Abstract
The purpose of this article is to explain the strong diversification in the volume and structure of exports in Polish regions, using a set of potential determinants originating from different foreign trade theories used in country level studies. Two sets of panel models of exports are estimated for 16 regions of Poland in the years 1999–2008. Model I shows that regional export performance is positively dependent on labour productivity, share of foreign-owned companies in employment, education level of population, location in the country's border region and access to the sea, and negatively on the importance of agriculture in the region's economy and labour costs. Model II indicates that exports of agricultural and food products are positively correlated with the importance of agriculture, labour productivity in agriculture and the economy of the region as a whole, availability of employees with an appropriate level of practical skills and access to the sea, and negatively with population density and location in the country's border region. Growth of this type of export is important for improvement of living conditions in many underdeveloped regions of Poland.
Notes
1. In countries with a low or moderate development level, transfer of technology explains – according to certain estimates – as much as 90% of technological progress. An important source of such progress is international trade exchange, but not so much exports as first and foremost imports (Keller Citation2004). It should be emphasised, however, that exports provide funds thanks to which imports are possible.
2. More recent research includes, inter alia, Paluzie (Citation2001) and Shin et al. (Citation2006).
3. Based on the principles of territorial division, they constitute NUTS-2 level.
4. The data on exports in 2009–11 are not available, which prevents us from extending the sample.
5. According to the Österreichisches Institut für Wirtschaftsforschung (Austrian Institute of Economic Research) OIFW1 taxonomy, industrial sectors are classified in five broader groups: mainstream industries, labour-intensive, capital-intensive, marketing-led and technology-led industries.
6. Kao (1999) observed, however, that for panel model estimation with the least squares method, the estimation of the structural parameter binding two independent non-stationary variables converges to zero in the case of panel data, whereas in the case of time series it is a random variable. This means that although non-stationary panel data may lead to biased standard errors, the point estimations of the value of parameters are consistent.
7. Available from http://www.stat.gov.pl/bdl/app/strona.html?p_name = indeks.
8. The estimator applied in the case of correlation between explanatory variables and individual effects is the estimator proposed by Hausman and Taylor (Citation1981). The condition for applying the estimator based on the instrumental variables method is, however, the determination of an adequate set of instrumental variables that are not correlated with the individual effects.
9. Both in PCSE-AR and GLS estimations of the two models (I and II), high values of residual autocorrelation coefficients were found, which means that estimations obtained with the use of the methods are more reliable.