasts.

Change in Scrappage in Response to Tax
     The decrease in new car sales does not,  however,  imply
a corresponding decrease in the stock of cars.   The increase
in prices of new cars has two immediate effects:   first,
some people, instead of trading in their old car and buying
a new car, will make repairs to their old car and postpone
purchase of a new car (or simply drive the old one longer);
second, some people who might have bought a new car will,
                             211

-------
instead, buy a used car.  This increase in the demand for
used cars leads to an increase in used car prices.  Conse-
quently, dealers and other owners of used cars will repair
them more frequently instead of scrapping them.  The scrap-
page rate of all vintages of cars will decline, and used
cars will account for a relatively larger proportion of
the car population.1
     The equation used to estimate the response of scrap-
page to changes in new car sales is as follows:2

     LOG(S) = 1.7Q899 + 0.742378*LOG(E) - 0.912141*LOG(P)
where:
     S = ratio of actual scrappage to scrappage expected
         on the basis of age alone;
     R = ratio of new car sales during the year to stock
         of cars at beginning of year;
     P = ratio of Consumer Price Index for used cars to
         Consumer Price Index for automotive  repair and
         maintenance.
     The equation implies that a 10 percent decrease
in new car sales leads to about a 7 percent decrease in
the scrappage rate, while a 10 percent increase in the
price of used cars (relative to the price of maintenance)
leads to about a 9 percent decrease in the scrappage rate.
These parameters suggest, therefore, that scrappage is

   *A recent unpublished paper on scrappage, rates by
Richard Parks  I "Automobile Scrappage Rates:   the Choice
Between Maintenance and Built-in Durability,"  University
of Washington, October 1974) finds chat, in recent years,
scrappage rates have declined.  It is quite possible that
this decline has resulted from the installation of emission
control devices that are not desired by buyers of new cars.
    2Details on the specification and estirruation of this
 equation are contained in Appendix D.
                             212

-------
quite sensitive to new car  sales directly, and also to
new car prices indirectly,  through their effect on used
car prices.
     As outlined above, one of the effects of an increase
in new car prices is to bid up the prices of used cars.  Thus,
we need another equation to link used car prices to new
car prices.  Common sense suggests two properties that this
relationship ought to have.  First, it seems clear that an
increase in new car prices ought to lead to an increase
in used car prices, as used cars are a substitute for new
cars.
     Second, it seems plausible that, over the very long
run (say 12 to 13 years), the prices of used cars must move
proportionately to those of new cars.  That is, suppose that
the price of new cars rises by  10  percent  in this year  and
stays constant for 13 years.  Eventually, therefore, all
cars on the road will, when new, have been sold at the
increased prices.  If the relative structure of vintage
prices remains the same, all new and used car prices will
have risen by 10 percent once all previously existing cars
have been retired.

   *The relationship between new cars and used cars has
been investigated by, among others, Gregory C. Chow,
Demand for Automobiles in the United States (Amsterdam:  North
Holland Publishing Company, 1957), and Frank C. Wykoff,
"A User Cost Approach to New Automobile Purchases," Review
of Economic Studies, Vol. 40, No. 123 (July 1973), pp. 377-
390.   Both of these researchers found that used cars are
imperfect substitutes for new cars.  That is, their
prices move together, but are not perfectly correlated.
                             213

-------
     The following equation reflects these constraints:1
     LOG(CPIUC) = -0.1194 + 0.223*LOG(CPINC)  + 0. ?8?*LOG(CPIUC(-1) )

where:
     CPIUC    = Consumer Price Index for used cars;
     CPINC    = Consumer Price Index for new cars; and
     CPIUC(-l) = Consumer Price Index for used cars, lagged
               one year.
     The estimated equation suggests that the impact of a
change in the new car price index is only partially felt
in the first year.  That is, an increase in the new car
price of 10 percent implies only a 2 percent increase in
used car prices this year.  After 12 years, the used car
price index will have increased by 9.5 percent in response
to a sustained 10 percent increase in new car prices.2
This result is quite consistent with the average life of
a car .
     This equation translates a percentage change in the
price of new cars into percentage changes in the used car price
index in the years following the imposition of the excise
tax.  We then substitute these percentage changes, along with
the changes in new car registrations,  into the scrappage
equation, to determine the change in the scrappage rate implied
both directly and indirectly by the change in new car prices
resulting from the excise tax.  This procedure implies that
scrappage rates for different vintages will all change in the
         equation is discussed in detail in Appendix D.
   2For convenience, we assume that new car prices do not
change (after the initial 10 percent increase) over this
12-year period.  This assumption allows us to focus on the
dynamic relationship between new and used car prices.
                             214

-------
same proportion.  That is, it implies that the relative

scrappage rates of, say, one-year-old to two-year old cars  do
not change, even though the scrappage rates for both of

these vintages have decreased.


Change in Gasoline Consumption

     At this point, we have all the necessary elements to

determine the change in gasoline consumption that will result

from an excise tax on new cars.  From earlier steps, we

calculated the change in average fuel economy of the stock

of cars in each of the years of the forecast period.1  From

the series of steps just described, we calculated the change

in the stock of automobiles.  These new values are then sub-

stituted into the short-run demand equation for gasoline.

The short-run equation for gasoline, it will be recalled,

predicts gasoline consumption as a function of, among

other factors, the stock of registered automobiles and the

average fuel economy of the stock of cars.  When new values

of these variables are substituted into this equation (hold-

ing other factors constant), new predictions of gasoline

consumption result.  These predictions are then compared


   1An increase in fuel economy reduces the per-mile cost
of operating an automobile.  Gasoline costs per mile,
that is, are equal to the price of gasoline (in dollars
per gallon) times the number of gallons per mile (the
inverse of miles per gallon).  This effect was implicitly
taken into account in the short-run demand equation for
gasoline, since the coefficient on MPG  (fuel economy of
the stock of cars) reflects both the effect on operating
costs and the direct effect on consumption.  (Had the
equation been specified to include per-mile operating
costs — i.e., price of gasoline divided by fuel economy
— then the new fuel economy would also have affected
gasoline consumption through the coefficient of per-mile
operating costs.)
                            215

-------
with those of the corresponding baseline forecasts to esti-
mate the percentage change in gasoline consumption due to a
given policy.

Change in Emissions
     It remains to calculate the change in emissions resulting
from this set of policies.  Emissions are affected in two
ways.  First, the change in gasoline consumption clearly affects
emissions.  Second, because these policies change the age
distribution of the stock of automobiles, they also change
the future emission factors for the stock of cars on the
road in the forecast period.  Consequently, using the
emission factors of each individual model year, we recompute
the emission factors corresponding to the stock of cars
on the road in each year under the different policies.  These
emission factors, along with the new forecast of gasoline
consumption, are then used to generate new forecasts of
emissions under the different policies.  Implicit in this
calculation is the assumption that in spite of the changes
in the number of cars, their average fuel economy, and the
fewer vehicle miles of travel, the distribution of trips
within different city sizes and between the different times
of day is the same as in the base forecast.  This assumption
is not, probably,  strictly accurate, as the change in the
number of cars might be expected to alter somewhat the uses
to which cars are  put and, hence, trip  lengths and frequencies.
As we have no way  to quantify  these effects, however, we
have  chosen  an  assumption which simplifies interpretation  of
the results.  As  the percentage change  in the size qf the  car
stock is quite  small  (because  of the  substitution of used
cars  for new cars),  the error  we introduce by this assumption
                                216

-------
is also likely to be quite small.  The results of Chapter 4,

in which we took explicit account of changes in trip length

and frequencies, suggest that a more elaborate assumption

would yield negligibly different results.


Restrictions  on  the Fuel Economy of New  Cars

     We have  described  in detail above the  procedures used

to analyze the effects  of an  imposition  of  an  excise tax

on new cars.  Many of the same  steps are also  used  in

analyzing the effects of a policy restricting  the fuel

economy of new cars.  There are, however, some important

differences in the analysis of  these latter policies.

     First, the  form of these policies implies directly

what the fuel economy of new  cars will be.1 That is, since
the policies  we  analyze require  each manufacturer's sales-

weighted, average cars to achieve a certain  number of miles per

gallon  (and assuming that manufacturers  would  not have chosen

such a high standard in the absence of the  policy), it follows

directly that the fuel  economy  of new cars  will be  precisely

that required by the policy.  Given the  age distribution


   JWe have assumed that imported cars are  exempt from
this set of policies (although not from  the previous set
of policies)  and have taken them into account  in the com-
putation of the  results.  We  have made this assumption
because of difficulties in determining what jurisdiction
U.S. statutes might have over foreign-based manufacturers
and how the statutes might be applied.   For example,
would the regulations apply to  the sales-weighted average
of all sales  by  a foreign manufacturer,  or  only to U.S.
imports?  Many imports  already meet these standards.  For
example, based on 1974  market shares, almost 90 percent
of the top 10 imports achieved  20 miles  per gallon or
better, so that  the costs of  compliance  for these manu-
facturers would  be small.  Thus, altering the  assumption
would not affect the results  significantly.
                             217

-------
of the car stock in each year of the forecast period,  it is
straightforward to use the fuel economy of new cars and the
fuel economy of cars on the road for each year of the  fore-
cast period.  It is, however, much more difficult to deter-
mine the age distribution and the total number of cars on
the road in each year, and this is the second area in  which
the analysis differs substantially from that for the poli-
cies discussed above.
     This second area is more difficult because the analysis
depends critically not only on consumer demand but on  the cost
to the automobile manufacturers of improving fuel economy.
There is a strong presumption that these costs will imply
increases in prices that are greater than the value consumers
attach to the improved fuel economy.  This presumption
follows from the view that, if the automobile manufacturers
know how much it will cost to improve fuel economy, and if
they also know that consumers will be willing to pay a price
increase at least as great as that implied by the cost,
then the automobile manufacturers would have already had
an incentive to improve fuel economy up to this level.
The response of the automobile makers to the recent increase
in gasoline prices seems to bear out this line of reasoning.
For example, if the share of sales in the first 7 months
of 1974 are used to weight the fuel economy of U.S. cars,
the sales-weighted average fuel economy of 1974 models was
13.1 miles per gallon.  Partly because of the sharp increase
in gasoline prices between the 1974 models and the introduction
of the 1975 models, the sales-weighted average of  1975 model cars
increased to 15.5 miles per gallon, an increase of 18 percent.
Table  5-2 shows the sales and market share data for different
periods of  1974, and  it also shows the 1974 and 1975  fuel  economy
                               218

-------
Q-

s;
in
/*—
CTi

0
CL
Z
1 —
cn
Uv
*O iS?*^
C* •*"* , K O^
to -*sW
— I
. — ro'
ID f— -C|
>«_<<_ -*?•
CM o o ^j-9 r~-
U> I/I >, JU. r-1
OJ S
UJ s_ a
C3 -C Q
< co G ,
t— UJ
40 |
o c: ""*["' r~i
*~ " v^
to  ci
O^ Qpy CO ^^ f"* ' 1 *^" ^3 C3 CO r*^ ^~ U3 »~*

.


o

in-vin^jcsiocvj CDCTXCDCNC^CO in^-

i/> co c?\ CT* c^ r^* *"^ r^* Gs w^ i^ K^
(sj i^ u~\ p^ ^5 f^ pj »^f C) *~" *•" ^y \Q O^ *••
,- -w —



v«f • — • \o co in c? co r^ f^> •— f^ ^5
— *O %O vo O •— fNI •CO'— C»^t»O O>^-
.- to —


f^~ fO ^D »^ O^ C^ f**" CM CO O^ C3
K> C* in O «— CM CM K> O — OO T O CO'—
— — to —




to r^. r^- *•* ^r co m fO K*I ** r^ o f^ oo co
C^- — OTC>^rO>K1 OtOlAKIOCM - in r^
fO F*-



mr^-'— oy^CMi — ro^r r*- CM O o CM vo
1-^iT.tnlA --^- CM CMCOT in
CM
O U
-=. o
ro c ex c 4- c
4- O SO (DO
i- cn o) 01 E cn
OIOL zi_n oro
Q-SO oyaj CJ3:
CT -c u jc 4- —
C L. t. ro c 0> l_ C 4- — — >-O
— +-4-OOU14— — o O c — — t-cn
— QJ-- — o — so
E C C • 4- — O >-. O 4-
^j ^ I_ 4~ 4~ -*2 ^ L- ^ ^ •— -i— L_ t f~
L. O O fl 'r? £ t/>u- ss


i





o
0

o
•~




o




^
*~





.-
in






CO





r~








^».
o
L.
(a
4-
C



1





•-
CO

03





O>




O






en
fo
CN





*-
*,.




0*
CM








in
n
CT

1


•—





to
Ox

CO
CO




KV




VO
CO





fO
fT
^





CO
r-




o
in









(O
'01
D
O
o
                                                 0
                                                 a.
                                                O
                                                 O
                                                 o

                                                 c

                                                4-
                                                 c
                                                 Q
                                                 U

                                                 e>
219