(This essay was published in Hong Kong Economic Journal on 7 May 2014)
Last week’s essay showed that there has been no decline in the relative prospects of intergenerational mobility for people born in the 1970s and 1980s in both the US and in Hong Kong. The estimated rank-rank correlations in intergenerational income mobility in the US averaged 0.34. The estimated rank-rank correlations in intergenerational schooling mobility in Hong Kong (which is the best available measure we have) averaged 0.31 – in fact this level of intergenerational schooling mobility stretches back even to people born since 1956.
Public perception in both the US and Hong Kong that intergenerational mobility has declined for those born in the 1970s and 1980s is therefore mistaken. This misperception stems from the belief that since income inequality has grown over time, therefore intergenerational mobility must have declined. What is especially noteworthy is that the estimated rank-rank correlations of 0.31 and 0.34 imply from a statistical point of view, descendants lose inherited advantages and disadvantages after three generations. There is an old saying that wealth does not stay beyond three generations and this is corroborated by the evidence from the US and Hong Kong.
Although upward intergenerational mobility has remained constant on average in recent decades, factors within the family or community where a child is brought up may still have an influence. There are many such factors. It is not always easy to establish what causal relationships exist between these factors and intergenerational mobility. Sometimes they may correlate with each other in complex ways that are not fully understood. But some correlations are clearly stronger than others; and we shall discuss them.
Factors Affecting Upward Mobility in the US
In the US, Chetty et al. showed that upward mobility varied considerably across different geographical areas in the US. They discovered that five factors accounted for most of the differences in mobility: (1) segregation, (2) inequality, (3) quality of education, (4) social capital, and (5) family structure.
With the first factor, upward mobility was significantly lower in areas with large, heavily segregated African-American populations. However, white individuals in areas with large African-American populations also had lower rates of upward mobility. This implies that racial shares affected mobility at the community rather than individual level through the effects of segregation.
Segregation was also related to geographic sprawl. Upward mobility was higher in cities with less geographic sprawl as people spent less time in commuting to work. Geographic sprawl implies the poor usually have to live far away from where jobs are concentrated and have to endure long daily commutes. This leads to more segregation with remote residential communities populated by mostly low-income families. Strong positive role models to inspire children tend to be fewer in these isolated communities. This is collaborated by evidence of a strong negative correlation between upward mobility and income segregation.
Segregation is therefore a mechanism for how race and poverty generate community-level effects on upward mobility in the US. These results imply that cities with more integrated neighborhoods in racial and income terms, with shorter commuting times between work and residence, will be positively correlated with upward mobility. Hong Kong today is likely to be much more segregated than say forty years ago as a growing fraction of our population now lives in the New Territories even though jobs are still concentrated in the old urban areas.
The second factor affecting intergenerational mobility is inequality. Geographic regions in the US with higher income inequality were found to have less upward mobility, a phenomenon consistent with the “Great Gatsby curve” documented across countries.
In contrast, top 1% income shares are not highly correlated with intergenerational mobility both across geographic areas within the US and across countries. Although one cannot draw definitive conclusions from such correlations, they suggest that the factors that erode middle class incomes hamper intergenerational mobility more than the factors that lead to income growth in the upper tail of the income distribution.
The third factor that was identified was the effect of the quality of the kindergarten and school system on upward mobility. Areas with higher test scores, lower dropout rates, and smaller class sizes had higher rates of upward mobility. In addition, areas with higher local tax rates, which are predominantly used to finance public schools, had higher rates of mobility.
Interestingly, the spatial patterns of intergenerational income mobility were also very similar to patterns across geographic areas of how parental income affects college attendance and teenage birth rates. The fact that much of the spatial variation in children’s outcomes emerged before they entered the labor market suggested that the differences in mobility were driven by factors that affected children while they were growing up.
The fourth factor was based on social capital indices, which are proxies for the strength of social networks and community involvement in an area. These were very strongly correlated with mobility. For instance, high upward mobility areas tended to have higher fractions of religious individuals and greater participation in local civic organizations.
The fifth and strongest predictor of upward mobility was family structure, for example, measures such as the fraction of single parents in an area. The correlation of upward mobility with children raised by single mothers in the community was -0.764 and this alone accounted for almost all the variations in mobility from area to area geographically. Moreover, children often experienced higher rates of upward mobility if they grow up in communities with fewer single parents.
These results suggest that understanding why the US family is breaking down and how to tackle that challenge would probably be a very productive way of improving the upward mobility of children from poor families. Indeed we know that in the US, as well as in Hong Kong, divorce rates are higher among low-income families.
The findings highlight the importance of community effects, showing that the impacts of such things as concentrated poverty, residential segregation, social capital, and local school quality, on intergenerational mobility. Of course, these correlations alone do not shed light on whether the differences in outcomes across areas are due to the causal effect of communities or differences in the characteristics of people living in those communities.
Factors Affecting Upward Mobility in Hong Kong
What factors explain intergenerational mobility in Hong Kong? Using population census data I was able to piece together a number of significant drivers, although it must be noted that this work was limited by the amount of information available in the data sets I worked with. So what I describe below should not be construed as being complete. A richer quantitative analysis would involve conducting more detailed surveys and linking them with administrative data. For the purposes of this essay, I shall focus on the level of schooling as the indicator of intergenerational mobility.
I examine the factors that account for a person’s schooling after controlling for parent’s schooling (both measures of schooling are in terms of percentile ranks estimated within their own birth cohorts). The factors were examined included: (1) gender, (2) whether a person was a recent immigrant in the previous 5 years, (3) whether any parent was a recent immigrant in the previous 5 years, (4) whether a person had a single parent, either father or mother, (5) whether the parent or parents owned their home (either private or subsidized), and (6) whether the parent or parents lived in a public rental housing estate. I performed separate analyses using first father’s schooling and subsequently mother’s schooling as measures of parent’s schooling.
The estimated effects are presented in Table 1, which controls for father’s schooling,
and Table 2, which controls for mother’s schooling.
The estimated effect of father or mother’s schooling on a person’s schooling reveals a common pattern over time. In particular, intergenerational schooling mobility for the cohort born in 1976-81 has not declined at all (and may even be slightly improving). Schooling mobility opportunities for the cohort born in 1961-71 were particularly good, but this was mainly the result of improving schooling opportunities because of the expansion of tertiary education in the 1980s and the emigration of the middle class in the decade leading up to 1997.