Processes of firm growth and diversification: theory and evidence

For example, diversification might correspond to imitation rather than innovation, and furthermore firms might diversify as part of a managerial ‘empire-building’ strategy, or perhaps to minimize risk, which are not attempts at innovation. We are grateful to a referee for pointing this out.

In our context, diversification refers to the commercialization stage of a new product introduction, i.e., offering a new product. This new product will certainly be new to the firm, but does not necessarily have to be new to the industry in which the firm operates.

Our implicit benchmark growth model for non-diversification events is Gibrat’s model of random growth shocks, according to which a firm’s growth rate has expected value μ and standard deviation σ. We therefore take as the growth rate of the control group the growth rate of all firms that are not undergoing a diversification event (i.e., the growth rate of the control group is μ).

It may be the case that further hires are needed to help execute the diversification plans, however, if the firm was ill-prepared for the growth event, or if the firm seeks to recruit employees at the lower ‘blue-collar’ levels, where new hires can be hired more easily (less distraction for managers) and can quickly learn their tasks (and quickly reach a high state of productivity).

This source and the corresponding classification scheme have already proven to be suitable for studying the market, regional and firm-level dynamics within this industry for the post-war era until the end of the last century (Fornahl and Guenther 2010; Buenstorf and Guenther 2011; Falck et al. 2013; Coad and Guenther 2013).

If a firm re-enters a product market in which it was previously active, this is not counted as a new product introduction in our data. Moreover, we used the database M&A Mergerstat to rule out the possibility of mergers and acquisitions for the considered companies in this time period (which might give rise to diversification events following other dynamics than outlined in the theory section).

Note that we do not investigate other modes of diversification, such as international expansion. For an investigation of firm growth in multinational corporate groups, see Oberhofer and Pfaffermayr (2013).

We also had information on other higher education institutes (that is, the number of universities, and music and arts colleges), but since these latter institutes were less theoretically relevant than universities of applied science, and furthermore were not significant in our preliminary analyses, we did not include them in our regressions reported here.

This is indicated by the fact that the number of observations is essentially the same as the number of firms in the lower panel of Table 2 that corresponds to ≥1 product introduction events.

Our variable for product introduction is a count data variable (non-negative integers corresponding to the number of new product introductions: most firms have zero product introductions, and the vast majority of those firms with positive product introductions have just one, although the maximum number of product introductions in 1 year was 10). When we estimated the Poisson, negative binomial and zero-inflated negative binomial models, they did not converge and we did not obtain meaningful output. Therefore, we focus instead on converting our product introduction variable into a binary variable: taking the value 0 if there was no product introduction and 1 if there was one or more product introduction.

The attentive reader will have noticed that Table 3 reports different numbers of observations for each of our three growth indicators (employees, assets and sales). We were therefore concerned that different numbers of observations for each growth indicator might invalidate our results somewhat, because they would not be directly comparable. To investigate this issue, we re-estimated Table 3 with the extra condition that we only focus on firm-year observations where we have (non-missing) observations for growth of employees AND growth of sales AND growth of assets. This analysis (available from the authors upon request) gave similar results, although slightly less significant in some cases (no doubt because of the smaller number of observations). We are grateful to an anonymous referee for drawing this to our attention.