You are on page 1of 12

International Tax and Public Finance, 7, 247258, 2000.

c 2000 Kluwer Academic Publishers. Printed in The Netherlands.

Local Public Funding of Higher Education when


Skilled Labor is Imperfectly Mobile
MOSHE JUSTMAN
justman@bgumail.bgu.ac.il
Department of Economics and Monaster Center for Economic Research, Ben Gurion University of the Negev,
Beer Sheva, Israel
JACQUES-FRANCOIS

THISSE
CORE, Universite Catholique de Louvain, Louvain-la-Neuve, Belgium and CERAS, Ecole Nationale des Ponts et
Chaussees, Paris, France

Abstract
If higher education is publicly funded by local (sub-federal) jurisdictions, while skilled labor is heterogeneous in
responding to wage differentials between jurisdictions, the spillovers that result give rise to a disparity between the
centralized output-maximizing allocation of resources to higher education and decentralized equilibria. Generally,
decentralization leads to under-provision, which can be offset by inter-jurisdictional subsidies based on gross
migration flows. But the extent of the discrepancy depends on the local balance of political forces. Indeed,
when the welfare of native-born emigrants is highly valued while new immigrants carry little political weight,
over-provision in equilibrium is possible.
Keywords: education, labor mobility, local public finance, fiscal federalism
JEL Code: I22, H77, J61

1.

Introduction

The greater mobility of labor in many parts of the world, and especially the increased
integration of the European Union, have motivated extensive re-examination of economic
and social policies that were designed in more stable times to provide long-term benefits
to well defined geographic constituencies. Increased labor mobility undermines the ability
of those who pay for these policies to capture their full benefits, thus possibly eroding
essential political support.1 The present paper explores this issue with regard to higher
education, a major category of public spending that may be similarly affected by increased
labor mobility.
Education, in general, accounts for as much as 5% of GNP, and 10% or more of public
spending in advanced industrialized countries, with public funding covering, on average,
almost 90% of education costs in these countries. Higher education typically accounts for
1520% of overall education expenditures. In some countries it is freely provided to all
qualified applicants (France is a prominent example); in others it is heavily subsidized and
closely regulated with regard to enrollment, salaries, construction, and degree programs
(e.g., Great Britain, Israel). In the U.S., public universities are subsidized at the state
level.

248

JUSTMAN AND THISSE

Migration of skilled labor implies that those who pay the bill for public higher education
may find it difficult to fully capture its benefits. In the past, such concerns were directed
primarily to brain drainsuni-directional flows of highly skilled labor from Third-World
countries. More recently, increased integration of labor markets, especially within the European Union (EU), has drawn attention to problems that arise from bi-directional movement
of skilled labor between similarly developed countries. Gordon and Jallade (1996) report
that in 1993/4 mobile foreign students in EU countries (i.e., students not previously resident in the country where they were studying) numbered over 95,000. In the United States
inter-state mobility has a similar effect: 5% of young adults, in the 2029 age group, move
from one state to another each year. In both cases, the impact of tertiary education transcends the boundaries of the local (sub-federal) jurisdictions where it is funded and where
its budgets are politically determined.2 In the absence of some form of inter-jurisdictional
compensation for these trans-local benefits, large flows of skilled labor can be expected to
erode political support for local public funding of higher education.3 Yet autonomous local
education is often valued in its own right as a key element of cultural and political autonomy.
This raises the question of how it can be efficiently reconciled with the movement of labor
within a federation of local jurisdictions.
In this paper we examine these issues in the context of a simple formal model in which we
assume free public provision of higher education. To fix ideas, we consider a two-region,
one-sector economy with two factors of production: skilled (i.e., educated) labor, assumed
to move between regions with a probability that depends on wage and amenity differentials; and a composite immobile factor of production, locally available in fixed aggregate
supply.4 By focusing on a probability of moving, we explicitly account for the fact that
some skilled workers have a higher mobility than others. Such a migration pattern provides
a much more realistic description of real world migration than the one assumed under the
standard assumption of perfect mobility. We assume that human capital is acquired solely
through publicly funded higher education, and compare the centralized, output-maximizing
allocation of resources to higher education with decentralized outcomes, characterized as
Nash equilibria among local authorities, each addressing a balance of local political interests. We show that the local balance of forces, particularly the relative political weight of
immigrants and emigrants, has a substantive impact on public education in a decentralized
equilibrium;5 and, in contrast to a brain drain, where one side gains and the other loses,
bi-directional flows offer clear scope for mutual benefits through coordination.
This approach is closely related to an earlier literature that similarly asked whether decentralization in the provision of local public goods leads to under-provision. Thus we
generally agree with Weisbrods (1964) claim that decentralization should give rise to
under-investment, but qualify it with a more detailed analysis that specifically describes the
conditions under which it is likely to occur.6 This also confirms and elaborates on Williams
(1966) intuition that whether under- or oversupply holds depends on the shape of the reaction curves. It is also related to much of the recent literature that examines the balance
between local and state funding of K12 education in the United States, which similarly
incorporates explicit modeling of decentralized political decisions on public education.
However, this literature largely focuses on the tradeoff between equity and variety, on the
interaction between public schools and housing markets, and on spillovers within school

LOCAL PUBLIC FUNDING OF HIGHER EDUCATION

249

districts (see, e.g., Hoxby (1996) for a wide-ranging survey). Our focus on higher education
leads us to emphasize efficiency considerations that relate to inter-jurisdictional spillovers.
Finally, our work follows previous models of labor migration that look specifically at the
relation between training and wage-based migration (Leach, 1996). We extend this work
in our more explicit treatment of its political dimension, recognizing that there are diverse
interests which may affect local education expenditures.7
The remainder of the paper is organized as follows. The model is presented in Section 2.
The centralized, output-maximizing allocation is characterized in Section 3. This is followed by a positive analysis of decentralized political equilibria in Section 4. Section 5
draws implications from the analysis for possible policy responses to disparities between
centralized and decentralized outcomes. Section 6 concludes.
2.

The Model and Some Preliminary Results

Consider a two-region, one-sector economy in which production is local, and each region
i = 1, 2 has access to the same technology. The production of the representative firm in
region i is a function of two factors:
yi = f (h i , m i ) i = 1, 2

(1)

where h i denotes human capital (skilled workers), and m i is the regionally immobile factor
of production. We assume that f has constant returns to scale, satisfies the Inada conditions
in each of its arguments, and has positive mixed derivatives. We further assume that all
factor and product markets are competitive, and that production is carried out by a unit
measure of identical, perfectly competitive firms, so that in competitive equilibrium h i and
m i will represent both firm level quantities and local aggregates. Aggregate local supply of
the composite immobile factor of production m i is assumed to be fixed in each region.
The supply of human capital, h i , available for production in region i is determined through
education and migration. To simplify the analysis we assume that education is exclusively
provided by the public sector in each region, an assumption that is more accurate in some
contexts (e.g., the EU) than in others (the US). We posit that the marginal cost of education
per unit of human capital, c, is constant and identical in all regions. Denote the amount of
education (the number of school places) provided in region i by si . Then the local education
budget in region i is csi . We refer to si as region is education policy and assume that it is
financed by means of lump-sum taxes; these are levied within each region at decentralized
political equilibria, and within the federation at the centralized solution.
The possibility of migration implies that si is not necessarily the amount of human capital
available for production in region i. We allow that after they have been trained, skilled
workers may decide not to remain in their native region, choosing to work in the alternative
locale in response to inter-regional differences in wages (as in the neoclassical model of
factor mobility) and amenities (as in the theory of hedonic migration). Consequently,
production in each region involves skilled workers trained in both regions, and there are
reciprocal spilloversspillins and spilloutsbetween regions.8
We describe the inclinations of skilled workers to migrate using a discrete choice model,
in much the same way that consumer preferences for differentiated products are modeled

250

JUSTMAN AND THISSE

(Anderson et al., 1992). Denote the endogenously determined probability of a worker


trained in region i migrating to region j 6= i by pi j , where
pii = 1 pi j

i, j = 1, 2 and i 6= j

(2)

is the probability of not migrating. Then h i is determined by these probabilities and by the
education policies in the two regions:
h i = si pii + s j p ji

i, j = 1, 2 and i 6= j

(3)

The probabilities of migration derive from the following specification of individual utility.
Assume the utility of skilled worker k born in region i and working in region j is
Uki j = w j + ai j + ki j

(4)

where 0 is a common parameter that reflects the responsiveness of skilled workers


to wage differences: when the wage response is overwhelming and skilled labor
relocates until wages are equal in the two regions; and when = 0 mobility responds only
to amenity differentials, and the probability of moving is exogenous with respect to wages.
The parameters ai j represent amenity differentials and moving costs, and in keeping with
the symmetric setting of the model we assume that a12 = a21 and a11 = a22 . Furthermore,
because moving costs are positive, and individuals are likely to have more affinity with
their region of origin, we assume that a11 > a12 . Finally, skilled workers are heterogeneous
in their perception of the features associated with a given region, and hence have different
matching values. However these values are not observable by local governments. They
are represented by the random variable ki j that accounts for the specific matching value
between skilled worker k and region j.
Following a well-established tradition in migration theory (see, e.g., Anderson and Papageorgiou, 1994a,b), we assume that, for given i and j, the ki j are identically and independently distributed across individuals according to the double exponential with zero mean
and variance equal to 6/ 2 (McFadden, 1974) so that the probabilities pi j are given by the
logit:9
pi j =

ewj +ai j
ewi +aii + ewj +ai j

(5)

or, equivalently,
pi j =

e(wj wi )
D + e(wj wi )

(6)

where logD = aii ai j, represents the advantages of remaining working where one was
trained, and by our assumption of a natural proclivity for ones native region, D > 1. A
similar relationship holds for pii :
pii =

D
D + e(wj wi )

(7)

LOCAL PUBLIC FUNDING OF HIGHER EDUCATION

251

Competitive equilibrium in the factor markets gives us the following equations:


f 1 (h i , m i ) = wi
f 2 (h i , m i ) = ri

i = 1, 2

(8)

i = 1, 2

(9)

where the subscripts of f denote its partial derivatives, wi is the factor price of human capital,
and ri is the factor price of the immobile factor. Given endowments of the immobile factor
{m 1 , m 2 } and education policies {s1 , s2 }, we shall say that an allocation of human capital
{h 1 , h 2 } is a competitive migration equilibrium if there exist non-negative factor prices such
that equations (1)(3) and (6)(9) are satisfied.
The lemma below provides a characterization of the migration probabilities at the symmetric outcome.
Lemma 1 Assume the two regions have identical endowments of the immobile factor,
m 1 = m 2 = m, and the economy is in a competitive migration equilibrium for a symmetric
pair of education policies, s1 = s2 = s. Then dh i /dsi is between 1/2 and p D/(1 + D),
and decreasing in : if = 0 then dh i /dsi = p, and if then dh i /dsi 1/2.
Proof: From (1)
h 1 = p11 s1 + p21 s2

(10)

h 2 = p12 s1 + p22 s2

(11)

so that
dh 1 /ds1 = p11 + s1 (dp11 /ds1 ) + s2 (d p21 /ds1 )

(12)

dh 2 /ds1 = p12 + s1 (dp12 /ds1 ) + s2 (d p22 /ds1 )

(13)

Thus the effect of local schooling on the local supply of skilled labor comprises a direct
effect, the proportion of skilled workers who would remain in the area given prevailing
wage differentials; and an indirect effect on the probability of migration, through its impact
on wages in the two regions. This can be further elaborated as:
dp11 /ds1 = ( p11 /w1 )(dw1 /ds1 ) + ( p11 /w2 )(dw2 /ds1 )

(14)

and similarly for dp12 /ds1 , dp21 /ds1 , and d p22 /ds1 .
Symmetry (s1 = s2 ) implies that also h 1 = h 2 and w1 = w2 ,10 and therefore that net
migration is zero in equilibrium. Hence we have p11 = p22 = p = D/(1 + D), while
p12 = p21 = 1 p = 1/(1 + D). By straightforward calculation, the derivative of the
probability of migration with respect to wage levels, at a symmetric equilibrium, equals:
pi j /w1 = p(1 p)

if j = l, and

pi j /w1 = p(1 p) if j 6= l.

(15)
(16)

The derivative of the local wage level with respect to local schooling equals:
dwi /dsi = f 11 (h i , m i )(dh i /dsi )

for i = 1, 2

(17)

252

JUSTMAN AND THISSE

and with respect to schooling in the other region it equals


dwi /ds j = f 11 (h i , m i )dh i /ds j

for i 6= j

(18)

Combining equations (12)(18), we obtain the following characterization of dh i /dsi in


symmetric equilibrium:
dh i /dsi =

1 + 2(1 p)( f 11 )s
(1/ p) + 4(1 p)( f 11 )s

(19)

If, as we assume, p > 1/2, then dh i /dsi is decreasing in , ranging between p, when
= 0, and 1/2, when .
3.

The Centralized Solution

We take as a benchmark the choice of education policies {s1 , s2 } and factor allocations
{h 1 , h 2 } that maximizes total (federal) output net of education costs, and focus exclusively
on the symmetric case (m 1 = m 2 = m). The necessary first-order conditions (FOC) set the
marginal product of human capital equal to the cost of training in each region:
f 1 (h i , m) = c

(20)

for i = 1, 2. This implies that h 1 = h 2 = h from the concavity of f in h. Moreover,


as wages are the same in both regions p11 = p22 = p, and therefore we must also have
s1 = s2 = h, i.e., no net migration. Thus, in a symmetric centralized solution s is determined
by:
f 1 (s, m) = c

(21)

It is readily verified that our assumptions on f preclude corner solutions, and imply that
second-order conditions hold and that the output-maximizing value of s is unique. This
will serve as a benchmark against which decentralized equilibria are compared.

4.

Decentralized Political Equilibria

We now consider decentralized Nash equilibria in which public investment in education is


determined by independent local governments. Such an equilibrium is a pair of education
policies {s1 , s2 } such that for i = 1, 2 each si maximizes the politically conditioned objective
function of region i, taking the education policy of the other region, s j , as given. The
sequence of interaction between local governments and factor markets is as follows: first,
each local government determines its education policy; then, skilled workers choose where
to live and work; finally, firms acquire their factors of production, produce their output, and
pay the factors of production their marginal products.

LOCAL PUBLIC FUNDING OF HIGHER EDUCATION

253

In determining its education policy, each local government anticipates the outcome of the
subsequent stages, weighing and balancing the diverse interests it represents. We identify
the following principal interests (for locale i):

the wages of local native skilled labor, wi pii si ;

the wages of local migrant skilled labor, wi p ji s j ;

the wages of native skilled workers who have emigrated, w j pi j si ;

the returns to local immobile assets, ri m i ;

Following Grossman and Helpman (1996), competition between these interests will result
in each local government maximizing an objective function which is a weighted sum of
these terms, less local education costs:
Bi = 1i wi pii si + 2i wi p ji s j + 3i w j pi j si + 4i ri m i csi

(22)

Specific weights {1I , . . . , 4I } will reflect the balance of political forces in particular
circumstances, varying from one region to the other. In what follows, we examine two polar
cases that demonstrate how the disparity between centralized and decentralized outcomes
depends on the political objective function. As before, we focus on symmetric settings,
which allows us to drop location subscripts. The two cases are:
1. Maximization of returns to local factors of production net of education costs (equal to
local output): 1 = 2 = 4 = 1 and 3 = 0.
2. Maximization of native output net of education costs, i.e., maximizing the sum of
returns on local immobile assets and the earnings of native skilled labor educated in the
region, whether remaining in the region or emigrating, while ignoring the earnings of
immigrant skilled labor: 1 = 3 = 4 = 1 and 2 = 0.
4.1.

Maximizing Local Output

The objective function in this case can be written simply as:


B1i = f (h i , m) csi

(23)

maximized over si for i = 1, 2. Assuming that a Nash equilibrium exists, the choice of the
local government i must therefore satisfy the first-order conditions:11
f 1 (h i , m)dh i /dsi = c

i = 1, 2

(24)

It then follows from Lemma 1, from which dh i /dsi < 1, and the declining marginal
productivity of human capital, that decentralized maximization of output leads to less local
expenditure on education than global output maximization. In two extreme cases, when

254

JUSTMAN AND THISSE

migration is either entirely insensitive or fully responsive to wage differentials, we can


derive explicit terms for the equilibrium wage.
(1) When migration is insensitive to wage differentials ( = 0), we have from Lemma 1
that dh i /dsi = p, where p = D/(1 + D) in symmetric equilibrium. Hence
f 1 (s, m) = c/ p

(25)

(2) When wages dominate the migration process overwhelmingly ( ), we have


from Lemma 1 that dh i /dsi = 1/2. Hence
f 1 (s, m) = 2c

(26)

which is higher than the equilibrium wage when migration is insensitive to wage differentials, as p > 1/2. Thus decentralized maximization of local output leads to underinvestment in education, and the shortfall is greater the more strongly migration responds
to pecuniary incentives.
4.2.

Maximization of Native Product

We use the term native product of region i to refer to the factor payments that accrue to
native-born skilled labor, whether resident in the region of not, and to the immobile local
asset. The objective function of region i is then:
B3i = pii si f 1 (h i , m) + pi j si f 1 (h j , m) + m i f 2 (h i , m) csi

(27)

maximized over si , where j 6= i. In this case, local governments ignore the negative external
effect of their education expenditures on the earnings of non-native skilled labor and the
positive external effect of these expenditures on the earnings of the immobile asset in the
other region. Under the current specification, the net result is unambiguous: maximization
of native output leads to over-investment in education. The extent of such over-investment
depends inversely on the responsiveness of migration to wage differentials: it is smallest
when migration is most responsive to wage differentials (large ), vanishing entirely in the
limit as , in which case the decentralized outcome coincides with centralized output
maximization. (A detailed derivation is provided in the Appendix.) This captures some
of the political motivation behind Third-World over-investment in higher education (Dore,
1976): local political elites value economic opportunity for their offspring, who have the
option of emigrating, far above the welfare of immobile low-skilled labor.12
5.

Federal Intervention

The preceding analysis identified disparities between the level of higher education indicated
by centralized federal maximization of total (federal) output and that which would be
chosen by local governments acting independently to pursue diverse political-economic
goals while interacting strategically. Such disparities indicate the possibility of improving
on the decentralized outcome through federal participation in local education budgets.

LOCAL PUBLIC FUNDING OF HIGHER EDUCATION

255

Our analysis indicates several factors that bear on the design of such a system of fiscal
intervention. It highlights the importance of understanding the political balance of local
interests that shapes education policy at the sub-federal level. If political interests are predominantly defined in geographic terms, and local governments represent the full spectrum
of local interests, so that they act to maximize regional output, our analysis indicates federal subsidization of regional education costs in proportion to gross inter-regional migration
rates of graduates.13 But federal subsidization of regional budgets is not called for if the
interests of native-born highly skilled emigrants are politically well cared for, in which case
investment in higher education may be excessive rather than deficient.
There are, of course, alternatives to such cross-subsidization, but not without serious
drawbacks. Centralized coordination of education budgets threatens the autonomy of subfederal education systems which may be valued for their role in preserving distinct cultural
identities. Alternatively, tuition could be financed by a government loan, payment of which
would be forgiven in proportion to the number of years worked in the jurisdiction one
has studied, but this raises problems of enforcing repayment by those who leave. Some
totalitarian regimes have conditioned immigration of educated persons on payment of an
exit fee, but this would not be consistent with prevailing norms in many countries.
6.

Concluding Remarks

The increasing mobility of skilled labor, notably within the European Union, implies farreaching changes in the geographic incidence of the costs and benefits of publicly-funded
higher education. The present paper explores the political-economic dimensions of these
changes in the context of a simple formal model in which education is publicly provided
and the educated are imperfectly mobile. It highlights the need for coordination between
sub-federal jurisdictions in this context, as well as its complexity.
The extent and direction in which decentralized political equilibria depart from coordinated output maximization will depend on the balance of political forces at the local level,
as well as on the underlying parameters that shape both inter-regional patterns of migration
and returns to schooling. Decentralization leads to under-investment in education when
political interests are predominantly defined in geographic terms and local governments act
to maximize regional output, but may lead to over-investment when the political interests
of native-born highly educated emigrants are well represented.
Several important avenues for further work are indicated. We view education in isolation, ignoring its interaction with other semi-local public goods, such as social insurance,
which may have an added effect on the costs and benefits of coordination. For example,
disparities in the level of human capital between two locales may encumber coordination
of their social insurance systems. We focus on symmetric equilibria although the most
interesting problems may arise in asymmetric cases, especially when one locale is a major
agglomeration, as in new models of economic geography (Fujita and Thisse, 1996). We
do not consider the possibly of supplanting or supplementing public education by private
expenditures on education (Gradstein and Justman, 1996). And we use a static framework
to analyze issues that have important dynamic dimensions. These avenues of research invite
further elaboration.

256

JUSTMAN AND THISSE

Appendix: Maximization of Native Product Leads to Over-Investment


Proof: The objective function in this case is given by (31):
B3i = pii si f 1 (h i , m i ) + pi j si f 1 (h j , m j ) + m i f 2 (h i , m i ) csi
The first order condition for region i is then:
d B3i /dsi = (dpii /dsi )si f 1 (h i , m i ) + pii f 1 (h i , m i ) + pii si f 11 (h i , m i )dh i /dsi
+(dpi j /dsi )si f 1 (h j , m j ) + pi j f 1 (h j , m j ) + pi j si f 11 (h j , m j )dh j /dsi
+m i f 21 (h i , m i )dh i /dsi c = 0
In symmetric equilibrium we can write:
d B3i /dsi = (dpii /dsi )s f 1 (s, m) + p f 1 (s, m) + ps f 11 (s, m)dh i /dsi
+(dpi j /dsi )s f 1 (s, m) + (1 p) f 1 (s, m) + (1 p)s f 11 (s, m)dh j /dsi
+m f 21 (h i , m)dh i /dsi c
As pii + pi j 1, dpii /dsi +dpii /dsi = 0; and as h 1 +h 2 = s1 +s2 , dh i /dsi = 1dh j /dsi .
Moreover, f 21 = f 12 ; and as f is CRS, f 1 is homogeneous of degree 0, so that h f 11 +m f 12 =
0 from which it follows that m f 21 = s f 11 . Therefore:
d B3i /dsi = f 1 + s f 11 [ pdh i /dsi + (1 p)(1 dh i /dsi ) dh i /dsi ] c
= f 1 + s( f 11 )[(1 p)(2dh i /dsi 1)] c
from which it follows that f 1 c, with equality only if p = 1 or = .

Acknowledgments
We gratefully acknowledge the helpful comments of Jack Mintz, Pierre Pestieau and the
referees. Financial help from Ben Gurion University (Israel) and Fonds National de la
Recherche Scientifique is also acknowledged.
Notes
1. Redistributive policies have recently received increased attention in this regard. See, e.g., Wildasin (1991,
1995, 1997), Cremer and Pestieau (1996), LeiteMontero (1997) and the survey by Cremer et al. (1996).
2. See De Ville et al. (1996) for further discussion of the costs and benefits of student mobility. Throughout the
paper we use the term local in this generic sense that refers to the sub-federal jurisdictions where political
decisions are taken on the funding of higher education, and between which there is free movement of labor. It
includes the sovereign nations of the European Union as well as the states of the United States.
3. In Austria and Belgium, for example, where about 5% of the student population comes from other EU countries,
there has been vigorous political debate on how they should be financed. When Belgium required full tuition
from students coming from other EU countries but not from its own nationals, the European Court of Justice
condemned its action (in its Gravier Judgment of 1985), adding a further dimension to the problem.

LOCAL PUBLIC FUNDING OF HIGHER EDUCATION

257

4. Implicit in this formulation are unspecified factors of production, such as capital, which are assumed to be fully
mobile, i.e., available in perfectly elastic supply on world markets. Though clearly an extreme assumption,
this distinction between perfectly mobile factors of production and immobile local assets is a simple way
of capturing observed differences in mobility between different factors of production. Thus in a preliminary
version of this paper, we explicitly assumed that unskilled labor is immobile, exaggerating empirically observed
differences between skilled and unskilled labor (Shields and Shields, 1989, pp. 279287). Of course, other
interpretations are possible.
5. The difficulty of determining a single appropriate government objective function when the population is
mobile underlies Mansoorian and Myers (1997) analysis.
6. Our explicit modeling of the effect of education policies on migration flows undermines the arguments that
support Weisbrods claim.
7. Issues of international coordination, which it raises, are a frequent theme in trade theory. See, for example,
Spencer and Branders (1983) analysis of R&D subsidies, or the more recent work of Devereux and Mansoorian
(1992) on fiscal policy and growth. Gradstein and Justman (1995) consider the implications of capital mobility
for coordination of higher education. As noted above, the present analysis is also related to recent work on
the implications of labor mobility for national redistributive policies.
8. The model could be extended in an obvious way to allow for migration in or out of the economy as a whole.
9. The axiomatic foundations of the logit are discussed in Anderson et al. (1992, Ch. 2).
10. If h 1 > h 2 then w1 < w2 and so p11 < p22 and p21 < p12 , implying h 1 < h 2 , a contradiction.
11. The second-order condition is f 11 (dhi/dsi )2 + f 1 (d 2 h/dsi2 ) < 0. We assume that it holds.
12. A third case of possible interest is maximization of local property values, a standard approach in regional
economics (cf. Wildasin, 1986). In the present framework, this implies maximization of local returns to
immobile assets, net of education costs, subject to the constraint of free public education, i.e., 4 = 1 and
1 = 2 = 3 = 0. Then, at a symmetric equilibrium we would have f 1 (s, m) = (1/ )[c/(dh i /dsi )] (a
detailed derivation is available on request). Comparing this with maximization of local output, we find that
it further compounds under-investment in higher education under decentralization (assuming is less than
1). However, this result only applies if property owners hold sufficient political power to determine public
investment in higher education, but not enough to impose tuition fees and maximize property values without
the constraint of free education.
13. Further analysis indicates that when the political balance of forces at the sub-federal level is dominated by the
owners of real property, possibly because local tax revenues draw on these assets disproportionately, a higher
rate of subsidization is called for (see note 12, above).

References
Anderson, S. P., A. de Palma and J.-F. Thisse. (1992). Discrete Choice Theory of Product Differentiation.
Cambridge, MA: MIT Press.
Anderson, W. P. and Y. Y. Papageorgiou. (1994a). An Analysis of Migration Streams for the Canadian Regional
System, 19521983. 1. Migration Probabilities. Geographical Analysis 26, 1536.
Anderson, W. P. and Y. Y. Papageorgiou. (1994b). An Analysis of Migration Streams for the Canadian Regional
System, 19521983. 2. Disequilibrium. Geographical Analysis 26, 110123.
Cremer, H., V. Fourgeaud, M. Leite Monteiro, and P. Pestieau. (1996). Mobility and Redistribution: A Survey.
Public Finance 51, 325352.
Cremer, H. and P. Pestieau. (1996). Distributive Implications of European Integration. European Economic
Review 40, 747757.
Devereux, M. and A. Mansoorian. (1992). International Fiscal Policy Coordination and Economic Growth.
International Economic Review 33, 24968.
De Ville, P., F. Martou, and V. Vanderberghe. (1996). Cost-Benefit Analysis and Regulatory Issues of Student
Mobility in the European Union. European Journal of Education 31, 205222.
Dore, R. (1976). The Diploma Disease: Education, Qualification and Development. Berkeley and Los Angeles:
University of California Press.
Fujita, M. and J.-F. Thisse. (1996). Economics of Agglomeration. Journal of the Japanese and International
Economies 10, 339378.

258

JUSTMAN AND THISSE

Gordon, J. and J.-P. Jallade. (1996). Spontaneous Student Mobility in the European Union: A Statistical
Survey. European Journal of Education 31, 133152.
Gradstein, M. and M. Justman. (1995). Competitive Investment in Higher Education: The Need for Policy
Coordination. Economics Letters 47, 393400.
Gradstein, M. and M. Justman. (1996). The Political Economy of Mixed Public and Private Schooling: A
Dynamic Analysis. International Tax and Public Finance 3, 297310.
Grossman, G. and E. Helpman. (1996). Electoral Competition and Special Interest Groups. Review of Economic
Studies 63, 265286.
Hoxby, C. M. (1996). Are Efficiency and Equity in School Finance Substitutes or Complements? Journal of
Economic Perspectives 10, 5172.
Leach, J. (1996). Training, Migration, and Regional Income Disparities. Journal of Public Economics 61,
429443.
LeiteMontero, M. (1997). Redistributive Policy with Labor Mobility Across Countries. Journal of Public
Economics 65, 229244.
Mansoorian, A. and G. M. Myers. (1997). On the Consequences of Government Objectives for Economies with
Mobile Populations. Journal of Public Economics 63, 265281.
McFadden, D. (1974). Conditional Logit Analysis of Qualitative Choice Behavior. In P. Zabramka (ed.),
Frontiers in Econometrics. New York: Academic Press, 105142.
Shields, G. M. and M. P. Shields. (1989). The Emergence of Migration Theory and a Suggested New Direction.
Journal of Economic Surveys 3, 277304.
Spencer, B. and J. Brander. (1983). International R&D Rivalry and Industrial Strategy. Review of Economic
Studies 50, 707722.
Weisbrod, B. (1964). External Benefits of Public Education: An Economic Analysis. Princeton, NJ: Princeton
University Industrial Relations Center.
Williams, A. (1966). The Optimal Provision of Public Goods in a System of Local Government. Journal of
Political Economy 74, 1833.
Wildasin, D. E. (1986). Urban Public Finance. Chur: Harwood Academic Publishers. (Fundamental of Pure and
Applied Economics 10).
Wildasin, D. E. (1991). Income Redistribution in a Common Labor Market. American Economic Review 81,
757774.
Wildasin, D. E. (1995). Factor Mobility, Risk, and Redistribution in the Welfare State. Scandinavian Journal
of Economics 97, 527546.
Wildasin, D. E. (1997). Income Distribution and Redistribution within Federations. Annales dEconomie et de
Statistique 45, 291313.

You might also like