Institutional environments for entrepreneurship: evidence from emerging economies in Eastern Europe

Entrepreneurship: Theory and Practice, Jan, 2008 by Tatiana S. Manolova, Rangamohan V. Eunni, Bojidar S. Gyoshev

[FIGURE 1 OMITTED]

In the model fit summary, we reported the fit measures for the Amos default model as well as two additional models--saturated and independence models. While the former is one with no constraints placed on the population moments and therefore would fit any data perfectly, the latter goes to the other extreme and assumes that the observed variables are not correlated at all, and as a result of the severe constraints provides a poor fit even when the data contains interesting relationships. The default model specified in Amos needs to be viewed and interpreted in the context of these two extremes. NPAR is the number of distinct parameters being estimated, CMIN is the minimum value of the discrepancy and P is the p-value for testing the hypothesis that the model fits the population perfectly. CMIN divided by its degrees of freedom (df) is suggested as a measure of fit. The closer it is to 1, the better the fit. Following the recommendation of Marsh and Hocevar (1985) that a ratio between 2 and 5 indicates a reasonable fit, we conclude that with a ratio of 2.33 the default model appears to be an acceptable fit for the variables measured by our data.

The Bentler and Bonett (1980) normed fit index (NFI) is a measure of the minimum discrepancy of the model being evaluated as a proportion of the baseline model while Bollen's (1986) relative fit index factors in the df. Although in our study both these indexes are slightly less than .90, the model fit is acceptable considering that Bollen's incremental fit index (IFI) is .92 and the Tucker-Lewis coefficient (TLI) is .90 (As a rule of thumb, the closer all these indexes are to 1, the better is the model's fit). This inference is reinforced by the fact that the comparative fit index (CFI) of our model is .92, well over the acceptable level of .90 in explaining a significant proportion of the variance. The root mean square error of approximation (RMSEA) of .07 is below the benchmark .08 and indicates a reasonable error of approximation (Browne & Cudeck, 1993).

Our model compares favorably with Busenitz et al.'s (2000) model in terms of factor loadings, scale reliabilities and goodness of fit indicators. Table 4 summarizes the relevant parameters of the two studies. We also assessed the model equivalency across the three country samples, performing factor analysis (1) on each country's sample separately. In each case, the loading patterns were similar, indicating identical factor structures across the three countries. These results suggest that Busenitz et al.'s instrument profiling institutional environments in industrialized countries is also valid for emerging economies in Eastern Europe, positively answering our first research question.

Our second research question referred to the favorability in the institutional profiles of the three emerging economies with respect to entrepreneurship. To answer this question, we performed analyses of variance. The results are presented in Table 5. They show that while there are no significant differences in their overall institutional profile scores, there are important differences among the three countries along each of the three individual dimensions: regulatory, cognitive, and normative. We therefore inferred that although somewhat masked by the aggregate profiles, significant differences do exist in the underlying dimensions of the institutional environment for entrepreneurship in the emerging economies of Eastern Europe.