Turbulence in growing and declining industries

“… by letting many flowers bloom and ensuring only the best survive…” (Nickell 1996, p. 741).

For instance, barriers to entry, such as a commitment to industry-specific investments, are found to act also as barriers to exit (Caves and Porter 1976).

Data are recorded annually, with steady 12-month periods in-between observations.

According to the GEM definition, an enterprise is classified as a young business if it has paid salaries and wages for more than 3 months but for less than 42 months, and as an established business if it has paid salaries and wages for more than 42 months (Acs et al. 2005).

We drop industry and time indices for the rest of our discussion of explanatory variables. Values for the variables change both across industries and over time, except for the variable accounting for the business cycle, which is the same for all industries.

An important drawback of our growth and volatility measure is that we use employment data instead of sales, due to the unreliability of the sales data at our disposal. However, estimations using growth rates and volatility indexes based on corrected sales data, which are not reported here but which are available from the authors upon request, yielded very similar results to the ones reported here.

In this way we are able to appraise the influence on turbulence of the evolution of the share represented by both the largest and the smallest firms in the market. The greater the combined market share of the largest firms, the lower the turbulence rate should be; the greater the combined market share of very small firms, the higher the turbulence rate should be.

This procedure is comparable to the one used by Birch (1987).

We realize that the 205 sectors included in this ‘low growth’ or ‘mature’ group are too different in terms of structure and growth stage to be classified under only one heading. The purpose of this ‘midway’ group is solely to act as a control for the high growth and declining industry groups.

The same tests were conducted for all remaining explanatory variables used in our study. The results are available from the authors upon request. Equality of means and variances between the high growth and declining industry groups, and the whole sample was rejected at the 5% significance level or lower for all variables, except for the mean value of the four-firm concentration ratio in high growth sectors.

In defining the MES with reference to firms and not plants we attempted to capture scale economies that are not restricted to production technologies, but which include company-level cost components, such as advertising, distribution, sales and R&D.

The number of firms in each of the 319 six-digit industrial sectors being examined varies from one to 13,014, with a mean of 409.5 and a standard deviation of 1017.7, so we expect significant variance in the values of explanatory variables.

A noteworthy empirical issue concerns the possibility of endogeneity of the dependent variable (TURBULENCE) and the explanatory variables measuring lagged entry and exit rates. The sum of three of the explanatory variables—lagged entry (ENTRY), lagged exit of young firms (EXIT3) and lagged exit of established incumbents (EXIT4)—corresponds to the lagged value of the dependent variable. However, the correlation coefficients between the dependent variable and these three lagged explanatory variables are quite low. According to Table 3, the correlation coefficients with the explanatory variable (TURBULENCE) are: 0.08 for EXIT4; 0.09 for EXIT3; 0.11 for ENTRY. Therefore, we do not believe that endogeneity is significant enough to seriously affect results.

The fact that the Portuguese economy went through significant deregulation changes in the period 1986–1989 (immediately after entering the EU) very likely also plays a role in the significantly positive coefficient for YEARDM.