Article of the Week: The Ecological Fallacy in National Cultural Research by Brewer and Venaik

I'll try a new experiment, a short note about an interesting paper I've read in the past week or so. Hopefully in an entertaining way, but I can't promise anything: it's really hard to punch up regression tables or structural equation models. I can't even promise it'll stay an article of the week, I'll be lucky if I can keep to an article a fortnight. Paul Brewer and Sunil Venaik's new paper in Organizational Studies provides a real valuable service to entrepreneurship and organizational studies people who want to look at culture. I think you'd be hard pressed to find anyone who doesn't at least agree that culture has a huge impact on how organizations operate. But the chief stumbling block in studying it is quantification. How power oriented is the culture? Can we measure it to the second decimal place?

Geert Hofstede came up with a great solution to this problem. Throughout the 1970s, he surveyed thousands, eventually over 100,000, IBM workers about psychological and leadership issues. Through the magic of component analysis and (one of my favourite mathematical terms) verimax rotations, he came up with 5 dominant cultural factors and generated scores of them for over seventy countries. Researchers love using this as a basis for work to explain national differences in things such as startup rates, innovation, and economic development. It's a really neat way of connecting something as nebulous of culture with something as important as economic growth through rigorous statistical methods.

So now, Hofstede's work has been criticized from a number of sources — the basic theoretical underpinnings, the methodology, the results, but one aspect has been really overlooked but is present in almost everything that uses this work (and other similar measurements of cultural attitudes). That is the ecological fallacy. This is when you use statistical evidence about an entire population to make judgements about an individual. The paper has a great table showing the problems this causes.

In essence, the authors argue that there are huge problems with taking the 'average' cultural values of an entire country (a country which might have a very heterogeneous population made up of natives and immigrants) and then saying that everyone in that society has cultural values exactly identical to this average. This hides all the variation within a country's cultural outlooks, leaving little room for nuance or variation.

At the end of the day, this kind of outlook contributes to what Duncan (1980) called the reification of culture. This was a criticism of how cultural geographers viewed culture as a real, objective 'thing' that could be examined, measured, and understood. But it's not: it's a weird thing that is maybe part of our basic neurochemistry and is also maybe a complete social construct that is produced and reproduced by the powerful echelons of society. But when we reduce it to a number for a nation, we forget all this complexity in return for empirical parsimony. Brewer and Venaik's paper is a good and timely reminder that it's very hard to simply very complex things.