Monday, June 27, 2011

How closely is well-being related to per capita GDP?

The relationship between a composite well-being index and per capita GDP in OECD countries is shown in the chart below. The well-being index has been derived by modifying and combining OECD indicators as described in previous posts (here and here). As might be expected, the chart suggests that well-being is generally higher in countries with high per capita GDP. For most countries, including the United States and Australia, there is not much difference between the well-being index and the picture of well-being presented by per capita GDP. There are some countries, however, in which well-being seems to be higher than would be expected (most notably New Zealand) and some in which well-being seems to be lower than would be expected (e.g. Luxembourg, Greece and Korea). In this post I want to explore whether some of those apparent anomalies may be attributable to the past income history or the countries concerned.

Note to the chart: Per capita GDP data is from Penn World Tables (rgdpl). The per capita income data is presented in log form because previous studies have suggested that this is appropriate in considering the relationship between well-being and income (for example, Stevenson and Wolfers, 2008).

Why might past income matter for current well-being? Past income is relevant because well-being is affected by wealth as well as current income. Some components of wealth are incorporated in the well-being index (e.g. the quality of housing) and others e.g. public infrastructure could affect several well-being indicators. Wealth may also provide peace of mind to individuals as a cushion against loss of income - for example as a result of ill health or unemployment. A study by Bruce Headey and Mark Wooden has shown, using Australian data, that wealth is at least as important to subjective well-being as is income (IZA Discussion Paper 1032, Feb. 2004).

In order to assess the extent to which past incomes matter I have used regression analysis to explain the well-being index in terms of two components of current per capita incomes: per capita incomes in 1970 and the growth in per capita income from 1970 to 2009. If income history is irrelevant to current well-being the estimated coefficients on the two components of income would be expected to be similar. In fact, the estimated coefficient on per capita income in 1970 is much higher (more than 1.8 times) the estimated coefficient on the growth component. (The standard errors of the estimated coefficients are fairly low and the difference between them is statistically significant at the 95% level. The regression explains about 76% of the variation in the well-being index. Anyone who wants further information on the regression results is welcome to contact me.)

The regression results have been used to decompose the well-being index in order to prepare the following chart.

The story that a comparison of the two charts tells me is that a past history of relatively high incomes helps to explain why New Zealanders score relatively highly on the well-being index. A past history of relatively low incomes also helps explain why Korea has a relatively low well-being score. However, the size of the relevant residuals suggests that past history doesn’t help explain the disparity between well-being and per capita GDP levels for Luxembourg and Greece.

The general picture that emerges is that the well-being index and per capita GDP generally convey similar information about relative well-being levels in OECD countries. In some of the countries where this is not so, the disparity can be attributed largely to income history. There do not appear to be any OECD countries with high levels of well-being that do not have either high current per capita GDP levels or a history of relatively high per capita GDP levels about 40 years ago. This suggests to me that over the longer term there can be no escaping the links between wealth creation and progress in improvement of well-being.


Dave Giles said...

I think you're right in your general conclusion, and I expect this would be borne out if we had an equivalent of the OECD Better Life Index for non-OECD, less developed, countries. One way to approximate this would be to replicate your analysis using the World Bank's Human Development Index. The correlation between the latter and the BLI for the OECD countries is 0.74. See my blog post here.
Nice job!

Winton Bates said...

Thanks Dave. It might be worth doing the exercise for the HDI as you suggest - or perhaps for components of the HDI i.e. life expectancy and mean years of schooling.
I'll think about it.