Sample and research design
This study is based on survey responses from 1606 Kosovan entrepreneurs who completed questionnaire in face-to-face interviews during December-January 2002-2003, 2003-2004, 2004-2005 (thereafter 2002, 2003 and 2004). SME surveys were conducted by Riinvest Institute for Development Research and one of the authors had a leading role in this research project. Samples were selected randomly from the business register kept at the Statistical Office of Kosova and the Ministry of Trade and Industry (2005). Interviews were conducted by experienced and trained final year students from the Faculty of Economics at the University of Prishtina and monitored carefully by the Riinvest research team. Respondents were key informants, mainly the owner or the general of companies. The response rate was very high with an average of 96.4 percent. The exclusion of these observations might lead to sample selection bias, but in our case this is not a concern as the non-response rate is very small (around 3.6 per cent of total cases) and also our sample remains still representative (see, Cameron and Trivedi, 2005).
Pooling data. The pooled data technique is used to test set of econometric models of firm growth. The pooled data contains information from three independent surveys from the period 2002, 2003 2004. An ‘‘independently pooled cross-section’’ technique is obtained by pooling randomly sampled cross-sections at different points in time (Wooldridge 2006).