J Popul Econ (2004) 17:157–175
DOI 10.1007/s00148-003-0131-8
Determinants of family migration: short moves vs.
long moves
Satu Nivalainen
Pellervo Economic Research Institute (PTT), Eerikinkatu 28A, FIN-00180 Helsinki,
Finland (Fax: þ358-9-34 888 500; e-mail: satu.nivalainen@ptt.fi)
Received: 18 January 2001/Accepted: 22 May 2002
Abstract. This paper examines factors underlying family migration. Based on
a sample of stable Finnish families, both short- and long-distance migration
is investigated. The empirical analysis carried out using multinomial logit
modelling shows a strong negative association between the family life-cycle
and migration. The findings indicate that migration takes place mainly due
to the demands of the husband’s career, resulting in the wives being tied
migrants. Two-earner families are less migratory, and in that sense the
husbands are tied stayers. Distance matters; several differences are noticed
between short- and long-distance migrants.
JEL-classification: C31, J61, R23
Key words: Family migration, ties, distance
1. Introduction
Migration can be defined as the movement of a population from one geographical
area to another. In fact, mobility has always been one of the fundamental
characteristics of the human species. However, while unattached
people are free to move, family relations may restrict the mobility of family
members. Hence, the factors underlying migration are likely to differ between
attached and unattached people. Therefore, the present study examines the
determinants of family out-migration in Finland. A family as defined here
Financial support from the Yrjo¨ Jahnsson Foundation (project no. 4271) is gratefully acknowledged.
I would like to thank two anonymous referees for their helpful comments on this paper. I
also wish to thank Kari Ha¨ ma¨ la¨ inen, Sari Pekkala and Hannu Tervo for their comments on an
earlier version of this paper. Responsible editor: John F. Ermisch.
consists of two married or cohabiting adults of opposite sexes, with or without
children. As the motives underlying different types of moves are likely to
vary, a distinction is made between short- and long-distance migration. Short
moves occur between municipalities but within a province, whereas long
moves are made between provinces1
.
During the last few years, the rate of migration in Finland has risen rapidly
in conjunction with the continuing urbanisation of the country. As a
consequence, increasing interest has been shown in migration research. In
spite of this, micro-economic analyses of migration have been in short supply
in Finland. Migration at the micro-economic level has recently been researched
(see Tervo 1997; Laakso 1998; Ritsila¨ and Tervo 1999; Ritsila¨ 2001;
Haapanen 2002; Pekkala 2002), but these studies have concentrated on individual
persons or workers. Family migration has not been investigated,
even though families account for about 80% of the Finnish population (Statistics
Finland 1995b). Furthermore, earlier studies have mainly investigated
long-distance migration, and have not dealt in particular with short moves.
Earlier Finnish research has shown that the most eager migrant is an
unmarried, educated, young adult (Korkiasaari 1991; Tervo 1997; Laakso
1998; Ritsila¨ and Tervo 1999). It has been noted that family status and children
affect migration propensities (Laakso 1998; Haapanen 1998), and that
the likelihood of moving decreases with the size of the household (Tervo
1997; Ritsila¨ and Tervo 1999). Longer moves seem to take place mainly for
job-related reasons, whereas housing and family matters are more important
in shorter moves (Korkiasaari 1991).
Migration is directed towards a few large towns located mainly in southern
Finland (Laakso 1998), and in-migration has become highly focused on
urban areas, with even middle-sized towns experiencing negative net migration
(Vartiainen 1997). Evidence shows that the migration process in Finland
has the feature of cumulative causation (Tervo 1997; Ritsila¨ and Tervo 1999),
which may lead to increasing disparities between areas. As in many European
countries, the trend in Finland seems to be towards a greater concentration
of the population and economic activity.
The aim of the present study is to answer two questions. First, what are
the factors underlying family migration? Specifically, are the family life-cycle
and family ties important in migration decisions? And second, are there any
differences between the determinants of short- and long-distance migration?
The study uses unique family data containing all the relevant characteristics
of both spouses. The empirical analysis is carried out with multinomial logit
models. The results show that, in Finland, young educated families are the
most eager to move, and that there is a strong negative association between
the family life-cycle and migration. The evidence indicates that more weight
is given to the husband’s career and, consequently, that wives are the tied
parties in family migration. At the same time, two-earner families are less
migratory, and in that sense the husbands of working wives are tied to certain
locations. Moreover, there are differences between short- and long-distance
migration, and these relate to both individual and regional characteristics.
The remainder of the paper is organised so that the second section
introduces the theoretical background and provides a short review of earlier
research relevant to the objectives of this study. The data, model and
variables used are presented in Sect. 3. Section 4 discusses the empirical
findings and Sect. 5 concludes the study.
158 S. Nivalainen
2. Theoretical framework
The theory of family migration introduced here draws mainly on Mincer
(1978), who uses Sjaastad’s (1962) human capital framework as a starting
point. The framework suggests that by devoting time to activities whose
benefits accrue in the future, individuals are making investments in their human
resources. Education and training are examples of such investments;
migration can also be regarded as an investment in human capital.
As families aim at maximising their total lifetime utility, those facing a
migration decision have to evaluate the profitability of this investment. In the
human capital setting, this is done by comparing the difference in returns
and costs that arise as a consequence of migration. The returns may include
a higher income in the destination, or a more pleasant environment (social
or physical), or some other non-monetary gains. In addition to the direct
expenses of moving2
, costs may derive from the psychological difficulties of
changing one’s environment and from uncertainty. Despite their potentially
great importance to some migrants, the non-monetary components are, for
simplicity, commonly excluded from the basic framework.
Thus, when a family is deciding whether to move or not, net family gain
from moving from location i to j is assessed. That is, the family weighs up the
present value of expected changes in future family income, net of the discounted
migration costs. The net family gain from relocation is the sum of
the personal net gains of the family members. Migration takes place only if
this sum is greater than zero, i.e., if the family’s utility increases as a result of
moving. From the set of J possible locations, the family chooses the one that
maximises the gains of the family as a whole.
The complexity of the migration decision becomes clear as soon as one
realizes that net gains may differ between the spouses. Firstly, the signs may
differ. If one spouse moves together with the other, even if s/he would be
better off in the current location, s/he becomes a tied mover. Conversely, if
one spouse’s potential loss exceeds the potential gain of the other, the family
will not move and the result is a family with one tied stayer. Secondly, even if
both spouses stand to benefit from moving, the destination that maximizes the
wife’s gain need not to be the same as that one which maximizes the husband’s.
This may result in the family not moving, or moving to a destination
where neither of the spouses’ personal gains is maximized but where the sum
of both is greatest. In this sense both spouses can be tied stayers or movers.
However, in this case, too, one spouse can suffer more than the other.
2.1. The family life-cycle and other determinants of migration
A variety of factors affect migration. Among these are individual-specific characteristics
such as age, education, accumulated job skills, earnings, unemployment
experience, and migration history (see Greenwood 1985, for a survey;
see also DaVanzo 1978; Tervo 1997; Ritsila¨ and Tervo 1999). In addition,
several studies have demonstrated the importance of family relations. According
to Rossi (1955), migration that takes place due to life-cycle changes
constitutes an important part of all geographic mobility. In addition, age
variations in migration rates are shown to reflect the effect of both work careers
and life-cycle stages (Carter and Glick 1970; Sandefur and Scott 1981).
Determinants of family migration 159
The family life-cycle has conventionally been divided into several phases
(see Grundy 1992), and the highest probability of moving is associated with
the beginning of married life and the arrival of children. After a more stable
phase, mobility again increases while the children are still at a pre-school age.
There tends to be greater stability when the children are at school and the
parents are consolidating their careers (Sandell 1977; Mincer 1978). Mobility
often increases again when the children leave home and less living space is
required (Cadwallader 1992).
With regard to family ties, a nearly unanimous finding is that families migrate
in response to economic motivations on the part of the husband (Duncan
and Perrucci 1976; Long 1974; Snaith 1990; Shihadeh 1991; Battu et al.
1998; Gardner et al. 2001), and that the wife’s employment considerations are
of minor importance in migration decisions (Bielby and Bielby 1992). On the
other hand, families with working wives have been shown to be less migrationprone
(Long 1974; Sandell 1977; Mincer 1978; Lichter 1980; Holmlund 1984).
The characteristics of the origin and destination regions, such as wage
differentials, job opportunities, unemployment rates, the region’s economic
structure, conditions in housing markets and location-specific amenities (unpolluted
environment, climate, landscape etc.) may also provide an incentive
for moving (see Greenwood 1985, for a survey). Concerning the distance over
which relocation occurs, stronger ties to the current location (children at
school, contacts with friends and relatives, home-ownership etc.) have been
shown to discourage long-distance migration (Holmlund 1984; Westerlund
and Wyzan 1995; Antolin and Bover 1997; Gardner et al. 2001). Locationspecific
human capital (job experience) also tends to inhibit long moves,
while a higher amount of general human capital (education) improves the
ability to move over longer distances (see, for example, Holmlund 1984;
Shields and Shields 1993; Westerlund 1993). Regional characteristics, in turn,
seem to play a more important part in short moves. For example, the size and
diversity of the current location, cost-of-living differentials, local unemployment
and public sector attributes are evidenced to contribute to residential
choices (Widerstedt 1998; Westerlund and Wyzan 1995; Dahlberg and
Fredriksson 2001).
3. Data, model and variables
3.1. Data
The data are from the longitudinal census file of Statistics Finland, which
contains information collected in population and housing censuses, completed
with information from various official registers. Consequently, the
data offer rich information on the Finnish population, and cover the years
1970, 1975, 1980, 1985 and 1987–1996. This large data set contains information
on individuals’ characteristics (for example education, occupation,
socioeconomic status, economic activity, income) as well as on individuals’
family relations (for example type and size of family, number and ages of
children). Of especial interest for the purposes of this study is that all the
persons belonging to same family can be identified3
. Another virtue is that all
the characteristics of both spouses can be observed. A drawback, however, is
that there is no information on the motives for migration or on its specific
160 S. Nivalainen
destination. Moreover, the data only indicate the province of residence at the
end of year, and whether or not the person moved between municipalities
during the year. Hence, the specific timing of migration cannot be observed.
A basic sample4
was drawn from the census file containing information on
over 600 000 individuals. The data used in this study are a stratified subset of
this sample. First, only adults belonging to two-adult families in 19905
were
selected. Second, to obtain the family as an observation unit, men and
women belonging to same family were merged. In order to focus on economically
active families, the sample was restricted to families where the
husband was in the labour force at the end of 1993. To avoid cases of elderly
spouses unable to migrate for medical reasons, only families in which the
husband was under 63 years of age in 1994 were selected.
As the interest here is in husband-wife migration, only ‘permanent’ twoadult
families were selected, i.e., the sample was restricted to cases where the
man and woman had been living together in the same household during each
of the years 1990–1995. This means that every family in the data had been a
family for at least three years before the move and remained a family for at
least one year after the move6
. In addition, families migrating from and to
abroad were excluded from the analysis. After these restrictions the final
sample consisted of 77 340 families7
of whom 1 747 had moved during 1994.
As expected, the migration rates calculated from the data are below the
actual figures8
for the population as a whole. Firstly, we are investigating
families, not individuals. Families in general are less prone to migrate than
single individuals. Secondly, among families we are restricting ourselves to a
subset with certain characteristics. Young families (i.e., those starting their
married or cohabiting lives), who often are quite migration-prone, are not
included in the data. Moreover, a particularly mobile group, students,
constitutes only a small minority in the sample. Taking into account all
these restrictions the small number of migrants is, in fact, quite reasonable.
3.2. Model and variables
In the present study, the choice of the family relates to the question of
whether to remain in the current location, migrate between municipalities
inside the current province9
(short-distance migration), or migrate to another
province10
(long-distance migration). Even though this distinction is
not ideal, it is typical of empirical work on this topic (see, for example,
Widerstedt 1998; Westerlund and Wyzan 1995; Shields and Shields 1993).
Migration only refers to joint moves by the two spouses, and the rest of the
province/country is treated as a single destination for migrants from a given
municipality/province.
As Finnish provinces are comparatively large, long-distance migration
most likely means a change of local labour market and a change of job11. With
short-distance moves, the labour market usually does not change, and although
a change of job is possible, short moves are more likely to be associated
with housing needs and family reasons (for Finnish evidence, see Korkiasaari
1991; see also Lansing and Mueller 1967; Reitsma and Vergoosen 1988).
The probability of family migration is a function of family and regional
variables, and the multinomial logit model12
is utilized in exploring the
effect of these variables on migration. The use of this model requires
Determinants of family migration 161
the Independence of Irrelevant Alternatives (IIA) condition to hold, which
means that the ratio of any two alternatives may not be influenced by any
other alternative. This was checked with a test developed by Hausman and
McFadden (1984), and the IIA condition was found to hold13
(see Table
3). In addition, the appropriateness of the distinction between short- and
long-distance migration was checked with a pooling test described in Cramer
and Ridder (1991)14
, according to which these two migration classes cannot
be pooled (see Table 3).
Variables used in the analyses are described in Table 1 and categorical
means are presented in Table 2. Except for the variables relating to age,
children and migration history, all independent variables are measured in
1993 (i.e., before migration).
Family migration studies have traditionally only concentrated on the
husband’s age, which mainly reflects the effect of his work career on migration.
However, the age of the wife is also important, as it is closely associated
with stages in the life-cycle of the family. Therefore, the present study utilizes
the average age of the spouses, and examines age in terms of age groups.
Earlier studies have investigated the effect of school-aged children, but the
present study goes a bit further as families are separated into several groups
on the basis of children. These groups correspond to the stages in the
family life-cycle.
In principle, at least five life-cycle stages can be identified. The first stage
ends with the arrival of the first dependant. Stage 2 lasts until one dependant
reaches school age. Stage 3 includes couples with two or more dependants,
where the ages are mixed (i.e., pre-school age and school age children). Stage
4 includes couples with school-aged children only. The final stage begins
when the last dependant living at home reaches the age of 18. Due to limitations
of the data, the first and last stage could not be distinguished, thus 4
stages are actually used. Moreover, the existence and ages of children are,
as an exception, measured in 1995, a year after the move15
, but this is
unlikely to affect the results.
Family ties are investigated through the variables describing education of
the spouses and participation in the labour force of the wife. The remaining
variables are selected on theoretical grounds and/or on the basis of their significance
in earlier migration studies.
A number of empirical studies suggest that the determinants of migration
differ with respect to the distance moved (see, for example, Holmlund 1984;
Molho 1986; Reitsma and Vergoosen 1988; Widerstedt 1998). Expected effects
of independent variables in short and long moves are shown in Table 1.
As regards personal or family characteristics, higher education is expected to
boost the likelihood of long moves. The process of schooling expands an
individual’s general human capital, which is easily transferable to distant
locations. Moreover, those with higher education often have narrower career
opportunities in certain locations and moving may be the only way to
develop their career. In turn, specific human capital resulting from job experience
and training is usually tied to a certain workplace. As this form of
human capital tends to accumulate with age, older individuals should be
more resistant to longer moves. The shorter time interval to enjoy the benefits
together with higher costs of moving also tend to reduce older people’s
incentives for migration.
162 S. Nivalainen
Table 1. Definitions of the variables and the expected effects of the independent variables
Variable Expected effect Definition
Dependent variable
Migration 1 if the municipality changed during 1994
but the province remained the same, 2 if
the province changed during 1994, 0
otherwise
Independent
variables
Short dist. Long dist.
Age À/À/À ÀÀ/ÀÀ/ÀÀ Dummy variables designating the average
age of the spouses in 1994: up to 35 (ref.);
35–44; 45–54; 55 upwards
Children ?/À/À À/À/À Dummy variables indicating the existence of
children in 1995: no children under 18
(ref.); children under 7 years only; children
0–17 years; children 7–17 years only
Husband’s
education
þ/þ þ/þþ Dummy variables indicating whether the
husband has less than upper level of upper
secondary education (ref.); upper level of
upper secondary education; higher
education
Wife’s education þ þþ 1 if the wife has higher than upper level of
upper secondary education; otherwise 0
Wife’s labour force
participation
À ÀÀ 1 if the wife participates in the labour force;
otherwise 0
Migration history
1990–1993
þ þ 1 if the family changed municipalities at least
once during 1990–1993; otherwise 0
In-migration in
1993
þþ þ 1 if the family moved into the province in
1993; otherwise 0
Home ownership À À 1 if the family owns their own home;
otherwise 0
Family income ? À Husband’s and wife’s income subject to state
taxation/FIM 1000
Commuting þþ þ 1 if the husband and/or the wife were
commuters in 1993; otherwise 0
Unemployment
experience
þ þþ 1 if the husband and/or the wife were
unemployed (over 2 weeks) during 1993;
otherwise 0
Area unemployment
rate
þ þþ Unemployment rate in the travel-to-work
area where the family lives
Size of municipality ?/?/þ ?/?/À Dummy variables indicating the number of
inhabitants in the municipality: up to
15 000; 15–39 999; 40–69 999 (ref.); 70 000
and above
Share of agriculture À þ The share of employed labour force in
agriculture and forestry
(0–9.99% ¼ 0; . . . ; 90–100% ¼ 9)
Share of industry ? þ The share of employed labour force in
industry (0–9.99% ¼ 0; . . . ; 90–100% ¼ 9)
Notes: (1) all variables are measured in 1993 if not otherwise stated (2) þ ¼ positive effect,
þþ ¼ stronger positive effect, À ¼ negative effect, ÀÀ ¼ stronger negative effect, ? ¼ positive or
negative effect (3) (ref.) indicates the reference group (4) all regional variables refer to the region
of origin.
Determinants of family migration 163
A negative association is expected between family income and longdistance
migration: the lower the family income, the lower the opportunity
costs of moving and the larger the number of attractive job offers. In short
moves the effect of income is less evident, and could be positive or negative.
A number of empirical studies show that personal (or family) unemployment
augments migration (see, for example, DaVanzo 1978; Schlottmann and
Herzog 1981; Van Dijk et al. 1989; Ritsila¨ and Tervo 1999). The unemployed,
most likely, move for labour market reasons, and therefore the probability of
long-distance migration should be higher for them.
Short moves are often related to life-cycle events: by moving, families
adjust to their changing needs. Due to larger moving costs, families with
children are generally less eager to move, but the need for a larger house or
the prospective schooling of children might accelerate short-distance mobility
Table 2. Means of the independent variables according to migration category
Variables Stayers
(mean)
Short-distance
migrants (mean)
Long-distance
migrants (mean)
Family characteristics
Age
–35 years 0.27 0.56 0.51
35–44 0.46 0.33 0.34
45–54 0.22 0.09 0.11
55– 0.05 0.02 0.04
Children
no children under 18 years 0.30 0.24 0.24
all under 7 years 0.17 0.40 0.37
0–17 years 0.17 0.19 0.20
all 7–17 years 0.36 0.17 0.19
Husband’s education
lower than upp. lev. of upp. sec. 0.67 0.57 0.49
upper level of upper secondary 0.17 0.22 0.17
higher 0.16 0.21 0.34
Wife has higher education 0.16 0.19 0.22
Wife in the labour force 0.84 0.76 0.67
Migrated 1990–1993 0.07 0.26 0.30
Migrated into province 1993 0.01 0.04 0.08
Home owner 0.84 0.58 0.57
Family income/FIM 1000 217 215 202
Commuting 0.34 0.51 0.41
Unemployment experience 0.35 0.45 0.53
Regional characteristics
Area unemployment rate 22.5 21.3 22.5
Size of municipality
–15 000 0.40 0.28 0.35
15 000–39 999 0.26 0.27 0.23
40 000–69 999 0.06 0.04 0.09
70 000– 0.28 0.41 0.33
Share of agriculture 0.58 0.34 0.46
Share of industry 2.18 2.04 2.21
Number of observations 75 593 1 176 571
164 S. Nivalainen
when the children are at preschool age16
. When children are at school, locational
ties are stronger, and migration is less likely to occur. Ties to the
current locality are also stronger in families where both spouses work (see
Mincer 1978; Lichter 1980; Holmlund 1984). Thus the wife’s labour force
participation is expected to deter migration, and this effect should be accentuated
in long moves.
The distance to work influences migration decisions (see Clark and Burt
1980), and therefore commuters should be more prone to short moves. Prior
mobility is also an important determinant of migration (Krumm and Kelly
1988; Widerstedt 1998; Bo¨ heim and Taylor 2000), and several studies have
concluded that the longer is the distance of the original move, the higher is
the likelihood of a subsequent (adjustment) move (see, for example, DaVanzo
1983; Yezer and Thurston 1976). Hence the probability of short-distance
mobility might be augmented shortly after the family has moved into
the province, since the final municipality of residence is not necessarily
chosen at the time of in-migration, but only after the location has become
familiar enough.
With regard to regional characteristics, a high area unemployment rate
is thought to reflect diminished labour market opportunities. Therefore, the
likelihood of moving, especially over longer distances, is expected to increase
with the unemployment rate. Inhabitation in larger cities with more varied
economic structures and more opportunities for employment should prevent
long-distance migration. However, at a certain stage of life families tend to
prefer more peaceful neighbourhoods within a reasonable commuting distance
(Green 1997). As the largest centres are surrounded by closely situated
smaller municipalities with varying characteristics (differences in costs of living,
tax rates, public services etc.), families living in the biggest towns are
expected to have a higher likelihood of short moves. In the countryside, job
opportunities are fewer, distances are longer and commuting between municipalities
is relatively limited. Therefore, the higher the share of agriculture,
the more likely long-distance migration should be in relation to short moves.
4. Results
The results17
of the basic model (model 1) are presented in Table 3. Since
interpretation of the estimated parameters of the multinomial logit model
would be difficult, marginal effects are reported. The parameter estimates
and odds-ratios are presented in the Appendix (Table A1).
When using both macro- and micro-level variables in the study of microunits,
the random disturbances within groups may be correlated, and doubt
has been cast on the reliability of such results (see Moulton 1990). To avoid this
pitfall, the standard errors of the models were adjusted for the general correlation
of disturbances18
. Even though no large changes emerged in relation to
the unadjusted models19
, the results presented here refer to adjusted models.
In general, the results show that almost all the coefficients are statistically
significant and have the expected signs, and are thus in accordance with the
theory and earlier empirical findings. Looking at the figures one notices that
the marginal effects are relatively small, which stems from the small number
of migrants. Therefore, it is not so much the magnitude, but rather the signs
that we are interested in.
Determinants of family migration 165
Table 3. Determinants of family migration; marginal effects of the multinomial logit model
(model 1)
Variables Staying Short-distance
migration
Long-distance
migration
Marginal
effect
t-ratio Marginal
effect
t-ratio Marginal
effect
t-ratio
Family characteristics
Age
35–44 0.00571*** 6.38 À0.00432*** À5.96 À0.00140*** À2.76
45–54 0.01286*** 9.51 À0.01020*** À8.85 À0.00267*** À3.54
55– 0.01224*** 5.41 À0.01053*** À5.39 À0.00171 À1.50
Children
all under 7 years À0.00054 À0.50 0.00024 0.28 0.00030 0.48
0–17 years 0.00549*** 4.78 À0.00412*** À4.37 À0.00137** À2.12
all 7–17 years 0.00888*** 8.10 À0.00680*** À7.33 À0.00208*** À3.35
Husband’s
education
upper lev. of upp.
sec.
À0.00287*** À3.10 0.00160** 2.21 0.00127** 2.29
higher À0.00589*** À5.92 0.00130 1.62 0.00459*** 8.86
Wife has higher
education
À0.00096 À1.00 0.00033 0.42 0.00064 1.23
Wife in the labour
force
0.00465*** 5.58 À0.00203*** À2.98 À0.00262*** À5.82
Migrated 1990–
1993
À0.00975*** À10.04 0.00619*** 8.26 0.00356*** 6.54
Migrated into
province 1993
À0.01178*** À5.80 0.00654*** 3.96 0.00525*** 5.62
Home owner 0.01290*** 16.41 À0.00868*** À13.98 À0.00422*** À9.37
Family income À6.2E À07 À0.19 3.8E À06** 2.18 À3.2E À06 À1.17
Commuting À0.00820*** À10.82 0.00630*** 10.44 0.00190*** 4.31
Unemployment
experience
À0.00611*** À8.42 0.00340*** 5.89 0.00271*** 6.49
Regional characteristics
Area unemployment
rate
0.00023** 2.35 À0.00034*** À4.28 0.00011** 2.02
Size of municipality
–15 000 0.00108 0.58 0.00208 1.24 À0.00316*** À3.59
15 000–39 999 0.00029 0.17 0.00312** 2.01 À0.00341*** À4.34
70 000– À0.00303* À1.74 0.00456*** 2.94 À0.00153** À1.98
Share of agriculture 0.00104 1.58 À0.00131** À2.41 0.00027 0.72
Share of industry 0.00004 0.09 À0.00067* À1.72 0.00062** 2.48
Number of observations 77.340 Number of migrants 1 747
Log likelihood À8 435.14 Restricted log likelihood À9 452.76
Model: 12
(44) 2 404.2 Likelihood ratio index 0.11
Pooling: 12
(22) 168.2a
*,
**,
*** significant at the 10,5,
1% level
IIA: 12
(23) 23.2a a
test values refer to unadjusted
model
166 S. Nivalainen
4.1. Determinants of family migration
The family life-cycle influences migration. Children, in general, have an
inhibiting effect (result not shown here), but families with only under 7-yearold
children are as migration prone as those without children. The presence
of school-aged children significantly reduces migration propensities (cf.,
Long 1974; Sandell 1977; Mincer 1978). On the other hand, there are differences
in migration propensities between families with school-aged children,
too: the older the children, the less likely the family to move.
As expected, the younger the spouses, the more prone they are to move.
The difference between the youngest and oldest age group is insignificant in
longer moves, though. In addition, the age variables become stronger when
the children variables are removed from the analysis (result not shown here),
which implies that the ages of parents also partly reflect the stages of the
family life-cycle.
The probability of long-distance migration increases with the husband’s
education (cf., Sandell 1977; Mincer 1978), and short-distance migration is
most likely to occur when the husband has completed the upper level of
upper secondary education. Surprisingly, the wife’s education appears to be
an insignificant, although positive, determinant of family migration. This
corroborates Lichter (1982) and Axelsson and Westerlund (1998), but differs
from the findings of Holmlund (1984) and Shields and Shields (1993), who
found the wife’s education to significantly increase migration propensities.
Contradictory results in separate studies concerning the effect of the wife’s
education might reflect problems with collinearity, i.e., a correlation between
spouses’ education due to assortative mating. In our sample the correlation
between the educational level of the spouses is about 0.4 (significant at the
0.01 level). In a sample this large, significance is not surprising. A closer
inspection reveals that among long-distance migrants both spouses have a
high educational level in 15% of the cases. In turn, the husband alone is
highly educated in as many as 18% of long-distance migrant families,
while the respective figure for the wives is only 7%. This compares with
10/11/9 for short-distance migrants and to 8/9/8 for non-migrants. Thus,
while long-distance migrants are more often highly educated, the difference
between spouses’ education also tends to be much larger. In this sense,
selective mating does not appear to be a major problem.
Hence, the result suggests that family migration is more often associated
with the development of the husband’s career. Even if this implication is
consistent with many previous studies (see, for example, Long 1974; Shihadeh
1991), it is a little surprising in Finland, where men and women are more
equal and women even tend to be more highly educated20
than men. On the
other hand, a rationale underlying men’s domination might be the gender
wage gap: despite theoretical gender equality, Finnish men still earn more
than women, and might have more weight in migration decisions due to their
higher earnings capacity21
.
To further examine the husband-wife relations in migration decisions,
a measure of intra-family income dispersion was formed22
. Model 1 was then
re-run with this variable (and excluding the wife’s characteristics). The income
dispersion is positive in all moves, but significant only in long ones (see
Appendix, Table A2). As the husband’s income is higher in 75% of the
Determinants of family migration 167
sample families, in practice this means that families with a larger husband/
wife income ratio tend to be more inclined towards long-distance migration.
Participation of the wife in the labour force reduces the likelihood of moving.
Hence two-earner families are less eager to move (cf., Long 1974; Sandell
1977; Mincer 1978; Lichter 1980). In addition, the deterrent effect of the
wife’s work increases with distance, with the odds for short and long moves
(vs. staying) of 0.80 and 0.56, respectively. The negative association between
the wife’s participation in the labour force and family migration suggests that
the husbands of working wives tend to be tied to certain locations.
As hypothesized, the experience of unemployment boosts long-distance
migration (on Finland, see Ritsila¨ and Tervo 1999; see also DaVanzo 1978;
Herzog and Schlottmann 1984; Hughes and McCormick 1989). On the other
hand, the positive effect of personal unemployment on shorter moves differs
somewhat from findings in other countries (cf., DaVanzo 1978; Westerlund
1993), and is most probably explained by the large size of Finnish provinces.
In line with earlier studies (Tervo 1997; Ritsila¨ and Tervo 1999; Widerstedt
1998), the variable indicating migration history is positive, and suggests
that previous migration experience facilitates migration and encourages
families to move again. Recent in-migration also increases migration probabilities,
thus lending support to the existence of imperfect information and
unpleasant surprises. Disappointment at the outcome of one move may become
the cause of the next.
On the other hand, the effect of migration history may also reflect
unobserved heterogeneity, which means that certain individuals are migrants
because of their (unobserved) personal characteristics, and therefore move
more often than others. This being the case, previous migration per se does
not have any effect on subsequent migration propensities, but appears to be a
significant determinant of migration simply because it serves as a proxy for
the unmeasured variables. The literature often refers to ‘‘true state dependence’’
and to ‘‘spurious state dependence’’ (for further discussion see Heckman
1981; Hsiao 1986; Baltagi 1995).
Homeowners are less likely to move (cf., Haapanen 1998; Ritsila¨ 2001).
They may have stronger locational ties, but the finding may also reflect higher
transaction costs for owner-occupants; even though in 1994 Finland started to
recover from the recession, considerable difficulties in selling properties at
reasonable prices continued to exist. As expected, commuters are migrationprone.
Family income seems to play no part in long-distance migration, but
it is positively related to short moves. At a certain stage of life, families often
want to move to more congenial neighbourhoods away from city centres.
Families living in middle-sized towns are the most eager to undertake long
moves, and the likelihood of short moves is greatest in the biggest towns. The
probability of long-distance migration increases with the share of labour
force working in industry, which is in line with a recently observed development;
one-sided industrial regions are losing population in Finland (see
Vartiainen 1997; Laakso 1998). The effect on short moves is the opposite. As
distinct from earlier Finnish findings (Tervo 1997; Ritsila¨ and Tervo 1999),
the share of agriculture has no importance in long moves, but is negatively
related to short-distance migration.
Higher area unemployment augments long-distance migration. This
supports earlier Finnish findings (Tervo 1997; Ritsila¨ and Tervo 1999), but
at the same time differs from results obtained in several other countries
168 S. Nivalainen
(for example, Van Dijk et al. 1989; Hughes and McCormick 1989). Somewhat
surprisingly, there is a negative association between unemployment
rates and short moves. A family that migrates out of a municipality because
of the bad unemployment situation is unlikely to move a short distance, as
the circumstances will be the same in all the surrounding areas. Hence high
unemployment rates reduce mobility inside a province, and increase migration
between provinces.
4.2. Short moves vs. long moves
The basic model indicated that differences might exist between different
moves. For example, the signs for family income and the characteristics of
areas changed in short and long moves. To examine the differences more
carefully, a multinomial logit model was run with short-distance migration as
a basic category (model 2). The parameter estimates and odds-ratios are
presented in the Appendix (Tables A1 and A2).
The estimates verify the differences suggested by the basic model. Longdistance
migrants do indeed have lower incomes than short-distance migrants.
Furthermore, the poorer the economic situation in the travel-to-work
area, i.e., the higher the unemployment rate, the more probable the longdistance
move (odds-ratio 1.51). There is also a clear relation between a region’s
economic structure and long-distance migration. That is, families who
migrate out of municipalities dominated by agriculture or industry are likely
to move across provincial borders (odds-ratios 1.22 and 1.23, respectively).
Against expectations, the older the spouses are, the more likely they are to
move over longer distances. The odds on a long vs. short move are twice as
high for those over 55 as for those under 35 years of age. This phenomenon
may relate to return migration, but can also be caused by older people’s
relatively weak employment opportunities23
. Those with experience of unemployment
are inclined to move between provinces. Recent in-migrants also
tend to move farther away. As hypothesized, the probability of long moves
increases with the husband’s education: those with higher qualifications are
more than twice as likely to change province as those with the lowest ones
(odds-ratio 2.41). In addition, larger intra-family income dispersion augments
migration across provincial borders (odds-ratio 1.48).
As expected, migration to another province becomes less likely if there
is a working wife or a commuter in the family (odds-ratios 0.69 and 0.78,
respectively). Finally, the wife’s education, ages of children and home ownership
are non-significant factors in determining the migratory distance.
5. Summary and conclusions
The aim of the present study was to investigate the factors affecting Finnish
family migration, especially the effects of the family life-cycle and family ties.
Furthermore, to explore the differences between different types of moves, a
distinction was drawn between short- and long-distance migration. Unique
family data were used, and empirical analysis was carried out using multinomial
logit models.
Determinants of family migration 169
In general, the findings are in line with those reported earlier; family
migration seems to behave quite similarly around the world. The results show
a strong negative association between the family life-cycle and migration.
Childless couples and those with only pre-school-aged children are the most
eager to move. When their children are at school, mothers are likely to
work, which, in addition to children’s ties to their schools and friends, creates
stronger ties to current locations.
Rather surprisingly, the findings lend support to the existence of the traditional
pattern of migration – the husband leads and the wife follows – in
Finland as well. That is, families more likely migrate due to the demands of
the husband’s career, and their wives move along with them because of family
ties24
. As a result, wives may become unemployed, underemployed or exit
the labour force at the destination. On the other hand, two-earner families
are less migration-prone, and the deterrent effect of the wife’s work status
increases with distance. This implies that the tied stayer in the family is more
often the husband. The restriction of the husband’s mobility to short distances
may reduce his opportunities for career advancement and increase the
probability of job mismatch or even unemployment.
There are differences between short and long moves. Stronger ties to the
current location counteract, and general human capital (of the husband)
boosts inter-provincial migration. Larger income dispersion between the
spouses increases the likelihood of long-distance migration. The unemployed
also more likely move between provinces. In addition, the size and the economic
structure of the existing region influence the distance of the move. Local
economic conditions are also important: a high area unemployment rate augments
long-distance migration, and has an inhibiting effect on short moves.
The findings lend support to Ritsila¨ and Tervo (1999), who suggested that
in Finland both personal and area unemployment operate towards reducing
regional unemployment differentials. However, at the same time, it is not
only families with experience of unemployment who are moving out of high
unemployment regions, but also other families. As young couples are the
most eager to move, the demographic structure of depressed regions skews
towards the older age groups. In addition to having a direct negative effect
on population structure, migration also decreases population size through
reduced birth rates. Provinces with high unemployment may lose their valuable
human capital in the form of whole families. This in turn brings about
even more unfavourable effects, eventually leading to a widening gap
between successful and poor areas. Therefore, the effect of migration on the
composition of regional population structure in the longer run is a topic
deserving more attention in future research.
I also feel that our knowledge of family migration is far from complete. The
use of estimation methods that utilize the panel nature of the data would allow
family migration to be analyzed more closely. Despite the fact that the educational
level, participation in the labour force and earning power of women
have increased, and men and women have become more equal, the effect of
family ties seems to have remained unchanged for decades; it continues to be
the human capital of the husband that rules. Hence, the interaction between
the wife’s locational ties, her general human capital and family migration
decisions is clearly an issue that merits further investigation. Moreover, both
short- and long-term consequences of family migration deserve examination.
Shedding light on these issues is the major challenge facing future research.
170 S. Nivalainen
Endnotes
1
In 1994, there were 19 provinces (‘‘maakunta’’) in Finland.
2
These include the actual moving costs, as well as a potential increase in living costs, transportation
etc., necessitated by migration.
3
All persons belonging to the same family have identical household-dwelling unit codes and
family numbers.
4
The basic sample is a 1% sample drawn from the longitudinal data, complemented with
individuals belonging to same household dwelling unit as the sample individuals.
5
Families can be identified every fifth year (family number is given in 1990 and 1995). For the
purposes of this study the most convenient year was 1990.
6
In families where both spouses move simultaneously, and also stay together after the move,
the complexity of migration decision-making is most probably the greatest, and many compromises
have to be made. It is, however, recognized that decision to move can result in
family dissolution if the gain from being a couple is less than the cost of family ties (see
Mincer 1978). Family dissolution can result in moving, too. These events, however, cannot be
investigated with the data set at hand.
7
In 1994 there were nearly 1.2 million two-adult families in Finland, constituting 87% of all
Finnish families (Statistics Finland 1995b).
8
In 1994 about 4.2% of the Finnish population migrated between municipalities (Statistics
Finland 1995a). The corresponding figure in the sample was 2.3%.
9
The data set does not contain information on intra-municipality moves, and therefore those
moving inside a municipality are not classified as migrants.
10
A minor shortcoming of the classification is that those moving between neighbouring provinces
are categorised as long-distance migrants. However, the number of these moves is minimal,
and the results are unlikely to be affected by this.
11
For example, Korkiasaari (1991), in the Finnish case, found that about 50% of all longer
distance moves took place for job reasons. Similar evidence has also been obtained in other
countries (see Harkman 1989).
12
For discussion of the multinomial logit model see Greene 1997.
13
The omitted category was short-distance migration. The test requires estimation of both the
unrestricted and restricted (smaller choice set) model. For further details, see Hausman and
McFadden 1984.
14
Pooling vs. non-pooling can be tested by the likelihood ratio test. To carry out the test, both
the pooled and non-pooled model have to be estimated. For further details, see Cramer and
Ridder 1991.
15
Information about children was available only from the years 1990 and 1995. Since the year
of interest is 1994, the situation in 1995 gives the best available approximation of the ages of
children.
16
Among children aged 0–17 years, those aged 0–6 years have the highest propensity to migrate
from one municipality to another. In over 50% of children’s moves the moving distance is
under 50 kilometres. (Kaartovaara and Sauli 2000)
17
All results referred to but not shown here are available from the author on request.
18
See STATA 7 manual for additional information.
19
The basic model was also estimated without the regional variables. The effects of family
characteristics remained almost unchanged (no significant variable changed its sign or became
insignificant), thus no major problems seem to exist.
20
Finnish women under 50 have a higher level of education than Finnish men, if the proportion
of the population with at least an upper secondary education is used as a criterion (European
Commission 1996).
21
Average earnings of Finnish women are about 80% of the average earnings of men.
22
Intra-family income dispersion is measured by the difference (in absolute value) between the
husband’s and the wife’s income, divided by the sum of their income.
23
In 1994, 25.6% of those aged 55 or more and in the labour force were unemployed, while
the average unemployment rate for all age groups was 18.4% (European Commission 1996).
24
Due to the sampling method, this result may be a feature of our data and not of the population
as a whole, and the finding will be tested in subsequent studies.
Determinants of family migration 171
Appendix
Table A1. Determinants of family migration; coefficients of the two multinomial logit models
(model 1 and 2) and the respective odds-ratios
Variables Short-dist. migration
vs. staying (model 1)
Long-distance
migration vs.
staying (model 1)
Long- vs. short-dist.
migration (model 2)
Coefficient Odds-
ratio
Coefficient Odds-
ratio
Coefficient Odds-
ratio
Constant À2.665*** À4.451*** À1.786***
Family characteristics
Age
35–44 À0.469*** 0.63 À0.316*** 0.72 0.153 1.17
45–54 À1.107*** 0.33 À0.604*** 0.54 0.503** 1.65
55– À1.142*** 0.32 À0.392 0.66 0.750** 2.11
Children
all under 7 years 0.026 1.03 0.067 1.06 0.041 1.04
0–17 years À0.448*** 0.64 À0.309** 0.73 0.139 1.15
all 7–17 years À0.739*** 0.48 À0.471*** 0.62 0.268 1.31
Husband’s
education
upper lev. of upp.
sec.
0.175** 1.19 0.284** 1.33 0.109 1.12
higher 0.146 1.16 1.025*** 2.79 0.879*** 2.41
Wife has higher
education
0.036 1.04 0.142 1.15 0.106 1.11
Wife in the labour
force
À0.223*** 0.80 À0.585*** 0.56 À0.362*** 0.69
Migrated 1990–
1993
0.674*** 1.96 0.801*** 2.23 0.127 1.13
Migrated into
province 1993
0.713*** 2.04 1.176*** 3.24 0.463* 1.59
Home owner À0.944*** 0.39 À0.949*** 0.39 À0.005 0.99
Commuting 0.684*** 1.98 0.432*** 1.54 À0.252** 0.78
Unemployment
experience
0.371*** 1.45 0.607*** 1.83 0.236** 1.27
Family incomea
4.1E À04** 1.04 À7.1E À04 0.93 À1.1E À03* 0.89
Regional characteristics
Area unemployment
ratea
À0.036*** 0.78 0.025** 1.19 0.062*** 1.51
Size of municipality
–15 000 0.221 1.24 À0.705*** 0.49 À0.925*** 0.39
15 000–39 999 0.333** 1.39 À0.760*** 0.47 À1.093*** 0.33
70 000– 0.491*** 1.63 À0.337** 0.71 À0.828*** 0.43
Share of
agriculturea
À0.142** 0.87 0.057 1.06 0.199** 1.22
Share of industrya
À0.071* 0.93 0.137** 1.15 0.208*** 1.23
Number of
observations
77 340
*,
**,
*** Significant at the 10, 5, 1% level.
a
Odds-ratios for continuous variables are calculated at the values of the 25th
and 75th
percentiles.
172 S. Nivalainen
References
Antolin P, Bover O (1997) Regional Migration in Spain: The Effect of Personal Characteristics
and of Unemployment, Wage and House Price Differentials Using Pooled Cross-Sections.
Oxford Bulletin of Economics and Statistics 59(2):215–235
Axelsson R, Westerlund O (1998) A Panel Study of Migration, Household Real Earnings and
Self-Selection. Journal of Population Economics 11:113–126
Baltagi B (1995) Econometric Analysis of Panel Data. Wiley, Chisester, UK
Battu H, Seaman PT, Sloane PJ (1998) Are Married Women Spatially Constrained? A Test of
Gender Differentials in Labour Market Outcomes. Paper presented at the European Regional
Science Association Conference in Vienna in August 1998
Bielby WT, Bielby DD (1992) I Will Follow Him: Family Ties, Gender-role Beliefs, and
Reluctance to Relocate for a Better Job. American Journal of Sociology 97:1241–1267
Bo¨ heim R, Taylor M (2000) Residential mobility, housing tenure and labour market in Britain.
The Institute for Labour Research Discussion Paper Series 99/35
Cadwallader M (1992) Migration and Residential Mobility. The University of Wisconsin Press,
Wisconsin
Carter H, Glick PC (1970) Marriage and Divorce: A Social and Economic Study. Harvard University
Press, Cambridge MA
Clark WAV, Burt JE (1980) The Impact of Workplace on Residential Relocation. Annals of the
Association of American Geographers 70:59–67
Cramer JS, Ridder G (1991) Pooling States in the Multinomial Logit Model. Journal of Econometrics
47:267–272
Dahlberg M, Fredriksson P (2001) Migration and Local Public Services. Uppsala University
Working Papers 2001:12, Uppsala
DaVanzo J (1978) Does Unemployment Affect Migration? Evidence from Micro Data. The
Review of Economics and Statistics 60:504–514
DaVanzo J (1983) Repeat Migration in the United States: Who Moves Back and Who Moves
On? The Review of Economics and Statistics 65:552–559
Duncan RP, Perrucci CC (1976) Dual Occupation Families and Migration. American Sociological
Review 41:252–261
European Commission (1996) Labour Market Studies, Finland. Luxembourg
Gardner J, Pierre G, Oswald AJ (2001) Moving for Job Reasons. (http://www.warwick.ac.uk/fac/
soc/Economics/oswald/GPOeconomica01.pdf )
Green AE (1997) A Question of Compromise? Case Study Evidence on the Location and Mobility
Strategies of Dual Career Households. Regional Studies 31(7):641–657
Greene W (1997) Econometric Analysis, 3rd ed. Prentice-Hall Inc, New Jersey
Greenwood MJ (1985) Human Migration: Theory, Models and Empirical Studies. Journal of
Regional Science 25:521–544
Grundy E (1992) The Household Dimension in Migration Research. In: Champion T, Fielding T
(eds) Migration Processes & Patterns, Vol 1, Research Progress & Prospects. Belhaven,
London, 165–174
Haapanen M (1998) Internal Migration and Labour Market Transitions of Unemployed Workers.
Government Institute for Economic Research, Helsinki
Table A2. The effect of intra-family income dispersion on migration propensities
Variable Short-dist. migration
vs. staying
Long-distance
migration vs. staying
Long- vs. short-dist.
migration
Coefficient Odds-
ratio
Coefficient Odds-
ratio
Coefficient Odds-
ratio
Income dispersion 0.173 1.19 0.563*** 1.76 0.390* 1.48
Note: Variables ‘‘Wife in the labour force’’ and ‘‘Wife has high education’’ are not included;
other variables as in Table A1.
Determinants of family migration 173
Haapanen M (2002) Labour Migration and Wages. University of Jyva¨ skyla¨ , Reports from the
School of Business and Economics N:o 29/2002 (licentiate thesis). Jyva¨ skyla¨
Harkman A (1989) Migration Behaviour Among the Unemployed and the Role of Unemployment
Benefits. Papers of Regional Science Association 66:143–150
Hausman J, McFadden D (1984) Specification Tests for the Multinomial Logit Model. Econometrica
52:1219–1240
Heckman J (1981) Heterogeneity and State Dependence. In: Rosen S (ed) Studies in Labour
Markets. University of Chicago Press, 91–139
Herzog HW Jr, Schlottmann AM (1984) Labour Force Mobility in the United States: Migration,
Unemployment and Remigration. International Regional Science Review 9:43–58
Holmlund B (1984) Labour Mobility: Studies of Labour Turnover and Migration in the Swedish
Labour Market. The Industrial Institute for Economic and Social Research, Stockholm
Hsiao C (1986) Analysis of Panel Data. Cambridge, UK, Cambridge University Press
Hughes G, McCormick B (1989) Does Migration Reduce Differentials in Regional Unemployment
Rates? In: Van Dijk J, Folmer H, Herzog HW Jr, Schlottmann AM (eds) Migration and
Labour Market Adjustment. Kluwer Academic, Dordrecht, 85–108
Kaartovaara L, Sauli H (2000) Finnish Child. Statistics Finland, Va¨ esto¨ 2000:1
Korkiasaari J (1991) Mobility and Structural Change. Tyo¨ ministerio¨ , Tyo¨ poliittinen tutkimus
11, Helsinki (in Finnish)
Krumm RJ, Kelly A (1988) Multiperiod Migration Patterns: The Timing and Frequency of
Household Response. Journal of Regional Science 28:255–270
Laakso S (1998) Inter-regional Migration in Finland. City of Helsinki Urban Facts, Research
Series 4 (in Finnish, Abstract in English)
Lansing JB, Mueller E (1967) The Geographic Mobility of Labor. Michigan Survey Research
Center, Ann Arbor
Lichter D (1980) Household Migration and the Labour Market Position of Married Women.
Social Science Research 9:83–97
Lichter D (1982) The Migration of Dual-Worker Families: Does the Wife’s Job Matter? Social
Science Quarterly 63:48–57
Long LH (1974) Women’s Labour Force Participation and the Residential Mobility of Families.
Social Forces 52:342–348
Mincer J (1978) Family Migration Decisions. Journal of Political Economy 86:749–773
Molho I (1986) Theories of Migration: a Review. Scottish Journal of Political Economy 33:396–
419
Moulton BJ (1990) An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables
on Micro Units. The Review of Economics and Statistics 72:334–338
Pekkala S (2002) Migration and Individual Earnings in Finland: A Regional Perspective.
Regional Studies 36(1):13–24
Reitsma RF, Vergoosen D (1988) A causal typology of migration: the role of commuting.
Regional Studies 22:331–340
Ritsila¨ J, Tervo H (1999) Regional Differences in the Role of Migration in Labour Market
Adjustment: The Case of Finland. In: Crampton G (ed) Regional Unemployment, Job
Matching and Migration, Series on European Research in Regional Science. Pion, London,
166–182
Ritsila¨ J (2001) Studies on the Spatial Concentration of Human Capital. Jyva¨ skyla¨ Studies in
Business and Economics, Doctoral Dissertation, University of Jyva¨ skyla¨
Rossi PH (1955) Why Families Move. Glencoe, Illinois
Sandefur GD, Scott WJ (1981) A Dynamic Analysis of Migration: An Assesment of the Effects
of Age, Family and Career Variables. Demography 18:355–368
Sandell SH (1977) Women and the Economics of Family Migration. The Review of Economics
and Statistics 59:406–414
Schlottmann AM, Herzog HW Jr (1981) Employment Status and the Decision to Migrate.
Review of Economics and Statistics 63:590–598
Shields MP, Shields GM (1993) A Theoretical and Empirical Analysis of Family Migration and
Household Production: U.S. 1980–1985. Southern Economic Journal 59:768–782
Shihadeh ES (1991) The Prevalence of Husband-Centered Migration: Employment Consequences
for Married Mothers. Journal of Marriage and the Family 53:432–444
174 S. Nivalainen
Sjaastad LA (1962) The Costs and Returns of Human Migration. Journal of Political Economy
70:80–93
Snaith J (1990) Migration and Dual Career Households. In: Johnson J, Salt J (eds) Labour
Migration. David Fulton, London, 155–171
Stata Reference Manual Release 7 (2001) College Station, Texas, Stata Corporation
Statistics Finland (1995a) Vital Statistics by Municipality 1994. Va¨ esto¨ 1995:10
Statistics Finland (1995b) Families 1994. Va¨ esto¨ 1995:14
Tervo H (1997) Long-Distance Migration and Labour Market Adjustment: Empirical Evidence
from Finland 1970–1990. University of Jyva¨ skyla¨ , Working Paper 168, Jyva¨ skyla¨
Van Dijk J, Folmer H, Herzog HW Jr, Schlottmann AM (1989) Labour Market Institutions and
the Efficiency of Interregional Migration: a Cross-nation Comparison. In: Van Dijk J,
Folmer H, Herzog HW Jr, Schlottmann AM (eds) Migration and Labour Market Adjustment.
Kluwer Academic, Dordrecht, 61–83
Vartiainen P (1997) Muuttoliikkeen uusi kuva. Sisa¨ asiainministerio¨ , Aluekehitys-osaston julkaisu
4 (in Finnish)
Westerlund O (1993) Internal Migration in Sweden – The Role of Fiscal Variables and Labour
Market Conditions. Umea˚ Economic Studies No. 293, University of Umea˚
Westerlund O, Wyzan ML (1995) Household Migration and the Local Public Sector: Evidence
from Sweden, 1981–1984. Regional Studies 29(2):145–157
Widerstedt B (1998) Moving or Staying? Job Mobility as a Sorting Process. Umea˚ Economic
Studies No. 464, University of Umea˚
Yezer AM, Thurston L (1976) Migration Patterns and Income Change: Implications for the
Human Capital Approach to Migration. Southern Economic Journal 42:693–702
Determinants of family migration 175