The Effect of Emigration from Poland on Polish Wages

• Author: Anna Rosso,Tommaso Frattini,Christian Dustmann
• Date: 01 April 2015
• Published date: 01 April 2015
• DOI: http://doi.org/10.1111/sjoe.12102

Scand. J. of Economics 117(2), 522–564, 2015

DOI: 10.1111/sjoe.12102

The Effect of Emigration from Poland on

Polish Wages∗

Christian Dustmann

University College London, London WC1H 0AX, UK

c.dustmann@ucl.ac.uk

Tommaso Frattini

University of Milan, 20122 Milan, Italy

tommaso.frattini@unimi.it

Anna Rosso

National Institute for Economic and Social Research, London SW1P 3HE, UK

a.rosso@niesr.ac.uk

Abstract

In this paper, we analyse the effect of emigration from Poland on Polish wages. Focusing

on the 1998–2007 period for Poland, we use a unique dataset that contains information

about household members who are currently living abroad, which allows us to develop

region-specific emigration rates and to estimate the effect of emigration on wages using

within-region variation. Our findings show that emigration led to a slight increase in wages

for high- and medium-skilled workers, which are the two groups with the largest relative

outmigration rates. Workers at the low end of the skill distribution might have experienced

wage decreases.

Keywords: Impact; migration; sending countries

JEL classification:J31; J61

I. Introduction

Since the late 1990s, Poland has experienced a dramatic increase in emi-

gration. Whereas in 1998, the share of emigrants in the overall population

was about 0.50 percent, by only a decade later, it had increased to 2.3 per-

cent.1Nevertheless, there is large regional variation in emigration rates,

with a 2007 share of emigrants that ranges between 1 and 5.6 percent

∗We gratefully acknowledge financial support from the Rockwool Foundation, and we are

grateful to Torben Tranaes for his support throughout the project and for many helpful

discussions. We also thank Iga Magda for her invaluable help with the data, and Bernd

Fitzenberger for helpful comments. C. Dustmann acknowledges support from the NORFACE

migration programme.

1See Table 1, based on the Polish Labour Force Survey.

CThe editors of The Scandinavian Journal of Economics 2015.

C. Dustmann, T. Frattini, and A. Rosso 523

across Poland’s 16 provinces (see Table 2). This decade also saw a change

in the composition of the emigration flow: emigrants became increasingly

younger and were better educated than non-emigrants. These large increases

in emigration, together with the variation in emigrant skill composition, are

likely to have had a notable impact on the Polish labour market and, in

particular, on the wages of those who stayed behind. It is this question that

we address in this paper.

Specifically, we investigate the impact of emigration on wages over a

period of 10 years (1998–2007) when emigration from Poland was at its

highest. Because our dataset includes rare detailed information on emigrants

and their education and age structure, it allows us to assign emigration rates

to local labour markets and to determine the emigration-induced changes in

skill ratios within local labour markets. We use the variation in emigration

rates within Poland’s regions to identify the effects of emigration on the

wages of non-emigrants.

Although our paper is related to the body of literature on the impact

of migration on wages, rather than concentrating, as most studies do, on

the impacts on wages in the countries of destination,2it is part of only

a small body of work that investigates the impact of emigration on the

labour markets of sending countries. One reason such studies are scarce

is the difficulty of obtaining information on emigrants, a problem that

Aydemir and Borjas (2007) and Mishra (2007) overcome by exploiting

the fact that over 95 percent of emigrants from Mexico go to the US.

After first measuring the size and composition of Mexican emigrants from

US censuses and wages in Mexico from Mexican censuses, these authors

follow the identification strategy proposed by Borjas (2003) and correlate

the wages of different skill groups in Mexico, defined in terms of age

and education, to the proportion of emigrants from the same skill group

in the US. Elsner (2013a) uses a similar approach to study Lithuanian

emigration, but he must rely on a number of simplifying assumptions

to reconstruct the size of Lithuanian emigration based on Irish and UK

data.3

2See, for instance, early work by Altonji and Card (1991), Card (2001), Angrist and Kugler

(2003), Borjas (2003), Dustmann et al. (2005), Card and Lewis (2007), Jaeger (2007),

and more recent papers by D’Amuri et al. (2010), Glitz (2012), Manacorda et al. (2012),

Ottaviano and Peri (2012), and Dustmann et al. (2013).

3Other papers on the labour market effects of emigration include: Hanson (2007), who

compares changes in labour market outcomes between 1990 and 2000 in Mexican states with

high and low historical levels of migration (measured in the 1950s); Docquier et al. (2014),

who use an aggregate production function model to simulate the effect of immigration and

emigration on wages and employment in OECD countries; and Elsner (2013b), who uses a

calibrated structural model of labour demand to simulate the effect of Lithuanian emigration

on the wages of non-emigrant workers.

CThe editors of The Scandinavian Journal of Economics 2015.

524 The effect of emigration from Poland on Polish wages

We contribute to this body of literature by focusing on one large Euro-

pean country, Poland, which, although locked away behind the Iron Curtain

for more than four decades, experienced a large amount of emigration from

the late 1990s onwards. Rather than identifying emigrants based on census

data and survey information from the destination countries, however (as

did the aforementioned studies), we have access to detailed information

(including age and education) on all emigrants measured in the source

country, which allows precise computation of the regional distribution of

emigrants in the country of origin. We are therefore able to estimate the

effect of emigration on resident wages using an identification strategy that

relies on regional variation in emigration, rather than on variation across

skill groups, as in previous work. This allows us to estimate the overall

effect of emigration on wages rather than its partial effects; see Ottaviano

and Peri (2012) for an insightful discussion. The availability in the dataset

of wage information for a subset of emigrants before they left the country

also helps us to address the possible change in the composition in the

non-emigrant population due to selective outmigration.

To better structure our empirical analysis and interpret our parameter es-

timates, we first present a model in which output is produced by combining

capital with a constant elasticity of substitution (CES) labour composite.

This model shows that wage effects are positive for skill groups in which

outmigration falls above a weighted overall average along the skill distri-

bution, and that – if capital is insufficiently mobile in the short run – the

overall wage effects can be expected to be positive. Our empirical results,

based on estimations using within-region variation, suggest that, overall,

emigration had a positive effect on the wages of those who did not em-

igrate. Across skill groups, it is those in the middle of the educational

distribution particularly that experienced the largest gains from emigration.

The effect on the highly educated is likewise positive, but smaller, while

the effect on the wages of those with a low level of education is negative,

albeit mostly not significantly different from zero. This result is thus in line

with emigration being more concentrated among individuals in the middle

and upper parts of the educational distribution.

Because emigration from a particular region might be induced by neg-

ative wage shocks, we argue that such estimates are likely to constitute a

lower bound on the effect of emigration on wages, and have therefore a

meaningful interpretation. However, as a robustness check, we also develop

an instrumental variable (IV) estimation strategy based on the detailed in-

formation we have available on the emigrants’ destination countries. We

combine these data with both the variations in economic conditions in the

main destination countries (Ireland, Germany, the UK, and the US) and the

large exchange rate fluctuations over this period, and we employ various

strategies that exploit regional differences in destination preferences.

CThe editors of The Scandinavian Journal of Economics 2015.