Annex I. Blanchard and Katz (1992) Methodology
This annex describes the indirect approach to quantify regional labor mobility in response to a regional labor demand shock proposed by Blanchard and Katz (1992).
The approach is based on the following identity:
where E indicates employment, LF is the labor force (the sum of employed and unemployed), POP is the working age population (the sum of the labor force and out of labor force working age population), E/LF is the employment rate, and LF/POP is the labor force participation rate.
This identity can be used to predict evolution of a regional labor market following the labor demand shock and decompose the response into changes in regional employment and participation rates and changes in the working age population. As most of the changes in the working age population are due to in- and out-migration rather than demographics, Δln(POP) could be interpreted as regional labor mobility.
Blanchard and Katz (1992) draw on the above identity and estimate the following VAR:
where Δe = Δln(E) is the employment growth, le = ln(E/LF) is the log employment rate, lp = ln(LF/POP) is the log participation rate, the parameter L is the lag operator, and ∊’s are the stochastic error terms. It is assumed that Δe, le, and lp are stationary I(0) variables. National means are differenced out from each variable before the estimation to isolate region-specific shocks from national shocks.
The estimates obtained are used to trace the fluctuations of Δe, le, and lp in response to a regional labor demand shock ∊e. The intuition is as follows (see picture). A labor demand shock has only a temporary effect on employment growth, employment and participation rates, but a permanent effect on the level of employment. Therefore, a positive labor demand shock will permanently increase the level of employment. Given the above identity, higher employment could only be achieved due to higher employment rates (or lower unemployment rates), higher labor force participation rates, and/or increase in the working age population due to in-migration. Since employment and labor force participation rates are stationary, they return to their long-run equilibrium in response to the labor demand shock. Therefore, in-migration by construction explains any permanent change in the level of employment following the positive shock.
Annex II. The Gravity Model
The empirical specification takes the following form:
where i denotes the origin region, j denotes the destination region, t denotes time, Mijt is the number of people that have moved from region i to region j in period t, Dij is the geographical distance between origin and destination regions, Xi are the factors in the origin region affecting labor mobility, Xj are the factors in the destination region affecting labor mobility, di are the origin region fixed effects, dj are the destination region fixed effects, dt are the time fixed effects, and ∊ is the residual. All variables are expressed in logs, so that the coefficients can be interpreted as elasticities1
Various variables were used in the literature as determinants of labor mobility. These variables include:
Distance. Larger distances across regional pairs are expected to reduce labor mobility as larger distance implies higher moving costs.
Population. Out-migration is expected to be larger from more populous regions. Also, more populous regions are expected to attract more in-migrants. These are the gravitational or demographic forces that are equivalent of GDP widely used in the trade literature.
GDP per capita. Out-migration is expected to be larger from poorer regions. Conversely, richer regions are expected to be an attractive destination for in-migrants.
Unemployment rate. Out-migration is expected to be larger from regions with higher unemployment rate. Conversely, regions with lower unemployment rates are expected to be an attractive destination for in-migrants.
Real wages. Out-migration is expected to be larger from regions with relatively lower wages. Conversely, regions with higher wages are expected to attract more in-migrants.
House prices. Relative house prices reflect cost of living differentials between regions (Bover and others 1989). Labor is expected to move out from regions with more expensive housing to regions with more affordable housing, ceteris paribus.
Arpaia, A., Kiss, A., Palvolgyi, B., and A. Turrini, 2018, “The Effects of European Integration and the Business Cycle on Migration Flows: A Gravity Analysis,” Review of World Economics (forthcoming).
Beine, M., Bourgeon, P., and J-C. Bricongne, 2018, “Aggregate Fluctuations and International Migration,” The Scandinavian Journal of Economics (forthcoming).
Bockerman, P., Kosonen, T., Maczulskij, T., and H. Keranen, 2017, “Job Market (In)Flexibility,” Labour Institute for Economic Research Report 34.
Bover, O., Muellbauer, J., and A. Murphy, 1989, “Housing, Wages, and U.K. Labour Markets,” Oxford Bulletin of Economics and Statistics, 51(2): pp. 97–136.
Böckerman, P., T. Kosonen, T. Maczulskij, and H. Keränen, 2017, “Job Market (In)Flexibility,” Report 34, Labor Institute for Economic Research (Helsinki: Finland).
Choy, W. K., D. Maré and P. Mawson, 2002, “Modelling Regional Labour Market Adjustment in New Zealand,” New Zealand Treasury Working Paper No. 02–01.
Debelle, G. and J. Vickery, 1999, “Labour Market Adjustment: Evidence on Interstate Labour Market Mobility,” Australian Economic Review, 32(3): pp. 249–63.
International Monetary Fund (IMF), 2014, “Housing Recoveries: Cluster Report on Denmark, Ireland, Kingdom of the Netherlands—The Netherlands, and Spain,” IMF Country Report No. 15/1.
International Monetary Fund (IMF), 2015, “Filling the Gap: Labor Market Reforms to Promote Jobs and Growth,” IMF Country Report No. 15/312.
Greenaway-McGrevy, R. and K. Hood, 2016, “Worker Migration or Job Creation? Persistent Shocks and Regional Recoveries,” Journal of Urban Economics, 96: pp. 1–16.
Greenwood, M. J. 1997, “Internal Migration in Developed Countries,” In Handbook of Population and Family Economics (M. R. Rosenzweig and O. Stark, eds.): pp. 647–720 (Amsterdam: Elsevier Science).
Greenwood, M. J., 2005, “Modeling Migration,” In K. Kempf-Leonard (Ed.), Encyclopedia of social measurement, Vol. 2: pp. 725–34 (New York: Elsevier).
Maczulskij, T., Böckerman, P., and T. Kosonen, 2018, “Job Displacement, Inter-Regional Mobility and Long-Term Earnings,” Labour Institute for Economic Research Working Paper 323.
Molloy, R., C. Smith, and A. Wozniak (2011), “Internal Migration in the United States,” Journal of Economic Perspectives, 3: pp. 173–96.
Moretti, E., 2011, “Local Labor Markets,” in Handbook of Labor Economics, Vol. 4b, David Card, and Orley Ashenfelter, eds. (New York: Elsevier).
Marques, H., 2010, “Migration Creation and Diversion in the European Union: Is Central and Eastern Europe a ‘Natural’ Member of the Single Market for Labour?” Journal of Common Market Studies, 48(2): pp. 265–91.
Pytliková, M., 2014, “The Effect of EU Enlargements and Labour Market Openings on Migration,” (Working Paper). Prague: CERGE-EI.
Vainiomäki, J., 2016, The development of wage dispersion and wage rigidity in Finland: The Report to the Economic Policy Council, December 2016.
Visser, J., 2016, “What Happened to Collective Bargaining during the Great Recession?” IZA Journal of Labor Policy, 5(1): pp. 1–35.
In addition to the IMF’s Finland team, I would like to thank Romain Duval, Christian Henn, Netta Hiitola, Tuomas Kosonen, Annaliina Kotilainen, Davide Malacrino, Jaakko Pehkonen, Matti Saari, and seminar participants at the IMF’s European Department and Bank of Finland for helpful comments and suggestions. Ryan Greenaway-McGrevy kindly shared the data for the U.S. Lubai Yang provided excellent research assistance.
In Finland, minimum wages are sector-specific and are determined by collective agreements. They also depend on work experience, individual characteristics, job complexity, and region.
The NUTS (Nomenclature of Territorial Units for Statistics) classification has been developed by Eurostat. NUTS 2 regions are the largest in size, while NUTS 4 regions are the smallest.
A similar positive association emerges when using other time periods.
While comparison with other countries is difficult in the absence of standardized cross-country datasets, the mobility is low when comparing to the U.S., where about 9 percent of the population moved between states and 18.6 percent of population moved between counties in 2000s (Molloy and others 2011).
The Blanchard and Katz (1992) methodology relies on the assumption that the labor force participation rate is a stationary variable. We tested the stationarity of the labor force participation rate variable using the Im-Pesaran-Shin panel unit root test. The null hypothesis of a unit root in all series is strongly rejected (p-value = 0.0007).
The decision to migrate may be based on observed changes in dependent variables, so we checked the robustness to the lag structure of the dependent variables — the results are qualitatively similar.
Using the square of distance produces similar results, but the coefficient of the distance variable is twice smaller given the logarithmic transformation.
The correlation between total population and working age population mobility rates is very high and close to 1 for the total sample and for 5-year rolling window subsamples. The scatterplot between the two also suggests a very close association. These results are available upon request.
Given that in some pairs of regions there may be 0 labor mobility in some time periods, a small positive number is added to Mijt to allow for a logarithmic transformation. There are about 200 observations with no labor mobility in our sample and dropping them out of the sample does not have qualitative effect on the results.