Journal of Public Economics 6 (1976) 191104. @ North-Holland Publishing Company
A NEW APPROACH TO THE DEMAND FOR PUBLIC GOODS
RobertP. STRAUSS*
Economist,Joint Committeeon Internal ReoenueTaxation, U.S. Congress
Uniaersityof North Carolina, ChapelHill, NC 27514, U.S.A.
G. David HUGHES
Uniuersityof North Carolina, ChapelHill, NC 27514, U.S.A.
ReceivedOctober 1975,revisedversion receivedApril 1976
The individual demand for public goods is measuredby a new survey research instrument
which permits respondentsto make hypothetical expenditure and tax recommendationswith
moveable penny coupons. Becauseeach respondent faces an identical coupon budget constraint, it may be expected that observed expenditure and tax recommendations represent
individual maxirnum utility. The instrument was applied to a random sampleof 1000residents
of North Carolina. Statisticalanalysisof the resultant data indicatessignificantsocio-economic
differencesin the preferencesfor particular spending and tax categories.
1. Introiluction
The demand for public goods has generally been analyzed at the theoretical
level in terms of the choice betweenpublic and private goods, either in a partial
or generalequilibrium setting. Empirical evidenceon the individual demand for
public goods, or the demand for public versus private goods is quite limited,
although Mueller (1963) and Akin, Fields, and Neenar (1973) have examined
public attitudes toward particular public expenditures.l
The absenceof an extensiveempirical literature on the demand for public
goods is somewhat surprising in view of the size of the public sector in most
industrial democracies.This shortfall in our knowledge of what individuals
prefer in the public sector may be due to the absenceof convenient, market
generated demand data as well as the generally recognized problem of the
'free
rider.' That is, empirical investigation of citizen preferencesfor public
goods may be limited, becauseit is generallythought that a rational person will
not reveal his true demand for and willingnessto finance public sector activities.
An individual may understate his true willingness to pay, since he may be able
to participate in certain public serviceswithout financing them.
*The authorswish to thank S. KennethHoward, StateBudgetOfficer,Stateof North
Carolina,for his interestand supportin the research,
and BurtonA. Weisbrod,Universityof
Wisconsin,for his helpfulcomments.Responsibility
for opinionsand errorsin the research
restswith the authors.
lSeealsoACIR (1974).
192
R.P. Straussand G,D. Hughes, The demandfor public goods
We seek in this paper to examine individual preferencesfor public goods (and
tax reductions) with the use of a new survey research instrument which permits
the respondentto make constrained,hypothetical budget allocations.With such
a measurementdevice,we may be able to overcome the measurementproblems
usually encountered in public sector demand analysis. Section 2 describesthe
surveyinstrument, its rationale, and comparesit to thoseusedin previousstudies.
Section 3 discussesthe basic response obtained to the instrument. Section 4
explores the determinants of an individual's choice to spend and/or reduce
taxes as a whole as well as the determinants of individual preferencestoward
particular spending and tax categories.Section 5 concludes.
2. A new survey instrument to assessthe demand for public goods
Survey research typically relies on either multiple choice responsesto inquiries about likes or dislikes of particular options or open-endedresponses
to broadly worded questions. Scaling techniques2 such as multidimensional
scaling or coded interpretation of the open-ended verbal responsesare then
applied to obtain an individual profile. Besidesthe usual problems of miscoding
and nonresponsewhich accompanyeither researchstrategy,additional problems
usually occur and are pertinent to the particular issue of demand measurement.
Mueller noted in the 1961study that respondentinterest in program expansion
dependedcritically on the wording of the question. Second,there was substantial
evidenceof inconsistencyand instability of attitudes through the questionnaire.
That is, many who favored expansion of six or more (the largest category)
government.programs at a later point favored financing such an expansion by
'spending less
on other things.'
Such respondent unreliability reflectsthe physical limitations of a traditional
questionnaire.Usually, the respondent answersthe questions serially with little
thought of the overall profile being developed.Also, becausethe respondent
merely indicates likes and dislikes for particular policy options, it is reasonable
to assume that he will not make constrained recommendations.That is. the
probing ofpreferences may not be specificenough to force the subject to spending and taxing recommendationsthat in effect maximize his utility.
A novel solution to these problems developed by one of the authors3 is to
allow the respondent to make a constant sum seriesof recommendationson a
questionnaire which is sufficiently large to permit consideration of all options
at once. In effect, the 'coupon scale' survey instrument allows each respondent
to simulate his solution to the actual aggregatetaxing-spending decision.
Each questionnairea when unfolded provides the basic twelve expenditure
2A'representative
sampleof the scalingliteraturemay be found in Guilford (1954),Hughes
(1971),andGreenandTull (1975).
3SeeHughes(1971)for a discussion
of a constantsumscale.
aA copyof the questionnaire
is availablefrom the authorsuponrequest.
R.P. Strauss and G.D. Hughes, The demandfor public goods
193
programs in the North Carolina State Budget and their historical percent
distribution.Also, the four major sourcesof Statetax revenuesare provided.
Affixed to the questionnaireare fifteen moveable 1l coupons which can
be 'spent' on expenditureincreasesin any of the programsand/or reductions
in the major taxes.The respondent
is askedto allocatethe couponsto expenditure or tax reductioncategoriesto match his preferences.
Each major line item
containsa list of currentprogramdescriptions
and alsowhat programswould
be expandedby additionalspending.It shouldbe noted that the instrument
deviseddoesnot allow for expenditurereductions,althoughthis could be handled
by modifyingthe questionnaire.
The major advantageof the coupon scaleis that the resultantdata forms an
interval scale,becausethe absolutedistancebetweentwo pointsis well understood.Eachcouponhasan intervalinterpretationin comparisonto thehistorical
patternof resourceallocation.As a consequence,
metricstatisticaldevicessuch
asmultipleregression
maybe usedthat aremorepowerfuland richerin interpretation than rank correlationtechniques.
Becausethe respondents
are urgedto
move the couponsabout until they reflecttheir opinion, it is reasonableto
assumethat the final coupon allocationreflectsindividualmaximumutility.s
This overcomes
theproblemsof orderingthe questionsand subsequent
response
biasaswell aspossibleinstabilityof preferences
found in traditionaltechniques.
Finally, becausethe coupon scaleis self-administered,
the surveycan be obtained via mailout without personalinterview.In terms of allocatinga fixed
researchbudget,this permitsone to trade interviewcostsfor a larger sample
size.The mail surveyalsoeliminatesthe bias of the interviewer's
presence.
After a pretestof the instrumentamongdifferentsocio-economic
groups,the
survey was mailed out in early November, 1973.Followup reminderswere
mailed one weeklater. A month after the followup, 1001werereceivedof the
3517which went to viableaddresses.
A five-weekcutoff wasusedto assurethat
the respondents
would not reactto the Governor'sbudgetmessage
the following
January.Overall,a28.5/, response
rate was achieved.6
It is worth noting that
sWhile the use of moveable coupons may suggest that the responsesreflect maximum
individual utility, there remains the possibility that the responsesdo not reflect actual or true
preferences.There are, however, several reasons why observed responsesmay be accurate.
First, the sampleof individualswas such that it is unlikely that there was any collusion or
discussionamong the respondents.Thus, eachresponseis most probably independent.Second,
the questionnairewas presentedas a study rather than a subsequenttool for decisionmaking.
It would appear unlikely that in such a purely informational context that the respondents
would take the trouble to under or overstatetheir preferences.However, whether or not this
in fact occurred can not be ascertainedwithout additional independentinformation which is
not available.
6A return rate of 281is reasonablefor a questionnaireto the generalpublic
[Kanuk and
Berenson(19751.Two follow-up lettersgenerallyincreasethe return to over 60\(ibid).,and,
the inclusion of monetary incentivesmay increaseit even more. However, extensivefollowup was not used becauseof an agreementto protect the anonymity of the respondents,and
the likelihood that such later responsesmight be materially biased.
t94
R,P. Straussand G.D. Hughes, The demandfor public goods
the entire project was achievedfor lessthan $6,000;this small cost reflectsin
main the self-administered
nature of the questionnaire.
3. BasicsurYeyresponses
More than three-quartersof the respondentsused their coupons for both
spendingincreasesand tax reductions.Those favoring expenditureincreases
only numberedabout one-fifth, those favoring only tax reduction numbered
4of, and those for whom the resultswere unusablebecauseof a misunderstandingof the instructionsnumbered0.8\. The idea of a self-administered
instrumentwhich permits metric scalingthen waswell communicated.
The interagencyallocationof the couponsis displayedin table 1. Note that
68.1%of the couponsrvent for increasedspendingand 31.9f for decreased
wishes
taxes.Put anotherway, facedwith a I5l statesurplus,the respondents
to use 10.2%of it for more programsand the remainderfor lower taxes.The
ayerageper capita coupon outlays display substantialvariability. Primary and
secondary
education(essentially
statefinancedin North Carolina),medicalcare
wheremore expendiand hospitals,and public healthwereclearlythe agencies
the smallestallocation.Clearly,
turesweredesired.Motor vehiclesexperienced
but
the respondents
did not wish an acrossthe board increasein all agencies,
distinguished
amongthem.For example,the meanexpenditureon medicalcare
and highereducaor public schoolswastwicethat on highways,socialservices,
tion takenseparately.
of the
A rather different picture emergesof the budgetaryrecommendations
populationwhenwe relatethe averageper capitaoutlayto the existingbudget
allocationtheyfaced(column5 of table l). Thus,whilepublic schoolswerethe
most desiredagency,the impliedbudgetincreaseis quitemodest:2.8ft. Onthe
other hand, the absolutelysmalleragencycategories(highereducation,highas a
largerproportionalincreases
ways,and socialservices)would experience
would grow by 13.2\ while
of public preferences.
Socialservices
consequence
higher educationand highwayswould grow by 4.81 and 4.9f, respectively.
Two of the three health areas,public health and medicalcare and hospitals,
evidencebudgetaryincreasesin excessof 60\.
Among the six tax reduction categories,the state personalincome tax,
followed by the local propertytax were the taxesmost frequentlydecreased.
of taxes
In fact, meanoutlaysof couponsfor reductionof thesetwo categories
exceededmean outlaysfor any agencyprograms.Becausesubstantialnumbers
(the tax
of couponswere usedfor items toward the end of the questionnaire
it is apparentthat there were no
categoriesfollowed the agencycategories),
order effects,and that respondentsdid not simply randomlydistributetheir
coupons.
We may analyzethe intensity of attitudes of the respondentsby examining
the fraction who spent unusuallylarge absoluteamountson particular items.
Table I
Citizens' allocation of a state surplus.
Historical
pattern
(1)
Mean
per capita
coupon use
(2)
I of coupons
allocated to
category
(3)
I ofrespondents using
4 coupons
(4)
F
k
d:
46(
6
13
1
6
5
2
2
1
2
13
J
1.295
0.915
0.617
1.122
0.9il
0.658
1.302
0.881
0.837
0.261
o.637
0.666
8(,
l7
25
v
8.7%
6.2
4.2
7.6
6.1
4.4
8.8
5.9
5.6
1.8
4.3
4.5
7.7%
2.2
1.2
3.1
1.9
2.0
3.8
1.5
1.7
o.l
1.8
0.6
2.8%
15.3
4.8
112.2
15.5
13.2
65.1
4.1
83.7
13.1
4.9
22.2
0.4%
3.9
3.8
12.5
11.1
0.9
0.8%
4.9
2.9
0.987
0.738
0.377
0.984b
0.771
0.387
1.I 56b
4.721
o.7a
Q.149
0.353
0.481
4
;;
fis.
S
R
s:
68.r%
100
Tax reduction categorrf
TSf : State corp. income tax
TS2: State gasolinetax
TS3: State salestax
TS4: State personal inc. tax
TLl: Local property tax
TA: Local inventory tax
Median
gouponuse
(6)
(2y(1)I
t(3y15I
Spendtng category
El : Public schools
E2:, Community colleges
E3: Higher education
E4:. Public health
E5: Mental health
E6: Social services
E7:. Med. carc & hospitals
E8: Resourcedevelopment
E9: Agriculture
E10: Motor vehicles
Ell: Highways
E12: Corections system
lincrease
in budget
(5)
0.069
0.825
o.713
1.534
1.370
0.218
100
ulnformation not directly provided to respondent.
bModal coupons were 1 for these classifications, and 0 for all others.
o.5%
5.6
4.8
10.3
9.2
1.5
:'
0.020
0.348
Q.276
0.801
0.968
0.075
:.:
\
G..
*
oa
*
3r.e%
\o
196
R.P. Straussand G.D, Hughes, The demandfor public goods
For this purpose,we definefour or more coupons as'unusual' and note that only
public schoolsand personal and property tax reductions had as many as 5% or
more of the respondents spending unusual amounts. These are the three areas
of the public sector which contact the highest fraction of the population.
Virtually all of the population goes to public school and pays income and
property taxes. It is of course surprising that the sales tax does not arouse
analogous interest. Note the dissatisfactionwith the local property tax vis-a-vis
the state salestax; this is consistent with other studies, e.g. ACIR (1974).
4. Determinants of individual taxing and spendingdecisions
Two complementary hypotheseshave been forwarded to explain individual
preferencesfor public goods. The first hypothesis arguesthat public goods are
substitutesfor private goods, which in turn suggeststhe application of ordinary
demand theory to data on spendingpreferences.The secondtheory suggeststhat
people will prefer those public goods which benefit them the most. The first
theory simply encouragesthe estimation of such typical demand parameters
as income elasticities.The second theory provides some a priori expectations
about the signs of various income effects in the demand for particular public
servicesand/or tax changes.
The presence of metric data on desired expendituresTshould provide an
opportunity for the estimation of a complete system of demand equations (e.g.
the Stone-Geary linear expenditure system or, were time dependent data
available, the Theil-Barten differential demand model) which would be a complete operational counterpart to the second theory just noted. Unfortunately,
due to insufficient variance in our data, such models proved to be unworkable.
As table I indicates, the modal responsein all but two categorieswas not to
spend,and the median responsewas in all but one casean outlay of lessthan one
coupon.
As a consequenceof the essentially qualitative but nonethelesscardinal
nature ofthe data (the number ofcategories does exceedthe number ofcoupons
so a constrained choice is being made), we shall investigate the qualitative
demand functions for public goods by means of the multinomial logit model.s
TOurcouponapproachto measuring
demandfor publicgoodsparallelsin spirit the recent
work on tokeneconomies.
SeeAyllon andAzrin (1968),Kagel(1972),and,Kagelet al. (1975)
for work on tokeneconomies.
sSeeTheil (1969)for its development
and Schmidtand Strauss(1975)for an analogous
development
in the contextof continuousright handsidevariables.
There is somecorrespondence
betweenour multinomiallogit specificationand various
modelsof systemsof demandequationswhich are basedon maximizationof an underlying
(1963)andParks(1969)for a discussion
utility function.[SeeGoldberger
of variousdemand
modelsandtheirestimation.l
It is closest
in spiritto Stone's(1954)linearexpenditure
system.
Recallthat in sucha system,all supernumerary
incomeis spenton Ncommodities.Theincome
sharedevotedto eachcommoditysumsto unity. In our experimental
couponcontext,each
respondent
hasthesameadditional'income'toallocate,andweareconcerned
in thecontextof
R.P. Strauss and G.D. Hughes, The demandfor public goods
Let us considerfirst a model of the spendingvs. taxing decisionunder the
Then a multipresumptionthat public and private goods are interchangeable.
nomial logit specificationwould be:
log,f,Pr2lPrllt: XtBr,
(l)
log"lPr3lPrllt : XrBr,
Q)
log"lPr.rlPr3ft: Xt(Br-Bz) : XrBz,
(3)
wherePrr is the odds of only spending,Pr, is the odds of only reducingtaxes,
Pr3 is the odds of both spendingand reducingtaxes,Xis a TxM matrix of
- 1,...,3) is a Mxl matrix of response
coefpersonalcharacteristics,
\(j
indexfrom l, ...,7.
ficients,and t is an observation
Thus, giventhat an individual will only spend,only reducetaxes,or both spend
and reducetaxes,we may predict the relative odds of an individual so doing
on the basisof his incomeand other factors.Maximum likelihood estimationof
(l)-(3) can be pursued,althoughit shouldbe noted that (3) is not independent.
(l) and (2). The logicalextensionof (l)-(3) is to provide
Onein effectestimates
l8 basiccomparisons.
Sucha completesystemof qualitativedemandequations
would be of the form:
log.lProlPrrlt : XtBn,
k : 1,. . ., 18,
(4)
where the k are spending and tax reduction categories.However, because
allocationsof morethan onecouponper categoryoccurred,Pr1,is not bounded
per person.A completespecification
would entail l5 (numberof coupons)x 18
(number of categories): 270 separategroups.The readerwill note, however,
that the resultingcomputationalburdenwith a reasonablenumberof right hand
sidevariablesbecomesquite onerous.Thus,with M : 15,the currentcircumthe storagecapacity
stances,4050parametersneedto be obtained,which exceeds
of currentcomputersystems.We propose,therefore,a compromisemodelwhich
yieldssomeusefulinformation:
log,lPr2lPrlltk : X,Boo,
(5)
k : l, . . ., I 8
Pr, is the oddsof not spending,
wherePr 2 is the oddsof spending,
(l){3) with which type of allocation (all spending,all tax reduction, or some combination of
the two) each person rnakes,given that he must make an allocation of the coupon income. The
place of characteristic income, or the income of the respondent, is then an attribute which
affects the type of allocation made. Put another way, the adding up condition common to
systemsof demand equations is in our context a characteristicof the sample,namely that they
allocate the entire budget. The logit specificationnormalizesto ensure that the odds of being
of one of the three types necessarilyadd to unity.
r98
R.P, Strauss and G.D, Hughes, The demandfor public goods
independent categories. That is, we estimate l8 separate logit equations and do
not, because of physical computer limitations, impose the system-wide constraints implied by (4).
Table 2 displays the maximum likelihood estimatesof (l)-(3) and the l-ratios
in parentheses.Note that the numeraire in the first two equations is to only
spend. Thus, the positive effect of age in (1) may be interpreted to mean that
as the respondents grow older, they prefer to reduce taxes rather than to increase
spending.Interestingly, there is a distinct male-female differencein thesepreferences.Males systematicallyprefer to just reduce taxes rather than just increase
spending. Yet, higher income suggestsa greater desirefor more public goods.
The measuresof demographic status and region do not exhibit by and large any
systematic relations to the question of tax reduction or increased spending;
however, residents of medium size cities in North Carolina (10,000 to 49,999
population) distinctly prefer more spending to tax reduction.
The derived equation which compares both spending and tax reduction to
just tax reduction evidencesfew systematicrelationships.Males prefer to reduce
taxesrather than to increasespendingsomewhatand also reducetaxes,consistent
with the above results, and, residentsof the largest cities prefer spending and
tax reduction to just tax reduction.
In table 3, the individual spending and tax reduction equations are displayed
in the top and bottom panels respectively.Recall that the numeraire in each
equation is to not spend a coupon. Accordingly, we may interpret the effect of a
regressorin terms of its effect on the odds to spend rather than not spending a
coupon.
Older persons tend to prefer to spend on community collegesand on motor
vehicles,and not to spend on resourcedevelopment.Of the l2 spending equations, age has a signiflcant effect only 3 times.
More education tends to encouragespending on more education at all three
levels: public schools, community colleges,and higher education. Also, more
highly educated people tend to want to spend more on public health, mental
health, natural resources,and the corrections systems.Also, they systematically
prefer not to spend more on agriculture, motor vehicles,and highways. Men
systematically prefer to spend on resource development, motor vehicles, and
highways, and not to spendon social services(for the poor) and corrections.
In some respects our results for income are disappointing in view of the
infrequency of significant effects of income ( of l2). This may in part be due to
possible interdependenciesbetween income and education (their simple correlation is 0.4042 in the sample); however, where income is signifi.cant,so is
education. We find that income significantly affbctsthe demand for community
collegesand higher education, and has a depressingor inverseeffect on spending
for resource development and corrections. In a sense,then, resource development and corrections are 'inferior public goods' while community collegesand
higher education are 'normal public goods.'
F
ts
k
Tabb2
$
Determioantsof total expenditureand tax reductions.'
F&c
FsXt
x 1O-.
FcXs
Fils
Bcxz
O.l21O
(0.?lp2,
0.4386
(r22s,
FgXc
Fnztg
0.4786
(r;5r9)
0.3214
(0.?9rs)
9tXn
Fn)ts
Fnxrz
R
\
FlrXs
o
p
(l)
Pt,
.
tog.E;
O.5D9
(o.lryzD
0.01228 O.UI,37
(0.sr93)
(r.669)
O.qlO -0214
(r.?s23) (-zr$)
O.lOl
(0.3,rs3)
(2)
.tosc Ptt
pr-'
-O.6nO
(-0.46s9)
-0,3796
0.2450
0.01653 -0.(X19 -0.2980 -0.3988 -0.3116 -1.3250
(-0.5439) (-0.3530
(r.rsr) (-0.6102) (-0.s8lr) (-r.58)
(-o.s7a2, (-l.zw)
(3)
Pr"
to
rt
-1.1569
(-o.e53a
0.00425 -0.0715
@.s362t (-r.Alq
O.3M
(0.8,{4)
0.O1695 -0.1547
(0.02s7) (-o.4rrsq
-0.66.24 -o.4m
(-2.s4,
(-r.6D
-1572
0.2330 -0.615
(0.3034)(-0.?497) (-r.334,
0.8U1 -0.8057
(r.36!2) (-r.nw,
-0.7710 -0.18,18 -O.4ll7
-1.1459 -0.8182 -0.7236 -0.0884 -0.9409 -1.5989
Gr.rlrrD Go.rxtD (-r.23st (-r.399o
GL642i) (-0.?6el) (-0.833e) (-r.3630) (-r.n4D
0.9588 -0.1433
(r.s5l3) (-0.2384)
0.6595
o.2s6g,
1.087
(2.26,
.ftlis
.&2 i! the odds of iust redwhg ta*es, Pre is thc odds of t oth spending md reducing tues, xr fu agc itr yers, x2 is yeaf,8 of completed sc.h@ling
thc oddr of justspending
income in 1974 in dollas, xr is spouse of had of househol4 x6 is rlative of head of househol4 (head of houschold is eupprcEsed etegory),
x3 is sex (malc: I, fcmale:0), xr is mey
x?isN.Piedmo[tre8ion,xaisS.Pi€drntregion'x9isCmstalPlains'(thcmoultainsisthe3upprca3edcate8pry)'xroisS.Coasta|Plain3,x1ristomsupto2,499
(rual is thc supprcssed category),
cities Z5q) to 9,999, x13 is cities 10,000 to 49,999, 11a is cities 50,flD*
H
$
s
$
R.
l!
Et
b
s
F
B
\0
\o
N
5
>U
Table 3
q
d
Determinants of expenditure and tax reduction categories.'
!a
Fr
9zXt
F*z
F+Xt
F sX+
Faxs
Fils
BsXz
FgXa
BroXg
FtXto
FtzXtt
Bttxrz
Fuxtz
9rcxt+
Spending e4wriow
EI
E2
E3
EA
E5
E6
w
E8
E9
E10
Etl
F,t2
-0.8761 -0.00391
-0.r2rs
-0.3826 -0.09331
0.1139 -0.1566
o.2r77
0.2808
0.4406
0.0,1409 0.0s604 0.2017
0.0&61 0.1559
(-1.57r) (-0.6160
(0.5200) (0.2,rot (0.7182) (-o.420s) (- 1.3467)(-O.29Os) (0.6853) (r.r282) (0.0185) (0.7480) (2.252)
@rs3r) (-0.7367)
-1.683
0 . 1 0 1 5 0 . 1 1 1 7 0 . 3 2 s 7 0 . 1 5 1 4 0 . 1 1 s 0 0 . 5 1 5 3 -0.1447 -0.3869
0.00672 -0.1609
0.2136 -0.5659 -0.06s41 0.1093
-3.028)
(1.8463) (4.111) (1.5643) (1.7530) (0.5063e)(1.833) (-0.4e82) (-1.3529) (0.02078) -0.5130) (0.86s2,(-2.266) (-0.3151) (0.s/143)
-r.798
0.001659 0.1124 -0.08682 0.1340 -0.3510 0.6884 -0.07037 -o2s96 - O.4s69
0.2975 -0.1243 -0.1365 -0.04899 -0.01195
(-3.2532) (0.3062) (4.0e3) (-0.418e4) (r.657) (-1.53D (as158) (-o.2467t (-0.e178) (r.4397) (0.9703)(-0.5049) (-0.s365)(- 0.2337)(-0.0601)
-0.2887 -0.5483
r.030 -0.007355 0.0570 -0.08579 -0.1218 -o.zM -0.2053
0.s053 -0.2361 -0.080t9 -0.1073 -0.1964 -0.3299
(r.7s37) (-r.28s7)
(2.046,+)(-0.3789) (-1.468) (-0.ee93) (0.6flO (-0.9rs2) (- r.7715, (- 1.4552) (0.6657)(-0.3066) (-0/0,42' (0.8658 (-1.5785)
-O.r2
-0.1307 -0.10fl
-0.3098 -0.03865 -0.5722
0.1578
0.001823 0.05438 -0.16{
0.2681 0.146l. -0.1751 -0.,1025 -0.s52
(0.2869) (0.3365) (2.0560 (-0.7900 (-0.15033) (1.17D) (0.5363) (-0.600iD (- r.3e92) (- 1.7e08) (- 0.41le) (-0.43 14) (- r.244) (-0. 1840)(-2.884)
-0.01313 -0.3595 -0.07281 -0.1961 0.3034 -0.2690 -0.6002 -0.9222
-0.3662
0.3559 0.005894
0.3111
0.04271 0.2620 -0.02679
(0.66s4) (1.1089)(-0.50s7)'. -rJ699) (-0.9211) (-0.882s) (r.128) (-0.e748) (-2.re48' (-2.e2e) (-1.224',) (1.304) (O.r732) (1.2805)(-0.1367
0.8343
0.00805 0.01314 -0.1681 -0.o39o1
0.022108
0.7715 -0.05112 -0.04588 -0.2670 -0.01446
0.487t -0.1452
0.2359 -0.4475
(1.331s) (r.2994) (0.4320)(-0.6934) (-0.4893) (O.O915'(2.232t (-o.r542) (-0.1391) (-O.72r3'
(0.03969) (1.5663) (-0.5172) (0.959s)(-2.048)
-0.5457 -0.01253 0.1156 0.3584 -0.1841
-o.o1322 -0.00756 -0.6056 -0.07424 -0.1362
0.057930.2515
0.1307
0.0@24 -0.2120
(-0.998e) (-2.3187) (4.26q
(1.7080)(-2.303)
(0.2e83) (0.8713) (0.4565) (0.2137)(-0.6664) (-0.04287)(-0.03087)(- 2.4347) (0.3523)(-0.67s3)
-0.05248
-0.06253
0.9866
0.00191
0.0645
0.1236
0.8756
0.1031 -O.O7332 0.1542
0.07216 -0.3208 -0.3193 -0.4184 -0.5474
(1.8145) (0.3578)(-1.982)
(0.5939) (0.8159)(-0.277s) (r.9267) (0.3630 (-0.2624) (o.4847t (0.23sO (- 1.3266)(-1.2749)(-2.024) (-2.7622)
-2.@5
-0.0669
-0.008
0.02487
0.5622
0.3989 03rn
0.1275
0.3152 0.6814
0.2207 -0.1487 -0.1933 -0.1234
0.2541
(-3.246)
(4.00.37'(-2.2s4)
(2.1708)(-0.086s) (1.419) (0.9378) (0.3704) (0.935r) (r.834)
(0.s961) (-0.5205) (-0.6366)(-0.s073) (1.1s5)
-0.3426
-1.05016
-0.2333
-0.C7605
-0.2991
-0.4598
0.00459
0.4334
0.Ol
0.3030 0.5648
0.4037 0.2169
Q.t627
0.3339
(-0.6284) (0.8747)(-1.909)
(2.0s7) (0.1408) (1.315) Q.0e3)
(0.s75)
(1.,149) (0.6973) (1.096) (-0.3218) (- 0.94827)
(- 1.4s7) (- 2.3162)
-1.072
-0.4347
-0.1662
-0.1528
-0.08339
-0.3376
-0.5547
-0.02680
0.00.2527 0.t141
0.2683
0.5056 -0.1461
0.3294 0.1880
(-r.97r7)
(0.47M) (4.2380)(-2.1044) (-2.109) (-0.6766) (0.e907) (- 0.2977)(- 1.2181)(- 1.7528) (-0.0880) (2.079) (-0.5372) (r.6043) (0.9613)
L
{
6
Sr
!b
N
Sg
G
*{
s
o'
:
s
F
qa
Tox rcduction equations
-3.104
TSI
TS2
TS3
TS4
TLI
TT2
0.09832 0.04951-0.3942 -0.6@2
0.2159 -0.1356 -1.2t7
0.2531 -0.00167 -O.4r77
0.00782 -0.01492
0.4tt9 -0.1019
(-2.3r,
(0.60t8) (-0.2388) (0.6871)(-0.,1489) (0..1063)(-0.00714X-0.5868) (0.14e4) (0.06833X-0.5060)(-0.e263) (0.3984)(-0.2624) (-1.8013)
0.2435
0.4048 0.1703 0.4568 0.1005 0.1667 -0.1598 -0.1033
0.7131 -0.00931 -o.044o7 -0.1736 -0.2489 -0.07461 -0.1197
(1.3082)(-r.724, (-r.672) (-0.8467) (-2.855) (-0.3328) (-0.,1420) (o.8r',92, (r.4330) (0.s3467) (1.4825) (O.42oZ' (0.6803)(-0.7700) (-0.5207)
-0.0697 -0.00513 -0.03033 -0.00024 -0.0978 -o.240s -0.1225
0.1543
0.2223 -0.1780
0.3020 -0.2729
0.1358 :0.1366
0.1038
(-0.1257) (-0.9931) (-1.130) (-0.0011t(-1.172) (-1,0449) (-0.4467) (0.5303) (0.7731)(-0.s342) (0.9623)(-1.0731) (0.5453)(-0.64102' (0.52081)
0.1267
0.2128
0.3186 0.3826
0.4481
0.2740 -0.02535
0.04011
0.3109 0.00103-0.03015 0.02578-0.05926 -0.1006 -0.3231
(1.1575) (l.,1066) (0.41159) (0.7145) (1.8484) (1.1141)(_0.1256) (0.m74)
(o.s846) (0.19s3)(-1.16rO
(0.r273)(-0.7654) (-0.4539) (-r.213)
0.08324 0.06145-0.1319 -0.3819 -0.1.tO9
0.01090 -0.02099
0.1087 -0.5020
0.3072 -0.009015-0.1100
0.4270
0.2357 -0.2018
(1.077) (-o.o322) (-0.3470
(0.2703) (0.2464)(-0.s25O (-1.8451) (-0.7o22t
(0.7764) (1.978) (-0.7748) (r.r337) (-2.487'
(0.4781)(-1.858)
-1.845
0.2073
0.5t42
0.s426 -0.1058
0.2198 -0.09856
0.06253
0.6048
0.2298
0.00679 -0.05965
0.1016 -0.02422 -0.08582
(-2.279)
(0.8557)(-1.5852) (0.3379)(-0.1968) (-0.2527) (1.6613) (0.5r5s) (0.4705) (1.0740) (r.1642) (-0.2941) (0.6412)(-0.3179) (o.2r32)
lEach eqution
Dcpendent
is the lo96 ofthe probability
variables
El
Publicschools
E2
Commity
colleges
E3
Hiaher education
V
Publichealth
E5
Mental health
E6
Social senices
E7
Medi€l ere and hosDitals
E8
Resouce development
E9
Agridlture
EIO Motor vehicles
El I llighways
El2
Conections
TSI State @rp. income tax
TS2 State gasoline tax
TS3 State sales tax
TS4 State personal inome tax
TLI
Local property t*
TL2 Local inventory tu
ofspending
to not spending for tbe dependent vuiable
€tegories:
Independent variables
is age in years
xr
yearsofcompletedschooling
xz
is sex (male: l, female:O)
x3
is money income in 1974 in dollars
x4
is spouse of head of household
x5
is relative of head of household (head of household is suppressed €tegor,
x6
is N. Piedmont Region
x1
is S, Piedmont Region
xs
is Crastal Plains
xe
xro is S. Coastal Plains (the moutrtains is the suppressed category)
is toms up to 2,499 population
xrr
xr2 is citi€s 2,500 ro 9,999
xr3 is cities 10,000to 49,999
x1a is cities 50,000+ (rual is the suppressed category)
F
;
^:
z
a
L
fis
<;
;sr
$
s
:'
\
s
;'
t
*
tt)
2U
R.P. Strauss and G.D. Hughes, The demandfor public goods
categoryof head
Our measuresof placein the family (vis-a-visthe suppressed
of household) do not indicate substantial systematicrelations to spending
preferences,although dependents(children and relatives of the head, e.g.
parents living with children) desire more post-secondaryeducation,medical
care,agriculture,and lesscorrections.Also, there do not seemto be substantial
regional differencesin the preferencesor demandfor variouspublic goodsthat
are not explainablein terms of the educationand income of the respondents.
We do note, however,a marked interestin terms of not spendingon social
(for the poor) in 2 of the 4 regionsascomparedto the Mountainsregion
services
in the westernpart of the state,which is the suppressedcategory.Thesetwo
particular regionshappento be perhapsthe most politically conservativein the
state.
While theredo not seemto be pervasiveregionaldifferencesin the demandfor
variouspublic goods,there doesseemto be a systematicdemandfor particular
servicesby the larger urban areas.Big city dwellerssystematicallyprefer more
spendingon public education,not spendingon publichealthand mentalhealth
and not spendingon medicalcare and hospitalsas well as on socialservices,
agriculture,and highways.The finding that those in the larger urban areasin
the State, as comparedto those in rural areas of the State (the suppressed
category)do not want to spendon variousforms ofhealth careperhapsreflects
the geo-imbalance
in termsof availabilityof healthcarewhich recentlyprompted
the stateto developregionalpublic health clinics to serveoutlying areas.This
problemof maldistributionof healthresources
is apparentlychronicnationally;
our resultsreflectrather strong citizenconcernon the matter.
taxesascompared
Because
spendinga couponon a tax reductionis to decrease
to not spendinga couponon a reductionofthat tax, somecarein interpretation
of the tax reductionpanelis required.The numeraireis to not reducethe tax.
Thus, we find that increasedageleadsto an increaseddesirenot to reducethe
gastax but to reducethe property tax. Perhapsthe most surprisingresultis the
absenceof any relation betweenincome and a reductionin the individual
income tax. That is, we know from table 1 that reductionof the individual
incometax had the highestayerageper capitacouponusageand alsothe highest
fraction of extreme(more than 4) outlays.Yet, there is no relation between
income and a desireto reducethe tax on it. We may conclude,then, that the
desireto reducethe income tax was widespreadand not explainableby any
particular characteristicsof the respondents.The only signiflcant relation
betweenincome and any of the tax reductioncategoriesis the apparentdesire
of thosewith higher incomesnot to reducethe local propertytax. This may
reflect an awarenessas one movesup the income ladder of the tax advantage
of such local property taxes vis-a-visthe Federal individual income tax (they
are deductible),or it may reflect a self-interestof higher income personswho
regard the property tax as regressiveand thereforein a narrow sensein their
favor. Interestingly,dependentsas comparedto the head of householdsharea
R.P. Straassand G.D. Hughes, The demandfor public goods
203
similar attitude toward the local property tax, for they prefer not to spend on
local property tax reduction. Perhapsanother explanation for this phenomenon
is ihat such a reduction would curtail certain immediate servicessuch as public
safety,fire protection, and sanitation - the bread and butter as it were ofpublic
sector activity.
Again, interregional differences in the demand for tax reduction are not
apparent, nor are there particularly marked urban-rural differences. We do
note that large city dwellers do not want to reducethe State corporation income
tax, and that small town dwellers do want to reduce the individual income tax.
5. Conclusions
In this paper we have developed and applied a new survey approach to the
measurementof the demand for public goods. As such, certain innovations in
methodology, the measurementof individual preferencesfor public and private
goods, and the measurement of preferencesfor particular public goods have
been accomplished.
In terms of methodology, the coupon scaleis a simple and inexpensivemeans
of measuring cilizen preferences for fiscal programs. While the research results
reported were primarily exploratory, it would appear that becausethe data are
fixed sum and refer to actually desiredallocations of resources,the outcome of
such survey researchcan be readily translated into public policy. Public officials
who are not familiar with scaling techniquesmay find data using the coupon
scale more understandable than data resulting from attitudinal scaling techniques usually employed. We should note that the coupon scale is not limited
to the fiscal surplus allocation question examined in this study; the technique
can be used for gauging a wide variety of attitudes towards fiscal programs, e.g.
uses of revenue sharing, allocation of school budgets, and zero-basebudget
procedures.
Without restating our empirical resultshere, we do note by way of conclusion
that systematicsocio-economicdifferencesin preferencesfor public goods were
apparent: strong differencesamong men and women were apparent in terms of
their demand for particular public goods, and education had a substantial
impact on the demand for particular public goods.
It would appear this approach to the measurementof the demand for public
goods overcomes sorne of the theoretical objections usually lodged against
inquiring hypothetically or in fact about what an individual desiresand is willing
to pay for public goods. To be sure, providing a set of choices and a budget
constraint to each person does not ensurethat he will not mislead in stating his
spending and taxing preferences.On the other hand, the long held concern that
rational individuals will never truly reveal their preferenceshas not been subject
to much empirical scrutiny. It may be that until we have completed surveys
c
204
R.P. Strauss and G.D. Hughes, The demandfor public goods
similar to the one reported here, and then examinedhow people in fact use
public facilities which result from their recommendations,we will not know
how seriousthis problemis.
References
Advisory Commission on Intergovernmental Relations, 1974, Changing public attitudes on
governmentsand taxes (U.S. Government Printing Office, Washington).
Akin, T.S., G.S. Fields and W.B. Neenan, 1973,A socioeconomicexplanation of demand for
public goods,Public FinanceQuarterly 7, no.2,169-189.
Ayllon, T. and N.H. Azrin, 1968,The token economy: A motivational systemfor therapy and
rehabilitation (Appleton-Century-Crofts, New York).
Birdsall,W.C., 1965,A study of the demandfor public goods,in: R.A. Musgrave,ed., Essays
in fiscal federalism (Brookings Institution, Washington, DC) 235-294.
Goldberger, A.S.G., 1963,Functional form and utility: A review of consumerdemandtheory,
Report 6703, Social SystemsResearchInstitute, University of Wisconsin.
Green, P. and D.S. Tull, 1975, Research for marketing decisions, 3rd ed. (Prentice-Hall,
Englewood Cliffs, NJ).
Guilford, J.P., 1954, Psychometricmethods (McGraw-Hill, New York).
Hughes, G.D., 1971,Attitude measurementfor marketing strategies(Scott-Foresman, Glenview, IL).
Kagel, J.H., R.C. Battalio, H. Rachlin, L. Green, R.I. Basmann and W.R. Klein, 1975,
Experimental studies of consumer demand behaviour using laboratory animals, Economic
Inquiry 13, no. 1,22-38.
Kagel, J.H., 1972,Token economiesand experimentaleconomics,Journal of Political Economy
80,779-185.
Kanuck, kslie and Conrad Berenson, 1975, Mail surveys and responserates: A literature
review, Journal of Marketing Research){ll,265-287.
Mueller, 8.,1963, Public attitudes toward fiscal programs, Quarterly Journal of Economics77,
210-235.
Mushkin, Selma, 1973,Public prices for public products (Urban Institute).
Parks, R.W., 1969, Systems of demand equations: An empirical comparison of alternative
functional forms, Econometrica 37, no. 4, 629-649.
Schmidt, P. and R.P. Strauss,1975,The prediction of occupation using multiple logit models,
International Economic Review 76, no. 2, 471486.
Schmoelders,G.,7970, Survey research in public finance: A behavioral approach to fiscal
theory, Public Finance25, no.2,300-306.
Stone, R.D., 1954, Linear expenditure systemsand demand analysis: An application to the
pattern of British demand, The Economic Journal, 64.
Theil, H., 1969,A multivariate extension of the logit model, International Economic Review
10,251-259.
Weisbrod, Burton A., 1975, Toward a theory of the voluntary non-profit sector in a three
sector economy, in: Edmund Phelps, ed., Altruism, morality, and economic theory
(Russell SageFoundation, New York).