The data is taken from a database built by the author on independent, indigenous manufacturing firms which were incorporated with limited liability in or after January 1973 and which employed 50 or more people (fulltime) in their country of origin in 1992. The sample is effectively a census of all such firms in Ireland, Scotland, and in 68, or threequarters, of the 89 departements (counties) of France. Though the coverage for France is not complete, it is geographically representative of France outside the core IledeFrance region.
The data was gathered in a threestep process. First, hardcopy and online or discbased national government and privatesector databases were crossreferenced to identify potential young growing firms with the above definition. For Ireland, a special list was kindly provided to the author by the Industrial Development Authority from its annual November employment survey for 1992. This was crosschecked against annual returns to the Irish Companys office for all firms and their subsidiaries. For Scotland, a series of private sector firm directories were crossreferenced (Jordans and ICC regional surveys; Dun and Bradstreet and Macmillan UK directories, Fame and Lotus OneSource discs). Most of these directories quote data contained in annual returns at the Companys Office; some crosscheck with the firms themselves. For France, Kompass France (hardcopy and CDRom version) and DIANE CDROM were crosschecked with VERIF, an online Minitel service provided by the French Chambers of Commerce, and INFOGREFFE, which supplied detailed financial and historical data on the evolution of the firms.
Second, historical press coverage (in library files and
online press databases) was studied on each potential
entrant to the database. Most mediumsized manufacturing
firms are sufficiently important in their local areas or national
markets to be reported in the regional or trade press. This data
helped to eliminate many ineligible firms. Third, when all other
had been exhausted, the firms themselves were contacted to verify their age, origin and where necessary, numbers employed in 1992. The status of a very small number of firms could not be confirmed, either because they had been completely liquidated or because they refused to confirm whether they fitted the definition. They were not included in the database.
This population set was supplemented with a separate population set obtained using a similar census methodology 5 years previously for Ireland and Scotland using 1987 as a cutoff point (see Levie, 1993; 1995a). By comparing this second population set with firms in the first population set which were founded in or after January 1978 (i.e. January 1973 plus 5 years), two similar cohorts can be compared of young growing firms at different times in the business cycle.
To test the hypotheses, valid construct categorical measures are required for the diversification type and mode constructs. Many studies utilise ISIC codes to measure relatedness. Ironically, Wrigley created the inductive relatedness construct to surmount a basic problem with ISIC codes in diversification research: many products have technological and market affinities, but are distant in terms of ISIC code. In this paper, firms were classified as related diversifiers if they had entered a new (i.e. noncompeting) product/market which was similar to their core product/market in terms of technology or markets, or both. Firms were classified as unrelated diversifiers if they had entered a new product/market which was in no way similar to their core product/market in terms of technology or markets. Firms were classified as vertical integrators if they entered a part of the value chain from raw materials supplier to endconsumer that they had previously contracted out. Firms were classified as acquirers if they (or the founder) had purchased a majority shareholding at least one other domestic company during their lifetime. Minority stakes and foreign acquisitions are not included in this paper. Classification was conducted using files which had been built up on each firm in the course of data collection. Interrater reliability has not been tested.
Technology level was determined by reference to Phillips (1991), who categorized high and low technology firms by their principal product types. Medium technology products were deemed to be all products not categorized as high or low technology, and a list of these was compiled using an ISIC manual (Office of Management and Budget, 1987). See Levie (1995b, appendix 3) for a copy of the full list.
The final construct measure is performance. The only performance variable (other than survival) that was universally available for all firms in this crossnational study was employment. A serious problem with the French data is that while comprehensive financial data is available for almost all firms, accounts are not consolidated, leaving the true finances of the most interesting firms?those that have a number of subsidiaries?impossible to discern because of internal asset and sales flows. A similar problem arose with the acquisitive Irish firms. They were able to exploit a legal right to small firm status for their core management companies, and thus avoid full disclosure of their accounts, while providing guarantees to their subsidiaries which enabled the latter to avoid disclosing their accounts.
Sales figures were also seen as sensitive data (in a small
nation such as Ireland, sales figures can be more sensitive than
profit figures). Employment figures were not generally seen as
sensitive data, and previous studies show correlations of 0.9
between sales and employment data in samples of young
manufacturing firms (McDougall et al., 1994). The author found
correlations of 0.8 between sales and employment in samples of a
19731987 cohort of Irish and
Scottish young growing firms for which sales data was available.
© 1997 Babson College All Rights Reserved
Last Updated 06/01/98