![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
This paper investigates the impact of human capital endowments on export intensity employing firm-level data for 29 transition economies. A particular focus is placed on comparing and contrasting Central and Eastern Europe countries (CEECs) with those from the former Soviet Union, the Commonwealth of Independent States (CIS). The impact of the share of employees with higher education, provision of on-the-job training, years of experience of the top manager and labour cost on export intensity is assessed. To test these relationships, Tobit and Fractional Logit approaches are adopted. The estimation results suggest that, overall, having a more educated workforce exerts a positive impact on the export intensity of firms in transition economies, the magnitude being larger for CEECs. Average labour cost, as an alternative measure, also turns out to exert a positive but stronger impact. Insufficient evidence is found of a role for training programmes and years of experience of the top manager.
Acknowledgements
We thank the anonymous referees for their helpful comments and suggestions.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Albania, Bosnia and Herzegovina, Bulgaria, Czech Republic, Estonia, Croatia, Hungary, Kosovo, Lithuania Latvia, FYR Macedonia, Montenegro, Poland, Romania, Serbia, Slovak Republic and Slovenia.
2. Armenia, Belarus, Kazakhstan, Kyrgyz Republic, Moldova, Russian Federation, Ukraine and Uzbekistan. The reported figure does not include Azerbaijan, Georgia, Tajikistan, and Turkmenistan due to missing data.
3. Due to missing data for 1995, the rate of change calculation here does not include Bosnia and Herzegovina, Montenegro, Kosovo and Albania.
4. Programme for International Student Assessment (PISA) is an international survey of the OECD that assesses the capabilities of 15-year-olds in reading, mathematics and science (see http://www.oecd.org/pisa for more information).
5. Trends in International Mathematics and Science Study (TIMSS) and Progress in International Reading Literacy Study (PIRLS) are administered by the International Association for the Evaluation of Educational Achievement. TIMMS is conducted every four years at the fourth and eighth grades while PIRLS is conducted at the fourth grade every five years (see https://timssandpirls.bc.edu for more information).
6. CEECs: Albania, Bosnia-Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Kosovo, Latvia, Lithuania, FYR Macedonia, Montenegro, Poland, Romania, Serbia, Slovak Republic, Slovenia; CIS: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Mongolia, Russia, Tajikistan, Ukraine, Uzbekistan.
7. One may argue that the direction of causality between education and a firm’s engagement in exporting activities might not only be from the former to the latter and that it is more likely for the exporting firms to employ highly educated workforce. However, we argue that even if such a feedback mechanism is likely to exist, it would take time for firms to adjust the skill level of their workforce in response to changes in their exports. Furthermore, the summary statistics reveal that exporting firms in our sample have, on average, lower shares of educated individuals than non-exporters, which suggests that reverse causation is not likely to be present in our model.
8. BEEPS data has two variables related to firm size: number of employees in the last fiscal year and three years prior to the survey. Four per cent of firms in the sample had not been in business three years prior to the survey; in order not to lose observations, these ‘missing’ values are replaced by the number of employees in the last fiscal year.
9. The low and medium-low technology intensive goods have been grouped into a single category given the similarities in the estimated coefficients, whereas, the latter two technology categories (medium-high and high) have been included separately.
10. The empirical analysis is implemented using Stata (Version 12.0). To facilitate replication Stata commands are available on request.
11. Referring to Maddala (Citation1991), Wagner (Citation2001, 231) states that Tobit ‘is appropriate when the value of the variable can be less than a lower limit but observations with such values of the variable are not observed because of censoring’. This is not the case for variables that are bounded by definition.
12. The estimations have been also replicated by Poisson modelling and the results are generally consistent with those from the other approaches. The results are available on request from the authors.
13. These variables are: the share of skilled production employees, a relative measure of technology, participation in a business association and, the share of foreign material inputs.
14. The set of plausible values for the missing observations is generated through an imputation model. This model is specified with a set of predictive variables that might contain potential information about missing observations. Schafer and Graham (Citation2002) used 20 imputations for a share of nearly 80% of incomplete data. On the other hand, White, Royston, and Wood (Citation2011) argue that number of imputations should be even higher, e.g. equal to the fraction of missing data.
15. Because of converting a continuous variable into a categorical one, this approach causes loss of information. It also is sensitive to the number of groups defined. However, it still provides a satisfactory indication of goodness-of-fit in the absence of alternative diagnostic checks for these two estimation approaches in literature.
16. During this comparison, the Tobit estimates have been adjusted by dividing the coefficient estimates by the standard error of the regression, σ (i.e. βj/σ). The Probit model treats exp_int as a binary variable: one for exporters and zero otherwise.
17. This result is obtained from estimations on a sub-sample of large firms. Estimation results are available on request.
18. The more experienced managers may have been managers before the start of transition and that experience may not be relevant to the current environment.
19. The marginsplots presented in the appendix for this and other variables of interest refer to the Tobit model. The corresponding plots for Fractional Logit are similar, hence are available on request.
20. According to their technology intensity level, the majority of exporting (manufacturing) firms seem to export low and medium low-tech goods (70.8%), followed by medium-high (23.3%), with only a very small proportion (5.9%) exporting high tech goods.
21. Given the relatively low share of missing data in the baseline model, 22 imputations are used for each missing observation.
22. The marginal effects could not be produced for the imputed models hence these results refer to parameter estimates.
23. The number of imputations used in this model is 45.
24. The Chow tests were run on a full sample of countries including explanatory variables as well as their interactions with the country-group dummy. The joint significance of the interaction terms was then tested by an F-test and Chi2 test.
25. The equality of these coefficients is rejected at 10% significance level: the 90% confidence interval for the CEECs is [0.00017, 0.00053] while it is [0.00005, 0.00015] for the CIS.