The employment impact of Poland’s special economic zones policy

ABSTRACT

In 1994, Poland launched a special economic zones experiment to maintain employment structures outside the major cities. Using difference-in-difference estimations, this paper evaluates whether the policy has been successful in its primary objective, which is to sustain employment in gminas (communes) more likely to be negatively affected by the economic transition. A significant and positive effect of the policy on employment is documented, but with some negative spillovers at the policy level as zones start to spread rapidly. Accounting for covariates, timing, policy spillovers and reducing imbalance through matching, the average treatment effect for employment is estimated to be 60%.

KEYWORDS:

• policy impact evaluation
• regional economic development
• special economic zones
• economic transition

JEL:

• C33
• L53
• P25
• R58

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed at https://doi.org10.1080/00343404.2017.1360477.

ORCiD

Camilla Jensen http://orcid.org/0000-0002-4367-2638

Notes

1. EUROSTAT’s Nomenclature of Territorial Units for Statistics (NUTS) classification has three levels regulated by the European Union. For example, at the NUTS-3 level the population size of an economic area is between 150,000 and 800,000. However, the LAU-1 and LAU-2 (formerly NUTS-4 and NUTS-5) are regulated individually by each country. In Poland, the LAU-2 level includes gminas varying in size from 300 to 43,000 permanent residents.

2. The employment rate reported here greatly differs from official employment statistics based on participation rates measured using labour force surveys. The official employment rate is typically much higher as a large part of the working-age population does not register as active in terms of participating in the labour force. This difference is not captured with the regional statistics available from the Polish local databank.

3. The exact effect for a semi-log regression when the dummy switches from 0 to 1 is 100(exp(0.0378) – 1) = 3.9%. The difference between the coefficient estimate and the exact effect is greater the greater the estimated coefficient (see also Giles, 2011).