A meta‐analysis of effect sizes in international marketing experiments

Abstract

PurposeEffect size is an important determinant of statistical power. However, very few experimental studies in international marketing (IM) report effect sizes and no meta‐analysis work in this regard has been done. The main objective of this paper, therefore, is to quantitatively document effect sizes of experiments in IM and to provide directions for further methodological improvement.Design/methodology/approachAll articles published in the top three marketing journals and the top six IM‐related journals during the period 1992‐2005 were screened; this yielded 35 experiment‐based papers within the domain of IM. For each study, ten methodological characteristics relevant to IM experimental designs were coded.FindingsThe 35 studies reported 68 experiments, which produced a total of 1, 074 observations. Results reveal that, on average, for experiments in international business marketing, about 2.89 percent of the variance in a dependent variable (DV) is accounted for by experimental treatments, and a variance of 3.61 percent is shared by the independent and DV for experiments in international consumer marketing. Sampling method, type of subjects, type of design and number of countries are found to have significant influences on effect sizes.Originality/valueThis paper provides a quantitative, state‐of‐the‐art review of effect sizes in IM experiments, points out problems such as inappropriate reliance on an overall effect size index, and further offers useful suggestions on how to report and improve effect sizes.