Cultural distance in international business and management: from mean-based to variance-based measures

Abstract

Extant practice in international management is to measure cultural distance as a nation-to-nation comparison of country means on cultural values, thereby ignoring the cultural variation that exists within countries. We argue that these traditional mean-based measures of cultural distance should take within-country cultural variation into account. Therefore, we propose the use of variance-based measures of cultural distance. To illustrate our argument, we examine total US foreign affiliate sales in more than 40 host countries over the 1983–2008 period, complemented with data from the World Values Survey. We analyze the effects of three cultural distance measures: the Kogut and Singh (1988) mean-based index of cultural distance, the Kogut and Singh (1988) index conditioned by host-country cultural variation and a variance-based measure that takes into account both home- and host-country cultural variation. Our findings indicate that, when within-country cultural variation is taken into account, the explanatory power of the Kogut and Singh (1988) index is substantially decreased. In addition, our variance-based measure of cultural distance outperforms the Kogut and Singh (1988) measure in the explanation of foreign US sales. We therefore suggest to move from mean-based to variance-based measures of cultural distance, thereby taking the cultural variation within countries into account.

Keywords::

• cultural distance
• IHRM
• intra-country variation
• measurement

Notes

^1. This can be calculated by dividing the difference on individualism (weighted by the variance on individualism) by the total cultural distance. The weighted distance on individualism is 8.11, and the total sum of all weighted difference dimensions is 11.8, yielding a share of 69%. To obtain the overall cultural distance score this 11.8 is divided by four yielding the 2.94.

^2. The website of the WVS, http://www.worldvaluessurvey.org/, provides more information.

^3. We note that the list of scores on our ICV measure shows some resemblance to a survey of response styles (Harzing 2006). Indeed, we would expect internally diverse societies to portray more extreme response styles in surveys.

^4. This is easy to calculate because ICV is measured with mean of 0 and a standard deviation of 1. For a mean level of ICV, the coefficient for the K-S index equals − 0.24 (see Model 2 in Table 2). When ICV increases by one standard deviation, the K-S index is corrected upwards to a new level of − 0.24+0.13*1 = − 0.11. In percentages, this is equal to a ( − 0.11 minus − 0.24)/ − 0.24 = 54% reduction of the effect of distance.

^5. To check whether these results are driven by the sample selection, we have done this analysis for Sample 2 as well. Our findings are robust for moving to Sample 2, in which case the reduction of the mean-based distance effect increases to 61% (results available on request).