Decomposing the difference in life expectancy between rich and poor

An under-recognised aspect of socioeconomic inequality is not measured in income or wealth, it is measured in disparities in length of life. There are essentially two ways that we can conceptually look at this issue. One is in terms of lifespan inequality - essentially how big are disparities in the length of life across the whole distribution of lifespan. I have a PhD student working on a number of research questions related to lifespan inequality and its relationship with income inequality. The second way is in terms of inequality in life expectancy - differences in the average length of life between different groups, such as the difference between the rich and poor. The two ways of looking at this issue are related, but they are not the same - one looks at differences group-level averages, while the other looks at individual differences.

So, with a PhD student working on this topic, I've had good reason to delve into the literature (which is something I have been wanting to do for some time). And we aren't the only ones looking at this issue - there's a growing literature on inequality in length of life. Most use fairly standard inequality measurements. However, this recent working paper by Gordon Dahl (University of California, San Diego), Claus Kreiner, Torben Nielsen, and Benjamin Serena (all University of Copenhagen) takes things in a different direction. They compare inequality in life expectancy between rich and poor for the US and Denmark, first noting that:

The gap in life expectancy between rich (top tertile) and poor (bottom tertile) males is around 8 years in both the US and Denmark in 2001. Over the short period from 2001 to 2014, this inequality increased by 1.7 years in the US and 0.9 years in Denmark. The gap between rich and poor females stayed constant in Denmark over this period, but also increased by about 1.8 years in the US.

Dahl et al. then decompose those differences in life expectancy into different mortality trends between rich and poor, and a common mortality trend for all. They refer to the common mortality trend as 'survivability', and it measures:

...the likelihood of surviving until a given age multiplied by the expected remaining life years after surviving this age.

Essentially, this decomposition recognises that differences in age-specific mortality have different effects on life expectancy, depending on what proportion of the population live to each age. Using this decomposition method, Dahl et al. find that:

In the US, half of the rise in inequality for forty-year old males... is due to larger reductions in mortality rates for the rich than the poor, while the other half is due to differences in their survivability. For Danish males... life expectancy inequality increased, even though mortality rates have fallen more for the poor. The explanation for this apparent puzzle is that survivability strongly favored the rich, more than offsetting the effect of differential mortality rate changes. For females in both countries, survivability plays a similarly important role.

They also find similar results to Denmark for nine other western European countries (Austria, Belgium, England/Wales, Finland, France, Italy, Spain, Sweden, and Switzerland). Why does all this matter? Dahl et al. argue that:

Trends in age-specific mortality rates of the rich and poor are informative about changes in underlying health status, while trends in life expectancy are a relevant measure of the associated welfare effects. Looking at each of these in isolation misses an important link - survivability - and, as demonstrated by our empirical results, can lead to misleading conclusions.

For example, say that there is a change in age-specific mortality at older ages arising from some new health procedure or treatment. This change appears to favour the poor because it affects the age-specific mortality of poor to a larger extent than the rich. However, this new treatment might lead to a larger increase in the life expectancy of the rich than the life expectancy of the poor, if a greater proportion of rich than poor live to the ages where the treatment has the most positive effects.

None of this is particularly surprising once you get your head around it, but nevertheless it is important to recognise. As we are finding in our own research, the greatest decreases in lifespan inequality (and likely in differences in life expectancy between groups) happen when mortality reduces at young ages.

[HT: Marginal Revolution, last year]