Researchers often fail to confirm cross-section associations between income and a variable of interest when they examine changes in the same variables over time. This illustrates vividly the old truism “Correlation is not causation”.
USC economist Richard Easterlin (born 1926), in a new paper, explains. In 1974 Professor Easterlin highlighted an example of this problem, now known as the “Easterlin Paradox“, the observation that happiness and income are positively associated between countries (or individuals within a country), but are typically not related in time series data.
International cross-section regressions on real gross domestic product (GDP) per capita are widespread in the social sciences. Findings of significant associations between GDP per capita and a multitude of economic, social, and political variables are commonplace, and these results are often read as demonstrating the effect of economic growth on the variables under study. For variables integral to production and consumption, such as material living standards or the rural–urban distribution of employment, such inferences are plausible. But economic growth is often viewed also as the main force responsible for such outcomes as the expansion of schooling, improved health, increased life expectancy, fertility decline, women’s empowerment, the extension of political and civil rights, and the like.
Moreover, these cross-section relationships are often taken to be predictive of time-series change, of what is likely to happen as a result of economic growth. Studies of the historical experience of individual countries, however, frequently fail to confirm expectations based on cross-section relationships. As one moves outside the purely economic realm to social and political variables, this lack of confirmation of an association with GDP per capita is especially apparent. Why, then, do we often find a significant cross-section relationship if the implied causal connection is not confirmed by time-series analysis? The answer suggested here is that cross-sections register the results of history, not insights into likely experience.
To focus the discussion, take as a demographic example the international cross-section regression of life expectancy at birth … on GDP per capita …. This relationship is sometimes thought to demonstrate the causal impact of economic growth on life expectancy ….
[Easterlin goes on to reason that economic growth does not cause increases in life expectancy, despite the positive cross-section relationship. Technological change has a profound effect on both variables, but advances in health technology spread to countries more quickly than advances in production technology. The result is a positive correlation between per capita GDP and life expectancy in cross-sections, but not in time-series.]
Richard A. Easterlin, “Cross-Sections Are History“, Population and Development Review 38 (Supplement, February 2013), pp. 302–308.
There is free access to the entire supplement at the link above. This issue of Population and Development Review contains 21 essays that celebrate the departure of Hungarian economist Paul Demeny (born 1932) as editor of the journal that he founded. Professor Demeny graduated from the University of Budapest in 1955 and received a PhD in economics from Princeton University in 1961.
NB: On reading my brief summary above, I realize that the explanation is more complex, and is difficult to explain in a few words. If you are interested, I recommend that you read the full article.
