Multivariate Data Analysis

Abstract

In our hypotheses, we stated that the various dimensions of internationalisation are expected to be a function of firm size, international experience of the founders, external fmance, technology intensity, innovativeness, the extent to which products are customised and the costs of commercialisation. Accordingly, we have to measure these influence factors. Firm size was operationalised in several ways. In the questionnaire, we asked respondents to state employees and sales both at start-up and at the time of the survey. The descriptive analysis revealed that the distributions are highly skewed to the right. Such a distribution is to be expected in a sample of start-up firms. Consequently, the direct operationalisations of size, i.e. the inclusion of the absolute sales volume or number of employees risk producing statistically insignificant parameter coefficients. Highly skewed distributions can be normalised using logarithmic transformations. We therefore first calculated log values of our various size measures. Second, for the regression models shown here, we constructed an index of the log values for firm size measured by number of employees and firm size by sales. The levels of the alpha coefficient (0.72 at start-up; 0.88 today) are above the recommended thresholds (Nunally and Bernstein 1994), therefore indicating construct validity. Since the actual firm size at the time of the survey could at the same time represent the cause and effect of international activities, we primarily used size at start-up in our regressions. We thus avoid possible effects of endogeneity in the models. Nonetheless, in most cases we also report parameter coefficients when size today is entered into the regression equation in order to explore whether the results are consistent. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.