2010 Review of "Cultures and Organizations"
20220318
Review of Geert Hofstede’s book, now in its third edition. As is - I wasn’t as serious about quality 12 years ago.
Graham
Five stars for ambition and importance, despite significant flaws
This entire book is dedicated to the statistical analysis of huge international surveys of people's of people's values, and an interpretation of the results of that statistical analysis.
Statisticians as a group are generally disdainful of people who use statistical products, and this would be no exception. The authors list the weaknesses of their approach in the beginning of the book, but then speak authoritatively, as if those limitations have been put behind them and could be safely ignored. Let's re-examine what they did:
They used massive surveys which depend on subjective questions such as, "on a scale of 1 to 7, where one is "love it" and two is "hate it,” how do you feel about the employee appraisal process? These five or seven point scales are called "Likert" scales. They are the best possible instrument, but as the authors point out, they have significant limitations. In particular, respondents may be culturally driven to provide answers they think the researchers want to hear, and they may be culturally driven to either select extremes or avoid extremes.
The authors used a process called factor analysis to determine what they find to be five different factors that define different cultures values. When you add up the effect of all the different factors, you can statistically explain a certain fraction of the pattern in which the survey findings varies systematically among different groups of respondents. There is always some "unexplained variance" that cannot be attributed to any factor. The relative importance of the factors depends on the order in which you consider them, but no matter how you slice it, the fourth and fifth factors are not likely to explain very much of the variance. In other words, even though they are statistically significant, their real-world significance may not be that vast. A researcher has to have humility.
Once they identify the questions which make up each factor, they were able to compare responses from all of the countries one factor at a time. In doing this they took some liberties, such as splitting the five or seven value Likert scale in two, for strong and weak. Where you make the split influences how strong and affect you find. Researchers have been known to pick the split that best supports their theories.
The five factors they find are:
Power distance
Individualism and collectivism
Masculinity and femininity
Uncertainty avoidance
Indulgence versus restraint
The authors have normalized the findings of all countries, putting them on a scale of roughly 1 to 100. This provides a convenient graphic for showing relative differences. This is well and good. They do not, however, ever discuss how statistically significant differences are. Some of them are certainly more significant than others, and some of the differences are quite probably trivial... but the authors never deviate from their tone of certainty about their findings.
The heart of the best book is a series of two-dimensional plots,