Author: Papuna Gogoladze
Have you ever heard of “The Occupy Movement”? If not, they are enthusiastic people, who fight against socio-economic inequality and try to establish contemporary forms of democracy. When we hear inequality, the first things we imagine, as economists, are Lorenz curve and Gini coefficient but do these concepts really make sense when we discuss inequality in terms of life Expectancy as an ultimate benchmark of any economic policy?
In this article, we will play roles of Myth Busters (popular science TV show) and deny some widely spread logical hypothesis about income inequality and life expectancy. Our discussion will mainly be based on “The Association Between Income and Life Expectancy in the United States, 2001-2014” (Raj Chetty et al.).
Even though, we, generally, associate high income with longer life expectancy[1] it would be interesting to dig deep into the data and through empirical evidence derive several aspects of the relationship. The study calculates period life expectancy conditional on income percentile at the age of 40, which is defined as “the expected length of life for a hypothetical individual who experiences mortality rates at each subsequent age that match those in the cross-section during a given year”. To put it simply, it implies that each hypothetical individual is supposed to expect probability of dying that match the data published by National Center for Health Statistics (NCHS).
To begin with, we will consider race and ethnicity adjusted life expectancy (for simplicity, in the future when we mention life expectancy, we imply race and ethnicity adjusted life expectancy) by household income percentile using pooled data from 2001 to 2014.
Figure 1. Race and Ethnicity adjusted Life Expectancy for 40-Year-Olds by Household Income Percentile, 2001-2014
| Source: The Association Between Income and Life Expectancy in the United States, 2001-2014 (Raj Chetty, et al.) |
The bars on the Figure 1 depict confidence intervals. As we observe, generally, men have lower life expectancy than women. Furthermore, at the bottom of the income Distribution the difference is substantial. At bottom 1% of income distribution, men tend to live 72.7 years, while women live 78.8 years on average, which is approximately 6 years more. The difference becomes less significant as we move to the higher percentiles of income distribution. At the top 1% of the income distribution, the difference of life expectancies between men and women decreases down to 1.5 years (87.3 for men and 88.9 for women).
Another finding in the paper is that relation between dollar income and life expectancy is concave. This basically means that at higher level of income, additional dollar contributes smaller increase in life expectancy. This finding matches the idea of Preston curve. The basic idea of Preston curve is that there are diminishing marginal returns to income in terms of life expectancy. [2]
Next, we move to the analysis of trends in life expectancy by income.
Figure 2. Changes in Race and Ethnicity Adjusted Life Expectancy by Income Group, 2001-2014
| Source: The Association Between Income and Life Expectancy in the United States, 2001-2014 (Raj Chetty, et al.) |
We observe that for both, men and women, the increase in life expectancy was larger for those from the higher quartile of income distribution. From 2001 to 2014, for men from the highest quartile life expectancy increased by 0.2 years while for those from the lowest quartile increased by 0.08 years. Similar trends appear in case of women. For women from the highest quartile o the income distribution life expectancy during this period increased by 0.23 years, while for women from the lowest quartile it increased only by 0.1 years.
To see how life expectancy varies across the states, let us consider following cities: New York (NY), San Francisco (CA), Dallas (TX) and Detroit (MI).
Figure 3. Race and Ethnicity Adjusted Life Expectancy by Income Ventile in New York, San Francisco, Dallas, and Detroit, 2001-2014
Figure 3. Race and Ethnicity Adjusted Life Expectancy by Income Ventile in New York, San Francisco, Dallas, and Detroit, 2001-2014
| Source: The Association Between Income and Life Expectancy in the United States, 2001-2014 (Raj Chetty, et al.) |
As we observe, there is a significant difference for men and women at the lower bottom of income distribution. For example, in Detroit, expected life expectancy for men at the bottom of the income distribution is 72.3 years, while in New York men tend to live 78.6 years on average. However, at the top of income distribution, the difference becomes less significant, implying that for a rich person it almost does not matter where they live. The life expectancy at the top of income distribution approximately converges to a single number. We observe the same trend in case of women.
Final and, in my opinion, the most important block of analysis, considers six group of factors and their correlation with life expectancy.
Figure 4. Correlations Between Life Expectancy in the Bottom (right) and Top (left) Income Quartile and Local Area Characteristics, 2001-2014.
The left graph depicts the correlation of six group of factors with bottom quartile of the income distribution, while the right graph depicts the correlation with top quartile of the income distribution (men and women are pooled). We see that not all factors are significant. The easiest way to determine whether a factor is significant or not is to have a look at their confidence intervals (horizontal bars). If it includes 0, we cannot decisively say whether the correlation if positive or negative. In this case we say that a factor is not statistically significantly associated with life expectancy.
We observe that for those people from bottom of income distribution (left graph), smoking and being obese are negatively correlated with life expectancy, while the effect of exercise rate is positive. The only significant factors from the second group is mortality rate, which is negatively correlated with life expectancy. The most important finding of this empirical study is that income segregation is positively correlated with life expectancy, which means that if a poor person lives in a highly segregated area i.e. along with rich people, he or she tends to live longer as well. Surprisingly, from income inequality and social cohesion group, only social capital is statistically significant, which is negatively correlated with life expectancy. What is even more surprising is that none of the local labor market conditions are significantly associated with life expectancy for the people in the bottom of income distribution. And finally, fraction of immigrants, median home value, local government expenditures, population density and share of college graduates are all positively correlated with expected life longevity.'
As for the people from top of income distribution, the relations change compared to people from the bottom of the distribution. For example, income segregation is not significantly associated with expected life longevity anymore, implying that if a person is rich, it does not matter whether he or she will live along poor or rich people. Furthermore, Gini index becomes also significant, which is negatively correlated with life expectancy.
From the conducted empirical analysis, we can draw four basic conclusions:
- Life expectancy increases continuously with income. There does not exist any threshold above or below which higher income was not associated with higher life expectancy.
- Inequality in life expectancy increased in recent years. From 2001 to 2014, individuals in top 5% of income distribution gained around 3 years of life expectancy while, those in bottom 5% had almost no gains.
- Life expectancy varies substantially across local areas.
- As we mentioned at the very beginning, we can deny four very popular hypothesis:
- Health and longevity are related to differences in medical care [3] - the denial of this hypothesis is easily seen from the Figure 4. We observe that for people with low income the relation between medical expenditure and life expectancy is not statistically significant.
- Gaps in longevity should be greater in areas with greater residential segregation by income [4] - we observe for low income individuals on Figure 4, as income segregation increases, poor people tend to live longer as well.
- Poor health is related to inequality or social cohesion – on Figure 4, we saw that none of the factors have significant correlation with life expectancy apart from index for social capital.
- Life expectancy is related to local labor market conditions [5] - for low income individuals we saw that none of the factors from local labor market conditions’ group are significantly associated with life expectancy.
Question for discussion: We observed that rich people tend to live longer compared to poor people, which suggests that Social Security program seems to be less redistributive than implied by its progressive benefit structure. Share your thoughts and possible solutions to this problem.
