Full text of DOI:10.1080/000368400425116

This article was downloaded by: [University of Otago]
On: 21 July 2015, At: 02:30
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick
Place, London, SW1P 1WG
Applied Economics
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/raec20
Geographic labour mobility and the cost of housing:
evidence from Italy
Luigi Cannari , Francesco Nucci & Paolo Sestito
Published online: 04 Oct 2010.
To cite this article: Luigi Cannari , Francesco Nucci & Paolo Sestito (2000) Geographic labour mobility and the cost of
housing: evidence from Italy, Applied Economics, 32:14, 1899-1906, DOI: 10.1080/000368400425116
To link to this article: http://dx.doi.org/10.1080/000368400425116
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose
of the Content. Any opinions and views expressed in this publication are the opinions and views of the
authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not
be relied upon and should be independently verified with primary sources of information. Taylor and Francis
shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and
other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation
to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

±
Applied Economics, 2000, 32, 1899 1906
Geographic labour mobility and the cost of
housing: evidence from Italy
LUIGI CA NNA R I, FR A NCESCO NUCCI * and PA OLO SESTITO
{
{
Bank of Italy, Research Department, Rome, University `L a Sapienza’, Faculty of
{
Statistics, Box 83, Roma 62, 00185, Rome, Italy and Ministry of L abour and Bank of
Italy, Rome, Italy
In recent years, internal migration in Italy has declined markedly, notwithstanding
±
the widening of the North South gap in terms of unemployment rates and real
income. Here, the extent to which the housing market has contributed to the decline
is examined. Preliminary to this analysis, di€ erentials in the cost of housing between
the macro-areas of the country are estimated using data on the market price of
houses located in 96 provincial capitals. Econometric evidence is provided support-
±
ing the view that the North South housing price di€ erential is a notable factor in
explaining the falling pattern of mobility. The positive impact on migration from the
South to the North of a wider gap in the two areas in terms of income and employ-
ment prospects has been o€ set by the housing price di€ erential, which has steadily
risen at least from the mid-1980s onwards. Yet, a considerable share of the decrease
in mobility remains unexplained, possibly owing to the heterogeneity in the compo-
sition of migration ¯ ows across di€ erent cohorts.
I. INTR ODUCTION
distance movements, i.e. those within an urban area.
Indeed, choosing both a house and a residential area to
match changes in life cycle and household composition is
a driving factor commonly taken into consideration (Hu€
and Clark, 1978; Onaka and Clark, 1983). On the contrary,
long-distance movements are generally linked to factors
related to the labour market and, speci®cally, to the
employment and income prospects o€ ered by di€ erent geo-
graphic areas (Blanchard and Katz, 1992; McCormick,
1997).
Less attention is paid to factors related to the housing
market when long-distance movements are examined
(Clark and Van Lierop, 1986).2 Yet, we argue that the
cost of housing is likely to represent an important disin-
centive to move and to a considerable extent it accounts for
the puzzling evidence of falling mobility levels in Italy.
Italian migration ¯ ows from the backward regions of
South to the North were large and persistent during the
1950s and 1960s, but they subsequently declined. Yet, there
was no parallel improvement in the secular mismatch
between the two areas in terms of real per capita income
and employment opportunities. On the contrary, these
±
North South di€ erentials have widened over the past dec-
ades, posing an empirical puzzle.1 The aim of this paper is
to analyse the role played by the housing market and, in
particular, to determine the extent to which increasing
housing price di€ erentials reduced geographic labour
mobility.
In the literature on residential mobility the role of hous-
ing is usually emphasized as a determinant of short-
*Corresponding author. E-mail: nucci@uniroma1.it
1 To account for this surprising trend Attanasio and Padoa Schioppa (1991) suggest that the `informal network’ provided by the family in
the South and the support of Government make the condition of being unemployed more a€ ordable, thus reducing the incentive to
move. Faini et al. (1997) argue that ine ciencies in the job matching process among di€ erent regions and high migration costs are key
determinants of the fall in mobility.
1899
#
±
±
Applied Economics ISSN 0003 6846 print/ISSN 1466 4283 online
2000 Taylor & Francis Ltd
http://www.tandf.co.uk/journals

1900
L . Cannari et al.
Therefore, in a standard framework where migration ¯ ows
between the South and the North depend on the di€ erent
prospects faced in terms of real permanent income and
employment, we investigate whether housing price di€ er-
entials have incremental predictive content in explaining
the falling pattern of geographic labour movements. The
underlying idea is that the price indices commonly utilized
for de¯ ating regional permanent income do not take
proper account of the housing costs (Cameron and
Muellbauer, 1998). Therefore, in order to avoid that a sub-
stantial cost of living e€ ect is omitted, house price di€ er-
entials should be explicitly considered.
in the backward regions of the South declined from 0.3%
in the 1960s to 0.19% in the early 1990s. The decline in
overall mobility was especially marked in the case of long-
±
distance movements. Net North Centre in¯ ows were
always positive but decreased signi®cantly over time, not-
withstanding a slight increase at the end of the 1980s
(Fig. 1).
In investigating how the housing market may a€ ect
long-distance mobility and, thus, partly explain these
facts, three di€ erent aspects ought to be considered. The
®rst is that the di€ erent user cost of house between two
areas `de¯ ate’ the permanent income prospects that a
household faces in the two areas, a€ ecting its decision to
move. An area hit by a negative (positive) shock may
not register a signi®cant out¯ ow (in¯ ow) of population
simply because the relative consumer wage, even if adjusted
for the decline (rise) in employment opportunities, has
not fallen (risen) enough when taking account of the
increase in the cost of housing in the area considered as a
possible destination. Second, in the presence of rationing
and, more generally, of imperfections in the housing
market, di culties of ®nding accommodations in the
destination place discourage mobility. Finally, the
propensity to move may be lower for the homeowners,
who have to liquidate their housing assets in a given
locality to buy a new ¯ at elsewhere, thus facing sizeable
transaction costs.
Indeed, in a properly-working housing market, the ®rst
aspect we mentioned, the di€ erentials in the user cost of
houses, should entirely re¯ ect di€ erences in rents.
However, two empirical problems are faced in the analysis.
First, the existence of housing shortages suggests that the
appropriate di€ erence in rent should not be computed
merely using rents paid by lessees, but should be adjusted
for the probability of ®nding an accommodation to rent at
any speci®c contract. In addition, reliable longitudinal data
on rents, comparable with those on price of houses, are not
available.4 Thus, it was decided to rely upon data on mar-
ket prices of dwellings and a further element lending sup-
port to our choice is the large and increasing proportion of
homeowners, which has reduced the importance of the ren-
tal market.
Preliminary to this analysis, time-varying measures of
housing price di€ erentials between the two macro-areas
of the country are constructed. These measures are
±
obtained for a large time span (1965 1995) by means of a
simple statistical model based on data on the market price
of houses located in the 96 Italian provincial capitals. As a
sensitivity inspection of the results, the estimated values are
then compared, for the year 1993 only, with those obtained
by applying the same method to data on housing prices
from the Bank of Italy’s Survey of Household Income and
W ealth (see Brandolini and Cannari, 1994 for a description
of these data). The information at the individual level
drawn from the Bank of Italy’s survey is very detailed
and it allows us to control for a large number of factors
which might be relevant for the market value of a house.
The paper is organized as follows: in Section II, the main
facts of geographic mobility in Italy are recalled and the
role of housing as an explanatory factor is discussed;
Section III illustrates a simple procedure for estimating
housing price di€ erentials; Section IV presents some econo-
metric evidence on the relationship between the cost of
housing and labour movements; Section V concludes.
II. BA CKGR OUND
Over the last 20 years internal migration in Italy has
declined markedly. In the early 1990s, annual changes of
residence from one town to another were less than 0.2% of
the total population, compared with an average of about
0.32% in the 1960s.3 In particular, movements originating
2 There are exceptions, however, which seek to integrate both aspects of the analysis. In Hughes and McCormick (1985), the implications
of housing policy on UK internal migration are examined. In Bover et al. (1989) the importance of regional housing price di€ erentials for
labour mobility in the UK is emphasized. Focusing on regional migration in Spain, Antolin and Bover (1997) examine house price
di€ erentials as an explanatory factor of mobility choices beyond personal characteristics, unemployment and wage. In addition, Jackman
and Savouri (1992) provide evidence for the impact of relative regional house prices on inter-regional migration.
3 These data are from the General Population Register (GPR). A shortcoming of this source is that there may be a time lag between
actual population movements and the GPR recording. Indeed, it may happen that some migration ¯ ows are not recorded in the GPR in
the year they take place but only later, when Census data are made available and the GPR is revised accordingly.
4 Data on consumer price indices, as released by the Italian Statistical Institute (ISTAT) for each provincial capital (capuluogo di
provincia), include rent disbursements in their calculation. However, it should be noted that rent contracts signed according to the
rent-control law tend to be largely over-represented in the sample with respect to contracts agreed upon in the free market. Hence,
housing costs are not properly incorporated in these price indices, thus omitting an important cost of living e€ ect.

L abour mobility and housing cost
1901
Fig. 1. Ratio of net North-Centre in¯ ows to population (rates per 1000)
III. MEA SUR ING GEOGR A PHIC
DIFFER ENTIA LS IN THE COST OF
HOUSING
(CI) reports data on dwellings located in provincial capi-
tals, with a statistical breakdown by location within the
urban area (town centre, between outskirts and town cen-
tre, outskirts). Data from this source are available for a
much longer period: from 1965 to the present.
The pattern of geographic price di€ erentials over time is
estimated using CI data, while SHIW data allow us to
verify, with respect to 1993, if the use of a large set of
explanatory variables related to the characteristics of
dwellings, results in geographic price di€ erentials which
are roughly as large as those estimated with CI data. If
this is the case, the estimates of geographic di€ erentials
are not a€ ected notably by di€ erences across regions in
the quality of dwellings.
An appropriate measure of regional di€ erences in housing
prices should re¯ ect prices of dwellings more than their
value. In other words, when house prices are compared
across geographic areas, di€ erences in value due to the
di€ erent quality of the houses ought to be set out. In
order to control for (and remove) the e€ ect of these quali-
tative factors, a statistical model is used whose simple
features will be illustrated after the description of the
data used in the analysis.
To measure housing price di€ erentials across geographic
areas, two di€ erent data sources are relied on: the Review
Il Consulente Immobiliare (henceforth, CI), which reports
average half-yearly market quotations of dwellings located
in the 96 provincial capitals and the Bank of Italy’s Survey
on Household Income and W ealth (henceforth, SHIW),
which collects data on the prices of dwellings according
to their tenants and homeowners’ subjective evaluation.
The latter has the merit of being representative of the
whole universe of Italian dwellings (owing to the nature of
the survey sample). In addition, it allows account to be
taken of many qualitative characteristics of dwellings.
However, most of this information was collected only by
the 1993 SHIW and, to a lesser extent, by the surveys
±
In measuring the North South house price di€ erentials a
statistical model is resorted to, to aggregate, at the level of
three macro-areas of Italy, information available for the 96
provincial capitals. This allows one to overcome some of
the problems that are faced when dealing with incomplete
longitudinal data. In particular, the CI data set does
not have complete time-series observations on all cross-
sectional units, as quotations of dwellings for some of the
towns included in the sample are sometimes missing within
the time periods, especially in the initial years. As a conse-
quence, if one calculates, for each period, a simple weighted
average of housing prices of a certain geographic area, a
serious bias is likely to occur when these aggregate
values are compared over time, because they are average
±
carried out in the period 1986 1991. The other source

1902
L . Cannari et al.
values related to nonhomogeneous groups. On the other
hand, comparing the subsample of towns for which house
prices are always recorded and constructing an index on
the base of them only may result in a considerable loss of
information. Conversely, the simple statistical model
resorted to allows control for possible over time heteroge-
neity of the towns representing a geographic area. Indeed,
even if the sample of towns making up an area may change
from one period to another, however, the housing price
component attributable to each geographic area can be
identi®ed net of the e€ ects determined by control factors
that vary over time because of changes in the sample struc-
ture. In particular, the statistical model is the following:
account to be taken of di€ erences in the quality of dwell-
ings. Here it su ces to say that results based on SHIW
data are very much comparable to ours.
IV. THE EFFECTS OF HOUSE PR ICES ON
LA BOUR MOVEMENTS
In this section, we evaluate the role of house price di€ er-
entials in discouraging geographic mobility. The analysis is
conducted on the basis of the mobility ¯ ows between dif-
ferent towns measured by the GPR (see Section II). Our
analysis mainly focuses on labour movements from the
South to the North-Centre. Later on, we extend our
focus by examining mobility within the North-Centre.
Variables related to the housing market are included
among the explanatory variables in a standard model
of geographic labour mobility, where migration is
ˆ
log Pt at ‡ btX ‡ ut
1
… †
t
where Pt is the price of a dwelling at time t (expressed as
thousands of liras per square metre) and X is a vector of
t
explanatory variables (mainly modelled as dummies),
related to: (a) geographic location; the country is divided
into four areas: North-West, North-East, Centre and
South; (b) location of dwellings within the urban area:
town centre, between outskirts and town centre and out-
skirts; (c) municipality size (in terms of population) and its
interaction with the location of dwellings in the town
centre.
The term at is a time-varying intercept and the error term
ut denotes the idiosyncratic component of the price of a
house.5 The statistical analysis has been conducted on
half-yearly data (spring and fall quotations) for new dwel-
lings located in 96 provincial capitals over the period
±
determined by the di€ erent chances in terms of income
±
and opportunities of employment characterizing the two
areas.
Unemployment rates have been used as a proxy for the
probability of obtaining a job. As a measure of the di€ erent
chances in terms of income, the di€ erences between areas in
per capita real private consumption expenditure have been
used. This indicator takes account not only of
discrepancies in wages and salaries but also of di€ erences
in other income sources and, in particular, transfer
payments. In addition, consumption expenditure, di€ er-
ently from disposable income, which also includes
transfers, captures the e€ ect of future income prospects
and as such it is close to permanent income. Di€ erentials
in per capita consumption expenditure between the
North and the South are expressed at constant prices
(base: 1985). Hence, the geographic discrepancies in prices
and, more in general, in the cost of living are taken into
consideration.
±
1965 1995. Model (1) was estimated separately on 57
cross-sections, one for each half-year from 1965:2 to the
end of 1995.6
The coe cients of the dummies for North-East, North-
West and Centre (each measuring the relative di€ erences in
housing prices with respect to the South) are almost always
positive and they increased progressively from the mid-
1980s to 1995.7
As mentioned above, these results, obtained throughout
a long time horizon, have been compared, for 1993 only,
with estimates of a model where the set of explanatory
variables is widened. We used 1993 SHIW data that allows
As to mobility within the North-Centre, the analysis
has been carried out using estimates of the degree of
5 House price di€ erentials, as measured in Equation 1 by the estimated coe cients of geographical dummies, are margin e€ ects, as they
single out the component of price di€ erences due to the geographical areas. These di€ erentials measure the cost of moving from one area
to another, holding constant municipality size and location of dwelling. In other words, what is exactly measured is not the generic higher
cost of housing that are faced when one moves from a town in the South to one in the North. Rather, the estimated coe cients of the
geographical dummies measure the price di€ erential between a house in the South, say in the town centre, and one in the North that is
also located in the centre of the municipality and has a similar population size.
6 Data for four half-years are not available (1966:1, 1975:2, 1980:1, 1981:2), hence estimation is conducted for 57 periods. We also
consider the subset of the sample made up of towns participating in the survey throughout the period (homogeneous sample): on average,
the results do not show a marked change. However, owing to the small size of the homogeneous sample, a higher instability characterizes
these estimates, thus lending support to the adoption of the procedure against the alternative chain-indexing techniques.
7 In 1985, for instance, a new house located in the North-West costs 15% more on average than one located in the South and one located
in the North-East 6% more. The percentage gap widened thereafter: in 1995, it was 35% in the North-West and 31% in the North-East.

L abour mobility and housing cost
1903
±
±
Fig. 2. North South housing price di€ erentials* (Italy: 1965 1995)
*The relative di€ erences between housing prices in the North-Centre and those in the South are yearly average of the semiannual
measures obtained from the estimations documented in Section III.
geographic variability in the relevant explanatory
variables.8
of homeowners is included among the explanatory vari-
ables, on the grounds that home-ownership may restrain
geographic transfers; unfortunately, satisfactory measures
of housing rationing are not available.
In the empirical speci®cation the dependent variable is
the ratio of transfers from the South to the North-Centre
(MSICN) to the total Southern population (PTOTSI). This
indicator can be viewed as the probability of moving from
the South to the North-Centre (see Bodo and Sestito,
1991) :10
However, the key feature of the investigation is examin-
ing the role of the housing market. The aim is to assess the
extent to which the widening of the gap in the price of
housing has restrained mobility. The relative di€ erences
in house prices between the North and the South measured
in the previous section are used.9 The time pro®le of the
geographical di€ erential in housing prices is presented in
Fig. 1. Direct estimates of transaction costs in the housing
market (for instance taxes on purchases and sales, legal
costs, registration dues) are not available. As a proxy for
these costs, which tend to restrain mobility, the general
index of housing prices at the national level is used,
de¯ ated by the cost-of-living index. In addition, the share
MSICN
PTOTSI
ˆ
MOB1
2
… †
The ®rst speci®cation (Table 1, column 1) does not take
account of the housing market. Coe cients have the
8 Clearly, these measures of variance of the variables will be referred to the North-Centre only. The argument is that large geographical
variability in both income and employment rates, no matter what its exact direction is, is expected to foster population localization in
other towns. Empirically, only the variance of income is used, because, with regard to unemployment, the available longitudinal data are
not homogeneous over time.
9
±
More precisely, a unique measure of the South North house price di€ erential is used for each period by calculating a weighted average
±
±
of the coe cients for North East, North West and Central Italy estimated in Equation 1. The weights re¯ ect population size in the
areas.
10 MOB1 belongs to the (0, 1) interval; thus, we use a logit model is used:
exp …- t0Zt†
ˆ
MOB1t
…3a†
0
‡
1
exp…- t Zt†
where Zt is a vector of explanatory variables and - are coe cients to be estimated. The model is estimated on the transformed equation:
MOB1t
1 ¡ MIB1t
0
ˆ
log
- t Zt
…3b†
The possible endogeneity of some of the regressors with respect to migration ¯ ows (house prices themselves may be sensitive to an
increased demand for existing houses due to in-migration) is tackled by considering labour movements at time t as a function of variables
lagged one period.

1904
L . Cannari et al.
Table 1. Estimated probability of migration from the South to the
North-Centre (logit models 1967 1992)
to the North, however, is `explained’ by a time trend
term.12
±
±
Housing price di€ erentials are included in the speci®ca-
tion of column 2. Their e€ ect has the expected sign and is
largely signi®cant (71.07; t-statistic 74.6), con®rming the
view that the housing market is a relevant explanatory
factor. Admittedly, the trend term continues to be signi®-
cant, suggesting that the decline in migration from the
South to the North is not completely explained by the
variables so far considered. In order to control for some
of these unobservable factors, we have tried to verify if the
proxies for the transaction costs of mobility mentioned
above contribute to explain the decline in mobility. In col-
umn 3 of Table 1 the general index of house prices and the
share of homeowners are included among the explanatory
variables. Yet, the coe cient on the share of homeowners
is not statistically signi®cant (and it has the wrong sign),
while the coe cient on the aggregate index of house prices
is negative but not signi®cant.
Variable
Constant
(1)
(2)
(3)
7
4.565
7
7
7
4.719
4.435
4.969)
0.280
(2.927)
4.232
(5.155)
7
7
(
11.248) ( 10.14)
(
=
‰
Š
log MOB1 …1 ¡ MOB1† t¡1
0.301
(3.076)
4.692
0.272
(2.963)
5.30
=
log…CN CS†t¡1
(4.699)
(5.73)
=
7
7
7
7
log…URCN URS†t¡1
(HPD)t¡1
0.413
2.905)
±
0.351
3.059)
1.07
4.577)
0.032
4.964)
±
0.338
2.924)
0.763
2.663)
0.0358
3.954)
0.009
7
7
7
7
7
7
7
7
7
7
(
(
(
(
(
(
(
(
7
0.0313
7
4.025)
±
trend
…PROPSI†t¡1
log…PRC†t¡1
Dcensus
(0.948)
±
±
7
7
0.085
0.996)
0.248
(8.677)
0.98
0.05
0.54
2.26
0.835
1.049
2.144
(
0.263
(7.908)
0.97
0.083
0.065
1.577
0.954
1.192
1.336
0.258
(9.447)
0.98
0.055
0.054
2.11
0.254
0.321
1.195
-
R2
SSR
SE
DW
ML M(1)
Q…1†
Q…2†
A simple out-of-sample numerical exercise, carried out
±
for the period 1993 1995, allowed the relative importance
of the di€ erence in house prices, unemployment rates and
income to be quanti®ed. First the estimated value of the
long run equilibrium of MOB1 in the equation of column 2
in Table 1 was computed, using the e€ ective values of all
regressors. After this benchmark calculation, the value of
the long run equilibrium in the same equation was again
computed but using the mid-1970s value for the house price
di€ erential and the actual values for all the other regres-
sors. The exercise was replicated under the hypothesis that
income and unemployment di€ erentials had stayed con-
stant at mid-1970s levels. If the house price di€ erential
between the North and the South were equal to that of
twenty years earlier (11.5% compared with 25.9% in the
Notes: The dependent variable is the logistic transformation of
MOB1. The explanatory variables are: the lagged dependent;
ˆ
CS and CN household per capita consumption in the South
and in the North-Centre, respectively; URCN and
ˆ
URS unemployment rate in the North-Centre and the South,
ˆ
respectively; HPD di€ erential in the price of houses;
ˆ
PROPsi share of homeowners in the South-Islands;
ˆ
ˆ
PRC price index for houses; Dcens dummies to take into
account the changes in the General Population Register (years
1961, 1962, 1971 and 1972) following the Population Census.
White t statistics on coe cients are reported in brackets.
-
±
period 1993 1995), then migration ¯ ows from the South to
R2 adjusted R-square; SSR sum of squares of residuals;
ˆ
ˆ
the North would increase by about 24% (the share of the
Southern population migrating every year towards other
regions of the country would increase from 0.37% to
0.46%). If the di€ erences in unemployment rates and per
capita consumption expenditure were equal to those of 20
years earlier, then migration ¯ ows would decrease from
0.37% to 0.29% (to 0.32% if only unemployment rates
were unchanged). Of course, these results have to be
taken cautiously, because, among other things, they do
not consider the backward e€ ect of migration ¯ ows on
house prices and the other explanatory variables.
±
ˆ
ˆ
SE standard error of regression; DW Durbin Watson test;
ˆ
ML M…1† Lagrange multiplier test for autocorrelation of ®rst
±
order; Q…1† and Q…2†: Ljung Box test for autocorrelation of ®rst
and second order.
expected sign: the lagged dependent variable shows that
migration reacts slowly to its determinants and underlines
the importance of the so-called migratory chains.11 The
di€ erence in per capita consumption expenditure between
the North-Centre and the South has a positive e€ ect on
transfers from the South to the North. Conversely, the
di€ erence in the unemployment rate has the expected
(negative) sign. A large share of mobility from the South
As mentioned earlier, to have a more complete picture of
the Italian case, the analysis was extended to the mobility
within the North-Centre, which represented (in the 1990s)
about 60% of the total migration ¯ ows across municipali-
11 In migratory chains the presence of emigrants in a place considered as a possible destination favours mobility of potential emigrants.
12 To control for the update of the GPR when Census data are released (in particular for 1961 and 1971 Census), regressions include a
dummy variable. In those years, in fact, the revisions of GPR following Census records improperly enhance the size of mobility ¯ ows,
because a considerable share of them is likely to have occurred in previous years.

L abour mobility and housing cost
1905
ties (another 15% refers to the bidirectional ¯ ows between
South and North-Centre and only about 25% were repre-
sented by ¯ ows within the South). To scale the variable,
the ratio of migration ¯ ows within the North-Centre to
the total population living in this area (MOB2) is
considered.
V. CONCLUDING R EMA R KS
Notwithstanding necessary caution due to some unpleasant
features of the data used and the di culty in accounting
for all explanatory factors, the widening of the di€ erence in
the cost of housing seems to have restrained internal migra-
tion ¯ ows between the South and the North of Italy. This
has o€ set the e€ ect of the widening of other gaps between
the North and the South, such as those in the real income
and the employment rates. Focusing on migration from the
South to the North, it has been shown that an important
role has been played by the di€ erence in house prices,
which has increased signi®cantly since the mid-1980s and
was equal to about 30% in 1995. Moreover, some more
general hints were found on the relevance of the housing
market in determining the propensity to move. There is in
particular some, albeit weak, evidence that the rise in the
share of homeowners experimented by the Italian economy
may have contributed to lessen the propensity to move
within the more aƒ uent regions.
In this case, the area of departure and the area of desti-
nation are not de®ned. As no reliable measure for unem-
ployment dispersion was available, to account for incentive
factors to mobility, the study is limited to a measure of the
variability in per capita real income across provinces: a
higher variability should increase migration ¯ ows towards
higher-income areas.13 As far as the housing market is
concerned, the increase in the share of homeowners and
the increase in the aggregate index of house prices should
restrain mobility. The lagged dependent variable should be
related to the slowness of the migration process. In Table 2
estimation results are displayed for an equation where the
house price variable, which did not turn out to be signi®-
cant, was excluded. The share of homeowners has a nega-
tive sign; on the contrary, the variability in per capita
income has a positive e€ ect on mobility. While the latter
is highly signi®cant, the former is however only weakly
signi®cant.
A non-negligible share of the decrease in the South-to-
North migration ¯ ows, however, remains unexplained.
Indeed, thoroughly explaining the decline of mobility pre-
sents some di culties due to over time heterogeneity of the
migration ¯ ows of di€ erent cohorts.
In particular, a relevant aspect is the migrants’ composi-
tion in terms of education endowment. Generally speaking,
movers tend, on average, to be characterized by a higher
level of education (for Italy, see Sestito, 1991). Using
SHIW data, it was veri®ed that for movements from the
South to the North this feature is con®rmed empirically
only for the youngest and oldest cohorts (individuals
born either between 1956 and 1965 or before 1930). Yet,
for people born between 1931 and 1956 the evidence is
Table 2. Estimated probabilities of mobility within North-Centre
±
±
(logit models 1967 1992)
Variable
Constant
=
7
7
1.273
2.442)
0.646
(
(
‰
Š
log MOB2 …1 ¡ MOB2† t¡1
±
reversed. This shows the peculiarity of the large South
(5.138)
7
7
…PROP†t¡1
…¼red†t¡1
0.003
1.745)
8.027
North ¯ ows occurred during the 1950s and the 1960s
(mostly composed by individuals born between 1931 and
1956). Incidentally, it has to be noticed that while those
large migration ¯ ows tended to reduce the gap between
the two macro-areas in terms of the human capital
endowment, the subsequent and smaller labour movements
appear to have been of a selective type, thus widening,
rather than reducing, the gap in the educational
endowment.
(2.899)
0.082
(1.770)
0.90
Dcensus
-
R2
SSR
0.052
0.050
2.0
SE
DW
ML M…1†
Q…1†
Q…2†
0.015
0.011
1.694
Notes: The dependent variable is the logistic transformation of
A CKNOWLEDGEMENTS
mobility within the North-Centre. The variable ¼
stands for
the variance of (log) provincial per capita income.reOdther abbre-
viations are described in Table 2.
The authors are grateful to Steve Cameron, Richard
Clarida, Ned Phelps, Francisco Rivera-Batiz and Lucio
13 To calculate the geographical variance of income, we use longitudinal data on GDP at constant price at the provincial level are used
(source: Istituto Tagliacarne).

1906
L . Cannari et al.
Guiso and I. Visco, Cambridge University Press, Cambridge,
pp. 369 86.
Cameron, G. and Muellbauer, J. (1998) The housing market and
regional commuting and migration choices, CEPR
Discussion Paper, no. 1945, August.
Sarno for very useful comments and discussions. We also
thank participants at the 37th European Congress of the
Regional Science Association and at seminars at the Bank
of Italy and Luiss for helpful suggestions. Responsibility
for any error is entirely our own. The views expressed in
this paper do not necessarily re¯ ect those of the institutions
to which the authors belong.
±
Clark, W. A. V. and Van Lierop, W. F. J. (1986) Residential
mobility and household location modelling, in (Ed.)
P. Nijkamp, Handbook of Regional and Urban Economics,
±
North-Holland, Amsterdam, pp. 97 132.
Faini, R., Galli, G., Gennari, P. and Rossi, F. (1997) An empirical
puzzle: falling migration and growing unemployment di€ er-
entials among Italian regions, European Economic Review,
R EFER ENCES
Antolin, P. and Bover, O. (1997) Regional migration in Spain: the
e€ ect of personal characteristics and of unemployment, wage
and house price di€ erentials using pooled cross-sections,
Oxford Bulletin of Economics & Statistics, 59(2), 215 35.
Attanasio, O. and Padoa Schioppa, F. (1991) Regional inequal-
±
41, 571 79.
Hu€ , J. O. and Clark, W. A. V. (1978) Cumulative stress and
cumulative inertia. A behavioural model of the decision to
±
±
move, Environment and Planning, A10, 1101 19.
Hughes, G. and McCormick, B. (1985) Migration intentions in
±
ities, migration and mismatch in Italy, 1960 1986, in (Ed.)
Padoa Schioppa, Mismatch and L abour Mobility, Cambridge
the UK, Economic Journal, 95 (Conference papers supple-
±
ment), 113 23.
±
University Press, Cambridge, pp. 237 320.
Jackman, R. and Sadourik, S. (1992) Regional migration in
Britain: an analysis of gross ¯ ows using NHS central register
Blanchard, O. and Katz, L. F. (1992) Regional evolutions,
±
Brookings Papers on Economic Activity, 0(1), 1 75.
Bodo, G. and Sestito, P. (1991) L e vie dello Sviluppo, Il Mulino,
±
data, Economic Journal, 102, 1433 50.
McCormick, B. (1997) Regional unemployment and labour mobi-
Bologna.
±
lity in the UK, European Economic Review, 41, 581 9.
Bover, O., Muellbauer, J. and Murphy, A. (1989) Housing, wages
and UK labour markets, Oxford Bulletin of Economics and
Onaka, J. and Clark, W. A. V. (1983) A disaggregate model of
residential mobility and housing choice, Geographic Analysis,
±
Statistics, 51(2), 97 136.
±
1(15), 287 304.
Brandolini, A. and Cannari, L. (1994) Methodological appendix:
the Bank of Italy’s survey of household income and wealth,
in Saving and the Accumulation of W ealth (Eds) A. Ando, L.
Sestito, P. (1991) Sviluppo del mezzogiorno e capitale umano,
±
Economia e lavoro, 25(4), 3 13.