Variables construction in the case of the IET survey

Instead of using Herfindahl-Hirschman index or import share as a measure of competition, in this model we are able to include as a right-hand side variable firms’ own assessment of the degree of competition it faces. Information on competition is a part of the regular IET surveys. The firms are asked to evaluate the level of competition using the four-score scale ranging from very strong (1 point) to none (4 points) for three groups of producers: domestic producers, producers from CIS countries, and producers from abroad. The index of competition intensity is constructed as the standardized inverted first principal component of firms’ assessment of competition with all three groups of competitors. That is, the higher is the size of the variable, the tougher is competition.

Other firms’ characteristics are the same as in the regressions, which use Goskomstat data. In some specifications we also use indicator of those firms, whose managers studied abroad. This indicator is obtained from another survey, which we conducted together with the IET. That survey contained the question about firm having managers with foreign MBA, and the question about presence of managers, who took short management courses and/or internships abroad. Positive answer to one of these questions was coded as presence of managers, which studied abroad. Of course, we can not control for the quality of education, which such managers received, so this variable gives only an imprecise measure of the quality of managers. Nonetheless it turns out that inclusion of this variable in the regression equation provides interesting information about behavior of firms.

Appendix 2 provides further details about construction of each variable, and some summary statistics.


Tables 4 and 5 report the results of regression estimation in Goscomstat and IET datasets respectively. As can be seen from the regression tables, major results are in line with the theory. Profitability is highly significant in all specifications, estimated using Goskomstat sample. In the case of IET sample, this variable is also always positive, but it gets less significant or even insignificant in the specifications, where we control for quality of corporate governance, and quality of managers. Profitability here is a predetermined variable, and we interpret profitability as a measure of credit constraints. Our results suggest that credit constraints is an important obstacles to innovations. Better corporate governance can relax credit constraints. Ability of a firm to attract managers, educated abroad, or send managers abroad, can be a signal of firm profitability, so, again, it is not surprising that inclusion of this variable in the regression decreases significance of profitability itself. We should notice also that the size of the sample is smaller in all these two cases, which can be an alternative explanation for insignificance of profitability.

In both datasets, when we compare firms, which innovate, with those, which do not innovate, size of the enterprise enters equations positively and significantly. This finding can have several explanations. Size of the firm can be another proxy for its credit constraints. Usually larger firms have better relationships with banks, which allows them to get credit. Large firms can also economize on scale, while doing R&D. Since we count not only large, but also small innovations, a technical explanation is also possible. Large enterprisers usually produce goods, which go through more stages of production, than the goods produced at smaller firms. Even if innovation rate in each stage of production is the same, the innovation rate on the large enterprise will be higher than on a small one.

Interestingly, size of the enterprise does not matter when we look at the changes in innovation rate since 1980s. In a sense this regression controls for size fixed effect, which becomes unimportant. Imagine, however, that firms, which innovate more often, grow faster. Than the positive relationship between the firm size and change in innovation rate should be positive. Since we do not find such relationships, our results tend to suggest that innovation activities of Russian firms affect size distribution of firms very little, i.e. that it is not necessarily true that firms with high innovation rate tend to increase their size, and slow innovators decrease their size. Absence of this effect can be explained either by the small size of most of innovations, or by ineffectiveness of most of innovations. It is also possible, that during most of the 1990s firms rarely innovated at all, and they only started to innovate recently. If this is the case, the size distribution of firms should change according to the intensiveness of innovation activities in the near future.

The regressions on the IET dataset show that competition has an inverted U-shape effect on innovations. This means that if competition is not severe, it actually forces firms to innovate. This effect is observed both in the comparison between different firms, and in the regressions, which looks on changes in innovation rate since 1980s. When we compare innovation rate across firms, the effect of competition gets insignificant in the specifications, where we control for the corporate governance quality, and for presence of managers, which received some training of education abroad. In both cases, when we include the quality of corporate governance, and education of managers, sample size is smaller than in the original specification. We estimated basic specifications on these smaller samples. Competition was insignificant already in this basic specification, so sample size explains most of the differences in the results.

Goscomstat sample is much larger than the IET one, so we can control separately for the effect of domestic and foreign competition on innovations. Foreign competition we measure as competition with imports. The effects of both of them have a U-shape form. It means that there is a threshold in the intensity of competition, after which innovative activities become less intense. Since linear coefficient is positive and significant, and all measures of competition, which we use, are standardized variables, we can say that maximum of the parabola is to the right from the median firm, and majority of firms are on the upward-sloping part of parabola. In the smaller samples, where we control for quality of corporate governance, domestic competition, measures by Herfindahl index, becomes insignificant.

On the Goscomstat dataset we can compare innovation rates of firms in different ownership. In the case of IET dataset this is impossible, because it contains almost no firms with high ownership stake, which belong either to the government or to the foreign firms. Most of firms in this dataset are privately owned, and not traded on the stock exchange. In the Goskomstat, both enterprisers in foreign ownership and in federal government ownership turned out to innovate with higher probability than other enterprisers. The former result differs from finding of Jefferson et al (2002a,b) regarding Chinese firms. That paper showed that foreign firms usually innovate less than the domestic ones. If they do innovate, though, they do it much more actively than the domestic firms. The Chinese paper concentrated more on R&D than on imitations. It interprets the finding regarding the foreign firms as if they rely more on innovations, made by the foreign investor.

Controlling for the quality of corporate governance and foreign education of managers produced positive, but insignificant results in all specifications in the IET samples. We did not run the regression with managers education on the Goskomstat sample, because this variable describes situation in 2003, while the information in the sample is relevant for 2001.

As we will see, though, these variables have different effects on different types of innovation activities. Therefore, it is not surprising that their effect becomes insignificant, is we treat all innovation activities equally.