Improving generalizations from multi-country comparisons in international business research

In this paper we address the problem of questionable generalizations from comparing small numbers of countries in international business (IB) research. We illustrate the misleading results that can arise from sparse samples, whether the relationship between national-level variables is strong (economic development and corruption) or weak (population density and trust). We show that 35% of recent international comparisons in leading IB journals examined just two or three countries, and present an exploratory analysis of 123 variables that reveals typical correlations across countries to be rather moderate (average r=0.24). To help interpret extant study findings, we provide formulas and graphs based on Bayesian analysis, and introduce a method of combining results from multiple international comparisons. We also describe methods for designing studies to give stronger evidence of relationships between variables. Our results suggest that a minimum of 7–10 countries may support credible international generalizations, but only when overall trends are very strong. A key strategy for improving IB generalizations is to use larger samples of countries, because research based on common sample and effect sizes may lead to generalizations that the findings do not justify.