Firm migration in the Swedish wholesale trade sector

Abstract

This article analyzes the determinants of firm migration in the Swedish wholesale trade sector using a unique dataset covering over 10,000 Swedish wholesale trade firms during the years 2000–2004. The results indicate that there are negative correlations between profits, firm age, and firm size and the probability of firm migration. There is a positive correlation between firm growth in the previous year and firm migration, indicating that growth opportunities that cannot be realized at the present location are an important motivation for migration.

Keywords:

• firm relocation
• firm entry
• firm demography
• hierarchical random effects
• logistic regression model

Acknowledgements

The authors would like to thank Sven-Olov Daunfeldt, Niklas Elert, and participants at the sixth annual HUI Research Workshop in Retailing for helpful comments and suggestions. Financial support from the Swedish Retail and Wholesale Development Council is gratefully acknowledged.

Notes

1. These three theoretical approaches have usually been developed as theories for the optimal location of firms, with the relocation of firms as a special case. These three types have, however, also been used as a theoretical basis for studying firm relocation decisions (e.g. Pellenbarg, van Wissen, and van Dijk 2002a).

2. Note that in the empirical model this component will capture all firm specific heterogeneity affecting firm relocations, and not only bounded rationality and imperfect information.

3. The potential endogeneity of independent variables will be addressed in the empirical part of the article.

4. In practice, the firm could of course use other strategies to achieve the target level of expected profits. However, in order to focus on the decision to relocate we disregard other potential strategies in the theoretical section of the article.

5. The annual financial data are aggregated to the main office (HQ) when reported to the PRV. For firms with more than one place of production, it is impossible to distinguish how each plant contributes to the final results.

6. A test based on the Akaike Information Criterion (AIC) clearly favors the random effects structure presented above as compared to models that do not take heterogeneity at different levels into account. Also, the results from the estimations of Equation (7) presented below suggest that leaving the random effects structure out of the estimations will cause missing variable bias.