household migration and the local public sector: evidence from sweden, 1981–1984

14
This article was downloaded by: [University of Illinois at Urbana-Champaign] On: 07 October 2014, At: 01:55 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Regional Studies Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/cres20 Household Migration and the Local Public Sector: Evidence from Sweden, 1981–1984 Olle Westerlund a & Michael L. Wyzan b a Department of Economics , Umeå University , S-901 87, Umeå, Sweden b Stockholm School of Economics , Box 6501, S-113 83, Stockholm, Sweden Published online: 16 Jul 2007. To cite this article: Olle Westerlund & Michael L. Wyzan (1995) Household Migration and the Local Public Sector: Evidence from Sweden, 1981–1984, Regional Studies, 29:2, 145-157, DOI: 10.1080/00343409512331348863 To link to this article: http://dx.doi.org/10.1080/00343409512331348863 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Upload: michael-l

Post on 23-Feb-2017

212 views

Category:

Documents


0 download

TRANSCRIPT

This article was downloaded by: [University of Illinois at Urbana-Champaign]On: 07 October 2014, At: 01:55Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Regional StudiesPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/cres20

Household Migration and the Local Public Sector:Evidence from Sweden, 1981–1984Olle Westerlund a & Michael L. Wyzan ba Department of Economics , Umeå University , S-901 87, Umeå, Swedenb Stockholm School of Economics , Box 6501, S-113 83, Stockholm, SwedenPublished online: 16 Jul 2007.

To cite this article: Olle Westerlund & Michael L. Wyzan (1995) Household Migration and the Local Public Sector: Evidencefrom Sweden, 1981–1984, Regional Studies, 29:2, 145-157, DOI: 10.1080/00343409512331348863

To link to this article: http://dx.doi.org/10.1080/00343409512331348863

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Regionul Studies, Vol. 29.2, pp. 145157.

Household Migration and the Local Public Sector: Evidence from Sweden, 1981-1984

OLLE WESTERLUND* and MICHAEL L. W Y Z A N t "Department of Economics, Ume2 University, S-901 87 Umea", Sweden

$Stockholm School of Economics, Box 6501, S-113 83 Stockholm, Sweden

(Received January 1993; in revised form January 1994)

WESTERLUND 0. and WYZAN M. L. (1995) Household migration and the local public sector: evidence from Sweden, 1981-1984, Reg. Studies 29, 145-157. In this paper, the relationship between public sector attributes and household character- istics, and Swedish household migration, 1981-4, is studied. We report results separately for metropolitan and non- metropolitan areas because the per capita levels of the tax base and intergovernmental grants are theoretically important migration determinants where population is sparse, whilst the tax rate may be more important where population is dense. Because fiscal influences are likely to be stronger for short-distance migration, we apply multinomial logit to a three-way choice set: staying and migrating short and long distances. Empirical results support our fiscal hypotheses and are consistent with previous findings on household characteristics.

Household migration Intergovernmental grants Multinomial logit Local public sector Tax base

WESTERLUND 0. et WYZAN M. L. (1995) La migration des mknages et le secteur public local: des preuves provenant de la Sutde de 1981 i 1984, Reg. Studies 29, 145-157. On ttudie le rapport entre les caracttristiques du secteur public et celles des menages, et la migration des mtnages en Sutde de 1981 i 1984. Les rksultats sont pri- sent& skparkment pour les zones urbaines et rurales parce que, quant i la migration, l'assiette fiscale par tEte et les primes intergouvernementales sont en principe des dkter- minants non nkgligeables dans les zones les moins peuplkes, tandis que le taux d'imposition est important dans les zones 2 forte densitk de population. Vu que la situation fiscale risque d'Etre plus importante pour la migration i faible distance, on applique un modtle du type logit polyn8me i un ensemble de choix tripartites: on reste ou on se dkplace i faible distance ou i grande distance. Les rksultats em- piriques viennent 1 I'appui des hypoth2ses fiscaux et sont en accord avec les preuves anttrieures au sujet des caractkri- stiques des mknages.

Migration des mknages Primes intergouvernementales Logit polynSme Secteur public local Assiette fiscale

INTRODUCTION

In recent years a substantial literature has appeared that deals with the effects of public sector attributes on human migration. Particular attention has been devoted to examining whether variation in local income or property taxes affects such behaviour. The

WESTERLUND 0. und WYZAN M. L. (1995) Haus- haltwanderung und der ortliche offentliche Sektor: Be- weise aus Schweden fur den Zeitraum 1981-84, Ref . Stud- ies 29, 145-157. Dieser Aufsatz untersucht die Beziehung zwischen Attributen des offentlichen Sektors und Eigen- schaften von Haushalten, und schwedischer Haushaltwan- derung im Zeitraum 1981-1984. Die Ergebnisse werden getrennt fur groBstadtische und nicht-groBstadtische Gebiete aufgefuhrt, weil bei dunner Besiedlung die Pro- Kopfhohen des SteuerfuBes und bei zwischenstaatlichen Beihilfen theoretisch bedeutsame Bestimmungsfaktoren der Wanderung sind, wahrend bei groBer Bevolkerungs- dichte der Steuersatz wichtig sein kann. Da finanzielle Ein- fliisse die Nahwanderung wahrscheinlich starker bestim- men, wurde ein multinominales Logit auf eine Situation angewandt, die drei Alternativen bot: an Ort und Stelle zu bleiben, sich zur Nahwanderug oder zur Fernwanderung zu entschlieRen. Empirische Ergebnisse bestatigen die finanzpolitische Hypothese der Autoren und stimmen mit fruheren Befunden bezuglich Eigenschaften von Haus- halten uberein.

Wanderung von Haushalten Zwischenstaatliche Beihilfen Multinominales Logit Orticher Offentlicher Sektor SteuerfuB

impacts of various forms of public expenditures have also been studied, as have the effects of unrestricted central government grants to localities.'

These studies have focused on the expenditure and/or taxation sides of the ledger, ignoring the role of the local per capita tax base (the per capita sum

0034-3404/95/020145-13 0 1995 Regional Studies Association

145

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

146 Olle Westertund and Michael L. Wyxan

of all income in the locality subject to local income taxation). This is despite the tax base’s theoretical importance as a determinant of household mobility (see especially WHEATON, 1975). Moreover, empir- ical evidence of a significant relationship between certain fiscal variables -for example, intergovern- mental grants- and migratory behaviour is difficult to interpret if the per capita tax base has not been taken into consideration.

Although one might expect a pviovi that variation in the local per capita tax base would affect the relat- ive desirability of living in various jurisdictions, such reasoning may not apply to densely populated areas. In such regions it is often possible to consume such local publicly provided goods and services as parks, sports facilities or cultural amenities, whilst living in another jurisdiction; one thus avoids paying taxes to support those activities. In such cases, the locational decision maker may select the jurisdiction with the lowest tax rate, irrespective of the level of that local- ity’s expenditures (and hence of its tax base). Accord- ingly, it is desirable to estimate the determinants of migratory behaviour separately for densely and lightly populated regions. To our knowledge, such a practice has not been followed in the literature to date.

Previous empirical work on the impact of fiscal attributes on migration has also failed to take into account the notion that the dominant motivation for moving may vary with distance moved (see, for example, MOLHO, 1986). Short-distance moves are less likely to be associated with changes in jobs than long-distance ones. Accordingly, variation in fiscal variables may have a particularly significant impact on short-distance migration, whilst long-distance migration may be especially susceptible to changes in labour market conditions. Distinguishing between these two types of migration may be necessary to obtain an accurate reading of the effects of public sector variables on human migration.

In this paper, we present the results of an empirical study of the role of public sector and other place of origin attributes, and of household characteristics, in explaining the migratory behaviour of Swedish households between 1981 and 1984. Our model is of the conventional ‘disequilibrium’ type, that is, it is based on the assumption that regional labour markets are relatively slow to clear and that at any given moment the effects of fiscal variables are not fully capitalized in land rents and wages.*

We employ the multinomial logit technique to analyse a three-way choice set, including staying, moving to another municipality (kornrnun) within the county (l;in) and moving to another county. Our models include, in addition to a number of household characteristics, unusually complete information on the fiscal characteristics of each local jurisdiction: its per capita tax base, income tax rate, per capita gen-

eral purpose grants received from the central govern- ment and a measure of the age and social structure of its population. In addition, because of the afore- mentioned potential differences in the effects on migratory behaviour of fiscal variables between densely and sparsely populated areas, we report results separately for metropolitan and non- metropolitan Sweden.

There are a number of reasons why Sweden is an appropriate locus for such an investigation. First, the Swedish spatial structure includes both such large metropolitan areas as Stockholm and Gothenburg and areas in the north with some of the world’s lowest population densities. Sweden is thus ideal for an investigation of how the determinants of house- hold migration vary between densely and lightly populated regions. Second, as is well known, the size of the Swedish public sector relative to GDP is amongst the highest in the world; in 1992 public expenditure exceeded 67% of GDP. If fiscal vari- ables exert an influence on individual decision making anywhere, it may well be in Sweden.

Third, excellent data suitable for an investigation of the effects of fiscal variables on human migration are available for Sweden. Highly detailed household survey data are contained in the Level of Living Surveys (Levnadsniviundevsokningarna: LNU) . More- over, information on the fiscal characteristics of each local jurisdiction in the country can be obtained from the national statistical authorities.

This paper is organized as follows. In the next sec- tion, we provide a theoretical discussion of the role of fiscal variables as determinants of household migration, with special reference to Swedish condi- tions. The third section describes our empirical strat- egy. The fourth section discusses our data set and the variables employed in the empirical work. The fifth section presents and discusses our empirical findings and the sixth provides a summary and some conclusions.

THE PUBLIC SECTOR A N D HOUSEHOLD MIGRATION:

THEORETICAL INSIGHTS A N D SWEDISH REALITY

One of the chief contributions of this study is its examination of the effects on locational decisions of regionally variable public sector attributes. Previous empirical studies of migration have examined a number of fiscal variables (a particularly complete set of such variables may be found in Fox et al., 1989). Missing, however, from virtually all such work is an appreciation of the fundamental role played by the local per capita tax base. In a seminal article, WHEATON, 1975, challenges TIEBOUT’S, 1956, notion of ‘voting with one’s feet’. TIEBOUT, 1956, holds that the establishment of a large number of

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 147

jurisdictions which assess local preferences and pro- vide public output efficiently will enable individuals to migrate to a stable equilibrium of optimal (and homogeneous) communities.

WHEATON, 1975, p. 377, demonstrates theoretic- ally that the ‘critical factor that determines migration preferences is the method by which each community finances its public consumption’. In a model with one private good and one public good, he investi- gates individual behaviour vis-2-vis moving to a new locality differing only at the margin from the one of current residence. His most relevant results for Sweden concern migration incentives under propor- tional income taxation. He finds that, unlike the case of a poll (i.e. equal per capita) tax, all individuals starting out in homogeneous communities will wish to move to wealthier ones. The same result holds if people start out in heterogeneous ones, at least if one assumes a logarithmic utility function. Thus, the per capita tax base is clearly an important local fiscal vari- able worth examining as a potential determinant of migratory behaviour.

Another important public sector attribute consists of transfers to lower level jurisdiction from higher ones. Much of the debate in the literature on the justifiability of such transfers has taken place in the context of an examination of the Canadian federal system. BOADWAY and FLATTERS, 1982, p. 622, demonstrate that ‘in a very simple federal economy . . . free migration will generally lead to an ineffi- cient allocation of labour over the federation’, that is, will not maximize the net benefit to the nation from the net migration of an individual between jurisdictions.

This inefficiency arises due, amongst other factors, to the differential ability of residents of different jur- isdictions to share in ‘land rents’, that is, rents gener- ated by immobile factors of production located in those jurisdictions. In the model of BOADWAY and FLATTERS, 1982, a sub-national government, wish- ing to maximize the utility of its current residents- and ignoring the effects of its actions on migration- determines how to divide output between the public and private sectors. Free migration of the population occurs until the population of each region is such as to equate maximum utility per capita across regions. At the equilibrium population levels determined by free migration, the sum of the net benefits to the regions from having an additional person move from one to another is not in general equal to zero. It can be shown that this inefficiency is the sum of two terms, one related to differences in per capita tax pay- ments- termed the ‘fiscal externality’- and the other to differences in per capita land rents.

Such inefficiency can be eliminated by a particular system of transfers from the residents on one juris- diction to those of another. Intergovernmental grants may be also be justified on equity grounds, if, as

noted by BOADWAY and FLATTERS, 1982, p. 631, ‘the federal government took the view that hori- zontal equity demanded that persons in identical positions before [the effects of sub-national] and fed- eral government budgets [were felt] should be so afterward^'.^

However, another school of thought, again con- cerned largely with the Canadian situation, views equalization payments and other fiscal programmes such as unemployment insurance as sources of ineffi- ciency. Such programmes are seen as preventing people from migrating on the basis of ‘traditional market variables’ (see, for example, COURCHENE, 1970; and SHAW, 1986).

The Swedish system of tax equalization grants (TEG) has received relatively little attention as a potential determinant of migratory behaviour (an exception is NELSON and WYZAN, 1989). There is considerable variation in the size of the per capita TEG received by municipalities: the mean level across our data set (in 1981) was 601 kronor, with a minimum of zero and a maximum of almost 3,800 kronor, amounts that would be approximately doubled if reckoned in today’s kronor. The size of these grants is substantial for many municipalities: in 1985, 51 of the 284 municipalities were receiving TEG constituting more than 25% of their tax bases.

The TEG system is officially designed to avoid the creation of ‘unreasonable differences in tax rates and/or standards [of service provision]. [Such differ- ences] would in the long run lead to serious con- sequences for settlement patterns and residential structure’ (SOU, 1985, p. 15). It is thus possible to regard TEG as a correction for deficiencies in the local per capita tax base. Formally, the per capita level of TEG received by a municipality (or county council) is positively related to certain measured deficiencies in its age and social structure, geograph- ical location and residential structure, along with other variables assumed to cause differences in the production cost of providing a given service. Another factor is whether a municipality experienced a population decrease of more than 5% during the preceding 10 years.

This interpretation suggests that in order to exam- ine adequately the effect of TEG on migratory pro- pensities, the local per capita tax base must be held constant. Suppose that, in a given country, localities receive general purpose grants from the central gov- ernment to make up for deficiencies in their per capita tax bases. Suppose further that the size of the local per capita tax base-whether or not augmented by such grants-is positively correlated with the pro- vision of amenities deemed desirable by individuals making locational decisions. Unless the capita tax base is held constant, a positive (ceteris paribus) rela- tionship between the per capita level of the grant received by a locality and migration into it may be

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

148 Olle WesterZund and Michael L. Wyzan

estimated as insignificant or even significantly nega- tive, since such localities are the ones with lower per capita tax bases.4

Another fiscal variable examined here is the local income tax rate, relatively h g h levels of which are expected, ceteris pauibus, to encourage out-migration. In Sweden, the most important tax on physical per- sons levied by sub-national governments is a propor- tional income tax; the rate of this tax is the sum of the county council ( landsting), municipal and parish forsumling) tax rates.5 The functions of the county council are largely in the medical sphere, whilst the municipalities provide such services as infrastructure, municipal services, social welfare and primary and secondary education.

In general, poorer communities will be forced to levy higher tax rates to pay for a particular level of public services. In the simplest case, where there are no differences in tastes for local public goods and services within or between municipalities, the tax rate would not be a significant determinant of migra- tion if the per capita tax base were held constant. In a small, ethnically homogeneous country such as Sweden, taste differences across municipalities are a less likely cause of varying tax rates than is variation in local per capita tax bases. Moreover, even in the presence of taste differences, most of the services provided by the municipalities are regulated by national legislation imposing minimum quality standards and cost levels. Under such conditions, such differences may well be of minor importance.6

An alternative procedure to examining the effects on migratory behaviour of the per capita tax base and per capita TEG would be to investigate the influence on that behaviour of both tax rates and vari- ous types of public expenditure. This is the approach followed, for example, by Fox et al., 1989, and DAY, 1992. In Sweden, however, the local public sector provides a great variety of goods, some of which are quasi-public or even private, at non- market prices-often necessitating queueing - and at varying quality levels. It thus seems extremely diffi- cult to characterize adequately the spending side of the ledger in each Swedish municipality.

Moreover, as pointed out by CHARNEY, 1993, if an inclusive set of expenditure categories and revenue categories are included as explanatory variables, the signs of their expected coefficients may be ambigu- ous. This arises from the fact that an identity links these categories through the budget-balance equa- tion. Note also that the coefficients on individual expenditure categories generally reflect spending to reduce ‘disamenities’ (e. g. crime) or increase ‘amenit- ies’ (e. g. recreational opportunities). Such coeffi- cients are difficult to interpret in the absence of information on the levels of the disamenities or amenities themselves, especially when tax rates are held constant. It seems to be ‘crime’-or the high

taxes required to fight it-that stimulates out- migration, not the expenditure level i t ~ e l f . ~

In view of the arguments of the previous two para- graphs, the best strategy seems to be to focus on the per capita tax base and related fiscal variables. A caveat is in order, however. In a densely populated urban area, it is often possible to consume local pub- licly provided goods and services whilst living in another jurisdiction, thus avoiding having to pay taxes to support those activities. This is an especially important phenomenon in Sweden. For example, local parks, libraries, sports arenas and opera com- panies charge very low fees because they are subsid- ized by municipal governments; no attempt is made to exclude residents of other jurisdictions from con- suming these services. This effect is more likely to be felt in densely populated areas, where the transaction costs of free-riding are relatively low, than in regions where municipalities are far apart.

Accordingly, tax rates may play a more important role in explaining migration in metropolitan areas than elsewhere. Moreover, given the possibility of free-riding, other fiscal variables aimed at capturing the extent of local public service provision may per- form worse for such areas. However, because there are certain expenditure categories for which free- riding is impossible, and for comparability with ear- lier work, we include the total per capita expendit- ures of the municipality in our specifications that contain the tax rate. The caveats presented above concerning public expenditures also apply in this case, so that it is difficult a puiovi to predict its sign. In any case, the per capita tax base-the measure of the overall state of the local public sector-is not expected to exert an independent influence on migratory behaviour in metropolitan areas.

Sweden, with its considerable variation in popula- tion densities, is an ideal locus for testing the hypo- thesis of the differential relative importance of the per capita tax base and the tax rate under varying such densities.

EMPIRICAL STRATEGY

In our empirical work, the household is assumed to maximize a utility function composed of determin- istic terms describing its own characteristics and those of its current locations, as well as a stochastic term. Most previous work exploring the determin- ants of the probability of migrating using household survey data has employed the binary-logit tech- nique.8 This approach suffers from the limitation that it does not allow one to distinguish between the determinants of short- and long-distance migration whilst making use of the entire data set.

As noted by MOLHO, 1986, p. 41 1, there are ‘sharp disjunctures in migration patterns across different distance ranges, [which can be] attributed to different

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 149

motives for moving, [since] there are inherent differ- ences between migrants who merely relocate residen- tially, and those who also change their workplace’ (emphasis in original). It seems reasonable that relatively small variations in fiscal variables will be more significant as migration determinants for short- distance moves where job changes are not involved.

Recent theoretical models of sequential migration by MCCALL and MCCALL, 1987, and BHATTACH- ARYA, 1990, posit workers as searching for jobs in a given locality in the absence of complete knowledge of the characteristics of that locality, which can be ascertained only by living there. Neither paper deals with purely residential location decisions. However, as an extension of their work one might imagine the migrator choosing amongst alternative housing sites once he has decided to accept a job in a given larger locality. Such a theoretical departure might be based on work by geographers such as REITSMA and VERGOOSSEN, 1988, who carefully distinguish between residential and labour migration. They find that, in the western Netherlands, labour migrants tend to originate from much farther away from their destinations than do residential migrants (ibid., p. 336).

There is thus considerable merit in attempting to distinguish between the determinants of short- and long-distance migration. Following previous work on household migration in Sweden employing the LNU (HOLMLUND, 1984, pp. 121-5), we define the former as a move between municipalities in the same county and the latter as a move between counties. Although any such definition is necessarily arbitrary, this ‘geopolitical’ subdivision of the country has con- siderable appeal when fiscal variables are the ones of primary interest. Clearly, a move between municip- alities in a county results in a change in the municipal income tax rate, which would not result from migra- tion within the same municipality. On the other hand, in densely populated parts of the country labour markets may comprise several counties or parts thereof. Accordingly, it is not self-evident that long-distance migration, as we define it, coincides closely with labour market migration. Thus, the determinants of this type of migration may not be related to labour market characteristics to a statistic- ally significant degree.

We estimate, via the multinomial logit technique, a discrete choice model in which the potential migrant has three choices: staying; moving to another municipality within the county; and moving to another county. Formally, this technique entails the estimation of the following model of the choices facing the ith individual (or household):

$;.,

1 + C t4 Prob( Y = j) = for j = 1 , 2

and

for j = O (1) 1

Prob(Y=j)= 1 + 2 t4eP;xz

where:

Y = the dependent variable j = 0 if the respondent elects to stay in his present

location, 1 if he moves to another municipal- ity within the county and 2 if he moves to another county

x, = a vector of attributes, both personal and those of its place of origin, characterizing the ith household

J; = a vector of coefficients on those attributes specific to the case where the jth alternative is chosen.

Multinomial logit estimation thus produces two coefficients for each independent variable, with the first option, to stay, being the reference alternative to which these two coefficients relate (for further dis- cussion, see GREEN, 1990, pp. 697-8).

Also worth mentioning is the fact that the decision process preceding migration may differ based on the number of adult household members. For singlc- adult households, migration may commonly be related to such life-cycle events as meeting a partner, or entering or leaving the university, which could blunt the estimated responsiveness of migration to measurable individual or locational characteristics. On the other hand, for the multiple-adult household optimal locational choice is based on the preferences of several individuals, whose utility functions are likely to be interdependent. In the results reported in this paper, we follow the practice of HOLMLUND, 1984, chapter 5, and confine ourselves to multiple- adult households; results for single-adult households are reported in WESTERLUND and WYZAN, 1993.9 We restrict the sample to the population aged 18 to 65-the age span of the entire sample is from 15 to 76 years-in order to obtain a sample corresponding most closely to the set oflocational decision makers.

In closing this section, a word is in order on our selection of discrete choice technique. It is clear from the foregoing that with our method we are able to examine only the decision to migrate and the separate determinants of short- and long-distance migration as we have defined them; the decision to which muni- cipality to move is not considered. This characteristic is typical of empirical work on migration employing micro data (see, for example, all of the studies reviewed by HERZOG et al., 1993).

There is a small number of articles that employ either conditional-logit methods that relate locational decisions to the characteristics of the potential loca- tions, mixed-logit methods combining both the con- ditional- and multiple-logit techniques (e. g. HENLEY et al., 1989) or two-stage nested-logit models (e.g.

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Olle Westerlund and Michael L. Wyzan

FALARIS, 1987, 1988 HANSEN, 1987). Each of these studies either divides the relevant geographical space into a small number of large regions (FALARIS, 1987, 1988; HENLEY et al., 1989) or deals with a space that is itself quite small (HANSEN, 1987). In the present context, each household is confronted with a large number of possible alternatives (283 municipalities) and the small number of total migration events. Any form of aggregation over these municipalities would result in the loss of the fiscal variation over a small geographical space that is the primary focus of atten- tion.” Accordingly, no attempt is made in this paper to explain the specific locational choices of households.

DATA AND VARIABLES

The data that we employ may be divided into two types: data on household characteristics and those on the characteristics of the place of origin. The first type of data are found in the 1981 LNU, as well as in so-called ‘register data’, which contain follow-up information on the surveyed individuals over the period after the survey. We have observations on 4,261 individuals, of whom 365 changed municipal- ity of residence during the 1981-4 period.” Of these, 2,981 were members of multiple-adult households, of whom 194 migrated. The second type of data con- sists of the official figures contained in various data bases created by Statistiska Centralbyrin, the Swedish national statistical agency. Table 1 contains

brief descriptions of each variable and Table 2 their respective sample means and standard deviations.

Our fiscal variables include the municipal income tax rate, the municipal per capita tax base, total per capita expenditures by the municipality, per capita tax equalization grants received by the municipality and a measure of the age and social structure of the local population. We define the ‘local’ tax rate (YTAX) as the sum of the proportinal rates levied by the relevant municipality and county council divided by the national average for this variable. The local per capita tax base in 1981 (TAXBA SEJ is also measured relative to the national average per capita tax base, with the latter set equal to 100. Similarly, per capita tax equalization grants (TEG) received in 1981 by the municipality and municipal per capita expenditures (EXPCAP) are expressed in terms of the relevant national average.

We have also made an attempt to control for differ- ences in age or social structure that cause variation in municipal expenditure on mandatory local public service provision. A variable that we have termed the ‘age factor’ (AGEFAC), which the Swedish tax authorities have calculated based on data for 1983 and 1984, is included in our migration equations. This variable is constructed by deriving indexes for the incidence within the municipality of a given factor known to engender increased mandatory municipal expenditure (e.g. children of school age, persons over 65 or number of recipients of social benefits). The additional expenditures thus required are related

Table 1. Definitions of independent variables included in estimated equations

Variable Definition

Household characteristics‘ A G E Age of respondent CHLD FR QC T C H S E O WN MIG7481 S E C S C H

Dummy variable’ indicating whether household has school-age children Dummy variable indicating whether respondent has frequent contact with relatives Dummy variable indicating whether respondent owns hidher place of residence Dummy variable indicating whether household moved during 1974-81 Dummy variable indicating whether respondent has at least a completed secondary education

Characteristics A C E F A C

E X P C A P HSEPRI TA X B A SE

T E G

U N E M P Y T A X

of municipality’ A measure of how structure of expenditure in municipality is affected by age distribution and social structure of local population, expressed in terms of national average tax base per capita in 1983 and 1984 (scaled so that it attains a higher value when age structure is less favourable) Total expenditures per capita by municipal government divided by national average for this variable Average level of house prices in municipality divided by national average for this variable Municipality’s tax base (the sum of all income in the municipality subject to municipal income taxation) per capita divided by national average for this variable Level of tax equalization grants per capita received by municipality from central government divided by national average for this variable Unemployment rate in municipality divided by national average for this variable Local (i,e. sum of municipal and county council) income tax rate divided by national average for this variabie

Notes: 1. All household characteristics refer to the situation in 1981 (the survey year) except where indicated otherwise. 2. All dummy variables are equal to unity if the relevant characteristic is present and zero otherwise. 3. All municipality characteristics refer to the situation in 1981 except where indicated otherwise.

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 151

Table 2. Sample means and standard deviations of inde- pendent variables

Variable Mean Standard deviation

A G E 42.602 12.256 AGEFAC’ 0.535 2-922 CHLD’ 0.398 EXPCAP‘ 1.076 0.180 F R QC TC’ 0.250 H S E O W e 0.723 HSEPRI’ 1.056 0.332 MIG748 1 0-269 S E C S C P 0.190 T A X B A S E 3 0.988 0.173 T E G 3 1.044 1.081 UNEMP‘ 0.897 0.444 Y T A X 1.006 0.043

Notes: 1. See text and Table 1 for explanation. 2. Dummy variable equal to unity if the relevant character-

3. Index formed by dividing value of variable in relevant istic is present and zero otherwise.

municipality by national average of variable.

to the national average per capita tax base in forming AGEFAC (for further details, see WESTERLUND and WYZAN, 1993, pp. 43-5). AGEFAC is scaled so that it attains a higher value when the age structure is less favourable.

We also consider two non-fiscal characteristics of the municipality of origin: the local unemployment rate ( U N E M P ) and the average level of house prices (HSEPRI), both in 1981 and both expressed relative to the national average value of the relevant variable. These variables are included to control for variation in income opportunities and the cost of living, respectively. In Sweden, regional net (i.e. after tax payments to the central government) wage disper- sion is relatively small. This is due, inter alia, to the highly progressive central government income tax in place during 1981-4 and to the ‘solidaristic wage policy’ (e.g. FLANAGAN, 1987) pursued by the trade unions. l2 Under such conditions, it seems reasonable to use employment opportunity, as indicated by the unemployment rate, as the sole indicator of local income opportunity. The local unemployment rate is defined as the number of unemployed at the place of origin normalized with respect to the size of the local population aged 16 to 64 in 1981.

Regional average house prices are typically employed in empirical studies of the Swedish eco- nomy as a proxy for regional variation in the cost of living, data on which are not available. As noted by SKEDXNGER and MCCORMICK, 1991, p. 158, who follow this practice in their examination of the deter- minants of local unemployment rates, ‘variations in house prices . . . account for a great deal of the vari- ation in consumer prices’. On this interpretation, house prices should be positively associated with out-

migration. O n the other hand, land rents and thus house prices may have capitalized into them the util- ity value of local amenities and disamenities. This may lead to a negative estimated relationship between house prices and the migratory propensity, especially since we do not include such amenities in the specification.

The independent variables describing the house- hold’s characteristics are amongst the standard ones employed in empirical studies of migration based on survey data. Except where indicated otherwise, all respondent answers refer to the situation in 1981. Those with a more advanced education are likely to have better access to, and better ability to process information on, employment opportunities in other locations. Furthermore, there is a ‘thin’ local labour market for professionals, for whom job changes often require migration. We include a dummy vari- able equal to unity for those who have completed secondary school or higher ( S E C S C H ) and zero otherwise (a construction that applies to all dummy variables described here).

An important migration determinant in many ear- lier studies has been previous migration e~perience.’~ In line with recent theoretical models of sequential migration (e.g. MCCALL and MCCALL, 1987, and BHATTACHARYA, 1990), a certain fraction of in- migrants will become disenchanted with what they find and move on (or back). Another possibility is that certain individuals may be more prone to loca- tional change than others. Accordingly, we include a dummy variable (MIG7481) for whether the household moved during 1974-81 in our estimated equations.

The respondent’s age (AGE) is expected to be negatively related to his likelihood of migrating, since the number of remaining years over which to enjoy the return on one’s ‘investment’ in migration declines with age. In addition, moving costs, both psychic and actual, presumably rise with age. Having school-age children is expected to act as a deterrent to migration, one probably accentuated for the long- distance type, where friendships may have to be broken; CHLD is a dummy variable equal to unity when the household has school-age children. Having frequent contact with relatives should act similarly. FRQCTC is a dummy variable equal to unity when the respondent averred that she has such contacts.

Owning one’s place of residence is a factor that may decrease migratory propensities. Given the transaction costs and illiquidity of real estate in muni- cipalities in economic decline, such ownership may reduce the probability of migrating. Moreover, pur- chasing real estate may also reflect a decision to stay at the present location long term. H S E O W N is a dummy variable equal to unity if the respondent household owns the dwelling in which it 1 i ~ e d . l ~

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

152 Olle Westerlund and Michael L. Wyxan

EMPIRICAL RESULTS

The two migration equations that we estimate are as follows:

Prob(Y = j) = A T A X B A S E , T E G , S E C S C H , M IG7481, A G E , C H L D, FRQCTC,HSEO W N , UNEMP, HSEPRI) (2)

and:

Prob(Y = j ) = g(YTAX,SECSCH,MZG7481, A G E , C H L D , F R Q C T C , H S E O W N , UNEMP, HSEPRI) (3)

where Y andj reflect the trichotomous choice defined in equation (1). As noted above, multinomial logit estimation of this equation produces two coefficients for each independent variable, 0 1 which relates to the choice of to stay or to make a short-distance move, and 0 2 , which relates to the choice of to stay or to make a long-distance move.15

As discussed previously, we expect the model described by equation (1)- the ‘tax base variant’- to apply to sparsely populated areas and that described by equation (2)-the ‘tax rate variant’--to apply to densely populated ones. Table 3 contains separate multinomial logit results of the tax base variant for multiple-adult households for two geographical sub- samples-metropolitan and non-metropolitan. The most striking result from the table is that for short- distance migration the coefficients on the per capita tax base and per capita TEG are both significant (the latter at 0-05) and have the expected signs for non- metropolitan areas, but not for metropolitan areas. No fiscal variables are significant determinants of long-distance migratory behaviour. These findings are congruent with our expectations, as outlined above, with the exception of the significantly nega- tive (at 0-05) coefficient on TEG for short-distance migration in non-metropolitan areas.“

With respect to the non-fiscal variables, the factors that appear as important determinants of short- distance migration for both metropolitan and non-metropolitan areas include previous migration experience, age and the local unemployment rate. The coefficients on the variables indicating the pres- ence of school-age children, local house prices and house ownership are significant with the expected signs only for metropolitan areas. The significantly positive coefficients on house prices suggest that this variable functions as a measure of the cost of living and does not have amenity effects capitalized in it to a significant degree.

For long-distance migration, fewer variables have significant coefficients than for short-distance moves. Significance is achieved only for previous migration experience, having frequent contact with relatives,

having a secondary education or higher (both for non-metropolitan areas only) and home ownership (for both metropolitan and non-metropolitan areas). That SECSHH encourages long-distance migration only is consistent with the notion that such migration is more likely to be related to the labour market. As discussed above, the more highly educated face a thinner local labour market and have a better ability to process information on distant job opportunities. In general, however, there is rather little evidence favouring the idea of a close association between long-distance migration and the labour market. Recall that the local unemployment rate is significant only for short-distance migration.

It is not always possible to explain why a certain variable matters more in one type of region than in the other. However, that having school-age children deters migration only in metropolitan areas may be because households in such areas have commuting options available as a substitute for migration, options that their non-metropolitan counterparts lack. The presence of such options in metropolitan areas may also explain why house prices matter for migratory decisions only in such areas. One can move without changing jobs, unlike in more sparsely populated regions. On the other hand, that having frequent contacts with relatives matters only in non- metropolitan areas may be explained as follows: res- idents of metropolitan areas are more likely to be migrants from another part of the country and less likely to have such contacts in the first place.

Table 4 contains our estimates of the ‘tax rate vari- ant’ of the model, that governed by equation (3). The most interesting finding is that the local tax rate is a significant determinant of short-distance migration in metropolitan areas (only). This result is consonant with our expectations. An increase in per capita municipal expenditures does not affect migration in metropolitan areas but discourages short-distance migration from non-metropolitan areas. It is useful in this context to recall our earlier discussion on the difficulty of interpreting estimated coefficients on this variable. l7

The results for the non-fiscal variables are highly similar to those contained in Table 3. Having school- age children and house ownership are no longer sig- nificant for short-distance migration in metropolitan areas, although the latter variable is now significant for such migration in non-metropolitan areas (it was not for the tax base specification). Otherwise, the results of estimating the tax rate model require no further elaboration, except perhaps to add that the robustness of the coefficient estimates to the change in model is reassuring.

SUMMARY A N D CONCLUSIONS

One of this study’s chief empirical findings is that, in line with our expectations, there is a significantly

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 153

Table 3. Multiple-logit estimates of migration determinants, multiple-adult households, metropolitan and non-metropolitan Sweden: tax base variant

Metropolitan areas Non-metropolitan areas

Variable Coefficient t-value' Coefficient t-value

D1 constant 0 2 constant

D1 T A X B A S E 0 2 T A X B A S E

D1 T E G 0 2 T E G

D1 A G E F A C 0 2 A G E F A C

D1 S E C S C H 0 2 S E C S C H

D1 MIG7481 0 2 ME7481

D1 A G E 0 2 A G E

D1 C H L D 0 2 C H L D

D1 F R Q C T C 0 2 F R Q C T C

D1 HSEU WN 0 2 H S E O WN

Dl U N E M P 0 2 U N E M P

D1 HSEPRI 0 2 HSEPRI

Sample size

Number of migrators

Log-likelihood

- 8.349** -0.779

1.375 -3.371

1.358** -0.355

0.0003 -0.098

-0.146 0.438

0.974** 2.119**

-0.046** -0.014

-0-653* -0.071

-0.362 0.241

-0.563* - 0.864*

2.079** 0.470

2.889** 0.478

1,184

104

-346.43

-4.85 -0-37

1.64 -1.40

4-26 -0.77

0.01 - 1.42

-0.49 1.15

3.35 4.58

-3.56 -0.74

-2.10 -0.17

-1.09 0.60

- 1-98 -2.23

4.60 0-70

3-27 0.32

7.310 -6.874*

-10.433* 4.614

- 0.726* 0.066

-0.138 0.082

-0.101 0.701*

0.919* 2.432**

-0.048* -0.019

0.222 -0-455

-0.591 - 1.366**

-0.765 -0.848**

1.126* -0.266

0.145 -0.829

1,797

90

-335.08

1.77 -2.19

-2.47 1.38

-2-22 0.31

- 1.43 1.05

-0.20 2.19

2.19 6-09

-2.48 - 1.32

0.57 -1.35

-1.27 -2.80

- 1-94 -2.76

2.44 -0.53

0.28 -0.87

Notes: 1. The t-values presented here refer to a distribution that is asymptotically equivalent to the t-distribution. * Denotes significance of the relevant coefficient at the 0-05 level; ** denotes significance at the 0.01 level.

negative relationship between the local per capita tax base and per capita TEG, on the one hand, and the probability of short-distance migration, on the other, for non-metropolitan areas. The significance of TEG- which has performed erratically in previous work on Swedish migration (NELSON and WYZAN, 1989) -illustrates the importance of holding the per capita tax base constant when examining the effects on migration of intergovernmental grants. On the other hand, also confirming our expectations, the tax rate is a significant determinant of migratory behavi- our only in metropolitan areas.

These results are consistent with the notion that the local per capita tax base may be a significant determinant of migration from sparsely populated areas, where it is difficult to free-ride on services pro- vided by other localities. Our findings are also in line

with the view that the tax rate should matter in densely populated areas where such free-riding is possible. Thus, valuable insights into the migratory process have been obtained by splitting the data up into metropolitan and non-metropolitan sub- samples. This demonstrates the importance of defin- ing the problem carefully before estimating migra- tion equations with household survey data.

Similarly, by adopting the multinomial logit tech- nique, we have been able to distinguish between the determinants of short- and long-distance migration, whilst making use of the entire data set. The latter is more difficult to explain in general; moreover, we have been able to detect no influence from our fiscal variables on such migration. The insignificant impact of fiscal attributes on long-distance migration seems reasonable, given that, as suggested, for example, by

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

154 Olle Westevlund and Michael L. Wyxan

Table 1. hfultiple-logit estimates of migration determinants, multiple-adult households, metropolitan and non-metropolitan Sweden: tax rate variant

Metropolitan areas Non-metropolitan areas

Variable Coefficient t-value' Coefficient t-value

D1 constant 0 2 constant

D1 YTAX 0 2 YTAX

D1 EXPCAP 0 2 EXPCAP

D1 SECSCH 0 2 SECSCH

DI ~ 1 ~ x 8 1 0 2 MIG7481

D1 AGE 0 2 AGE

D1 CHLD 0 2 CHLD

D1 FRQCTC 0 2 FRQCTC

D1 HSEOWN 0 2 HSEOWN

D1 UNEMP 0 2 UNEMP

D1 HSEPRI 0 2 HSEPRI

Sample size

Number of migrators

Log-likelihood

-14.732** -5.899

10.827* 2.385

- 13.588 4.519

-0.154 0.484

1.072** 2-078**

-0.041** -0.016

-0545 -0.177

-10.384 0.237

-0.477 -0.839*

1-1 39** 0.167

2-662** -0.486

1,184

104

-354.04

-3.01 - 1.06

2.40 0-45

- 1.43 0.40

-0.53 1.28

3.71 4.54

-3.24 -0.89

- 1.80 -0.43

-1.17 0.59

- 1.65 -2-20

2.90 0.28

3.90 -0.64

-5.568 -0.520

5.025 -2.691

-3.398* 0.482

0.211 0.765*

0.874* 2.429**

-0.049* -0.018

0.220 -0-459

-0.590 -1.342**

- 0.798* -0.852**

1445** -0.343

1.214 -0.273

1,797

90

- 336.66

~

-1.07 -0.12

0.98 -0.63

-2.33 0.53

0.42 2.40

2.22 6.09

-2.55 - 1.28

0.56 - 1.36

-1.27 -2-77

-2.00 -2.76

3.66 -0.82

1.10 -0.33

Notes: 1. The t-values presented here refer to a distribution that is asymptotically equivalent to the f-distribution. * Denotes significance of the relevant coefficient at the 0.05 level; ** denotes significance at the 0.01 level.

MOLHO, 1986, job market phenomena may domin- ate long-distance migration. On the other hand, con- trary to this view, we have also found variables related to the labour market to be insignificant for this type of migration. It should be noted, however, that the level of vacancies throughout Sweden was very low during the rather depressed period under study (e.g. EDIN et al., 1991; WESTERLUND, 1993).

This brings us to some suggestions for future work. First, note that the period covered by our study, 1981-4, was one during which the Swedish economy was depressed to a degree that was at the time unprecedented. Indeed, in 1982 the vacancy rate reached its lowest point in decades and it was not until 1984 that recovery on the labour market was evident. Under such conditions, migratory flows were very low by historical standards. 'Thus, there would be considerable merit in conducting the sort of study presented here with more recent data from a less depressed period.

Moreover, the period under examination was one during which the variation in the provision of local public services was relatively small. At present, budgetary stringency is resulting in cut-backs in the quantity and quality of such services in many, although not all, municipalities. Moreover, a number of localities are privatizing or placing on a fee basis services that they have long provided gratis or for a nominal charge. Accordingly, there are grounds to expect far more variation in local public service pro- vision in the coming years. Results such as those pre- sented here should be useful to central government decision makers as we enter such a period; repetition of the study with more recent data again seems in order.

Second, note that the methodology pursued in this paper has not enabled us to examine the potential simultaneous determination of the fiscal and non- fiscal characteristics of the place of origin. None the less, it is likely that local housing prices, the state of

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 155

the local labour market and the local fiscal variables are related to each other in a manner that is amenable to empirical analysis. Thus, a re-examination of the determinants of migration using more aggregate data-from, say, the municipal level-that explicitly models the interaction amongst these variables seems a promising undertaking.

Third, and lastly, an interesting research project would be replicating the present study using data for a single metropolitan area. This would enable us to look at the determinants of households’ specific loca- tional choices. Although the LNU would not yield sufficient data for such a study, observations from the Swedish census, Folk- och bostadsrakningen, seem well suited to this task.

Acknowledgements-The authors wish to acknowledge the generous financial support of Delegationen for Arbets- marknadsforskning (EFA) of the Swedish Ministry of Labour, and Institutionen for Social Forskning (SOFI) at Stockholm University for making available the survey data that we employ. We also wish to thank Kurt Brannas, Gerard Evers, Karl-Gustaf Lofgren, Michael Nelson and two anonymous referees of this journal for their com- ments. The authors alone assume responsibility for any errors.

NOTES

1. Empirical studies of the effects of fiscal variables on migratory behaviour include COURCHENE, 1970; ARONSON and SCHWARTZ, 1973; CEBULA, 1979; HOLMLUND, 1984, chapter 5; HERZOG and SCHLOTTMAN, 1986; SHAW, 1986; Fox et al., 1989; NELSON and WYZAN, 1989; and DAY, 1992.

2. There has, in recent years, been a large number of articles that address the issue of whether or not local labour markets are in equilibrium and the implications of this determination for empirical migration models. In his recent review of this literature, HUNT, 1993, states that the empirical evidence creates a ‘strong presumption in favour of the hypothesis that disequilibrium models that fail to specify amenity factors are misspecified’ (p. 345). CHARNEY, 1993, argues that, for similar reasons, public sector attributes must be controlled for in studies of migration, along with land rent and wages. In this paper we deal with fiscal variables but not with amenities. This practice is motivated both by our desire to focus on the former and by the considerably greater difficulty of coming up with meaningful amenity variables for Sweden than for the USA. An attempt to include in our estimated migration equations an amenity variable-local employment in the cultural sphere-failed to yield significant results. In any case, as discussed by HERZOG and SCHLOTTMAN, 1993, many amenities and disamenities are closely related to urban scale, and we present results separately for metropolitan and non-metropolitan Sweden.

3. The fact that both fiscal externality and differences in per capita land rents contribute to the inefficiency means that neither a head tax to eliminate the former nor a Henry George-type tax on land rents to eliminate the latter would suffice to do away with that inefficiency; for a comprehensive look at this issue, see MIESZKOWSKI and ZODROW, 1989. A clear discussion of the model of BOADWAY and FLATTERS, 1982, and an application to a Japanese underpopulated region can be found in HAGIHARA and HAGIHARA, 1991.

4. A case may be made for employing the sum of the per capita tax base and per capita TEG as a single independent variable rather than including each of them as a separate one. Such practice would impose an untested restriction on the effects of per capita TEG on migratory behaviour but would have the advantage of avoiding potential multicollinearity problems between the two variables. We return briefly to this issue in the section describing our empirical results.

5. Previous empirical work on Swedish migration that has included fiscal variables has focused almost exclusively on the local income tax rate (HOLMLUND, 1984, chapter 5; EDIN et al., 1991), although NELSON and WYZAN, 1989, also examine TEG, and WESTERLUND, 1993, includes the ‘starting assistance’ offered to previously unemployed migrants as they begin life in their new localities.

6. Differences in the age or social structure of the local population cause variation in municipal expenditures on mandatory local public service provision. Although municipalities receive conditional grants from the central government that contribute to the cost of providing such services, these grants do not fully cover those costs. In our empirical work, we include a measure of the local age and social structure in our migration equations in order to hold such factors constant. The manner in which we do so is described below.

7. Both of the considerations discussed in this paragraph apply with particular force to the results obtained by Fox et al., 1989.

8. For examples of studies employing binary logit that deal with either fiscal variables or Sweden, or both, see HOLMLUND, 1984, chapter 5; HERZOG and SCHLOTTMAN, 1986; Fox et al., 1989; and HARKMAN, 1989. The authors report elsewhere (WESTERLUND and WYZAN, 1993) the results of the estimation of binary-logit versions of the migration equations described in the present paper.

9. The results for single-adult households are, in general, fairly similar to those reported here for their multiple- adult counterparts. However, certain non-fiscal variables excluded here on grounds of insignificance, such as whether the respondent commuted to work or was unemployed, are sometimes significant for short-distance migration for single-adult households.

10. At the suggestion of an anonymous referee, we have estimated a nested-logit model, where the first stage concerns whether to move and the second stage whether to migrate a short or a long distance. The results of this specification, which do not differ markedly for the fiscal variables of interest from the ones reported below, are available upon request.

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Olle Westerlund and Michael L. Wyxan

11. Migration is ‘observed’ when the officially registered place of residence changes between these two points in time. There is thus a ‘hidden component’ consisting of repeat migration within the relevant time span. As a result, reported migration rates are understated.

12. As noted by JONSSON and SIVEN, 1986, p. 31, ‘regional wage differentials are small and correspond poorly with regional imbalances in labour demand’.

13. On Sweden, see HOLMLUND, 1984; and NELSON and WYZAN, 1989; see also HERZOG and SCHLOTTMAN, 1982; and DAVANZO, 1983.

14. A number of other explanatory variables were included in preliminary runs but never proved statistically significant, so that they were excluded from the specifications reported here. These include whether the respondent had a complete university education, the number of gainfully employed family members, whether the respondent or spouse received unemployment benefits during 1980-1 and whether he spent more than four hours per week commuting between his sites of residence and work. Thus, unlike virtually all of the studies reviewed by HERZOG et a l . , 1993, those who are unemployed do not exhibit different migratory behaviour from those who are not. It must be admitted, however, that our variable, aimed

at capturing whether the respondent received Unemployment benefits, leaves a great deal to be desired.

15. All empirical work described here was carried out with the LIMDEP statistical package, which employs Newton’s method to solve the non-linear optimization problems posed by logit estimation.

16. Note that A G E F A C is never close to significance and that removing it from the specification has no effect on the estimated coefficients on the other explanatory variables. In an effort to avoid possible multi- collinearity between the per capita tax base and per capita TEG, we have also estimated models in which the sum of these two variables is employed. In the case where both are significant individually (i.e. for short-distance migration in non-metropolitan areas) their sum is also significant. For the results of estimating migration equations containing a variety of different combinations and definitions of the fiscal variables, see WESTERLUND and WYZAN, 1993.

17. Note also that including per capita municipal expenditures in the specification does not affect the estimated coefficients on the other explanatory variables, including the tax rate. There is thus little evidence of the presence of an identity linking the two sides of the municipal budget.

REFERENCES

A R O N S O N J. R. and SCHWARTZ E. (1973) Financing public goods and the distribution of population in a system of local

BHATTACHARYA G. (1990) Migration under uncertainty about quality of locations, J . Econ. Dynamics G Coritvoll4, 721-39. BOADWAY R. and FLATTERS F. (1982) Efficiency and equalization payments in a federal system of government: a synthesis

CERULA R. J. (1979) A survey of the literature on the migration-impact of state and local government policies, Public

CHARNEY A. H. (1993) Migration and the public sector: a survey, Reg. Studies 27, 313-26. COURCHENE T. J. (1970) Interprovincial migration and economic adjustment, Can. J . Econ. 3, 550-76. DAVANZO J. (1983) Repeat migration in the United States: who moves back and who moves on?, Rev. Econ. Stat. 65,

DAY K. M. (1992) Interprovincial migration and local public goods, Can. J . Econ. 25, 12-4. EDIN P.-A., JANSSON M. and NILSSON C. (1991) Inter-regional migration in Sweden: the effects of labor market

conditions, labor market programs and the price of housing, Department of Economics, Uppsala University (mimeo). FALARIS E. M. (1987) A nested logit model with selectivity, Int. Econ. Rev. 28, 429-43. FALARIS E. M. (1988) Migration and wages of young men, J . Human Resources 23, 514-34. FLANAGAN R. J . (1987) Efficiency and equality in Swedish labor markets, in BOSWORTH B. P. and RIVLIN A. M. (Eds)

FOX W. F., HERZOG H. W. JR and SCHLOTTMAN A. M. (1989) Metropolitan fiscal structure and migration,J. Reg. Sci.

GREEN W. H. (1990) Econometric Analysis. Macmillan, New York. HAGIHARA K. and HAGIHARA Y. (1991) The role of intergovernmental grants in underpopulated regions, Reg. Studies 25,

HANSEN E. R. (1987) Industrial location choice in Siio Paulo, Brazil, Reg. Sci. G Urban Econ. 17, 89-108. HARKMAN A. (1989) Migration behaviour among the unemployed and the role of unemployment benefits, Papers Reg.

HENLEY A., CARRUTH A., THOMAS A. and VICKERMAN R. (1989) Location choice and labour market perceptions: a

HERZOG M. W. JR and SCHLOTTMAN A. M. (1982) Moving back versus moving on: the concept of home in the decision

HERzoG H. W. JR and SCHLOTTMAN A. M. (1986) State and local tax deductibility and metropolitan migration, National

HERZOG H. W. JR and SCHLOTTMAN A. M. (1993) Valuing amenities and disamenities of urban scale: can bigger be

government, National Tax J . 26, 137-60.

and extension of recent results, Can. J. Econ. 15, 613-33.

Finance 34, 69-84.

552-9.

The Swedish Economy, pp. 125-73. Brookings Institution, Washington, D.C.

29, 523-36.

163-72.

Sci. A s s . 66, 143-50.

discrete choice study, Reg. Studies 23, 431-45.

to remigrate, J. Reg. Sci. 22, 73-82.

Tax J. 39, 189-200.

better?,]. Reg. Sci. 33, 145-65.

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4

Household Migration and the Local Public Sector 157

HERZOG H. W. JR, SCHLOTTMAN A , M. and BOEHM T. P. (1993) Migration as spatial job-search: a survey of empirical

HOLMLUND B. (1984) Labor Mobility: Studies of Labor Turnover and Migration in the Swedish Labor Market. Industriens

HUNT G. L. (1993) Equilibrium and disequilibrium in migration modelling, Reg. Studies 27, 341-9. JONSSON L. and SIVEN C.-H. (1986) Varfdr loneskillnader! Norstedts Tryckeri, Stockholm. MCCALL B. P. and MCCALL J . J. (1987) A sequential study of migration and job search, J . Labor Econ. 5, 452-76. MIESZKOWSKI P. and ZODROW G. R. (1989) Taxation and the Tiebout model: the differential effects of head taxes, taxes

MOLHO I. (1986) Theories of migration: a review, Scot. J . Pol. Econ. 33, 396-419. NELSON M. A. and WYZAN M. L. (1989) Public policy, local labor demand, and migration in Sweden, 1979-84,J. Reg.

REITSMA R. F. and VERGOOSSEN D. (1988) A causal typology of migration: the role of commuting, Reg. Studies 22,331-40. SHAW R. P. (1986) Fiscal versus traditional market variables in Canadian migration, J. Pol. Econ. 94, 648-66. SKEDINGER P. and MCCORMICK B. (1991) Why do regional unemployment differentials persist?, in SKEDINGER P. (Ed)

Skatteutjiimningssystemet, Betunkande av skatteutjiimningskommitte'n (SOU) (1985) Stockholm (referred to as S O U 1985:61). TIEBOUT C. M. (1956) A pure theory of local expenditures, J. Pol. Econ. 64, 416-24. WESTERLUND 0. (1993) Internal gross migration in Sweden: the effects of variation in mobility stimuli and regional labor

market conditions, in WESTERLUND 0. (Ed) Internal Migration in Sweden: The Role of Fiscal Variables and Labor Market Conditions, Umei Economic Studies no. 293. Umei University, Umei.

WESTERLUND 0. and WYZAN M. L. (1993) The determinants of household migration in Sweden: the role of the public sector, in WESTERLUND 0. (Ed) Internal Migration in Sweden: The Role uf Fiscal Variables and Labor Market Conditions, Umei Economic Studies no. 293. Umei University, Umei.

WHEATON W. C. (1975) Consumer mobility and community tax bases: the financing of local public goods, J . Pidblic Econ.

findings, Reg. Studies 27, 327-40.

Utredningsinstitut, Stockholm.

on land rents, and property taxes, J . Econ. Lit. 27, 1,098-146.

Sci. 29, 247-64.

Essays on Wage Furmation, Employment, and Unemployment, pp. 145-72. Uppsala University, Uppsala, Sweden.

4, 377-84.

Dow

nloa

ded

by [

Uni

vers

ity o

f Il

linoi

s at

Urb

ana-

Cha

mpa

ign]

at 0

1:55

07

Oct

ober

201

4