Cross‐Country Determinants of Early‐Stage Necessity and Opportunity‐Motivated Entrepreneurship: Accounting for Model Uncertainty

Abstract

Model uncertainty is one of the most pervasive challenges in the social sciences. Cross‐country studies in entrepreneurship have largely ignored this issue. In this paper, we evaluate the robustness of 44 possible determinants of early‐stage opportunity‐motivated entrepreneurship (OME) and necessity‐motivated entrepreneurship (NME) that are broadly classified in four groups: (1) economic variables, (2) formal institutions, (3) cultural values, and (4) legal origins and geography. The results, which are based on a representative world sample of up to 73 countries, suggest that institutional variables associated with the principles of economic freedom are most robustly correlated with OME and NME. Our findings also identify net income inequality and Scandinavian legal origins as weakly robust predictors of both types of entrepreneurial activity. Furthermore, we find that log GDP per capita is only a weakly robust predictor of NME, but not OME. We discuss implications for future research.

Notes

1 Terjesen, Hessels, and Li (2016) provide an excellent review of the comparative international entrepreneurship literature and conclude that out of 259 articles published in 21 leadings entrepreneurship and management journals from 1989 to 2010, close to one‐fourth of the articles are based on cross‐country data.

2 By model uncertainty, we mean the uncertainty that arises when analysts choose the true causal model that is then used in classical statistical tests. In the course of classical statistical analysis, authors often estimate a large number of models, but report only a handful of empirical estimations. These preferred models reflect only one “ad hoc” path in the modeling space (146) and are likely to represent a non‐random and unrepresentative sample of all plausible models (146). This is because significant findings that support the main hypotheses in a study are more likely to be published and therefore more likely to be reported. Young and Holsteen (2015), for example, review replication results reported in footnotes in two major sociological journals and find that out of 164 cases, not a single one of them reported results that failed to support the main findings. This is important because classical statistical theory assumes that only one (true) model is applied to a sample of data. If researchers, however, are uncertain which is the true model, this produces a much wider range of estimates than suggested by standard errors or confidence intervals associated with classical statistical methods. According to Pinello (1999) sampling uncertainty accounts for only 11 percent of the total variance in estimates while the remaining 89 percent is due to model specification (i.e., the choice of control variables in a model). We provide a more technical explanation of the problem of model uncertainty in the Data and Methodology section where we relate it to classical statistical tests, which account only for the uncertainty associated with the data sampling.

3 We searched for articles that contained the words “legal origins” or some variation of the term in their titles, abstract, or keywords.

4 Tropical countries fall between 23.45° North and South latitudes.

5 Young and Holsteen (2015) find that out of 60 quantitative articles published in two of the top sociological journals, 85 percent contained at least one footnote (average 3.2) referencing additional robustness tests that were often unreported. Out of 164 footnotes, however, not a single one reported results that failed to support their findings.

Additional information

Notes on contributors

Boris N. Nikolaev

Boris N. Nikolaev is Research Professor in the Department of Entrepreneurship at Baylor University.

Christopher J. Boudreaux

Christopher J. Boudreaux is Assistant Professor in the Department of Economics at Florida Atlantic University.

Leslie Palich

Leslie Palich is Professor of Entrepreneurship in the Department of Entrepreneurship at Baylor University.