ronald c. fisher

8/13/2019 Ronald C. Fisher

1/30

onald C. Fisher

rofessor of Economics and Directorthe Honors College, Michigan Stateniversity. The author is grateful toffrey Guilfoyle for valuable researchsistance and to Timothy Bartik,arley Duncan, Therese McGuire,

nd Robert Tannenwald for helpfulmments.

The Effects of Stateand Local PublicServices on EconomicDevelopment

The purpose of this paper is to summarize the burgeoning literatureconcerning the relationship between public services, or at leastgovernment spending, and economic development in the jurisdic-

tion providing those services, as measured by changes in population,employment, income, or the number of firms. Much of this work,particularly the initial studies, arose as a reaction to research examiningthe effects of taxes on economic growth. Taxes, of course, are a means bywhich public services may be financed. If taxes reduce growth or inhibitdevelopment, then presumably no democratic government would collecttaxes unless there were (at least partially) offsetting benefits. Accurate

estimates of the possible negative effects of taxes require similar estimatesof the possible benefits from the public services financed by the taxes. Asthis work has advanced, however, the public policy focus has becomemore positive, as the research seeks to identify public service actions thatgovernments might use to stimulate various types of development.

Just as the focus of this research has evolved, so has the technicalsophistication of the underlying theoretical models and econometricmethods. The initial attempts to include government services often weresimply afterthoughts to research focused on tax effects. In manyinstances, one or two measures of government spending were included ina more or less ad hoc manner as controls. As the research has becomemore sophisticated and more concerned about service effects, attempts

have been made to include the entire governmental budget, to bettermeasure the quantity and quality of public services, to account for thepossible reverse causation of growth affecting public service choice, andto allow for possible differential effects of public services on differentindustries or factors of production.

The paper begins with a survey of the general results concerning theeconomic development or growth effects of transportation, public safety,and education spending or services as well as the results of studiesrelating public capital to national productivity. These results represent


2/30

e current conventional wisdom about the relation-hip between economic growth and the provision ofublic services, and they also illustrate the theoreticalnd econometric issues that are still unresolved or atast bring some uncertainty to the interpretation of

ose results. In the next section of the paper, each ofese difficult theoretical and empirical issues is con-dered and evaluated. One question is whether thesue has been dealt with adequately in the existingsearch; another is whether the issue appears to exert

n important influence on the statistical results. Someuggestions for additional research or a refocused

search direction are offered in the final section.It is important to note at the start that the topic of

is paper has been the subject of a very large numberresearch studies, and that many of these already

ave been surveyed and evaluated comprehensively.rominent reviews include Bartik (1991), Wasylenko

991), Munnell (1992), and Fox and Murray (1993),mong others. Fortunately, many of these people and

hers who have conducted specific studies of theevelopment effects of fiscal policy are participating inis symposium. Primarily a consumer of this researchther than a producer, I see my role not as one ofplicating or duplicating those prior reviews, butther as one of characterizing the existing state of

nowledge as a means of raising interesting newuestions and inducing important new research.

What Do We Know (or Think We Know)?

In many studies, government spending, publicpital, or public services are estimated to exert a

ositive and statistically significant effect on economicevelopment (Table 1). But the results vary greatly.erhaps the most that can be concluded is that someublic services clearly have a positive effect on someeasures of economic development in some cases. A

single partial-equilibrium resultdoes not appear to exist and,indeed, various reasons dis-cussed below suggest why oneshould not look for one.

Highways and TransportationFacilities

Of all the public servicesexamined for an influence oneconomic development, trans-portation services, and highway

facilities especially, show the most substantial evi-dence of a relationship. Of the 15 studies reviewed, apositive effect of highway facilities or spending oneconomic development is reported in 10 (or nearly 70percent), with that effect being statistically significant

in eight of the cases. A summary of these results isshown in Table 2.This significant and positive relationship arises in

studies of very different typeswhether the unit ofobservation is differences among states or differenceslocally, whether transportation service is measured byhighway spending or by a physical measure of facili-ties (miles of highways per area), for different mea-sures of economic development (including employ-ment, income, new investment, and the like), and instudies that attempt to correct for state or local fixedeffects as well as those that do not. Those studies thatmeasure highway service with a physical measure of

facilities (usually highway density) report positiveand significant results more often than those thatmeasure highway spending. It also seems importantthat while some studies report an insignificant rela-tionship, only one (Dalenberg and Partridge 1995)actually reports a significant negative relationship be-tween development and highway spending.

Often difficult to calculate because of the varietyof measures of government spending or facilities aswell as differences in the units of observation andmeasures of growth, the magnitudes of the estimatedeffects of highway spending on economic develop-

ment appear to be quite small, even when significantlypositive. Luce (1994), for example, suggests that a 90percent increase in public works/public safety spend-ing per household would translate into about a 30percent increase in manufacturing employment, butno increase in employment at all in retail trade andfinance.

The work by Dalenberg and Partridge (1995) is ofinterest partly for what it suggests about the major

Table 1

Summary of Results, Studies of Selected Public ServicesHighways

(Transportation) Public Safety Education

10/15 positive 5/9 positive 12/19 positive

8/15 positive andsignificant

4/9 positive and significant 6/19 positive and significant2/8 positive and negative, by

industry and functional form

5/19 highly varied results

March/April 1997 New England Economic Review


3/30

sues that need to be addressed in this area ofsearch. These authors examine the effects of high-

ay and other public spending on employment inggregate and for specific sectors among metropolitaneas, over a 15-year period. The research accounts fore entire set of government fiscal policies by includ-g measures of all taxes and spending categorieselative to one omitted category); it attempts to cor-ct for unobserved area-specific factors by usingree-year first differences for employment; it exam-es the disaggregated effects on different types of

industries; and it explores a possible influence of thestock of public capital separate from marginal changes

to the stock (new highway spending). In short, Dalen-berg and Partridge seem to have done all the rightthings theoretically and econometrically, and eventhey call their finding of a significant negative rela-tionship between employment and highway spendingsurprising.

Dalenberg and Partridge offer three possible ex-planations for this result. It may be that the averageeffect of highways is positive, but that the marginal

Table 2

Selected Studies of Highway (Transportation) EffectsAuthor(s) and Year Transportation Measure Development Measure Unit of Observation Results

Dalenberg and Partridge

(1995)

Highway spending/

personal income

Employment Metropolitan areas , significant

Evans and Karras (1994) Highway capital stock

and current

highway spending

Gross state product States , not significant for

capital stock; , not

significant for

spending

Luce (1994) Access to highway

and railroads

Employment, labor force Local governments , labor force

, significant,

employment

Garcia-Mila and McGuire

(1992)

Highway miles per

square mile

Gross state product States , significant

Coughlin, Terza, and

Arromdee (1991)

Highway miles per

square mile

Foreign investment States

Jones (1990) Highway spending per

capita

Employment, personal

income, investment

States , not significant

, some significant

Luce (1990) Highway spending/

personal income

States

Mofidi and Stone (1990) Highway spending/

personal income

Manufacturing

investment and

employment

States , jointly significant

Reynolds and Maki (1990) Highway spending per

capita

New plants Labor market

areas

insignificant

Bartik (1989) Highway miles per

square mile

Small business start-ups States , not significant

McGuire and Wasylenko

(1989)

Highway spending per

capita

Employment States , not significant

Carlino and Mills (1987) Highway miles per

square mile

Population, employment Counties , significant

Place (1986) Highway spending States , not signficant

Bartik (1985) Highway miles per

square mile

New plants States , significant

Helms (1985) Highway spending/

personal income

Personal income States , significant

arch/April 1997 New England Economic Review 55


4/30

fect is negative, given the current stock. Or it may beat greater highway expenditures are responses to

eteriorating highways and represent declining high-ay service. Or highway service may influence devel-

pment only at the state level, and not locally. Thelationship between government spending and the

tual quantity and quality of public service provideda crucial one. Similarly, spillovers can be substantialr many public services, suggesting both that onerisdictions service or spending can seriously under-timate the total service available and that the valueone jurisdictions public service may depend on the

ecisions of others. These are some of the issuesonsidered in the second section of this paper.

Of all the public services

examined for an influenceon economic development,transportation services, andhighway facilities especially,

show the most substantialevidence of a relationship.

It is also interesting to compare Dalenberg andartridges (1995) results to those reported by Luce

994). Both are relatively recent studies of highwayfects on local economic development, measured by

mployment in aggregate and for specific sectors. Bothse appropriate empirical techniques to test and cor-ct for possible endogeneity between governmentrvice and economic development. Both take accountthe entire public sector budget by including the

xhaustive set of government spending and taxingtegories. The studies differ in at least two importantays. Luce examines the effects of fiscal policy on

mployment among localities in a single metropolitanea, while Dalenberg and Partridge examine the

fects among different metropolitan areas. Luce mea-ures transportation service by bothpublic works (andublic safety) spending and a measure of physicalcess to transportation facilities, while Dalenberg and

artridge use highway spending as a fraction ofersonal income. Even in an area with relatively goodgreement about results, and with such care by thesearchers, dramatically different results arise: Luceports positive effects of public works spending (not

significant) and access (significant) on employment,while Dalenberg and Partridge find a significant neg-ative relationship between highway spending andemployment.1

Public Safety

The structures of the nine studies of the effect ofpublic safety services on development reported inTable 3 are more similar than those for the transpor-tation studies, but the results are slightly less consis-tent. Less than one-half of the study results indicate asignificant positive effect of public safety spending oneconomic development, as measured by changes inemployment or new investment. As with transporta-tion, the influence of public safety spending seems tobe more important for some industries than others;Luce (1994) and Papke (1991) report substantial differ-

ences in effects both within manufacturing and be-tween manufacturing and other sectors. (See Table 5,below). Among the difficult to explain or less thanobvious results, Luce (1994) finds public safety/publicworks spending very important for employment inwholesale trade but not important for retail employ-ment, while Papke (1991) reports police and fireprotection spending to be positively and significantlyrelated to new investment in publishing but nega-tively and significantly related to investment in elec-tronic component manufacturing.

One major difference from the transportationstudies is that public safety services are in each case

measured by government spending on public safetyand not at all by measures of public safety facilities oractivity. Because environmental circumstances arelikely to vary substantially among localities or metro-politan areas, the difficulty of translating spendinginto services or public safety itself may be severe inthis instance. Devising useful measures of publicsafety also may be problematic. Carlino and Mills(1987), in a study of changes in population and em-ployment among counties, proposed including crimerates as an explanatory variable, in addition to mea-sures of highway density, educational attainment, use

of industrial revenue bonds, and per capita taxes.However, they excluded crime rates from the reportedregressions, noting that Crime rates produced unsta-ble coefficients, presumably because of serious under-

1 I do not mean to criticize these authors specifically. Theirstudies merely illustrate the point that even with good agreementabout techniques and expected results, substantial differences inactual results still occur.



5/30

porting of crimes (Carlino and Mills 1987, p. 44).As with transportation, public safety is another

tegory where the interjurisdictional effects may beubstantial. Jones (1990) examines the effects

spending in various public services categories onhanges in employment, investment, and income

mong states and reports nearly uniformly positivend often statistically significant coefficients on policend fire spending. But Jones calls these results puz-ing. One of his concerns is the uncertain manner byhich police spending is translated into public safety.nother is the fact that most police and fire protection

pending is done by local governments, and spendingay vary substantially within states. He hypothesizesat state averages of police and fire spending may berving as a proxy for a broad set of local governmentrvices.

ducation

Of the three major public services categories re-ewed here, the evidence about a relationship be-

ween economic development and spending on edu-tion is least convincing. Of the 19 studies reviewed,

2 show a positive relationship, but only six studiesport asignificantpositive relationship, and in five ofe 19 the results are highly variable withineach study

(Table 4). One study even reports a significant nega-tive effect of education spending on development. Inshort, the empirical evidence about whether and howeducation influences economic development is quitecloudy.

One possible reason for the uncertainty in these

results is that education service or level is measuredby education spending, and education is one areawhere the relationship between spending and serviceis highly variable. Many differences in circumstancesaffect the educational outcomes that result from agiven level of spending. In addition, education spend-ing decisions primarily are made by local government,and most states show very substantial variation inspending levels (and education levels) among locali-ties. It is not clear, therefore, that the statewide aver-age spending can reflect the educational situation ineach of those local areas. The difference in educational

spending between two cities in any one state often isas great or even greater than the difference betweentwo state averages. Finally, educational outcomes arethe cumulative result of a series of spending decisions,not one. And the final outcomes may not be known formany years after secondary education is completed.

As in the case of public safety, Carlino and Mills(1987) proposed using a specific measure of educa-tional attainment, median years of schooling, rather

Table 3

Selected Studies of Public Safety (Police, Fire) EffectsAuthor(s) and Year Public Safety Measure Development Measure Unit of Observation Results

Tannenwald (1996) Per capita spending Capital spending States , significant

Tannenwald and Kendrick(1995) Per capital spending Capital spending States

and

, significant andnot depending on

functional form

Evans and Karras (1994) Current spending Gross state product States , not significant

Luce (1994) Public safety and

public works

spending per

household

Employment, labor force Local governments , significance varies

by industry

Papke (1991) Per capita spending Plant births States varies by industry

three , two

Jones (1990) Per capita spending Employment, personal

income, investment

States , significant

Bartik (1989) Per capita spending Small busines s star t-ups States , not significant

Deich (1989) Per capita spending Small business startsand branch plants

States , significant

Papke (1987) Per capita spending Capital spending States , not significant



6/30

an just education spending, to explain changes inounty population and employment. Again, however,

ey found it necessary to exclude that measure fromeir analysis, noting that median schooling andmily income cannot both be included in the samegressions (Carlino and Mills 1987, p. 44), most

kely because schooling affects income and income

affects the choice of schooling. Garcia-Mila andMcGuire (1992) used both education spending andmedian years of schooling and reported that both arepositive and statistically significant for gross stateproduct. Including years of schooling reduces themagnitude of the spending coefficient, as one mightexpect. Garcia-Mila and McGuire also noted that ed-

Table 4

Selected Studies of Education EffectsAuthor(s) and Year Education Measure Development Unit of Observation Results

Tannenwald and Kendrick

(1995)

Per capita s pending Capital spending States , significant and not

depending on

functional form

Dalenberg and Partridge

(1995)

Change in spending Employment Metropolitan areas , significant

Evans and Karras (1994) Current spending Gross state product States , significant

Luce (1994) Spending per pupi l Empl oyment, labor force Local gover nments , significant,

labor force

, not significant,

employment

Garcia-Mila and McGuire

(1992)

Spending and median

years of schooling

Gross state product States , significant

Jones (1990) Per capita spending Employment, personal

income, investment

States Varies by period and

development

measure

Luce (1990) Spending/personal

income

Employment States , significance varies

by year

Mofidi and Stone (1990) Spending/personal

income

Manufacturing

investment and

employment

States , significant and not

OhUallachain and

Satterthwaite (1990)

Education percentage

of local spending

Local monthly not significant

Reynolds and Maki (1990) Per capita spending New plants monthly not significant

Bartik (1989) Per capita spending Small business start-ups States , not significant

Carroll and Wasylenko

(1989)

Spending/personal

income

Employment change States , not significant

McGuire and Wasylenko

(1989)

Per capita spending Employment States /, not significant

Testa (1989) Spending per pupil , significant

Nakosteen and Zimmer

(1987)

Per capita spending Firm migration States , significant

Quan and Beck (1987) Per capita spending Wages, employment States , not significant

Place (1986) Per capita spending States , not significant

Helms (1985) Spending/personal

income

Personal income States , significant

Wasylenko and McGuire

(1985)

Spending/personal

income

Employment States , significant



7/30

cation spending may be endogenous to income. Thisbut one example of the possible endogeneity prob-

ms that must be dealt with in this research.

ational Productivity

A related but somewhat distinct body of researchas explored the possible relationship between theock of public capital or infrastructure and nationalctor productivity and resulting economic growth.

rominent in this work are studies by Aschauer (1989)howing that private and public capital are comple-

entary, by Munnell (1990) showing that labor andublic capital also are complementary, and by Garcia-ila and McGuire (1992), who report that both high-ay and education services contribute to economicowth, in addition to labor and private capital. The

mplication is that increases in public capital or infra-

ructure increase the productivity of both privatepital and labor, generating increases in output andcome. Aschauers work, particularly, was viewed as

ramatic and thus controversial, as it showed a veryrge return from investment in public capital. In thisork, what has come to be called the core infrastruc-retransportation and communication facilities,ectric and gas utilities, water and sewer systemsems to exert the greatest effect on productivity andowth, more than public buildings or other forms of

ublic capital.As noted, these results have been controversial

nd have come under attack from both policymakers

nd academics. A number of subsequent studies,cluding Holtz-Eakin (1994), Evans and Karras994), and Garcia-Mila, McGuire, and Porter (1996)

uggest that the previous results were spurious or atast biased by not controlling for important time- orea-specific effects correlated with the level or growth public capital. The results of studies that useatistical techniques to control for these unobservedxed characteristics that also influence growth findmaller or zero returns to additional marginal invest-

ent in public capital.Some analysts have suggested that public capital

creases the productivity of private capital and labor,ut with diminishing returns. Fox and Murray (1993)

particular cite a number of cross-section studieshowing diminishing returns from infrastructure in-estments. Thus, investments in public capital mayave contributed substantially to private productivity

the past, but as the stock of public capital hasown, the effects of additional public infrastructurevestments may be small or nonexistent. Similarly, in

declining areas or regions where the public capitalstock has been allowed to deteriorate substantially, thereturns from new investment or reinvestment may behigh, as reported by Deno (1988).

Finally, it seems entirely possible that time seriesstudies of aggregate private factor productivity using

national data might show a positive effect of publiccapital, even though public capital may not exert ameasurable effect on development in specific states orjurisdictions, what Fox and Murray (1993) call sitedevelopment. Again, if substantial spillovers or ex-ternalities are associated with specific types of infra-structure, then the cumulative effect on the nationaleconomy could be positive even if the effect is insig-nificant in one location. For instance, the installation ofa major communications tower may have little effecton development where the tower is located, but sub-stantial effect on the growth of industries that utilize

the towers services. Similarly, satellites have stimu-lated little site development in space but substantialproductivity growth and output in the communica-tions and entertainment businesses and in the nationaleconomy. Still, studies using national data and fixed-effects techniques (Tatom 1991) find no association.

The fixed-effects studies also have received somecriticism. Munnell argues that differencing the devel-opment measure over a short time period may missany underlying long-run relationship between publicservices and development, such as would seem to bethe case for education. In addition, the fixed-effectsstudies often report inaccurate estimates of labor and

private capital shares of output. Because the estimatedshares for these inputs are too large, some importantfactor still seems unaccounted for.

Differential Impact by Sector

As with the study of tax effects, a major and notunexpected result concerning public services is anapparent differential influence on different industriesor sectors. Unfortunately, the work to date does notappear to have shown any consistent pattern amongthe set of services and industries. For instance, in the

three major studies reported on in Table 5, the resultsare simply all over the place. Comparing the effects ofeducation spending in Dalenberg and Partridge(1995), who report an insignificant positive effect in alldisaggregated sectors, to those in Luce (1994), whoshows an insignificant negative effect in most sectors,one might reasonably conclude that no differentialeffects exist for those broad sectors.

The manufacturing sector in aggregate has been



8/30


9/30

fect of balanced-budget changes. One popular exam-e is a comparison of the effects of increased taxes

sed to finance additional transfer payments withcreased taxes used to finance additional publicorks, public safety, or education services. Another

xperiment simply substitutes public works, public

fety, or education spending for transfers, holdingverall taxes constant. And still another uses a simul-neous increase in all taxes and spending (or perhapsreduction in the surplus).

Several researchers have reportedthe possibility that higher

taxes, when used to increasespending on specific services,

could produce a net positiveeffect on economic development.

Several researchers conducting these statisticalxperiments have reported the possibility that higherxes, when used to increase spending on specificrvices, could produce a net positive effect on eco-

omic development. For instance, Helms (1985, pp.7879), one of the first to find such a result, reported

aising the property tax . . . to allow an increase in

ansfers . . . would lead to a first-year decrease inersonal income; by contrast, a short-run increment

income results from a property tax-financed in-ease in local educational expenditures. . . . After

xamining the effects by sector, however, Luce (1994)uggests that such stimulative effects may occur for

rtain industries but not all. His work suggests thatcreased taxes to fund additional public safety/pub-

c works spending would lead to a net increase inmployment for the manufacturing, service, andholesale trade sectors. But because the increased

pending would be financed by taxes paid by all

ctors, including finance and retail trade, where em-oyment was not estimated to increase in response to

ublic safety spending, the net effect on aggregatemployment would be negative. The similarity ofpending and tax effects is clear from the estimatedasticities: The median inter-area tax elasticity is ine range of0.25 to 0.50, while the positive publicrvices elasticities reviewed here vary from 0.02 to65.

Even if it is true, however, that an increase intaxes to finance some specific additional governmentspending properly estimated shows an increase inaggregate economic growth (however measured), it isnot clear what the policy conclusion should be. If agovernment is already providing the locally efficient

quantity of public services, then obviously simulta-neous increases in revenue and spending wouldreduceresidents welfare (and social welfare, in the absenceof spillover benefits). An empirical finding that simul-taneous increases in revenue and spending raise em-ployment or income, then, can be interpreted twoways. One is that the government is not providing thePareto-efficient quantity of public services, perhapsbecause of a disequilibrium or because of imperfec-tions in the public choice process. The other possibleinterpretation, stressed by Courant (1994), is that anyincreases in employment, investment, or even income

are not welfare-maximizing for the residents. Thus,Courant challenges the fundamental question posedby much of this research by asking: Should the focusbe on employment and investment or on social wel-fare (utility)?

Conceptual Issues in Measuringthe Impact of Public Services

Public Services as Inputs to Production

Bartik (1991) identifies three ways in which public

services might influence economic growth or develop-ment through an impact on business inputs. Publicservices may be an unpriced input to production;expansion of public services may reduce the pricespaid by business for those services; and some publicservices may work to reduce the cost of private inputsused by business. Examples seem obvious. Publichighways provide an input (usually unpriced) tomany businesses; expansion of public airports mayreduce the full price business must pay for air trans-port; and public education may reduce quality-adjusted prices of labor by increasing the supply of

workers of a given quality (either by increasing aver-age skills everywhere or by attracting additionalworkers to a specific location).

In all of these cases, the idea is that the publicinput reduces production costs directly or increasesthe productivity of a private input and thus increasesoutput. The underlying model is one in which firmsare passive recipients of public services. As explainedby Deno (1988, p. 400), firms do not purchase public



10/30

apital on a per unit basis. . . . local governmentsupply public capital to firms in return for a lump sumroperty tax payment. . . . An alternative perspective

one in which firms become demanders of publicrvices or capital and active participants in the public

hoice process. (Firms may not vote directly, but

wners and employees do vote, and firms may pro-de contributions for various purposes.) In that case,

rms do purchase publicly provided inputs at somex price, and those input prices may alter productionchnologies (relative input combinations) as well as

osts and output.Suppose, for example, that a firms output de-

ends on labor, L, private capital, K, as well as publicrvices, G, characterized by the production function

Q Q(L, K, G).

uppose also that public services are financed by a tax

n private capital (or labor), so that the private pro-uction cost is

C PL*L PK*K hG,

here h is the firms tax price of public services orhare of taxes. Given that the objective of the firm is to

inimize costs for every output, if h is assumedonstant the resulting conditions are:

QL/QK PL/PK,

QK/QG PK/h, and

QL

/QG PL/h.

If a firms tax price, h, is low enough the firm mayrefer public services or public capital to either labor

private capital,even if the public input is less produc-ve than the private inputs. In such a case, firms maytually prefer to receive more of the public input, sosubstitution effect is working against increases inrivate investment or employment. A correspondingale effect may be great enough to generate increasesK and L, but that is not guaranteed. The point is thatpublic inputs to business are subsidized sufficiently

nd if firms are given sufficient opportunity to substi-te those public inputs for private ones, output may

crease without an increase in private investment ormployment.The economic growth effect of the public

rvice would be captured by some measures, but notl.2

Public Services and Changes in Consumer Behavior

It also appears possible that provision of publicservices might alter private consumption patterns, assuggested by Clotfelter (1977) and others. For in-stance, it seems clear that creation and expansion of

the interstate highway system had substantial (posi-tive) effects on consumer demand for automobiles andon the automobile industry. To the extent that theproduction of automobiles is unevenly distributedgeographically, the expansion of the automobile in-dustry influenced by public highway investment cre-ated local or regional development benefits.

While highways may present the most dramaticexample, in numerous other less dramatic instancesgovernment services have altered private consump-tion, and indeed in some cases that was the primaryobjective of the government expenditure. A commu-

nity or state may invest heavily in recreational facili-ties, for instance, which will serve to attract visitorsfrom other jurisdictions. In addition, those facilitiesmay induce jurisdiction residents to alter their behav-ior toward increased use of the facilities and increasedconsumption of privately provided complements. Thepublic service or facility may not attract population inthe form of residents, but it may attract additionalprivate spending that contributes to growth of em-ployment or income. In short, public services or gov-ernment spending may influence economic develop-ment through demand-side as well as supply-sideadjustments.

Public Services versus Government Expenditures

One of the potentially most serious difficulties forstudies of the effects of public services on economicdevelopment is the problem of measuring both thequantity and the quality of those public services orfacilities. As already noted, the dominant practice isto use measures of government spending (per capitaor relative to income) to reflect public service levelsor public capital. But economists have long recog-nized that government spending does not necessar-

ily translate directly to additional public output orservice.

Bradford, Malt, and Oates (1969) suggested threealternative ways of characterizing public output:spending on inputs; the directly produced output orfacilities that result from those inputs (such as hours ofpolice patrols); and the consumer output or consump-tion service that results partly from the directly pro-duced output or facility. Suppose directly produced

2 A similar result arises if h is not constant but is a function ofher K or L as a consequence of a wage or property tax. In thatse, the private/public capital equilibrium condition is QK/QG

K dh/dK)/h.



11/30


12/30

ublic Choice Regarding Government Spending

Still another important theoretical and empiricalctor recognized in many of these studies is theossible endogeneity of the fiscal variables, as eco-omic growth is expected to influence the choice of

overnment service levels and thus taxes. This may beparticular concern with transfer payments, which

ave been the focus of several studies and whichearly are expected to be cyclical. A number ofaditional approaches have been used to deal withis potential problem, including two-stage estimationchniques (Dalenberg and Partridge 1995; Luce 1994),strumental techniques (Helms 1985), and use ofgged variables (Coughlin, Terza, and Arromdee

991; Quan and Beck 1987), as well testing for exoge-eity using the usual methods, such as the Hausmanst (Mofidi and Stone 1990).

One result from the literature examining demandr public services implies that some of the concernbout endogeneity may be alleviated. In addition to

e positive effect of economic growth on consumeremand for public services, Hamilton (1983) andhers have argued that income growth may be actorreducingthe marginal cost of public services. Forstance, the inputs necessary to achieve a given levelpublic safety (police and fire protection) may fall as

onsumers alter their private consumption with in-easing income. A similar case might be made for

ducation, as consumers purchase more educationalputs privately as income rises. If this is the case, then

sing income (or other forms of economic develop-ent) would increase the demand for services butduce the cost of providing those services; as a result,pendituresmay not rise (or at least not rise as muchotherwise).

xternalities and Public Goods

A final issue concerns the relationship betweene value- or productivity-enhancing effects of publicrvices provided by one locality or state and the

ublic services provided by other similar jurisdictions.

seems obvious that the value of an interstateghway in one state depends on the existence ofmilar and interconnected highways in other stateshus the idea for a federal grant program). Someuthors consider possible externalities; for example,alenberg and Partridge (1995, p. 631) note a possi-lity that positive infrastructure spillover effectsross government boundaries offset the negative re-tionship observed at the metropolitan level. How-

ever, if the possibility is handled empirically, usuallyit is only through some time-specific effect that allowsfor aggregate changes in public spending or otherfactors. This issue is also considered by Munnell (1990)and Holtz-Eakin (1994).

An Unfinished Research Agenda

Is there any systematic evidence that these ana-lytical issues affect the empirical results substantiallyand consistently? Or any reason to believe that certainof these analytical issues are more important thanothers? Phillips and Goss (1995) have conducted aregression analysis (called a meta-analysis) in whichthe estimated tax elasticities from some 84 studies are

More attention needs to be paid tothe normative, welfare issues

related to economic development.The key is to define precisely the

objectives of both developmentpolicy and public service provision.

regressed on variables representing the technical char-

acteristics of the studies. The analysis provides twoimportant findings relevant to the issues discussedabove. First, they report that estimated tax elasticitiesdiffer substantially, depending on whether public ser-vices effects are included. Combining this finding withthat of Mofidi and Stone (1990) suggests that allowingfor the full range of government fiscal decisions iscrucial. Second, they report that most differences inanalytical technique (other than the inclusion of publicservices) do not lead to substantial differences inresults, with the possible exception of failing to controlfor fixed effects.

These findings about technique by Phillips andGoss relate to the question: How does one obtainimproved estimates of the effects of public services oneconomic development? In addition to including thefull range of fiscal policies and controlling for fixedeffects, I would also suggest that more attention bepaid to the issue of interjurisdictional externalities.One can think of this from two perspectives. First, onemight explore how public services levels in one juris-



13/30

ction are affected by public spending or capital inher jurisdictions. Alternatively, one might argue thatonomic development in one jurisdiction depends onegional public services. If the focus really should be

n service level (rather than capital or spending), thenese interjurisdictional effects cannot be ignored. The

ducational level of the population in a specific juris-ction obviously has been influenced by educational

pending or services in the past in many jurisdictions.I would suggest, however, that the more impor-

nt research question may be different: What shouldne do with an estimate of the marginal effects ofublic services on economic development, assumingat a set of good estimates can be obtained? Sup-

ose, for example, that a high-quality study showsat simultaneous increases in taxes and in school

pending have no effect on employment in the state orcality. I suspect that very few would conclude that

hool spending should, because of this result, be elim-ated or reduced substantially. It might be that edu-tion ultimately affects earnings rather than employ-ent, or perhaps any employment increase occurs in afferent jurisdiction, or perhaps education increasesility, if not income.

More attention needs to be paid, then, to theormative, welfare issues related to economic devel-pment.3 In this, I wish to associate myself with theews of Paul Courant (1994) mentioned earlier. The

ey issue is to define precisely the objectives of bothevelopment policy and public services provision.uppose that a simultaneous increase in taxes and

pending does lead to economic development (mea-ured by employment or investment). As noted ear-er, this may reflect inefficiency in public provisionitially, in which case additional public services are

ppropriate. Alternatively, it may reflect a move awayom the efficient level of public services, even thoughrtain groups might benefit. For instance, it has

ometimes been suggested that at least an implicitoal of development policy is to increase land values.nd landowners may not be residents. Other possibil-es seem clear. If employment increases, who getsose jobs? Does an increase in employment occur

simultaneously with an increase in income (higherlabor demand), or does the increased employmentarise from increased supply and lower wages?

From this viewpoint, it seems very important tounderstand who benefits from the various forms ofeconomic development; that is, the distributional

implications of using fiscal policy for development, asopposed to allocative, purposes. I take this opportu-nity, then, to renew a research suggestion I haveadvanced before. It is important to develop betterestimates (or in some cases initial estimates) of theincidence of public sector expenditures or services. Itseems apparent, at least to this public finance econo-mist, that tax incidence has received a disproportion-ate amount of attention from scholars, compared toexpenditure incidence. Only when we have a clearunderstanding of who benefits from changes in publicservices will we be able to evaluate carefully the

welfare implications of economic development thatarises from attractive public services.Finally, as researchers, we also need to be sensi-

tive to how our work is interpreted and subsequentlyused by policymakers. In one study of the effects ofpublic services on development published by a veryrespectable academic journal in the 1990s, the authorincludes the following summary sentence: The esti-mates . . . suggest that educational services are the onlyproductive current government service.4 I suspect thatthe author was referring narrowly to the specificeconometric estimates, but in almost any broadercontext this statement seems silly. The author then

states: Without evidence that government capital doesindeed generate empirically important direct nonmarketconsumption services . . . , that is, evidence that localresidents benefit from local public capital, providingmore is not appropriate.What about the simple fact thatresidents often vote for more spending or facilities orthat individuals vote with their feet in favor of com-munities with good public services? In short, thereis ample opportunity for these econometric estimatesto be misunderstood and misused, and researchersshould not encourage that tendency.

3 Remember that many localities enact provisions to limitvelopment, explicitly.

4 As this is a general point rather than specific comment aboutthat paper, the identity is not disclosed here.



14/30

eferences

nderson, John E. and Robert W. Wassmer. 1995. The Decision toBid for Business: Municipal Behavior in Granting Property TaxAbatements.Regional Science and Urban Economics, vol. 25, no. 6(December), pp. 73957.chauer, David Alan. 1989. Is Public Expenditure Productive?

Journal of Monetary Economics, vol. 23, pp. 177200.rtik, Timothy J. 1985. Business Location Decisions in the UnitedStates: Estimates of the Effects of Unionization, Taxes, and OtherCharacteristics of States. Journal of Business and Economic Statis-tics, vol. 3, no. 1 (January), pp. 1422.. 1989. Small Business Start-Ups in the United States: Esti-mates of the Effects of Characteristics of States.Southern EconomicJournal, vol. 55, no. 4 (April), pp. 100418.. 1991.Who Benefits from State and Local Economic DevelopmentPolicies?Kalmazoo, MI: W.E. Upjohn Institute.. 1994. Jobs, Productivity, and Local Economic Develop-ment: What Implications Does Economic Research Have for theRole of Government? National Tax Journal, vol. 47 (December),pp. 84761.. 1995. Taxes and Local Economic Development: What DoWe Know and What Can We Know? In Proceedings of theEighty-Seventh Annual Conference. Columbus, Ohio: National TaxAssociation.adford, David F., R.A. Malt, and Wallace E. Oates. 1969. TheRising Cost of Local Public Services: Some Evidence and Reflec-tions.National Tax Journal, vol. 22 (June), pp. 185202.rlino, Gerald A. and Edwin S. Mills. 1987. The Determinants of

County Growth. Journal of Regional Science, vol. 27, no. 1, pp.3954.rroll, Robert and Michael Wasylenko. 1989. The Shifting Fate of

Fiscal Variables and Their Effect on Economic Development. InProceedings of the Eighty-Second Annual Conference on Taxation.Columbus, OH: National Tax Association.. 1994. Do State Business Climates Still Matter? Evidence ofa Structural Change.National Tax Journal, vol. 47, no. 1 (March),pp. 1937.otfelter, Charles T. 1977. Public Services, Private Substitutes, andthe Demand for Protection Against Crime. The American Eco-

nomic Review, vol. 67 (December), pp. 86777.oughlin, Cletus C., Joseph V. Terza, and Vachira Arromdee. 1991.State Characteristics and the Location of Foreign Direct Invest-ment in the United States. The Review of Economics and Statistics,vol. LXXIII, pp. 67583.

ourant, Paul N. 1994. How Would You Know a Good EconomicDevelopment Policy if You Tripped Over One? Hint: Dont JustCount Jobs.National Tax Journal, vol. 47 (December), pp. 86381.

alenberg, Douglas R. and Mark D. Partridge. 1995. The Effects ofTaxes, Expenditures, and Public Infrastructure on MetropolitanArea Employment. Journal of Regional Science, vol. 35, no. 4(November), pp. 61740.

eich, Michael. 1989. State Taxes and Manufacturing Plant Loca-tion. In Proceedings of the Eighty-Second Annual Conference onTaxation.Columbus, OH: National Tax Association.eno, Kevin T. 1988. The Effect of Public Capital on U.S. Manu-

facturing Activity: 1970 to 1978. Southern Economic Journal, vol.55, pp. 40011.uffy-Deno, Kevin T. and Randall W. Eberts. 1991. Public Infra-structure and Regional Economic Development: A SimultaneousEquations Approach. Journal of Urban Economics, vol. 30, pp.32943.erts, Randall W. 1991. Some Empirical Evidence on the Linkagebetween Public Infrastructure and Local Economic Develop-ment. In Henry Herzog and Alan Schlottmann, eds., IndustryLocation and Public Policy. Knoxville, TN: University of TennesseePress.erts, Randall W. and Joe A. Stone. 1992. Wage and Employment

Adjustment in Local Labor Markets. Kalamazoo, MI: The W. E.Upjohn Institute.

Evans, Paul D. and Georgios Karras. 1994. Are GovernmentActivities Productive? Evidence from a Panel of U.S. States. TheReview of Economics and Statistics, vol. 76, pp. 111.

Fisher, Ronald C. 1996. State and Local Public Finance. Chicago:Richard D. Irwin.

Fisher, Ronald C. and John C. Navin. 1992. State-Local FiscalBehavior: Analysis of Interjurisdictional Differences, 19621987.Public Finance Quarterly, vol. 20 (October), pp. 43349.

Fox, William F. and Matthew N. Murray. 1993. State and LocalGovernment Policies. In David Barkley, ed., Economic Adaptation:Alternatives for Rural America. Boulder, CO: Westview Press, Inc.

. 1997. Intergovernmental Aspects of Growth and Stabiliza-tion Policy. In Ronald C. Fisher, ed., Intergovernmental FiscalRelations. Boston: Kluwer Academic Publishers, forthcoming.

Garcia-Mila, Teresa and Therese J. McGuire. 1992. The Contribu-tion of Publicly Provided Inputs to States Economies. RegionalScience and Urban Economics, vol. 22, pp. 22941.

Garcia-Mila, Teresa, Therese J. McGuire, and Robert H. Porter. 1996.The Effect of Public Capital in State-Level Production FunctionsReconsidered.The Review of Economics and Statistics, vol. LXXVIII(February), pp. 177 80.

Hamilton, Bruce W. 1983. The Flypaper Effect and Other Anoma-lies. Journal of Public Economics, vol. 22 (December), pp. 34761.

Helms, L. Jay. 1985. The Effect of State and Local Taxes onEconomic Growth: A Time SeriesCross Section Approach. TheReview of Economic and Statistics, vol. LXVII (November), pp.57482.

Holtz-Eakin, Douglas. 1994. Public-Sector Capital and the Produc-tivity Puzzle. The Review of Economics and Statistics, vol. 76, pp.1221.

Johnson, George E. 1970. The Demand for Labor By EducationalCategory.Southern Economic Journal, vol. 37 (October), pp. 190204.

Jones, Bryan D. 1990. Public Policies and Economic Growth in theAmerican States.Journal of Politics, vol. 52, no. 1 (February), pp.21933.

Luce, Thomas F., Jr. 1990. The Determinants of Metropolitan AreaGrowth Disparities in High-Technology and Low-TechnologyIndustries. Working Paper. Department of Public Administra-tion, Pennsylvania State University.

. 1994. Local Taxes, Public Services, and the Intrametropoli-tan Location of Firms and Households. Public Finance Quarterly,vol. 22, no. 2 (April), 13967.

McGuire, Therese J. and Michael Wasylenko. 1987. EmploymentGrowth and State Government Fiscal Behavior: A Report onEconomic Development for States from 1974 to 1984. New JerseyState and Local Expenditure and Revenue Policy Commission,July.

Mofidi, Alaeddin and Joe A. Stone. 1990. Do State and Local TaxesAffect Economic Growth? The Review of Economics and Statistics,vol. 72, no. 4 (November), pp. 68691.

Munnell, Alicia H. 1990. Why Has Productivity Growth Declined?

Productivity and Public Investment. New England Economic Re-view, January/February, pp. 322.. 1992. Infrastructure Investment and Economic Growth.

Journal of Economic Perspectives, vol. 6, pp. 18998.Nakosteen, Robert A. and Michael A. Zimmer. 1987. Determinants

of Regional Migration by Manufacturing Firms. Economic In-quiry, vol. 25, no. 2 (April), pp. 35162.

Noiset, Luc and William Oakland. 1995. The Taxation of MobileCapital by Central Cities.Journal of Public Economics, vol. 57, no.2 (June), pp. 297316.

OhUallachain, Breandan and Mark A. Satterthwaite. 1990. SectoralGrowth Patterns at the Metropolitan Level: An Evaluation of



15/30

Economic Development Incentives. Discussion Paper No. 29, J.L.Kellogg Graduate School of Management, April.pke, Leslie. 1987. Subnational Taxation and Capital Mobility:Estimates of Tax-Price Elasticities. National Tax Journal, vol. 40(June), pp. 191203.. 1991. Interstate Business Tax Differentials and New FirmLocation: Evidence from Panel Data. Journal of Public Economics,vol. 45, pp. 4768.

. 1995. Discussion. In Proceedings of the Eighty-SeventhAnnual Conference.Columbus, OH: National Tax Association.rtridge, Mark. 1993. High-Tech Employment and State EconomicDevelopment Policies.The Review of Regional Studies, vol. 23, pp.287306.illips, Joseph M. and Ernest P. Goss. 1995. The Effect of State and

Local Taxes on Economic Development: A Meta-Analysis. South-ern Economic Journal, vol. 62, no. 2 (October), pp. 32033.ace, Frank. 1986. The Relationship of State and Local Govern-ment Spending and Taxing to Economic Performance: An Econo-metric Analysis of the States from 1972 to 1984. WisconsinDepartment of Development, Division of Policy Development,Bureau of Research Report No. RP-86-7, September.uan, Nguyen T. and John H. Beck. 1987. Public EducationExpenditures and State Economic Growth: Northeast and SunbeltRegions.Southern Economic Journal, vol. 54, pp. 36167.ynolds, Paul D. and Wilbur Maki. 1990. U.S. Regional Charac-teristics, New Firms, and Economic Growth. Working paperpresented to the Cross-National Workshop on the Role of Small,

Medium Enterprises in Regional Economic Growth at the Univer-sity of Warwick, Coventry, United Kingdom, March 28.

Tannenwald, Robert. 1996. State Business Tax Climate: HowShould It Be Measured and How Important Is It? New EnglandEconomic Review, January/February, pp. 2338.

Tannenwald, Robert and Christine Kendrick. 1995. Taxes andCapital Spending: Some New Evidence. In Proceedings of theEighty-Seventh Annual Conference. Columbus, OH: National Tax

Association.Tatom, John A. 1991. Public Capital and Private Sector Perfor-mance. Federal Reserve Bank of St. Louis Review, May/June, pp.315.

Testa, William A. 1989. Metro Area Growth from 1976 to 1985:Theory and Evidence. Working Paper. Federal Reserve Bank ofChicago, January.

Wasylenko, Michael J. 1991. Empirical Evidence on InterregionalBusiness Location Decisions and the Role of Fiscal Incentives inEconomic Development. In Henry Herzog and Alan Schlott-mann, eds., Industry Location and Public Policy. Knoxville, TN:University of Tennessee Press.

. 1995. Has the Relationship Changed Between Taxes andBusiness Location Decisions? In Proceedings of the Eighty-SeventhAnnual Conference.Columbus, OH: National Tax Association.

Wasylenko, Michael and Therese McGuire. 1985. Jobs and Taxes:The Effect of Business Climate on States Employment GrowthRates. National Tax Journal, vol. 38, no. 4 (December), pp.497512.

Discussionmothy J. Bartik, Senior Economist, W.E. Upjohn

stitute for Employment Research.

Both Michael Wasylenko and Ronald Fisher pro-vide good reviews of the research literature onhow state and local fiscal systems affect eco-

omic development. My comments will highlight theain themes in their reviews and give my own

erspective on these themes. My emphasis will be onow research can better inform policymakers.

Main Themes of Wasylenko and Fisher

Wasylenkos and Fishers reviews have fiveemes that I wish to highlight. First is the recurringeme that research on state and local fiscal systems

nd economic development often results in quiteagile results. Equally competent research projectsay get widely divergent estimates of the economic

evelopment effects of fiscal variables. Second, the

fiscal variables in this research are difficult to measure.The measurement difficulties are particularly acute forpublic services variables. Third, the research faceseconometric difficulties because measured fiscal vari-

ables are often endogenous, in that they might beaffected by economic development.

Fourth, despite these problems, they find someconsensus that tax and public services variables dohave some effects on state and local economic devel-opment. Wasylenko argues that tax studies suggest anelasticity of0.2 for interregional studies, and elastic-ities at least four times as great for intra-regionalstudies. Fisher points out that many studies find thatat least some public services do matter for state andlocal economic development.

Fifth, both Wasylenko and Fisher emphasize that

a fiscal systems effect on economic development is anincomplete measureof its true social benefit. A completecost-benefit analysis should consider who pays forpublic services and who benefits from them, who getsthe jobs, and the effects of economic development onthe local budget.

I will add a few of my own comments to thesevarious themes, and suggest how more research mighthelp policymakers.



16/30

he Fragility of Estimates andonsensus Findings

As mentioned by Wasylenko, my position is thate interregional tax elasticity is between 0.1 and0.6, which seems consistent with Wasylenkos con-

usion. I also agree with Wasylenko and most othersearchers that the intra-regional tax elasticity is

onsiderably greater, perhaps in the range from 1.03.0.

It is true that as interregional studies add inontrols for fixed regional effects and public services,

e estimated tax elasticity often increases in absolutealue. Fixed-effect controls are particularly influentialn estimates. Some might interpret this as establish-g that elasticities closer to 0.6 are more plausiblean elasticities close to 0.1. I have argued in the

ast that fixed-effect estimates have econometric ad-

antages. However, few studies have both fixed-effectnd public services controls, which means that oneust be wary about endorsing these greater elasticity

umbers.

To convince skeptics that taxesand public services affect economic

development, we need morestudies that rely on natural

experiments, examining theresponse to large changes in taxor expenditure regimes such asProposition 13 in California or

Kentuckys school reforms.

In addition, I have become more concerned incent years about measurement and endogeneity

roblems in many studies that use fixed-effect con-

ols. Studies that use fixed-effect controls often useoarser data on taxes and public services. Fixed-effectudies must generate multiple observations for eachcal economy in order to allow estimation. Thiseans that one must use the tax and public services

ata most readily available, which are data on reve-ues and expenditures. But current revenue and ex-enditure per capita, or as percentages of personalcome, are poor measures of tax rates or public

services. Furthermore, fiscal measures using currentrevenue and expenditure are probably highly endog-enous. I will discuss these measurement and endoge-neity issues later in these comments.

The fragility of estimates has led some to beskeptical of the consensus estimates of tax elasticities

(for example, see McGuire 1992). To convince skepticsthat taxes and public services affect economic devel-opment, we need more studies that rely on naturalexperimentsstudies that observe how similar localeconomies or firms respond to large, exogenouschanges in tax regimes or public services. With largechanges in tax rates or public services, measurementissues and endogeneity issues become less of a con-cern. If we can compare similar local economies orfirms, concerns about omitted variable bias are re-duced.

To be specific, we need studies that will examine

the response to large changes in tax or expenditureregimes, such as Proposition 13 in California, Propo-sition 212 in Massachusetts, the Engler administra-tions shift from property to sales taxation in Michi-gan, or Kentuckys school reforms. To avoid problemscaused by unobservable differences in state econo-mies, we could focus on comparing individual coun-ties economic development in states that experiencedthese regime shifts, to development in similar countiesin states that did not experience these regime shifts.Research by Isserman and his colleagues indicatesways in which similar counties might be identified(Isserman 1994). To control further for unobservables,

the research might analyze differences across indus-tries to see if they are consistent with the expectedimpacts of such shifts in fiscal regime. For example,one expects capital-intensive industries to be espe-cially responsive to property taxes, and industriesusing skilled labor to be especially responsive toschool quality.

The best examples of this kind of research are apaper by Hines (1996) and another by Holmes (1995).Hiness study is the most convincing evidence to datethat taxes matter to business location decisions. Hecompares the effects of taxes on business location for

firms from countries that provide full tax credits forU.S. taxes paidthus making state and local taxesirrelevant to the location decisionto the locationeffects of taxes for foreign firms from countries thatallow only a deduction for U.S. taxes before profits aretaxed in the home country. Holmess study comparesjob growth in counties on the border between right-to-work and non-right-to-work states with job growthin counties in the interiors of the two groups of states.



17/30

Hiness study finds that firms from countries thatnly allow deductions for U.S. taxes paid are morensitive to state and local taxes in their location

ecisions than firms from countries that allow taxedits for U.S. taxes. This finding is hard to explain

nless taxes, or some state characteristic that is highly

orrelated with taxes, really do affect location deci-ons. Holmess study finds large growth advantagesr right-to-work over non-right-to-work states for

ounties near the border between the two groups ofates, compared to counties in the interiors of the twooups. This finding is hard to explain unless right--work laws, or some state characteristic that isrongly correlated with right-to-work laws, really dofect economic development.

Measuring Public Services

Accurately measuring the quality and quantity ofublic services is difficult. Consider education. I sus-ect the difficulty in measuring education quality isne reason that some studies, as Fisher points out, doot find education to have significant effects on eco-omic development. It is difficult to believe that cur-nt expenditure on education is the measure of

ducation quality that is most relevant to businessecision-making. Education should be most relevant

businesses because it affects the productivity of theorkers that the business can hire. The current qualitythe work force depends on the quality of education

ver a lengthy history. Furthermore, education qualityepends only in part on public sector inputs; othereterminants include the socioeconomic characteris-cs of the students. Finally, the gross amounts ofublic sector inputs, such as money or number ofachers per student, are only part of the public sector

ontribution to educational quality; one also has toonsider the effectiveness with which these public

ctor resources are deployed.Fisher suggests that including fixed effects for

cal jurisdictions may control for the omitted vari-bles, other than current public spending, that affect

e quality of state and local public services. Fixedfects may help, but I am not confident that they re-

olve this problem. We need studies that measureublic service quality more carefully from a businesserspective. We need studies that focus on large exog-nous changes in public service quality. For example,ne might consider states that have undertaken sig-ficant education reforms, and carefully measure

ow work force quality has changed as a result.

The Endogeneity Issue

The endogeneity of state and local fiscal variablesis particularly acute for variables that are measuredusing actual revenues and expenditures, becauserevenues and expenditures are directly and immedi-

ately affected by local economic development. Endo-geneity problems are reduced when researchers usemeasures such as tax rates, or serious measures ofpublic service quality. Tax rates and public servicequality are of coursepolitically endogenous, in that localeconomic development trends will influence policy-makers choices about tax rates and public servicequality. But the response of tax rates and publicservice quality to local economic development maybe long delayed, making tax rates and public servicequality less endogenous than actual revenues andexpenditures.

Because of both the measurement problems andthe endogeneity problems, I have become increasinglyskeptical of the budget constraint approach used byHelms (1985) and other researchers (for example,Bartik 1996a). Defining all tax and public servicevariables as percentages of personal income, or percapita, almost ensures that all fiscal variables will bepoorly measured. It makes no sense to scale all tax andpublic service variables by the same denominator,whether the denominator is population or personalincome. For example, as Wasylenko points out,shouldnt education quality depend more on schoolspending per student than on spending per person or

as a proportion of personal income? In addition,public service quality may, as noted above, be onlyloosely related to current expenditure. Finally, all thecurrent revenue and expenditure variables in theHelms approach are highly endogenous. One canuse instrumental variables to try to correct for this, butit is difficult to find good instruments that will beconvincing to other researchers.

Why Do We Care About HowState and Local Fiscal VariablesAffect Economic Development?

As noted by Wasylenko and Fisher, we should beclear about why the effect of state and local fiscalvariables on economic development is important forsocial well-being. I will argue that the relationshipbetween fiscal variables and economic development isimportant for public policy for four reasons. Researchshould be more focused on how fiscal variables affect



18/30

onomic development in ways that are relevant forublic policy.

mployment Growth

I argue that we care about local employment

owth because wages often do not clear labor mar-ets. Wages often exceed the opportunity cost of labor;udies by Gordon (1973) and Jones (1989) suggest thate lowest wage at which the typical unemployeddividual is willing to workthe reservationageaverages about 85 percent to 90 percent of hisher previous wage. Because wages often exceed the

pportunity cost of labor, creating jobs in a local laborarket usually provides social benefits.

I have argued that the social benefits of jobeation probably persist in the long run (Bartik 1991).anchard and Katz (1992) argue that the effects of

mployment growth on local employment rates dieut after five years. Bartik (1993) uses Blanchard andatzs data to argue that their model is misspecified.

When their data are used in a more general model, thefects of local employment growth on labor forcearticipation persist for at least 17 years.

We care about local employmentgrowth because wages often

do not clear labor markets,

and creating jobs in alocal labor market usuallyprovides social benefits.

The social benefits of local employment growthrobably are higher in local economies with highnemployment, because reservation wages will bewer. These social benefits of job creation may be

onexistent in low unemployment areas, where reser-ation wages may be close to market wages. If the

osts of job creation are similar across different localonomies, then high unemployment areas shouldore vigorously pursue economic development. Bute know little about whether the costs of creatingbs through tax reductions or public service expan-ons are different in different local areas. As arguedy Courant (1994), we need more research on thelative effectiveness of development policies in dif-rent local economies.

What implications do the benefits of local em-ployment growth have for the principle that localpublic services levels should be chosen so that mar-ginal benefits of public services equal marginal costs?(This issue is raised by Ronald Fisher.) This principlestill applies, except that we must consider an addi-

tional category of benefits and costs. Traditionally wehave considered just the direct effects of public ser-vices and taxes on household well-being. We nowmust also consider their indirect effects on householdwell-being, through their influence on employmentgrowth.

As Wasylenko points out, a complete cost-benefitanalysis of local fiscal policy should also consider therevenue effects of local employment growth. I wouldadd that the analysis should consider the effects ofgrowth on public expenditures. Research suggests thatlocal employment growth is often a fiscal drain, once

one considers the marginal capital expenditure re-quired to deal with the resulting household in-migra-tion (Bartik 1996b).

Employment Growth by Industry

The benefits of local employment growth vary byindustry. Growth in high wage-premium industriesprobably has greater effects on labor force participa-tion and earnings (Bartik 1996c). We need more re-search on how the effects of state and local fiscalvariables differ across industries. My own experienceis that industry-specific estimates often are imprecise

or have implausible magnitudes. I suspect that simi-lar findings are buried in the unpublished computerprintouts of many other researchers.

Productivity

We want to know the effects of state and localfiscal variables on productivity, even if local incomeper capita is unchanged. The usual assumption ineconomics is that, in the long run, local or sector-specific increases in productivity are spread over theentire economy. We need more research that looks at

how local policies affect productivity.

Externalities

As highlighted by Fisher, a key issue is whetherthe public services of one state or local area providespillover benefits for nearby state or local areas. If so,then state and local fiscal decisions will likely beinefficient without federal intervention. This issue has



19/30

ecome even more important given recent trendsward a reduced federal role in providing publicrvices. To my knowledge, only Bartik (1996a) has

xamined this issue. This study found evidence thatate public services cause positive spillovers on man-facturing output in nearby states. More research

hould examine this issue.

onclusion

In sum, current research on fiscal policies andcal economic development suffers from three key

roblems: serious measurement difficulties; a lack of focus on the issues that are crucial for

public policy; and

endogeneity of the fiscal variables.New research is needed to address these prob-

lems. We need research that is more careful aboutmeasuring public service quality and tax rates. Weneed research that provides more specific evidence onhow fiscal variables affect different industries and

different local economies. We need more research thatlooks at effects of fiscal variables on productivity andat spillover effects on nearby state and local areas.Finally, research on fiscal variables and economicdevelopment will be more convincing if it can exploitlarge, natural experiments in varying state tax andpublic service policies.

These types of new research will require muchmore time spent on data construction. The payoff forthis extra time will be empirical results that are moreuseful for policymakers.

eferences (those not listed by Fisher or Wasylenko)

rtik, Timothy J. 1996a. Growing State Economies: How Taxes andPublic Services Affect Private-Sector Performance. Report. Washing-ton, DC: Economic Policy Institute.. 1996b. Strategies for Economic Development. In J. RichardAronson and Eli Schwartz, eds., Management Policies in LocalGovernment Finance (4th Edition), pp. 287312. Washington, DC:International City/County Management Association Press.. 1996c. The Distributional Effects of Local Labor Demandand Industrial Mix: Estimates Using Individual Panel Data.Journal of Urban Economics, vol. 40, pp. 15078.ordon, Robert J. 1973. The Welfare Cost of Higher Unemploy-

ment.Brookings Papers on Economic Activity, No. 1, pp. 133205.

Holmes, Thomas J. 1995. The Effects of State Policies on theLocation of Industry: Evidence from State Borders. FederalReserve Bank of Minneapolis, Staff Working Paper No. 205.

Isserman, Andrew. 1994. A Family of Geographical Control GroupMethods for Regional Research. Research Paper No. 9436, Re-gional Research Institute, West Virginia University.

Jones, Stephen R.G. 1989. Reservation Wages and the Cost ofUnemployment.Econometrica, vol. 56 (May), pp. 22546.

McGuire, Therese. 1992. Review of Who Benefits From State andLocal Economic Development Policies? National Tax Journal(December), pp. 45759.

Discussionarley T. Duncan, Executive Director of the FederationTax Administrators. The views reflected in thesemments are the authors and should not be construedreflecting the policies or opinions of the Federation

its members.

The papers by Michael Wasylenko and RonaldFisher have as their principal aim three objec-tives: first, to review trends in the research

xamining the relationships between economic devel-pment and state and local tax policy (Wasylenko)nd expenditure or fiscal policy (Fisher); second, to

identify the major conceptual and methodologicaldifficulties involved in such research; and third, tosummarize the key findings of the research in afashion that will be meaningful to policymakers.

In my estimation, as one not steeped deeply in theeconomics and measurement issues of such research,Wasylenko and Fisher have succeeded on all three

counts. In so doing, they have made the results ofextensive, sophisticated research understandable tothose in a position to act on its results. At the sametime, the research they describe has a strong tendencyto leave policymakers in a quandary about how toproceed. Wasylenko concludes that taxes, and partic-ularly specific incentives, do not have much impact onlevels of economic development. Yet, the hottesttopic in state legislatures is what sort of tax incen-



20/30


21/30

lan Peters in this volume is an excellent first step inis regard, but in the design of future research the

eeds of policymakers should be kept foremost inind.

aising the Discussion to the Next Level

The question then becomes what policy prescrip-ons (or at least suggestions) we can draw from theork of Wasylenko and others in trying to evaluatex policy and economic development. In my estima-

on, at least two policy implications can be drawn.With all due modesty, I would suggest that the

Wasylenko results buttress arguments I made in anarlier paper, in which I suggested that the manner inhich states compete on the tax front could be dividedto three categoriesThe Good, the Bad and thegly (Duncan 1992). The good category consists of

eneral competition among the states for the lowestvel of tax burden consistent with the provision of a

I agree . . . that competitionamong the states is self-regulatingto a considerable degree, without

any dumbing-down or race-to-the-bottom effect. States are more

than capable of regulating theirown activities and establishingtheir own policies.

esired level and quality of public services. And, inct, most of the studies reviewed do suggest thatwer overall tax levels are statistically related togher levels of economic development, although thesults are varied and generally small. Nonetheless,is finding argues that the general tax level is the

ppropriate area of competition for states, if there is toe competition. This conclusion is similar to the Con-oy Theory of competition espoused by John Shan-on (1994), a keen student of intergovernmental fiscallations.2 And as Wasylenko notes, the trend amonge states to lower both individual and corporatearginal tax rates is emblematic of this effect.

In the bad category, I placed tax competition ine form of various tax incentives aimed at stimulating

investment and job creation. I argued that such incen-tives were disruptive of horizontal equity and had avery high likelihood of subsidizing decisions thatwould otherwise be made anyway, rather than stim-ulating new activity. The research done to date sup-ports this thesis. The research goes further to indicate

that most jobs do not go to intended beneficiaries.In the final ugly category, I placed tax compe-

tition in the form of bidding wars among the states forindividual company and facility locations. In my view,such bidding wars meet no test of sound tax policyand breed mistrust of government institutions. Wa-sylenkos review of the research does not directlyaddress this form of competition.

In short, the Wasylenko paper effectively mar-shals the arguments supporting a position that statesare best off competing on the basis of overall taxpolicies and levels rather than through specific incen-

tives.3

Furthermore, his suggestion that states avoidusing specific tax incentives when in fact overall taxreform is required, is also a message to which thestates should pay attention.

The second policy prescription to be drawn fromthe analysis is the need for states to place tax incen-tives in the context of an overall development policyand to subject such incentives to periodic analysis andevaluation. Given the limited impact that can beexpected from such incentives, it seems prudent toreview their utility regularly. I suggested several partsof such a policy analysis in my earlier piece (1992).Likewise, Dabson, Rist, and Schweke (1996) have

suggested a broad policy framework in which taxincentives should be considered and placed.

Two Final Comments

In discussing tax incentives and economic devel-opment, I feel compelled to comment on two emerg-ing topics in the field. It strikes me that one of the mostinteresting questions is the potential constitutionalinfirmity of many state tax incentives. Walter Heller-stein, the constitutional scholar on state and local

2 Shannons thesis is that states avoid tax policies, rates, andlevels that cause them to stand out from the bulk of the statesbecause of their negative connotations from a development stand-point. In this fashion, tax competition acts as an invisible regulatorof state tax actions. J. Papke (1995) finds a similar effect of taxcompetition.

3 I have some concern that Wasylenkos conclusions about theimpact of taxes on intra-regional decisions as well as some of thesuggestions in Dabson, Rist, and Schweke (1996) might be inter-preted to suggest that offering incentives in such bidding wars canbe determinative.



22/30

xation, has developed a cogent line of argument thatoncludes that U.S. Supreme Court jurisprudence mayell cause a number of tax incentives to be found inolation of the antidiscrimination requirements of theommerce Clause (Hellerstein and Coenen 1996). Inarticular, Hellerstein finds suspect those incentives in

hich the coercive power of the state is involved (forxample, a credit lowering an existing liability because

in-state investment, as opposed to an outrightxemption of some in-state activity). Peter Enrich996) has reached a similar conclusion regarding the

onstitutionality of many incentive programs, albeitrough somewhat different reasoning. If Hellerstein

nd Enrich are correct, the potential for a seismicsturbance in the tax incentive landscape is relativelyeat and an item for students in the field to watch.

I disagree with the conclusion of Burstein andolnick (1996) about a potential federal role in regu-

ting tax competition among the states. I agree withhannon (1994) and J. Papke (1995) that competitionmong the states is self-regulating to a considerableegree, without any dumbing-down or race-to-the-ottom effect. The demand for quality services pre-udes such a race to the bottom, and states are morean capable of regulating their own activities andtablishing their own policies. As one modest exam-e of the ability of states to act together, I would pointrecent actions of the member states of the Northeast

ates Tax Officials Association (NESTOA) to agree onommon guidelines to govern issues of residency andomicile, to eliminate potential multiple taxation of

tangible income, and to resolve conflicting determi-ations of domicile among member states.

xpenditure Policy and Economicevelopment

Fishers paper is intended to achieve objectivesmilar to Wasylenkos; namely, to summarize the

mpirical work examining the relationship betweene provision of certain types of public services andonomic development; to identify the conceptual and

ethodological issues that are involved in such stud-s and which limit the precision and utility of such

research; and to identify some areas for further work.Fisher, too, does an excellent job in hitting thesetargets. Nonetheless, I must admit that I was less thansatisfied when I finished digesting Fishers paper thatI had a good idea what the research was telling me orwhich policy actions one should take in response to

the research. The fault is in no way Fishers. It is,instead, the fault of the difficulties involved withresearch of this nature and the mixed results it hasgenerated.

The research is all over the board and some-what inconsistent in its results as to whether invest-ments in public services can increase levels of eco-nomic development. For example, the relationshipbetween transportation investments and development

The research examining therelationship between the provisionof certain types of public servicesand economic development is all

over the board and somewhatinconsistent in its results as towhether investments in publicservices can increase levels of

economic development.

is generally positive (but not overpowering), and itis statistically significant only half of the time. Like-wise, the relationship between public safety expendi-tures and development is generally positive, but soinconsistent and variable across industry groups asto cause one to question the meaning of the relation-ship. Most disturbing, I suppose, is the less thanconvincing evidence about a positive relationship be-tween education services and economic development.

In addition, seemingly inexplicable differences existacross industry groups in the relationship with certainservices.

Simply put, if one tried to act on this research, onewould likely become very confused and uncertain. Ifanything, the tendency would be to underinvest ordisinvest in public services, because of the quandary.

As Fisher points out, it seems likely that theprimary reason for such mixed results is the difficulty

4 See Enrichs comments elsewhere in this issue.5 Enrich (1996) also argues that the Commerce Clause of the

S. Constitution can act as a brake on interstate competitionthout the need for federal involvement.

6 See NESTOA Domicile Working Group, Final Report toommissioners: Domicile Status in NESTOA States, May 8, 1996,ailable from the author.



23/30


24/30

Discussionherese J. McGuire, Associate Director of the Institute ofovernment and Public Affairs and Associate Professor ate College of Urban Planning and Public Affairs,

niversity of Illinois at Chicago.

It is with great pleasure that I comment on thesetwo survey pieces. While numerous surveys havebeen made of the literatures covered by Michael

Wasylenko and Ronald Fisher, a sense of confusionnd uneasiness remains about what we are to con-ude from the plethora of studies. And conclude weust, because the issue of the effect of taxes and

overnment services on economic development is ofeat importance to public policymakers and resi-

ents/voters across the country. I can think of no onehose assessment of the state of our art I trust morean Mikes and Rons.

The central question posed in the Wasylenkoaper is whether taxes are a significant determinant ofgional and local economic development. Relying

eavily on a previous survey of the empirical litera-re by Bartik, Wasylenko suggests that a consensus

as developed among researchers that taxes do mat-r, but that the effect is relatively small (an elasticity 0.2). I challenge whether a consensus that taxesatter really has been achieved, and I suspect that

Wasylenko himself is ambivalent about this conclu-

on. My qualms stem primarily from Wasylenkoswn work in this area, much of it with Robert Carroll

Therese McGuire, in which he finds that the empir-al results are sensitive to the choice of time period toe analyzed.

Thus, I suggest that a new survey of this well-troderature would be most useful if it only briefly

ummarized previous surveys of the literature andstead focused attention on three subsets of theerature: (1) the Wasylenko oeuvre; (2) the empiricalerature since the Bartik 1991 survey; and (3) a

and-picked selection of favorites. Wasylenko does

ention recent studies several times in his review, butese new studies are not clearly identified or critically

xamined, nor is a synthesis provided of what theost recent evidence has to offer. In particular, it is notade clear whether the most recent evidence (since

artiks 1991 survey) supports or is in conflict with thesults of the earlier literature.

Another cause of Wasylenkos seeming ambiva-nce about the consensus conclusion might be the

fact that the evidence from studies of specific taxincentives and enterprise zones, which appears toindicate that specific tax incentives have limited im-pact on business location decisions, contradicts theconclusion that taxes matter. It is imperative thatwe reconcile the findings from these two related, but

separate, literatures. It would improve the clarity ofour thinking on tax incentives to see a summary of theresults of the tax incentives literature presented in atable similar to those Wasylenko has provided sum-marizing the econometric studies.

Finally, I appeal to Wasylenko in this paper, andto all contributors to this literature, to provide guid-ance on what we should advise state policymakerswhen they turn to us for help in designing tax oreconomic development policy. Do we really believethat if Minnesota lowered its rather high tax burdensby 10 percent, the state would see a 2 percent or

greater increase in employment growth or invest-ment? And even if we decide we believe this to be

ronald c. fisher

Documents