Succession and firm growth: results from a non-parametric matching approach

Gersick et al. (1997) report that family firms account for 65–80% of all worldwide business, and for about 40% of the Fortune 500 companies. Although many family firms are small, in aggregate they represent about half of the US gross domestic product (Aronoff et al. 1997) and employ more than 80% of the work force (Neuberg and Lank 1998).

For a description of the institutional aspects in Austria, see Mandl and Voithofer (2005).

Succession is so central that Ward (1987) chooses to define family firms in terms of the potential for succession: ‘we define a family business as one that will be passed on for the family’s next generation to manage and control’ (p. 252).

For a more recent literature review on family firms, see also Chrisman et al. (2005).

The authors allude to Griliches and Regev (1995) who analyse firm exits and find firms that will exit in the future to have significantly lower growth rates in the years before exit compared to surviving firms. This phenomenon is called the ‘shadow of death’. Recent empirical evidence on this issue is provided by Almus (2004).

A large number of empirical studies, particularly for the farm sector, suggest a significant impact of farm characteristics as well as of characteristics of the farm operator’s family on the probability of (planned and actual) succession (Kimhi 1994; Stiglbauer and Weiss 2000; and Glauben et al. 2004; Bennedsen et al. 2007).

The institute annually collects economic performance data on a random sample of roughly 4,500 small- and medium-sized manufacturing firms in 19 industries.

To shed some additional light on the specific circumstances of the succession events a follow-up telephone survey was conducted by the Austrian Institute for SME Research in February 2007. From the 136 firms in the sample that experienced succession in the past, 94 were successfully contacted. Half of them (47 firms) refused to participate in the follow-up survey, 17 indicated that they are unable to recall details of the succession event (which took place about 10 years ago), and the remaining 30 firms provided additional information. Although this number is too small to carry out a detailed statistical analysis, the results nevertheless are helpful to make some first inferences. From these 30 firms, six companies (20%) experienced a transfer of management only (with ownership unchanged), another six companies (20%) experienced a transfer of ownership only (with management unchanged) and the remaining 18 companies (60%) reported both, a change in ownership and management. With respect to the differentiation between ‚family succession’ and ‚non-family succession’ the follow-up survey reveals that the transfer took place within the family in 24 succession cases (80%). Non-family members were involved in six cases (20%) only. This gives some justification to the assumptions made above.

A detailed discussion of the matching approach as well as a survey on its applications in labour-market evaluation studies is available in Heckman et al. (1999). An interesting empirical application of the matching-technique, which is closely related to the present issue, is reported in Almus (2004). The author investigates the growth rates of firms in the years before market drop-out and compares this to the performance of surviving firms.

The different propensity score matching schemes used in the empirical literature are discussed in more detail in Heckman et al. (1997), Smith and Todd (2005) as well as Becker and Ichino (2002). Since there is no consensus on the best matching estimator to adopt, we compare our results from 5-nearest neighbour matching with those obtained from one-to-one nearest neighbour matching, kernel matching and radius matching (the latter two applying kernel weights).

We use the unbounded \( x{'}_{i} \hat{\beta } \) rather than the bounded propensity score \( \Phi {\left( {x{'}_{i} \hat{\beta }} \right)} \) because of its preferable distribution properties (Hujer et al. 1997).

Authors from different fields of economics (and, in particular, from economic psychology and marketing) have challenged the usefulness of intention measures (such as succession plans) as a predictor for actual behaviour. Foxall (1983), for example, argues that a high intention-behaviour correspondence should be expected only under strictly limited (and unrealistic) conditions. Empirical support for the intention-behaviour discrepancy in succession plans is provided by Väre et al. (2005).