Data and Methodology of Regression Analysis

To analyze, whether factors, discussed in the previous section, have any effect on innovation activities of Russian firms, we use probit regressions techniques. We estimate whether probability of a firm to be involved into innovation activities depends on competition with domestic and foreign firms, credit constraints, ownership structure, state interventions in the regional economic activities, quality of corporate governance, and quality of managers. In the case of Goscomstat data, the dependent variable takes the value of 1 if the enterprise belongs to the list of firms, which conducted innovative activities in the last three years according to the 2001 publication. The remaining firms are the rest of the firms, included in the Registry of Russian Firms. We should note that this approach can suffer from mis-classification basis. It can happen in those cases, when a firm, which is included in the registry, and is being involved into innovating activities, did not participate in the Goskomstat innovation survey. However, since the survey is quite large, we don’t think that the number of firms misclassified in this way is large. A similar specification can be estimated using IET survey data.

IET questioner allows us to estimate several other specifications. We can use the answer to the question about change in the rate of innovative activities since 1980s as a dependent variable. Additionally, we can test whether there are any differences in factors, which influence firms with imitating and innovating development strategies.

Methodology of constructing dependent variables also differs slightly in the case of Goskomstat and IET data. In the case of Goskomstat data we have to rely on other sources, usually also produced by Goskomstat, to construct most of the variables. In the case of IET some of the variables can be obtained from other IET surveys, conducted using the same original sample of firms.

In addition to survey data and data from various Goskomstat publications, we use the so called Russian firms registry. This dataset contains firm level balance sheet statistics, and other statistical information, which Russian firms have to submit to statistical agencies. This dataset was constructed using information from GNOZIS, ALBA, Registry, and other datasets. These datasets are a traditional source of information on Russian firm, which was used in a number of other studies, among which are Yudaeva et al. (2003), Guriev and Rachinsky (2004), Brown and Earle (2001), etc.