dual-earner migration in britain. earnings gains...
TRANSCRIPT
Dual-earner migration in Britain.
Earnings gains, employment, and self-selection
Birgitta Rabe1
Ruhr-University Bochum
Institute for Social and Economic Research, Essex
email: [email protected]
26/1/2006
Abstract
This paper examines how spouses in dual-earner couples in Britain weigh each partner’s expected wage growth in the decision to migrate. Previous research suggests that husbands’ job prospects dominate the migration choice irrespective of their relative earnings potential. Based on British panel data this paper employs an endogenous switching model and estimates wage change differentials of migrating vs. staying for husbands and wives corrected for double selectivity of migration and employment. The analysis shows that dual-earner couples put roughly equal weights on each partner’s expected wage gains when deciding to migrate. Moreover, migrant wives’ employment declines temporarily and there are significant selection effects in migration and employment among non-migrants. Keywords: family migration, dual-earner couples, double selection, endogenous switching JEL Classification: J61, D19, C35
1. Introduction
This paper examines the factors influencing whether dual-earner couples migrate. In
particular it investigates the effects of expected wage gains of a husband and a wife on
the decision to migrate. Whether the effects are asymmetric has received much
attention and has important implications for discriminating between the unitary and
collective theory of family migration, but has rarely been explicitly analyzed. In 1 This paper was produced during a research visit to the Institute of Social and Economic Research whose hospitality I gratefully acknowledge. Thanks to John Brice, John Ermisch, Axel Heitmueller, Stephen P. Jenkins, Cheti Nicoletti, Mark Taylor and seminar participants at the University of Essex and the Policy Studies Institute, London, for helpful comments and suggestions.
1
addition, the paper studies the employment and wage consequences of migration for
couples while rigorously modelling selection effects in employment and migration.
The paper makes several contributions. It is one of the first to explicitly
examine the impact of each spouse’s wage differential between origin and destination
on family migration decision-making. The main finding is that spouses weigh each
partner’s expected wage growth roughly equally in the migration decision. Moreover,
it models the double selection into migration and employment of both spouses,
revealing that spouses who choose not to migrate and non-migrating wives who
choose to work possess comparative disadvantages. Furthermore it gives insight into
what happens to the employment of husbands and wives when families move. While
wives experience a temporary 8% reduction in labour force participation after
migration, husbands’ employment remains unaltered.
2. Previous research
Theories of migration generally assume that people relocate because they expect
higher utility in the new location. In most cases individuals are attracted to locations
where they can receive higher wages, and the earnings gains offset the social and
economic costs associated with migration.
Looking at migration of families, the common theoretical assumption is that
spouses seek to maximize joint family income when making the decision to migrate.
In the traditional unitary model of the household the spouses maximize a single utility
function. They are indifferent as to who contributes to joint family income and in
which proportion (Mincer 1978, Sandell 1977). Whenever maximization of family
income makes spouses stay (move) although they could individually receive higher
earnings by moving (staying), these spouses are tied stayers (movers). Mincer (1978)
predicts that greater market earnings power and more continuous labour force
participation potentially yields higher migration returns to husbands than to wives.
Wives are therefore likely to be tied movers who presumably experience reductions in
wages and working hours following the move, thus reinforcing the initial differences
in “earnings power” between the spouses. On the other hand, working wives - in
particular if they contribute a large share to family income and have a stable labour
force attachment - presumably deter family mobility, making husbands likely to be
tied stayers.
2
There is ample empirical evidence supporting the hypotheses proposed by
Mincer and colleagues in the 1970’s. Several papers have shown a reduced migration
in couples with a working wife (e.g. Long 1974; Sandell 1977; Mincer 1978; Juerges
1998; Nivalainen 2004). Here the husband may be a tied stayer who foregoes
potential earnings increases in order to secure the wife’s labour market position.
Married women have been shown to experience a decline in labour market
participation and earnings following a family move, thus classifying them as tied
movers (e.g. Morrison and Lichter 1988; Jacobsen and Levin 1997; Lee and Roseman
1999; Boyle et al. 2001). Research investigating the dynamics of employment and
earnings in the years following a move shows that, after an initial drop, women do
recover their pre-move positions after 1-3 years (Spitze 1984, Maxwell 1988, Lichter
1983). A more recent paper reveals no effect of family migration on wives’ earnings
and a dip in their labour force participation which gradually disappears after 3 years
(LeClere and McLaughlin 1997).
Husbands’ earnings are generally found to increase after migration (e.g.
Sandell 1977; Cooke 2003) while their employment remains unaffected. The effect on
household earnings is less clear. By expectation the household as a whole should
benefit from migration (Sandell 1977; Cooke 2003), but Axelsson and Westerlund
(1998) find no effect, and Jacobsen and Levin (1997) find a negative effect of family
migration on household income.
While these results generally support the unitary family model, there has been
research suggesting that expected earnings gains of husbands are weighted more
heavily than the earnings gains of wives in the family migration decision.
Specifically, several studies show that, contrary to expectation, personal and job
characteristics of the wife which should be indicative of their earnings potential, like
high occupational status, earnings, or job attachment, exert little or no influence on
family migration (e.g. Duncan and Perrucci 1976; Lichter 1982; Shihadeh 1991;
Juerges 1998; Nivalainen 2004; however: Bird and Bird 1985). This research usually
studies how migration probability is influenced by characteristics at origin, not at the
potential destination, and thus does not explicitly analyze the gains from migration.
By contrast, Jacobsen and Levin (2000) explicitly test the coefficients on a measure
of the spouses’ potential earnings gain in a migration equation. This measure is
created by estimating wage equations for each of the 37 locations available to each
household in their sample. They find that migration is positively related to potential
3
household earnings change regardless of who contributes to it. On the other hand,
using spouses’ pre-migration income as a measure of earning potential, Cooke (2003)
finds that post-migration income is correlated with the husband’s earning potential but
not with the wife’s, thus giving evidence of asymmetrical decision-making in favour
of the husband. This literature often concludes that a gender-role model does better in
predicting migration than human capital theory does (e.g. Bielby and Bielby 1992).
Another interpretation of this evolving literature is that it challenges the
unitary family model which underlies the theory of family migration. In recent years
household economics has seen a departure from the unitary family utility function in
favour of collective models of marriage which analyze interactions of spouses with
separate sets of preferences (e.g. Chiappori 1992; Fortin and Lacroix 1997).
Regarding the migration of dual-earner couples, a non-cooperative bargaining model
suggests that each spouse’s individual utility maximization may result in location and
employment decisions which are not efficient at the household level (Lundberg and
Pollak 2003). The main argument is that location decisions of dual-earner couples
may affect future bargaining power, for example via the assumption that individual
earnings determine distribution within marriage. In the absence of binding
commitments to prevent exploitation of future bargaining advantage, inefficient
outcomes are plausible and provide an alternative explanation of asymmetrical
decision-making.
Against this background this paper investigates how the prospective earnings
gains of husbands and wives are weighted when making the migration decision. Put
another way, it addresses the question of how the gains and losses resulting from
family migration are distributed between the spouses within individual families. The
analysis is based on the British Household Panel Survey, 1991-2003, which provides
a rich set of information about individual and household characteristics. The panel
nature of the data allows one to identify pre- and post-migration characteristics and to
account for individual heterogeneity. Migration is defined as a move across a Local
Authority District boundary. Unlike most other papers, the focus is on matched
partnerships rather than groups of husbands and wives that are not necessarily
associated to each other. Couples are defined to include legally married or cohabiting
adults of opposite sex, with or without children. These will be referred to as
‘husbands’ and ‘wives’, or ‘families’.
4
I estimate reduced and structural forms of the migration decision function as
well as selectivity-corrected earnings change equations for migrants and non-migrants
(separately for husbands and wives) in an endogenous switching model. The paper
accounts for selectivity both in the migration decision and in each spouse’s labour
force participation by employing a double-selection framework (Heckman 1979;
Tunali 1986). Many previous migration studies have refrained from modelling
selection effects. Those who have modelled them obtained results which suggest that
there is either positive selection into migration or that selection effects cannot be
uncovered (see Greenwood 1997 for a review). Only few papers have studied the
double selection into migration and employment (Vijverberg 1995; McNabb and
Vijverberg 1998).
3. Econometric methods
In line with the standard human capital formulation a family’s decision to migrate is
modelled as a function of, for each spouse, the difference between origin and
destination earnings, and migration costs. For simplicity I aggregate all potential
destinations into one so that migration is treated as a binary choice. A couple will
decide to migrate if the discounted net gain of moving, , is positive. This can be
written as:
cM
0* >−∆+∆≡ cwhc CYYM , (1)
where , are the present values of the expected lifetime earnings differentials
of migrating vs. not migrating for husbands and wives respectively. is the present
value of the couples’ mobility costs which include monetary and non-monetary costs
of moving like gathering information about alternative labour markets, leaving
networks of family and friends, house-price differentials, etc.
hY∆ wY∆
cC
In the empirical specification of this model I take the hourly wage gain of
migrating vs. staying between 1−t and t (around one year in the data) as a predictor
of lifetime earnings change. However, for migrants we only observe the change in
origin to destination wage, and for non-migrants we only observe the wage change in
the origin from one year to another. The alternative expected wage changes are not
observable. Neither are migration costs directly observable. Instead of the actual gain
from migrating we observe a binary random variable which is defined as *cM
5
⎩⎨⎧ ≥
=otherwise 0
0* if 1 cc
MM .
Assuming that migration costs are determined by a vector of exogenous household
and both spouse’s individual characteristics as well as house prices in the origin
locality, Z , and defining 1−−≡ tt www , a structural probit model of family migration
can be described such that:
ccwswmhshmc vZwwwwM ++−+−= ')()(* 321 γγγ , (2)
where the first two terms capture the predicted hourly wage change differential of
migrating vs. staying for husbands and wives respectively. The error
term, , is normally distributed with zero mean and unit variance.
),( sm ),( wh
cv
The predicted hourly wage change differential is obtained by estimating
earnings change equations for migrants and non-migrants respectively, separately for
husbands and wives. These are used to impute earnings changes for the counterfactual
situation, i.e. the change in earnings a migrant (stayer) would have received had he or
she stayed (migrated). Letting X be a vector of human capital and personal variables
that determine log hourly wages and defining 1−−≡ tt XXx and 1−−≡ tt εεε , the log
hourly wage change equations for migrating and non-migrating husbands and wives
respectively which complete the model are:
hmhhmhm xw εβ += 'ln , (3)
hshhshs xw εβ += 'ln , (4)
wmwwmwm xw εβ += 'ln , (5)
wswwsws xw εβ += 'ln . (6)
Taking first differences of the variables over a one year period eliminates any time-
invariant unobserved heterogeneity present in the error term. The wage-change
equations could consistently be estimated by OLS under the assumption
that 0)( 1 =− −ttE εε if wage data for migrants and non-migrants was randomly
selected. However, equations (3)-(6) may be subject to selectivity both in mobility
and labour force participation choice.
Self-selection. The selection into migrants and non-migrants may be non-
random as arguably migrants may differ from stayers in observed and unobserved
characteristics which also affect wages. Some forms of individual heterogeneity like
6
ability may be interpreted as time invariant and are thus eliminated by differencing the
variables in the outcome equation over time. However, there may be a transitory
component of the error term correlated with the migration choice. For example, the
cost-benefit calculation implicit in the migration decision will depend on the current
economic environment which is likely to change over time (Yankow 2003), and
motivation to invest into human capital can change with life-cycle events like having
children or getting married. Not controlling for this will lead to biased estimators.
Furthermore, post-migration wages are only observed for individuals who
participate in the labour market, and they may be a non-random sample of the
population. For example, individuals who have the greatest difficulties in finding new
employment post migration select out of the labour market. One obvious example of
time-varying factors which may influence such difficulties for women and also affect
their wages is family factors.
The equations which model these selection processes are
111 '* uXM c += δ family migration choice (7)
222 '* uMXP cw ++= µδ wife’s labour force participation (8)
333 '* uMXP ch ++= ηδ husband’s labour force participation. (9)
Equation (7) is in essence a reduced form of the structural model described by
equation (2). and are latent variables, *wP *hP 1,2,3)( =bX b are vectors of
explanatory variables assumed to determine mobility and labour force participation
respectively, 'mδ are vectors of unknown coefficients, and ηµ , are scalar unknown
coefficients. By including a migration status indicator variable into the
participation equations of husbands and wives I allow participation to vary according
to the migration choice made. This binary variable captures a level effect of migration
on labour force participation of each spouse.2 The potential endogeneity of migration
choice with respect to the participation equations is discussed below.
)( cM
By assumption, a family chooses to migrate if the utility of moving exceeds
that of staying. At the same time, the spouses make participation choices for the wife
and the husband that depend on both spouse’s personal and job-related characteristics
and which may be interdependent. I discuss the empirical specification of the models
2 Several authors have shown that labour force participation choices may differ between migrants and stayers, for example because female migrants might temporarily withdraw from the labour force in order to accommodate increased household needs in the first years following a move.
7
in the next section. The household will choose participation by a spouse in the labour
market if the household’s utility from participation exceeds that of non-participation.
The utilities of the three selection equations are not directly observable; we observe
only a dichotomous variable indicating whether or not an individual migrates and
whether or not the spouses are employed.
A two-step estimation approach in the spirit of Heckman (1979) is adopted to
deal with the self-selection problems. This procedure begins by estimating the
selection equations and constructing sample-selection correction terms based on these
estimates. Then the wage equations (3)-(6) can consistently be estimated by OLS,
including the correction terms as additional regressors. The wives’ wage change
equations for migrants and non-migrants are corrected for selectivity in migration and
the wives’ labour force participation. Likewise, the husbands’ wage change equations
are corrected for non-random selection into migration and husbands’ employment.
If the selection effects specified in equations (7)-(9) are independent, i.e. the
correlations of the error terms, 1221 ],[ ρ=uuCov ; 1331 ],[ ρ=uuCov ; 2332 ],[ ρ=uuCov ,
are all equal to zero, the correction terms can be derived from separate probit models
for migration and participation choices as in the single selection case. The correction
terms for selection into migration then are the inverse Mills ratios, namely:
)()(
11
11
XX
M δδφ
λΦ
= if and 1=cM)(1
)(
11
11
XX
M δδφ
λΦ−
−= if 0=cM (10)
where φ is the standard normal density function and Φ the corresponding standard
normal distribution function. The correction terms for selection into employment are
derived analogously.
As the migration and participation decisions are – by assumption – made
jointly, the selection processes may not be independent, i.e. some or all of the
correlations 231312 , , ρρρ may not be equal to zero even after controlling for
observed heterogeneity. If only one of the pair-wise correlations, say 23ρ , is different
from zero, it is possible to apply a double-selection framework (Tunali 1986; Fishe et
al. 1981; Ham 1982). This assumes a trivariate normal distribution of the error terms
of the outcome equations and, in our example, the two equations modelling the
spouses’ selection into employment. The selection processes are estimated using a
bivariate probit, and augmented selection correction terms are derived for migrants
who are working:
8
),ˆ,ˆ()ˆ()ˆ(
232
2
ρφ
λhw
wP PP
GPΦ
Φ= (wife) and
),ˆ,ˆ()ˆ()ˆ(
232
1
ρφ
λhw
hP PP
GPΦ
Φ= (husband). (11)
The correction terms for working non-migrants are:
),ˆ,ˆ()ˆ()ˆ(
232
2
ρφ
λ−−Φ
−Φ−=
hw
wP PP
GP (wife) and
),ˆ,ˆ()ˆ()ˆ(
232
1
ρφ
λ−−Φ
−Φ−=
hw
hP PP
GP (husband) (12)
where
223
2322
23
231
1
ˆˆˆ ,1
ˆˆˆρ
ρ
ρ
ρ
−
−=
−
−= whhw PP
GPP
G ,
and is the bivariate standard normal distribution function. 2Φ
Selection correction terms accounting for the interdependence among three
selection equations have – to my knowledge – not been derived in the literature. I
therefore adopt an evidence-based procedure as follows. First I estimate the selection
equations (7)-(9) jointly in a trivariate probit model using simulation methods based
on the GHK simulator (Gourieroux and Monfort 1996; Cappellari and Jenkins 2003),
and choosing 94 random draws which is the square root of the sample size. The
resulting model is a recursive, simultaneous equation model because the dependent
variable of equation (7) appears as independent variable in equations (8) and (9).
Identification of the equation system is assured by employing a set of exclusion
restrictions. In particular, each equation includes one exogenous variable assumed to
be relevant only for this equation which is excluded from the other equations. I
describe the variables used in the next section and provide test results for the validity
of these restrictions in the results section. In all estimates I adjust the standard errors
for having repeated observations on some couples.
These estimates yield values for 231312 , , ρρρ and can be seen as a test for
independence of the selection processes. Based on these results I am able to separate
the migration equation from the participation equations because I do not reject the
assumption 01312 == ρρ while I do reject 023 =ρ at 1% level of significance. This
also means that the migration dummy is exogenous in the participation equations.
Thus I continue to estimate a single migration probit and a bivariate probit of
husbands’ and wives’ labour market participation. I derive inverse Mills ratios to
account for selection into migration and augmented correction terms to account for
9
selection into employment as described above, equations (10)-(12). These then enter
into the wage change equations and allow consistent estimation.
The estimated coefficients on the migrants’ and non-migrants’ log hourly
wage change equations are used to predict each spouses’ expected wage change when
migrating and when staying in the original location, separately for husbands and
wives. The coefficients of the migrants’ wage change equations can be interpreted as
a weighted average of prices obtainable for individual and job-related characteristics
in potential destinations. The weights represent the locations actually chosen by
migrants. Conversely, the coefficients on the non-migrants’ equations reflect the
prices obtainable at an average origin, weighted by the origin locations of stayers.
Expected wage changes are derived from predicted using wln
)2ˆˆexp(ln)ˆ( 2εσ+= wwE . Differencing predicted wage changes in origin and
destination yields a wage change differential of migrating vs. staying for each
individual. The final step of the procedure is to estimate the structural version of the
migration probit (equation 3) using these differentials. The coefficients on the wage
change differentials reveal how the potential gains and losses of migration are
weighted between the spouses. In interpreting these coefficients the different expected
working times of husbands and wives have to be taken into account.
4. Data
The analysis uses the 13 currently available waves of the British Household Panel
Survey (BHPS), spanning 1991 to 2003. The BHPS is a nationally representative
sample of about 5,500 households recruited in 1991, containing approximately 10,000
adults who are interviewed each successive year. Data are collected on a broad range
of socio-economic characteristics at both the individual and the household level. The
use of panel data allows identification of both the pre- and post-migration situation.
From the BHPS I extract an estimation subsample. Following common
practice I restrict the analysis to couples, married or cohabiting, who remain intact and
living together at the same address over a two-year period.3 I choose couples where
3 Restricting the sample to include only couples who remained intact over a two-year period potentially introduces selection bias in the analysis because it could be differences in migration preferences that lead to couple dissolution. Indeed, in our dataset mobility is higher among recently separated individuals than among intact couples. However, divorced and separated persons are shown to have higher migration propensities compared with other individuals as a result of partnership break-up. With causality running in both directions it is difficult to disentangle the effects (Mincer 1978).
10
both spouses were in employment at 1−t , i.e. prior to a possible move and are aged
between 20 and 58 inclusive. The focus on dual-earners allows examination of wage
changes for both partners. The age band concentrates the analysis on a group most
likely to migrate for work-related reasons. Migration by individuals approaching
retirement age may reflect retirement location decisions. The inclusion of younger
individuals would potentially confound education-related moves with migration.
Same-sex couples, individuals who ever worked in the armed forces, and employed
individuals reporting zero working time in the previous week were dropped from the
dataset. Finally the bottom and top 0.3% of the hourly wage distribution were deleted
in order to eliminate extreme outliers. The resulting sample is an unbalanced panel
which includes all couples, for whom information is available at two consecutive
interviews, allowing the same couple to enter the sample several times. After
removing observations with missing data for any of the variables used in the analysis,
the sample comprises 8,557 observations with 1,838 unique couples.
Migration is defined in this paper as a change in a family’s address in the
period between two interviews which also involved crossing the boundaries of one of
Britain’s 278 local authority districts (Böheim and Taylor 2005). Any local move
within the boundaries of a local authority is thus considered to be residential mobility.
This definition identifies 196 migrating couples in the dataset. While the BHPS
attempts to follow all movers who remain in private households in Britain, attrition
rates are higher among individuals who move house than among those who remain in
the same residence (Buck 2000). However, in most cases it is possible to identify the
destination area and thus the migrant status of individuals who do not remain in the
sample. I have analysed non-response among dual-earner couples whose pre-attrition
characteristics were identical to the restrictions imposed on my estimating subsample.
Attrition is very low in this group (below 2% both among non-migrants and
migrants). Therefore I do not expect it to influence the results.
The dependent variable in the wage change equations is the change in the log
of usual real hourly gross earnings from 1−t to t in January 2000 prices. Hourly
wages were derived from usual monthly wages using hours worked and assuming that
paid overtime hours were associated with a 50% wage premium. Wages were
modelled as a function of changes in time-variant personal and job-related
characteristics. In particular, changes in occupational status (5 binary variables),
working time, hours of paid overtime, type of job contract (temporary/permanent),
11
sector (public/private), occupational pension, and union membership are included in
the wage change equations. As there is a close link between the returns to migration
and job mobility I also include a dummy indicating whether an individual has changed
jobs between two interviews in the wage change equations of migrants and non-
migrants. It would be desirable to further distinguish layoffs, quits and transfers
among job changes (Bartel 1979). However, information on this is collected
retrospectively in the BHPS and is known to suffer from problems of recall error and
ex-post rationalisation (Böheim and Taylor 2005). A variable for the change in job
tenure measures the time between two interviews for individuals who don’t change
job, and time elapsed between starting a new job and the next interview for job
changers. The wage change equations also contain correction terms for the selection
into migration vs. non-migration and for the selection into labour market participation.
Each equation also includes year-of-interview indicator variables.
The three equations modelling the selection effects are specified as follows.
The migration equation includes all personal and job-related variables of both spouses
which are relevant for the wage change equations. These are included to capture the
economic opportunities that individuals face in the labour market. In particular, they
model risk of job-loss (e.g. via information on fixed-term job contracts), labour
market attachment (e.g. job tenure), and job offer arrival rates. In addition I include
personal and household characteristics which act as proxies for the direct and indirect
costs of moving. Older individuals are expected to be less mobile than younger ones,
either because they reap the benefits of migration for a shorter period or because they
are more attached to their current location. Thus I include the husband’s age as well as
the age difference between husband and wife in the migration probit. A binary
variable coding whether each partner has received a higher qualification (degree,
teaching qualification or other higher qualification/other qualification) is assumed to
proxy their ability to perform effective job search. The size of the household is an
indicator of the direct costs of moving and of the network attached to any family. I
distinguish between the number of children under age five and over age five in the
household. Families with pre-school children are often found to migrate in search for
better environments for their children while the presence of school-children usually
deters relocation because of the difficulties involved in changing schools. I also
include a variable which indicates whether any of the pre-school children were newly
born between . Binary variables indicating outright home ownership vs. tt and 1−
12
mortgaged home ownership vs. rental accommodation approximate the costs of
relocation. Finally, because legally married spouses may attach a higher cost to
partnership dissolution and would therefore be more likely to migrate together than
those merely cohabiting, I include this variable in the migration equation. All of these
variables are measured at , i.e. the time when the migration decision is assumed
to be made.
1−t
The husbands’ and wives’ labour force participation equations each include
the same variables as the migration equation, i.e. personal and job-related variables of
both spouses. As the labour force participation of spouses is by assumption
interdependent, a husband’s job prospects may influence the wife’s participation
decision and vice versa. The job-related characteristics are measured at time ,
assuming that characteristics of each partner’s job determine the likelihood of labour
force participation for each spouse. The personal characteristics refer to time .
Labour force participation will depend on the partners’ ages, education, and marital
status – being married generally decreasing the labour force attachment of women and
increasing it for men. Likewise, the existence of children is expected to reduce a
woman’s involvement in market work, especially if the children are young, while it
will tend to increase a man’s labour market activity. Owning a home outright and
having more non-labour household income should reduce the dependence of the
family on earned wages and thus participation in the labour market.
1−t
t
The three equations modelling the selection process are, in a first step,
estimated simultaneously. Identification of the equation system is assured by a set of
exclusion restrictions, i.e. including an exogenous variable in each equation which is
not included in or relevant for any other equation. In particular, the migration equation
includes a time-varying measure of local house prices in the origin location, assuming
that families seek to move away from high house prices.4 For the spouses’
participation equations we make use of information on the employment status of each
individual’s parents. We assume that a woman will be influenced in her choice of
employment status by the status her mother chose when she was a youth. Arguably a
mother serves as a role model in terms of female labour market participation for her
daughter. Likewise we assume a man to be influenced by his father’s past 4 A price index was constructed using historical data provided by Halifax Housing Research on average annual house prices at UK posttown level from 1991-2003. These data were aggregated to Local Authority District (LAD) level, and the annual index in any LAD is the log of average house prices in that LAD divided by the national average.
13
employment status, particularly with respect to the choice between employment and
self-employment. We use a binary variable indicating whether an individual’s mother
(father) was an employee when she (he) was aged 14 to identify the wife’s
(husband’s) labour force participation equation.
5. Empirical results
Table 1 presents descriptive evidence on migration, wages, and labour market
participation over the period 1991-2003 for the analysis sample. About 9% of the
couples in the sample migrate some time over the 13 year period, which are 2.2% of
the couple-year observations. These migration rates are well below those found in the
BHPS for a sample of married and single men (15% and 3.9% over a 12 year period
according to Böheim and Taylor 2005), confirming the hypothesis of reduced
migration among dual-earner families. Migration is concentrated among couples with
pre-migration wages well above the average: hourly wages of migrating husbands
were 17%, and those of migrating wives 21% higher than those of their non-migrant
counterparts prior to migration. The table also shows that wage increases over a one-
year period are greater for migrants than non-migrants. Interestingly, on average both
migrating and non-migrating working women have wage increases that exceed those
of their husbands by about one percentage point. Hours normally worked per week are
slightly higher among migrants before they move and increase post-migration
particularly among employed women. However, participation in the labour market
declines more among migrants than non-migrants. While this difference is moderate
among men (1 percentage point), the participation rate of migrating wives is 11.9
percentage points lower than that of non-migrating wives. Finally, the table shows
that, as expected, job change is more prevalent for migrants than non-migrants. In
summary, these descriptive results indicate that couples migrate to get better paid jobs
of husbands and wives, but that migrating wives may stop working following a move.
Obviously we can only confirm these conjectures by controlling for individual and
couple heterogeneity which I do next.
The first step in the estimation procedure is to estimate the three selection
equations by multivariate probit in order to explore the error term correlations. The
upper panel of table 2 presents Wald test results for instrument validity and hence the
identification of the model. The hypothesis that the coefficients on the instruments
14
equal zero can be rejected at conventional levels of significance, both separately and
jointly. The lower panel of table 2 displays the estimates for the error term
correlations. It shows that there is no statistically significant correlation of the
unobservables of the migration equation with the two participation equations. Thus I
can reject endogeneity of the migration dummy in the participation equations. This
allows me to estimate a single migration probit and to derive the selection correction
terms from the results. The labour force participation of husbands and wives are,
however, clearly interdependent even after the influence of the observed factors is
accounted for. This interdependence must be taken into account when constructing the
selection correction terms. Augmented correction terms are derived from a bivariate
probit model of labour market participation by husbands and wives.
Table 3 shows the estimates. The migration probit confirms most of the
standard expectations about mobility choices. The presence of school-age children
and house ownership inhibit migration. Migration propensity also declines with age of
husband and wife, while high house prices raise this propensity. Marital status and
education do not seem to influence migration probabilities. The results pertaining to
the characteristics of the spouses’ job in 1−t are noteworthy, however. Unlike for
husbands, wives’ job characteristics at origin such as occupational status, type of job
contract, etc. (not displayed separately) play no role in the joint migration decision.
The Wald test does not reject the hypothesis that the coefficients on these variables
are jointly equal to zero for the wife. These estimates thus prima facie replicate the
findings cited in the literature review that it is the husbands’ job that drives the
migration choice.5 The estimates of the structural model containing the estimated
wage change differentials between origin and destination, presented below, will imply
that this conclusion should be revised.
Characteristics of the job in origin are clearly important in determining wives’
employment. In particular, public sector employees, skilled manual workers,
permanent contract and occupational pension holders as well as individuals working
many hours have a high probability of post-migration employment. New mothers are
5 The wife’s job characteristics might be unimportant in the migration decision if couples with dissimilar work attachment select into migration. I have tested this proposition by including relational variables comparing the spouses’ education and occupational status instead of separate variables for husbands and wives characteristics in the migration equation. The results rather suggest that couples with similar traits migrate together. This may suggest that collinearity of husbands and wives’ characteristics, which is likely to appear in the presence of assortative mating, may inflate standard errors. However, collinearity diagnostics do not support this.
15
less likely to work while mothers of school-age children have an increased likelihood
of employment. Having a mortgage rather than renting increases female participation
while higher levels of non-labour household income decrease it. Most important,
wives who migrate have lower employment rates at the interview following migration
than those who do not. Evaluated at the means of the other variables, a migrating wife
has a 7.6 percentage point (or 8%) lower probability of employment than a non-
migrating wife controlling for her personal, household, and job-related characteristics.
This reduction is lower than that found in several other studies predominantly based
on US data (e.g. LeClere and McLaughlin 1997: 13% (US); Boyle et al. 2001: 42.5%
(UK), 40.7% (US); Lee and Roseman 1999: around 10-15 percentage points (US)).
Probit estimates based on data for those 7,285 wives also interviewed in the following
year ) reveal that migration no longer has a negative impact on their
employment probability. The coefficient on the migration dummy is not significantly
different from zero at 1+t (not shown). Thus the decline in wives’ employment
probability is temporary and, in line with the findings in other papers, they recover
after a
1( +t
bout one year.
Also in accordance with past research, migrant husbands do not experience a
reduction in employment, compared with non-migrant husbands. The presence of
school-age children and being married enhance employment propensities, whereas
age and non-labour household income inhibit it. Moreover, husbands’ labour market
participation is directly linked to their wives’ past job characteristics. Finally, those
who worked in the private sector or in temporary job contracts at have lower
employment rates, while higher levels of paid overtime and occupational pension
coverage increase employment probability at t.
1−t
The estimates were used to derive the selection correction terms for time-
variant individual heterogeneity described by equations (10)-(12). These are used as
regressors in the log wage change equations in order to derive consistent estimates.
Table 4 displays the results. Wage changes were estimated for individuals employed
in . The estimates of the migrants’ wage change equations are based on relatively
small samples and thus have low levels of significance. The coefficients are quite
consistent across the four groups. Most notably, wage changes are negatively
associated with work hours and paid overtime, indicating that extended hours are not
rewarded with proportional pay rises. Jobs covered with occupational pension
t
16
schemes have a wage premium attached to them, and becoming a union member also
helps. Changing to a fixed-term contract has a negative effect on year-to-year wage
changes, especially for migrants. On the other hand, migrant wives are rewarded more
than non-migrants for taking on a new job. The coefficients on the occupational status
binary variables are a bit surprising. In contrast to non-migrants, wages are lower for
both migrating husbands and wives if they move into some of the higher status groups
compared to the reference group (partly and unskilled occupations). On average,
migrants improve their occupational status more than non-migrants, but there is no
wage premium associated with this.
The estimates also reveal statistically significant selection bias for non-
migrants. The signs on the coefficients on Mλ indicate that non-migrants had
unobserved characteristics which deter both migration and wage growth. Moreover,
non-migrant wives who chose to work receive lower wages than the female
population at large, as the negative coefficient on WPλ indicates. In contrast, non-
migrant husbands who chose employment possess comparative advantage at it, but
with a smaller effect on wage changes than the selection into non-migration. The
coefficients for migrants are not statistically significant, giving no evidence of
selection bias for them.
The coefficients on the wage change equations reflect the average price of
each individual and job characteristic in a hypothetical origin and destination
respectively. In order to explore the counterfactual wage changes I define the potential
destination for each non-migrant as the average of the locations chosen by migrants,
the hypothetical origin as the average of the locations chosen by non-migrants. Using
these estimates, wage changes were imputed for each husband and wife for the
counterfactual situation, i.e. the wage change a non-migrant would have received
when migrating and vice versa. The coefficients on the selection-correction terms
were included when imputing the counterfactuals. This implicitly assumes that the
unobserved (time-variant) differences between individuals captured in the selectivity
terms are equally useful in origin and destination. Deriving expected wage changes
from the log wage equations and differencing between origin and destination wage
change yields a wage change differential for each employed individual. Individuals
not employed in t will, by assumption, also evaluate lifetime earnings changes in
their decision to migrate which are approximated by the immediate expected returns.
17
For these individuals I derive a wage change differential by assuming their personal
and job characteristics in t to be equal to the average among the group of migrants and
non-migrants respectively. Any migration-related withdrawal from the labour market,
often associated for women with accommodating family needs post migration, is
interpreted as being temporary – indeed no effect of migration on participation choice
remains in . As there is no methodology available for deriving counterfactual
withdrawal information for stayers, the costs related to women’s temporary non-
participation will be picked up in the error terms of the structural migration probit.
1+t
Table 5 describes the distribution of predicted wage gains of migrating versus
staying by forming groups based on whether or not the husband, wife, and the couple
as a whole gain in terms of wage changes by moving. Column I gives counts of
migrants and non-migrants who could as a couple achieve a positive wage change by
migrating for different combinations among the spouses. Column II counts couples
who would fare worse in terms of joint income change by migrating. Below the
counts are, in parentheses, average estimated hourly wage change differentials in each
category, first for husbands and then for wives. Looking first at migrants, column I
shows that in 195 migrating couples (99.5% of migrants) both spouses gain a
considerable wage change by migrating: husbands gain an average of 48 pence per
hour in year 2000 prices, wives gain an average of 34 pence. Migration thus seems to
be a rational choice for these families. Only one couple migrates with a positive
family return to migration but minor losses for the wife. According to our estimates
there is thus only one case of a wife possibly being a tied mover. In the entire sample
we observe no migration with a negative family return.
Turning now to non-migrants, the first row in column I shows that 4,016
couple-observations (47% of observations) do not migrate although they would
apparently experience wage gains individually and as a couple by doing so. Compared
to the migrants in this category, however, their average expected gains are modest,
amounting to 16 pence for husbands and 12 pence for wives per hour worked. As the
figures do not include the unknown mobility costs, it is well possible that these
couples would in total not fare better by migrating. Likewise, it may well be that the
2,377 observations (28%) of couples seen not to be moving in spite of a predicted
family wage gain, would not benefit after taking mobility costs into consideration.
However it is possible that some husbands in the group are tied stayers who would
have fared better individually by moving net of migration costs. Likewise, the 606
18
observations listed in the row below could include wives who are tied stayers. Non-
migrants in column II have nothing to gain as a family from migration. While for 432
observations listed in the bottom row the decision was probably unambiguous, the
two middle rows may include a number of husbands and a few wives who are tied
stayers.
The comparison between migrants’ and non-migrants’ expected wage gains is
consistent with expectations. Although we cannot observe the earnings gain threshold
after which migration becomes profitable individually or as a family, the estimates
show that couples who expect higher wage gains migrate. Migrants’ expected wage
gains do not vary by distance moved, although information on this is only available up
to year 2002.
The last step in the estimation procedure is to estimate a structural migration
probit containing the predicted wage change differentials of husbands and wives.
Table 6 displays two different sets of results. Column (1) measures the hourly
expected wage change differentials in January 2000 pounds. Column (2) takes the
spouses’ log wage change differentials as independent variables. Bootstrapping was
used to derive confidence intervals in order to infer significance levels for the two-
stage procedure (1000 iterations). If the bias-corrected confidence interval did not
include zero, the effect was taken to be significant at the level indicated by the size of
the confidence interval.
The coefficients on the spouses wage change differentials are quite similar
between husbands and wives. The wives’ and husbands’ predicted wage change
differentials are weighed 41% to 59% in the family migration decision in specification
(1). Specification (2) yields a relationship of 46% to 54% in favour of the husband.
Considering that women work fewer hours than men, however, and that the wage
differentials refer to changes in hourly wages, it seems that both spouses wage gains
are roughly equally important. In answer to the main question driving this paper I
conclude that both spouses in dual-earner couples weigh their respective wage gains
about equally in the migration decision. Thus I reject earlier findings that it is the
husband’s job alone that motivates dual-earner migration. I discuss the implications of
these findings in the conclusion.
In line with most migration studies I also find that home owners and older
individuals are less likely to migrate. The same is true for families with pre-school
and school-age children. The influence of non-labour household income and house
19
prices on family migration is not consistent across specifications. The other factors
hypothesized to influence migration behaviour – marital status and possessing a
higher education which was thought to proxy the ability to perform effective job
search – did not turn out to be statistically significant, however.
In order to test the robustness of results to model specification I carried out
several sensitivity checks. In particular, I tested the effect of using different exclusion
restrictions (information about self-employment status of parents or whether parents
were working at all, different measures of house price differentials) on the weights
attached to each spouse’s wage gain. The weights remained much the same as those
presented here, although in most cases alternative instruments did not identify the
model. In some of these specifications the migration equation was not independent of
the participation equations, which is not surprising in a non-identified model. Further
tests included checking whether the inclusion of imputed wages changed results,
controlling for labour market interruption between dates of interview in the wage
equations, checking the influence of health status on wages and mobility and the role
of care responsibilities outside the household on migration choice. In most cases these
variables were insignificant, and they had little effect on the final results while
reducing sample size due to missing values. Confining the structural probit analysis to
legally married, and thus perhaps more ‘traditional’, couples did not change results
either.
6. Conclusion
This paper provides evidence that dual-earner couples weigh the expected wage
change differentials of husband and wife about equally when deciding whether or not
to migrate. In contrast to much previous research, the results of this paper clearly
reject the assertion that couples make migration decisions asymmetrically in favour of
the husband’s job prospects for the group of dual earners. In fact, the analysis of
couples’ expected wage change differentials has shown that while there may be a
substantial number of husbands who are tied stayers, there is but one case of a wife
possibly acting as a tied mover in the subsample of the BHPS (1991-2003) used for
the estimates.
Previous papers suggesting that decision-making is husband-centred have
often relied on estimates from migration equations where the wife’s personal and job
20
characteristics in origin proxy expected wage gains. Like many of these papers I do
not find the characteristics of the wife’s job in origin to significantly influence the
family migration decision in a reduced form probit equation. However, the wife’s
expected wage change differential, obtained in a switching regression model and
corrected for the double selection into migration and employment, is highly relevant
in a structural form of the migration equation. This suggests that characteristics of the
job in origin may not capture the opportunities that women face in destination when
migrating. Put another way, migrating women have realized higher wage gains than
could have been expected from their jobs at 1−t . This highlights the value of using
panel data on origin and destination characteristics in migration analysis to uncover
these opportunities.
The employment effects of migration found in this paper are less pronounced
than those reported in other papers: In the first year after migration wives’
employment was reduced temporarily by 8% after other factors have been accounted
for, and husbands’ employment was unaltered by migration. In the following year no
migration effect on female employment remained. There is evidence of significant
selection effects among non-migrants. Non-migrants have unobserved characteristics
which deter both migration and wage growth. Moreover, non-migrant wives who
chose to work possess comparative disadvantage at it, while in contrast non-migrant
husbands experience higher wage growth than the population at large. This
emphasises the importance of modelling self-selection not only into migration but also
into employment when analyzing the wage effects of migration. In this paper I
demonstrate how double selection - which is fairly involved when analyzed in a
family context - can be accounted for by following an evidence-based procedure.
What can be learnt from this paper in terms of family decision-making?
Regrettably the results do not allow us to discriminate between the traditional unitary
model and a bargaining model of the family. Symmetry between husband and wife
could equally support the unitary model which assumes equal weights on the partners’
wage gains as it could a bargaining model of the family in which both partners have
equal bargaining power. In fact the assumption of equal bargaining power does not
seem unlikely in dual-earner couples that according to this analysis have similar
educational and occupational backgrounds. A natural research agenda would be to
further test this issue in differing frameworks, for example by explicitly modelling
21
several destinations available to a couple rather than aggregating these options into
one destination.
22
References Axelsson, R. and O. Westerlund (1998). "A panel study of migration, self-selection
and household real income." Journal of Population Economics 11(1): 13-126. Bartel, A. P. (1979). "The Migration Decision: What Role Does Job Mobility Play?"
American Economic Review 69(5): 775-786. Bielby, W. T. and D. D. Bielby (1992). "I Will Follow Him: Family Ties, Gender-
Role Beliefs, and Reluctance to Relocate for a Better Job." American Journal of Sociology 97(5): 1241-1267.
Bird, G. A. and G. W. Bird (1985). "Determinants of Mobility in Two-earner Families: Does the Wife's Income Count?" Journal of Marriage and the Family 47(3): 753-758.
Böheim, R. and M. P. Taylor (2005). "From the dark end of the street to the bright side of the road? The wage returns to migration in Britain." Labour Economics, forthcoming.
Boyle, P., T. J. Cooke, et al. (2001). "A Cross-national Comparison of the Impact of Family Migration on Women's Employment Status." Demography 38(2): 201-213.
Buck, N. (2000). “Using panel surveys to study migration and residential mobility.” Researching social and economic change. D. Rose (ed.). London/New York, Routledge: 250-272.
Cappellari, L., S. P. Jenkins (2003). “Multivariate probit regression using simulated maximum likelihood.” Stata Journal 3(3): 278-294.
Cooke, T. J. (2003). "Family Migration and the Relative Earnings of Husbands and Wives." Annals of the Association of American Geographers 93(2): 338-349.
Duncan, R. P. and C. C. Perucci (1976). "Dual Occupation Families and Migration." American Sociological Review 41(April): 252-261.
Fishe, R. P., R. Trost, and P. M. Lurie (1981). “Labor Force Earnings and College Choice of Young Women: An Examination of Selectivity Bias and Comparative Advantage.” Economics of Education Review 1(2): 169-191.
Fortin, B. and G. Lacroix (1997). "A Test of the Unitary and Collective Models of Household Labour Supply." Economic Journal 107(443): 933-955.
Gourieroux, C. and A. Monfort (1996). Simulation Based Econometric Methods. Oxford, Oxford University Press.
Greenwood, M. J. (1997). “Internal Migration in Developed Countries.” Handbook of Population and Family Economics. M. R. Rosenzweig and O. Stark (eds.). Amsterdam, Elsevier. 1 B: 647-720.
Ham, J. C. (1982): “Estimation of a Labour Supply Model with Censoring Due to Unemployment and Underemployment.” The Review of Economic Studies 49(3), 335-345.
Heckman, J. J. (1979). "Sample Selection Bias as a Specification Error." Econometrica 47(1): 153-161.
Jacobsen, J. P. and L. M. Levin (1997). "Marriage and Migration: comparing Gains and Losses from Migration for Couples and Singles." Social Science Quarterly 78(3): 688-709.
Jacobsen, J. P. and L. M. Levin (2000). "The Effects of Internal Migration on the Relative Economic Status of Women and Men." Journal of Socio-Economics 29: 291-304.
23
Juerges, H. (1998). "Beruflich bedingte Umzuege von Doppelverdienern. Eine empirische Analyse mit Daten des SOEP." Zeitschrift fuer Soziologie 27(5): 358-377.
LeClere, F. B. and D. K. McLaughlin (1997). "Family migration and changes in women's earnings: A decomposition analysis." Population Research and Policy Review 16: 315-335.
Lee, S. and C. C. Roseman (1999). "Migration Determinants and Employment Consequences of White and Black Families, 1985-1990." Economic Geography 75(2): 109-133.
Lichter, D. T. (1982). "The Migration of Dual-Worker Families: Does the Wife's Job Matter?" Social Science Quarterly 63(1): 48-57.
Lichter, D. T. (1983). "Socioeconomic Returns to Migration Among Married Women." Social Forces 62(2): 487-503.
Long, L. H. (1974). "Women's Labor Force Participation and the Residential Mobility of Families." Social Forces 52(3): 342-348.
Lundberg, S. and R. A. Pollak (2003). "Efficiency in Marriage." Review of Economics of the Household 1(3): 153-167.
Maxwell, N. L. (1988). "Economic Returns to Migration: Marital Status and Gender Differences." Social Science Quarterly 69(1): 108-121.
McNabb, R. and W. P. M. Vijverberg (1998). “Household Migration and the Labour Market Behaviour of Married Women”. mimeo.
Mincer, J. (1978). "Family Migration Decisions." Journal of Political Economy 86(5): 749-773.
Morrison, D. R. and D. T. Lichter (1988). "Family Migration and Female Employment: The Problem of Underemployment among Migrant Married Women." Journal of Marriage and the Family 50: 161-172.
Nivalainen, S. (2004). "Determinants of family migration: short moves vs. long moves." Journal of Population Economics 17(1): 157-175.
Sandell, S. H. (1977). "Women and the Economics of Family Migration." Review of Economics and Statistics 59(4): 406-414.
Shihadeh, E. S. (1991). "The Prevalence of Husband-Centered migration: Employment Consequences for Married Mothers." Journal of Marriage and the Family 53(2): 432-444.
Spitze, G. (1984). "The effect of Family Migration on Wives' Employment: How Long Does It Last?" Social Science Quarterly 65(1): 21-36.
Tunali, I. (1986). “A General Structure for Models of Double-Selection and an Application to a Joint Migration/Earnings Process with Remigration”. Research in Labor Economics. Greenwich, JAI Press. 8, part B: 235-282.
Tunali, I. (2000). "Rationality of Migration." International Economic Review 41(4): 893-920.
Vijverberg, W. P. M. (1995). "Dual Selection Criteria with Multiple Alternatives: Migration, Work Status, and Wages." International Economic Review 36(1): 159-185.
Yankow, J. J. (2003). "Migration, Job Change, and Wage Growth: a New Perspective on the Pecuniary Return to Geographic Mobility." Journal of Regional Science 43(3): 483-516.
24
Table 1: Migration, Wages, and Labour Market Participation, BHPS 1991-2003
Non-migrants Migrants Observations 8,557 196 (2.2%)Couples 1,823 175 (8.8%) husbands wives husbands wives Mean hourly wage (£) 1−t 8.7 6.3 10.2 7.6Mean hourly wage t (£) 9.4 6.9 11.5 8.7Mean wage increase to t 1−t 8.0% 9.5% 12.7% 14.5%Mean weekly working hours 1−t 40 29 41 32Change in working hours to t 1−t 0% 0.2% 0.2% 1.2%Participation rate t 95.5% 92.0% 94.4% 80.1%Job changers to t 1−t 19.7% 20.5% 42.7% 45.2%Notes: Gross monthly wages in January 2000 prices. All values for time t refer to individuals employed in t only. Migration is defined as a move across Local Authority District boundaries.
Table 2: Selection model: tests for identification and instrument validity, and estimated correlations of error terms, BHPS 1991-2003
Instrument Equation d.f. 2χ p-value House price index t-1 M 1 3.78 0.052 Mother employee when respondent aged 14 wP 1 4.67 0.031 Father employee when respondent aged 14 hP 1 4.23 0.040 All instruments jointly 3 12.29 0.007 Correlations of error terms variable coefficient t-value Migration, participation wife 12ρ -0.249 (0.83) Migration, participation husband 13ρ -0.423 (1.35) Participation wife, participation husband 23ρ 0.152 (3.10)**Notes: ** significant at 1%. Estimated using GHK simulator (94 draws) and heteroskedasticity-robust standard errors. Top panel displays Wald test that the coefficient estimate for an instrument is zero in the listed equation. Bottom panel displays estimated correlations of error terms. A full set of results is available on request.
25
Table 3: Migration and participation probits, BHPS 1991-2003
Univariate Probit Bivariate Probit Migration LFP wives LFP husbands Migrant -0.483 (4.16)** -0.100 (0.64) Married -0.022 (0.27) -0.105 (1.48) 0.249 (3.27)**New baby -0.021 (0.11) -1.593 (16.09)** 0.182 (1.22) Number children < 5 -0.084 (0.94) -0.013 (0.20) 0.100 (1.26) Number children ≥ 5 -0.111 (2.71)** 0.164 (5.39)** 0.089 (2.55)**Home owner outright -0.537 (3.12)** 0.147 (1.35) -0.034 (0.30) Home owner mortg. -0.560 (6.72)** 0.188 (2.53)** 0.014 (0.16) Non-labour household-income
0.026 (1.15) -0.144 (8.57)** -0.166 (8.43)**
High education wife 0.008 (0.09) 0.021 (0.37) -0.043 (0.67) High education husb. -0.058 (0.76) -0.063 (1.22) 0.038 (0.66) Age husband -0.025 (5.69)** 0.0063 (1.79) -0.013 (3.48)**Age husband-wife 0.018 (2.01)* 0.0084 (1.41) 0.0010 (0.15) House price index 0.235 (2.29)* Mother employee 0.102 (2.16)* Father employee 0.131 (2.03)* Constant -1.473 (4.50)** 1.590 (6.06)** 2.739 (9.18)**Further variables: 2χ p-value 2χ p-value 2χ p-value Husband’s job characteristics (11)
1−t40.21
0.000 6.20
0.860
105.77
0.000
Wife’s job characteristics (11)
1−t12.93
0.298 123.28
0.000
23.90
0.013
Year indicators (11) 9.61 0.566 23.61 0.015 22.50 0.021 ρ 0.123 (2.59)** Log likelihood -831.49 -3422.59 Observations 8,753 8,753 Notes: * significant at 5%, ** significant at 1%. Absolute value of heteroskedasticity-robust t statistics in parentheses. All regressors in migration equation measured at time . Regressors in participation equations measured at time t except job characteristics which are measured at
1−t1−t . Reference categories for indicator
variables are non-migrant, cohabiting, renter, no high education wife/husband, mother/father self-employed or not working. Job characteristics at refer to 11 variables included in wage change equations (table 4) except new job indicator and
1−t
s'λ . 11 year indicator variables included.
26
Table 4: Selectivity-corrected wage change equations, BHPS 1991-2003
Non-migrants Migrants
Husbands Wives Husbands Wives
Constant -0.016 (0.82) 0.013 (0.53) 0.323 (1.90) -0.192 (0.69)Professional
0.065 (3.93)** 0.095 (3.09)** -0.153 (1.15) -0.130 (0.96)Manager/technician 0.056 (3.74)** 0.113 (5.49)** 0.0072 (0.06) -0.056 (0.46)Skilled non-manual 0.040 (2.62)** 0.062 (2.91)** 0.079 (0.50) 0.033 (0.30) Skilled manual 0.019 (1.79) 0.038 (2.27)* -0.083 (0.83) -0.214 (1.03) Working hours -0.014 (17.14)** -0.011 (10.87)** -0.017 (4.51)** -0.013 (3.14)**Paid overtime hours -0.018 (22.14)** -0.037 (24.81)** -0.012 (2.18)* -0.042 (4.85)**Job tenure 0.055 (3.20)** 0.064 (3.17)** -0.037 (0.50) 0.162 (1.17) Fixed-term contract -0.026 (1.05) -0.0047 (0.25) -0.234 (2.47)* -0.233 (1.79)Private sector -0.014 (0.62) -0.0035 (0.20) -0.042 (0.73) -0.0043 (0.04)Occupational pension
0.059 (4.75)** 0.053 (3.82)** -0.036 (0.42) 0.073 (1.17)
Union member 0.032 (2.74)** 0.042 (3.12)** -0.076 (1.26) 0.033 (0.40) New job 0.064 (5.52)** 0.074 (5.49)** 0.045 (0.87) 0.125 (1.70)
Mλ 0.241 (6.21)** 0.137 (3.02)** 0.043 (1.05) 0.006 (0.07)
wPλ -0.043 (2.68)** -0.016 (0.11)
hPλ 0.076
(2.59)** 0.014 (0.06) Observations 8,172 7,873 185 157
2R 0.23 0.25 0.32 0.37 Notes: * significant at 5%, ** significant at 1%. Absolute value of heteroskedasticity-robust t statistics in parentheses. Dependent variable is log of hourly gross wage in January 2000 prices. Independent variables except s'λ and new job indicator measured as change from 1−t to t. Reference categories are partly and unskilled occupation, permanent contract, and public sector. Year indicator variables included.
27
Table 5: Husbands’ and wives’ predicted wage change differentials and mobility
I: 0>∆+∆ wh ww II: 0≤∆+∆ wh ww migrants non-migrants migrants non-migrants
0 ,0 >∆>∆ wh ww 195 4,016 n/a n/a (0.48; 0.34) (0.16; 0.12)
0 ,0 ≤∆>∆ wh ww 1 2,377 0 768 (0.31;-001) (0.20; -0.06) (0.07, -0.13)
0 ,0 >∆≤∆ wh ww 0 606 0 358 (-0.04; 0.16) (-0.08; 0.03)
0 ,0 ≤∆≤∆ wh ww n/a n/a 0 432 (-0.07; -0.06) Notes: are the estimated hourly wage change differentials of migrating vs. not migrating of husbands and wives respectively in January 2000 pounds. The table displays the number of couples in each category. In parentheses are the average wage change differentials for husbands (left) and wives (right).
wh ww ∆∆ ,
Table 6: Structural migration probit, BHPS 1991-2003
(1) (2) Wage change differential wife 4.889 ** 14.281 ** Wage change differential husband 7.108 ** 16.945 ** Married -0.003 -0.468 New baby -0.490 -0.754 Number children < 5 -0.214 -0.633 * Number children ≥5 -0.275 ** -0.762 ** Home owner outright -0.997 ** -3.513 ** Home owner mortgage -0.507 * -1.817 ** High education wife 0.163 -0.056 High education husband 0.174 0.397 Age husband -0.035 ** -0.112 ** Age (husband-wife) 0.045 ** 0.103 ** House price index 0.422 * 0.528 Non-labour household income 0.063 0.233 ** Constant -2.369 ** -3.853 ** Log likelihood -239.53 -35.30 Observations 8,753 8,753 Notes: * significant at 5%, ** significant at 1%. Model (1) measures wage change differentials in pounds, model (2) in natural logs of pounds. All variables except wage change differentials measured at time 1−t . Bootstrapping (1000 replications, sampling couples) was used to derive confidence intervals. Significance of the effects was inferred if the bias-corrected confidence interval failed to include zero.
28