Are You Better Off Than Your Parents?

Part I of a two part series looking at intergenerational mobility. Part I focuses on national and global trends. Part II will focus on implications of the variations in intergenerational mobility within the US.

Summary:

• Evidence suggests that relative mobility has been stable in the US for children born from 1971 to 1993. While rank mobility is stable, the upward mobility income dollar payoff (and downward mobility penalty) increased because of growing income inequality over this period.

• Although relative mobility appears stable, a 2016 research paper documents a decline in absolute mobility since 1940. This result was widely covered by the press and was broadly interpreted to imply that children are “less well off” than their parents and that the American Dream (of a brighter future for succeeding generations) is dead.

• It is easy to show that aggregate US real income has continued to increase, but more slowly than GDP and at a declining rate.

• Decreasing absolute mobility is a function of both income growth and how income growth is distributed (i.e., income inequality). Decreasing income growth is related to declining GDP growth (which is not well understood), and direct and indirect increases in the supply of labor. Increasing income inequality is partly attributable to increasing returns to skill.

• While one can point to ways in which younger generations are clearly better off than their parents, it does appear that decreasing income growth has lead them to opt for lower rates of marriage, fertility, and household formation. 

• Declining US absolute mobility occurred as the US labor force became more inclusive, technology advanced, and global absolute mobility increased. These closely related trends raise interesting ethical questions for those who wish they could turn the clock back to a time of higher US absolute mobility.

Stable Relative Mobility

There are two types of intergenerational mobility: relative and absolute. Relative mobility is based on rank -- e.g., the odds that a child of parents in the bottom income quartile will make it into the top income quartile. Absolute mobility is measured in dollars and measures the likelihood that children will make more than their parents (adjusted for inflation).

Figure 1 illustrates the findings of the Chetty, Hendren, Kline, Saez, and Turner 2014 paper which documents stable relative mobility for children born from 1971 to 1993.

Figure 1: US Relative Mobility

In a society in which all children received identical upbringings and had equal access to social networks, the expected probability that a child rises to the top quintile (or any quintile) for all parent income quintiles would be equal at 20%. Thus the roughly 9% probability for children of bottom quintile parents versus the 32% probability for top quintile parents indicates strong variations in educational preparation and access, and in the relative power of social networks.

The stability of relative mobility may seem surprising given the increasing prevalence of college education, however the income of college graduates vary greatly based on the college they attend. Figure 2 from the 2014 Chetty et. al. relative mobility paper shows that the quality of the college students attend (and consequently their future income) is highly correlated with their parents’ income. Thus the mere presence of education has not significantly altered relative income ranks.

Figure 2: US College Quality vs. Parent Income Rank

Declining Absolute Mobility

Figure 3 is taken from The Fading American Dream: Trends In Absolute Income Mobility Since 1940 by Chetty, Grusky, Hell, Hendren, Manduca, and Narang (@2016).  It shows the probability that children's’ real income would exceed that of their parents (both measured when they are in their early 30s) based on their birth cohort and their parents’ relative income rank. If one’s parents make nothing, clearly the probability of earning more than they did is quite high. Similarly, if your father is Warren Buffett, the odds of outearning him are quite small. However one can see the probability of outearning one’s parents has declined uniformly for successive birth cohorts.

Figure 3: Absolute Mobility/Parent Income Curves By Birth Cohort

Figure 4 shows the steady decline of mean absolute mobility since 1940. It turns out that the true amount of decline is significantly dependent on the choices made by the researchers. Chetty et. al. discuss the impact of different assumptions in their paper.[1] Choices include which deflator is used to derive real income, how one chooses to account for benefits that are not captured in income, and how one adjusts for decreasing family size. Depending on how one compounds these choices, the true decline of mean absolute mobility might be from 90% to 72%, instead of 90% to 50% as reported by Chetty et. al.[2] But the qualitative observation of a declining trend remains.

Figure 4: Declining Mean Absolute Mobility By Birth Cohort

Declining Absolute Mobility, GDP Growth, Productivity, And Labor Supply

Figure 5 shows trailing 30 year real annualized growth rates in GDP per capita and in compensation/hour (30 years is the relevant period because Chetty et. al. measure children and parent income when they are 30). In aggregate, there is no question that succeeding generations are making more than prior ones because real income growth is positive. However, equally noticeable is the significant decline in that growth, as well as the sharper decline in compensation/hour than in GDP/capita.

Figure 5: Real Growth In US GDP/Capita And Compensation/Hour

As a stylized fact, if growth in GDP/capita were 0% and compensation/hour was a constant percentage of GDP/capita -- the average odds of any one worker outearning their parents would be 50%. Under the same circumstances, if GDP/capita growth was very high -- the odds of a worker outearning their parents would also be quite high. This is a general statement that absolute mobility increases as GDP/capita growth increases, as well as an observation that absolute mobility depends on the relationships between GDP and compensation. Thus as Chetty et. al. observe, the decline in the growth of GDP/capita is part of the story. There is no broad consensus as to why GDP growth has slowed.

But why did compensation/hour fall faster than GDP/capita? Absolute mobility is more directly tied to compensation/hour than it is to GDP and one can see the similar timing of the sharp drop in compensation/hour and the drop in mean absolute mobility shown in Figure 4. Figure 6 shows that part of the decline in compensation/hour was due to falling productivity. As with the related drop in GDP growth, the drop in output/hour is also not well understood and was observed in other G7 countries. If the growth rate of what workers are producing per hour declined, however, it is not surprising that the growth in their compensation/hour also slowed.

Figure 6: Non-Farm Business Sector Output/Hour Vs. Compensation/Hour

Nevertheless, the drop in US compensation per hour growth is still greater than the drop in output/hour. This greater decline can be partly attributed to the growth and changing nature of the labor supply.

Figure 7 depicts the strong growth in the labor force until about 1995, coinciding with the period of the greatest declines in absolute mobility and in compensation/hour. The net rise in labor came from the young and from women.

Figure 7: US Labor Force Participation

In addition to the downward wage pressure from a simple rise and deepening in the supply of labor, it is depressingly quite likely average wages were suppressed initially by gender-based pay discrimination.[2] Finally, compensation growth was also held down by the significant increase in “virtual” labor from increasing automation and global trade.

Declining Absolute Mobility And Income Inequality

Actual growth in compensation/hour has fallen to 1% -- not far from 0%. I stated earlier that average absolute mobility would be 50% if compensation growth was 0%. This is true regardless of how incomes are distributed. However, the rate at which absolute mobility improves with compensation growth is highly dependent on how income is distributed.

Figure 8 is a stylized illustration using lognormal distributions to approximate US income distribution. The two colored lines each represent constant, but different levels of income inequality. The bottom gold line represents dramatically more income inequality than the top green line. The Figure illustrates how much more quickly absolute mobility grows as compensation growth increases when income is more equally distributed.[4]

Figure 8: Theoretical Absolute Mobility Versus Growth And Income Distribution Shape

When absolute mobility was greater than 90% for the 1940s birth cohort with compensation/hour growth of about 2.4% (top black point), the distribution of income was broadly consistent with the less positively skewed green top line shown in Figure 8.

One can see from Figure 8 that how large a role rising income inequality played in declining absolute mobility will depend on what one believes the true absolute mobility number is now. If you believe the true absolute mobility number now is closer to 72% (based on different assumptions on how to deflate earnings, adjust for family size, etc. than Chetty et. al. use to arrive at their reported 50% number) shown as the topmost red point, than increasing income inequality played a small role in decreasing absolute mobility. In other words, you could get there just due to the drop in compensation/hour growth rates to 1% along the green top line with constant inequality. This particular path seems unlikely as we know income inequality increased over this time.

If absolute mobility now is truly about 50% (bottom red point) as Chetty et. al. report, increasing income inequality played a much larger role. Chetty et. al. suggest about 71% of the decline in absolute mobility to their reported value is due to increased income inequality and the balance is due to declining GDP growth. They seem to underestimate the role of declining compensation/hour, however, as their counterfactual calculation implicitly assumes the relationship between GDP and compensation remained constant. In fact it was not, as described above. If one compensated for the greater drop in compensation/hour growth, the non-distributional GDP-compensation changes might account for just under half of the drop in absolute mobility to their 50% value.[5]

Just as the 72% number seems too high because we know income inequality increased, the reported 50% value also seems too low because it is difficult to achieve 50% mobility without 0% growth (or negative income growth for some meaningful portion of the population). I think a fair characterization is that absolute mobility would have fallen anyway due to falling compensation/hour, but the drop was somewhat magnified by changes in the distribution of income growth (see below).

In thinking about the relationship between absolute mobility, income growth, and income inequality, it is worth noting that China has increasing income inequality and increasing absolute income mobility -- because they also have income growth. Figure 8 makes clear this is also a feasible path for any move to the upper right with a slope that is flatter than the slope of the green line.

Finally it is important to observe that increasing income inequality in the US was caused by increasing returns to skill (associated generally with education and increasing technological leverage), and decreasing returns to “medium skilled” jobs due to greater competition from global trade and automation. This is illustrated in Figure 9 taken from Autor and Dorn.

Figure 9: Changes In Employment Share And Compensation By “Skill” Required

So increasing US income inequality was not a case where the top income quintile simply “took” compensation away from lower quintiles. As I discuss below, it is not clear that increasing returns to skill (and decreasing returns to medium skill) could have been avoided without further impairing economic growth and/or raising the cost of living for all.    

Declining Absolute Mobility, Consumption Tradeoffs, And Fertility

Income is half of the answer to the question of whether you are better off than your parents were when you were both 30 years of age. Consumption is the other half of the answer. While your consumption obviously comes from your income, the price of (some) goods may be falling faster than your income growth is slowing. This dynamic often arises from global trade which suppresses income for some workers, but also makes goods available more cheaply for all consumers. One sees this for consumer durables, for example, as shown in Table 1 from Horowitz.[6]

Table 1: Labor Hours Required To Purchase Selected Consumer Durables

There are other ways in which life has unambiguously improved for younger generations. Life expectancy, for example, increased for all income quintiles.[7] However this result was likely supported by the subsidization of health care (e.g., Medicare, Medicaid and Emergency Room care). Slowing income growth did not help this outcome.

While the cost of durable goods and apparel have fallen, other significant costs have grown faster than income growth. The largest component of CPI is housing at 43% of total consumption. Housing costs have increased by about 3% over the last 30 years -- well above the growth in compensation/hour.

One way to preserve purchasing power with lower income growth is to avoid large costs, particularly those which are growing faster than your income. So for example, one would avoid paying for housing, choose not to marry (or postponing marriage to) someone with much lower income, and/or electing not to have children. While partly attributable to the Great Recession of 2008, it is true that unprecedented numbers of children chose to continue living with their parents, remained single or deferred marriage to a later age, and chose not to have (any or as many) children. This behavior was also seen in the 1930’s when income growth was low and many believed America would never recover from the Great Depression.

In a very broad context, declining demand for labor as reflected in diminishing wage growth appears to be causing recent birth cohorts to elect to provide lower amounts of future workers (i.e., children).

The Ethics Of Turning Back The Clock To A Time Of Higher US Absolute Mobility

Earlier I indicated that increased and more inclusive US labor force participation was one factor dampening income growth, and thus US absolute mobility. Although it played a role in holding down some US wage growth, strong technological progress also occurred. Beyond US borders, declining US absolute mobility occurred as global absolute mobility increased.

Figure 10 is excerpted from a Resolution Foundation paper and depicts varying income experiences by country. It is a busy chart: each dot represents an income decile for a color coded country or country set. The size of the dot represents the size of the population. Several things are observable as the authors note. First: A significant portion of the the US population comprise the highest incomes in the world. Secondly, income growth has been quite positive (and by implication, absolute mobility) for more than a billion people who earn less than the bottom US income decile.

Many studies have ascribed some part of declining US absolute mobility and worsening income inequality to increasing global trade. What are the ethics behind policies that would seek to reverse the absolute mobility development within the US by erecting barriers to trade?

Figure 10: Income Growth By Decile For Selected Country And Country Sets

Niño-Zarazúa, Roope, Tarp in their paper Income Inequality In A Globalizing World make an interesting point that takes us back to where this essay started with a discussion of relative mobility:

“Over the past 40 years, over one billion people around the world have been lifted out of poverty, driven largely by substantial growth in income in developing countries. While this growth has been accompanied by a striking rise in absolute inequality, it has also improved the lives of hundreds of millions of people. It is difficult to imagine how in practice such growth, and the associated poverty reduction, could have occurred without an increase in absolute inequality. There would be huge implications for the fight against global poverty if attempts were made to halt economic growth in order to appease absolute inequality. Instead, the policy emphasis should be on creating more inclusive growth with falling ‘relative’ inequality – these two goals are complementary.”

While they are talking about developing countries, their logic extends to the US and suggest the best set of policies to restore US income growth would be those which focus on increasing the 10% likelihood that children of poor parents can rise to the top (see Figure 1). Everyone wants higher GDP growth and the high absolute mobility that arises from it. However, GDP growth -- and thus absolute mobility -- cannot be increased by fiat.[8] But maximizing the utilization of potential talent can increase the likelihood of economic growth.

As Scott Winship has argued, ultimately the question of how we can raise the likelihood that poor children who are bright can rise to the top? is more important than the proportion of people making more than their parents. To pursue this question, Part II of this essay will examine 8 research papers which Chetty and his co-researchers have made available on the Equality of Opportunity website. These interesting papers document factors that are correlated to variations in upward mobility within the US.

[1] Scott Winship has also written extensively about the assumptions behind different absolute mobility numbers and earlier raised many of the same issues that I discuss in writing he did in late 2016. 
[2] See the slide 58 of Chetty, Hendren, Kline, Saez, and Turner's presentation slides for the 72% result.
[3] One can see this is one looks at unadjusted and adjusted gender normalized compensation (also normalized for age, education and occupation).

[4] For this hypothetical, I use the transition matrix from Table 1 of Where is the Land of Opportunity?The Geography of Intergenerational Mobility in the United States by Chetty, Hendren, Kline, and Saez @2014 that approximately correlates with the relative rank mobility shown in Figure 1. The transition matrix doesn't matter very much in this context: one could assume the child's income has no relation to the parents' and obtain nearly identical results.
[5] See slide 55 of Chetty, Hendren, Kline, Saez, and Turner's presentation slides. An extra 0.6% of annual compensation growth would add about 6% to the 62% absolute mobility result for their GDP growth counterfactual.

[6] One can also see this by plotting CPI - Durables against other CPI indices on FRED.

[7] For women more than men and not for men or women in the bottom 5% of income. See “The Association between Income and Life Expectancy in the United States, 2001 - 2014” by Chetty, Stepner, Abraham, Lin, Scuderi, Turner, Bergeron, and Cutler

[8] One could imagine a host of policies that tried to increase absolute mobility by redistributing income, but as Niño-Zarazúa, Roope, Tarp suggest, it is hard to imagine how simple redistribution on its own could be done without adverse effects on economic growth.

Transparent and reproducible: I generated Figures 5 and 6 from FRED data. I also generated Figure 8 using lognormal sampling functions within R. These figures can be generated by using the free, publicly-available R program and the R code available in “AbsoluteRelativeMobility.r" on github.