Introduction: Inequality of Economic Opportunity Katharine Bradbury and R Obert K

Introduction: Inequality of Economic Opportunity Katharine Bradbury and R obert K. T riest

In the United States, inequality in the distribu- part, on the underlying causal mechanisms. tion of economic rewards, such as income and Inequality of economic opportunity, a contribut- wealth, has widened greatly in recent decades. ing factor to overall economic inequality, re- This has led to a spirited debate regarding the flects inequality in an individual’s innate char- causes and consequences of increased eco- acteristics and the circumstances of birth and nomic inequality. A fundamental area of con- early environment. It is generally regarded tention is whether the increase is evidence of much more negatively than inequality of eco- underlying economic problems that need cor- nomic outcomes, because it is associated with rection. Increased inequality may result from characteristics and circumstances that are be- increased risk taking and entrepreneurship in yond an individual’s ability to control. Al- an environment of rapid technological change, though a degree of inequality in economic out- with some entrepreneurs producing better, or comes can have the positive effect of providing just luckier, innovations than others, and reap- incentives for entrepreneurship and work ef- ing greater rewards. It may also result from in- fort, inequality of economic opportunity can- creased disparities in work effort, with more not be defended on these grounds. And few, if industrious individuals earning higher in- any, would claim that inequality of opportunity comes as a result of their greater effort. In both is desirable on moral grounds. these cases, one could argue convincingly that The articles in this issue examine the causes the increase in inequality is justified and that of inequality of economic opportunity and an- no remedial changes in public policy are alyze the potential for public policy to reduce needed. On the other hand, if the increase in it. This essay provides an overview of the topic inequality results mostly from factors largely and aims to integrate the analyses provided in beyond the ability of individuals to control or the other articles. Several major themes counteract, then a strong case can be made for emerge from our reading of the research on a public policy response. inequality of opportunity: In other words, the extent to which economic inequality is viewed as an appropriate • Inequality of economic outcomes is both a matter for public policy concern depends, in consequence and a cause of inequality of

Katharine Bradbury is senior economist and policy advisor at the Federal Reserve Bank of Boston. Robert K. Triest is vice president and economist at the Federal Reserve Bank of Boston.

The views expressed in this article are those of the authors and do not necessarily represent the positions of the Federal Reserve Bank of Boston or the Federal Reserve System. The authors thank Suzanne Lorent and two anonymous reviewers for very helpful comments and Stephanie Bonds and Sam Richardson for expert research assistance. Direct correspondence to: Katharine Bradbury at [email protected], Research Depart- ment, Federal Reserve Bank of Boston, 600 Atlantic Ave., Boston, MA 02210; and Robert K. Triest at robert [email protected], Research Department, Federal Reserve Bank of Boston, 600 Atlantic Ave., Boston, MA 02210. 2 opportunity, mobility, and increased inequality

opportunity. Children growing up in low- least partly for early-life disadvantages. income families lack many of the develop- Other policies may improve the quality of mental and educational advantages enjoyed schools and neighborhood environments by children growing up in more affluent encountered by disadvantaged children, families. The barriers to opportunity that limiting the extent to which their environ- disadvantaged children face are often ment hinders their educational attainment amplified by the effects of other barriers to and early labor market outcomes. Labor schooling and labor market success later in market policies that offer opportunities for life, leading to a perpetuation of inequality skill development and career development of economic outcomes across generations. for workers with relatively modest formal • Intergenerational economic mobility, the credentials may also attenuate the negative ability of children to enjoy higher eco- effects of earlier disadvantages. nomic status than that of their parents, is • Reducing inequality of opportunity need closely linked to equality of opportunity. In- not come at the expense of economic growth tergenerational mobility appears to be rela- or efficiency. Barriers to economic opportu- tively low in the United States and shows nity generally interfere with the efficient -op no sign of improvement in recent history. eration of the economy and result in subop- Empirical studies point to a substantial risk timal development of human talent and of continued stagnation of intergenera- resources. Empirical evidence shows a pos- tional mobility unless public policy changes itive association between equality of oppor- in ways that break down barriers to eco- tunity and aggregate economic growth. nomic opportunity. Rates of intergenerational mobility vary substantially by the Concepts and Measures geographic location where one grows up, of Inequality of Economic suggesting that policies that change the Opportunity economic and social environment may be Although the term inequality of economic op- effective in promoting economic opportu- portunity is broadly understood to refer to in- nity and mobility. equality associated with an individual’s cir- • Barriers to economic opportunity occur cumstances at birth and during childhood, the throughout life, but many of the most dam- way the term is operationalized varies across aging obstacles to opportunity appear at researchers and commentators, and a wide very young ages. Some children are born range of measures are used to quantify its mag- into environments of economic deprivation nitude. and risk falling behind before they even en- To measure inequality of opportunity, and ter preschool or kindergarten. to develop polices to promote it, a common understanding of the term, at least conceptu- • Additional barriers appear as children age, ally, is critical. A key difficulty in arriving at a often amplifying the effects of disadvan- workable definition lies in specifying the cir- tages first encountered at very young ages. cumstances and outcomes for which individu- Disadvantaged children tend to live in areas als are not considered responsible. Equality of with relatively low-quality public schools opportunity intuitively requires that circum- and face obstacles to attaining postsecond- stances beyond one’s control, such as parents’ ary degrees. Barriers to opportunity in the education, should not affect one’s economic labor market further amplify the effects of outcomes, but allows other factors that are un- unequal access to opportunity during child- der one’s control, such as one’s own educa- hood and adolescence. tional attainment, to affect one’s economic • Public policy has the potential to reduce outcomes. But parents’ educational attain- substantially the effects of barriers to op- ment has been found to influence, although portunity. Some policies, such as targeted, not completely determine, the educational at- high-quality preschool, can compensate at tainment of their children. So how can we sep-

: introduction 3 arate the effects of circumstances beyond one’s not on one’s circumstances but instead on po- control from those that are influenced by these sition in the distribution of effort. This cap- circumstances but still at least partly under tures the notion that effort is influenced by cir- one’s control? cumstances, but still under individuals’ control. In our example, when parental educa- Circumstances and “Effort” tion is the only circumstance and income is John Roemer (1993, 1998) has been especially the outcome of interest, the income of some- influential among economic researchers and one at the 75th percentile of the effort distribu- has spawned a growing research literature on tion among people whose parents did not grad- the measurement of inequality of opportunity uate from high school would equal that of (for a clear exposition, see also Roemer and someone at the 75th percentile of the effort Trannoy, forthcoming). Roemer’s approach distribution among people whose parents both separates the population of interest into types graduated from four-year colleges.1 The re- defined by circumstances beyond one’s con- sponsible actions taken by someone at the 75th trol, such as parental education, parental eco- percentile of the effort distribution of these nomic status, race, ethnicity, gender, the two types would generally differ. Consider one neighborhood or labor market where one grew specific type of responsible action, investing up, and one’s family structure while growing in one’s schooling. Educational attainment of up. Within the group of people who experience people whose parents did not graduate from a specific set of circumstances, there will be a high school tends to be lower than educational distribution of an outcome variable of interest, attainment of people whose parents graduated such as income in adulthood. The distribution from college. This difference in the effort dis- of the outcome for a given circumstance type tribution across circumstance types is taken to is assumed to depend on responsible actions be due to the influence of circumstances on (such as the effort exerted in the labor market) effort, and full equality of opportunity in Ro- taken by individuals of that type. For example, emer’s approach requires eliminating (or com- suppose that the only circumstance variable we pensating for) differences in outcomes associ- choose to treat as not under one’s control is ated with differences in the distribution of parental education. There will be some distri- effort across circumstance types. bution of income among people whose parents Marc Fleurbaey and Vito Peragine (2013) dis- did not graduate from high school as well as tinguish this approach, which they call the ex other distributions for other circumstance post perspective, from a somewhat different types, such as those whose parents both re- approach to defining and measuring equality ceived four-y ear college degrees. The distribu- of opportunity, which they call the ex ante per- tion of income within a circumstance type is spective. In the ex ante perspective, equality of taken to be driven by differences in the respon- outcomes is not needed between people of dif- sible actions (which we often refer to as effort) ferent circumstance types at the same point in taken by people of that type, whereas differ- the effort distribution; instead, it requires only ences in the distribution of income between that the average outcomes of people of differ- types are taken to be the result of differences ent circumstance types be equated. In our ex- in opportunities available to people of differ- ample, equality of opportunity would require ent types. that the average income of people whose par- In Roemer’s approach, complete equality of ents did not complete high school would be opportunity requires equality of outcomes for equal to the average income of people whose people of a given percentile rank in the effort parents both graduated from four-y ear col- distributions of each type. In other words, leges. This is an ex ante perspective on equality equality of opportunity is attained when the of opportunity in the sense that it is based on chance of achieving a given outcome depends information available before we know where

1. This would hold true for any given percentile of the income distribution comparing any two circumstance types.

: 4 opportunity, mobility, and increased inequality people stand in the effort distribution for their and Paolo Brunori, Francisco Ferreira, and Vito circumstance type. In contrast, Roemer’s ap- Peragine (2013) review empirical applications proach is an ex post perspective in the sense of the approaches outlined, as well as some that equality of opportunity requires that out- other methods. Because of data limitations, comes for people of different circumstance the set of circumstance variables specified in types be equated for people with the same rank empirical studies is much more limited than in their effort distributions, which means that the complete set discussed previously (and one we need to know where people stand in the ef- could argue convincingly that even that set is fort distribution. incomplete). In practice, only a small subset of Whether one adopts the ex ante or the ex the full range of potential circumstance vari- post approach to defining and measuring in- ables is observed and specified in defining cir- equality of opportunity, specifying what range cumstance types. As a result, the portion of of factors are to be included in the set of cir- overall inequality of outcomes attributed to in- cumstances over which one has no control is equality of opportunity is undoubtedly an un- critical. Parental income is an obvious candi- derestimate of what a more complete set of date and is often used in empirical applica- circumstance variables would yield. Of the tions. This variable is a proxy for the differing Brunori, Ferreira, and Peragrine country esti- opportunities available to children growing mates, the share of total income inequality at- up in high-income families compared with tributed to inequality of opportunity ranges the opportunities of children growing up in from 2 percent to 34 percent. low- income families. Parental education is another relevant circumstance. It is closely re- Intergenerational Mobility and lated to parental income, but it also captures Equality of Opportunity how more educated parents may promote ed- Much of the empirical research related to in- ucational attainment and economic success equality of opportunity focuses on measures in their offspring. Other family characteris- of intergenerational mobility rather than using tics, such as the number of parents living with the approaches described. Intergenerational the child, also play an important role in chil- income mobility is closely related to measures dren’s development and the economic oppor- of inequality of opportunity in which parental tunities open to them later in life. The pos- income is the only variable used to classify peo- sible effect of discrimination on opportunities ple into circumstance types and income is the suggests that race, ethnicity, and gender outcome variable of interest. High intergener- should be included in defining the set of cir- ational income mobility implies that one’s cumstance types. Childhood geographic loca- childhood family income plays only a modest tion is another important circumstance that role in determining one’s income as an adult. affects economic opportunity in several ways. One would intuitively expect an increase in in- City and state play a large role in determining tergenerational income mobility to be associ- the availability and quality of preschool pro- ated with a decrease in inequality of opportu- grams open to low-income families and deter- nity. However, although intergenerational mine the quality of public K–12 education; the income mobility is closely related to inequality local labor market also plays an important of opportunity, the two concepts differ in im- role in creating economic opportunities and portant respects. providing incentives for investment in educa- One distinction is that measures of inter- tion and training. At a finer-grained level of generational income mobility typically do not geographic demarcation, one’s neighborhood capture the relationship between inequality in while growing up affects factors such as expo- parental income and the extent of inequality sure to crime and drugs, peer and role mod- in adult children’s income. However, this rela- els, and often the specific schools that one at- tionship is central to the concept of inequality tends. of opportunity. If variance in parental income Xavier Ramos and Dirk Van de Gaer (2012) is minimal, then income inequality associated

: introduction 5 with parental income types (inequality of op- tween types, so measured inequality of oppor- portunity) will also be minimal, even when in- tunity will decrease. tergenerational income mobility is low. Con- Another potential distinction between inversely, a high degree of parental income tergenerational income mobility and equality inequality may be a source of a substantial in- of opportunity concerns whether parental inequality of opportunity even if intergenera- come is a comprehensive summary measure tional income mobility is reasonably high. of circumstances, and also whether income is A second potential distinction between in- the outcome measure of greatest interest. In- tergenerational income mobility and equality come certainly may be used to convey advan- of opportunity revolves around whether we are tages to one’s children, but some parents considering absolute or relative income mobil- choose to forgo earnings opportunities and ity. Absolute income mobility refers to the instead use more of their time for parenting relationship between the amount of income activities. Such choices are facilitated by the received by someone as an adult and the availability of financial wealth or the presence inflation-adjusted income received by his or of a second parent. Thus, a case can be made her parents a generation earlier. In contrast, that parental income must be supplemented relative income mobility refers to the relation- by other variables, including parental educa- ship between a person’s position in the in- tion and family structure, to capture ade- come distribution and the position of his or quately the circumstances that determine the her parents in the income distribution a gen- opportunities open to children. One can make eration earlier. It is possible for a high degree a somewhat similar case for why income is not of absolute income mobility to be present even the only outcome variable of interest, and for when income rankings change little across the analysis of broader measures of well- generations. Consider an extreme hypotheti- being. cal case in which every adult’s income is ex- Isabel Sawhill and Richard Reeves (in this actly double that of his or her parents. In this issue) discuss the distinction between relative example, absolute mobility is high, but every- and absolute income mobility, and reach the one occupies the same position in the income same conclusion: relative intergenerational distribution as their parents a generation ear- mobility is more closely related to equality of lier. If we take parental income as an exoge- opportunity than absolute intergenerational nous circumstance, then all of the income in- mobility is. They also discuss the question of equality of the children’s generation represents “mobility of what?” and note that it is instruc- inequality of opportunity, because it is fully tive to examine mobility across a broad range determined by parental income. In general, no of outcomes beyond just income, including consistent theoretical relationship exists be- education, well-being, and educational status. tween absolute income mobility and equality However, they also note that income is a pow- of opportunity. erful indicator of other outcomes and of spe- In contrast, an increase in relative income cial interest. mobility will generally be associated with an In recent research, Brunori and his col- increase in equality of opportunity. As relative leagues (2013) compute an ex ante measure of income mobility increases, parental income inequality of opportunity and a measure of in- becomes a less important determinant of one’s tergenerational earnings mobility (the inter- position in the income distribution. Suppose generational earnings elasticity) for a large that members of the children’s generation are number of countries and find a robust cross- classified into types according to their parents’ country negative correlation between intergen- income (treating parental income as the sole erational earnings mobility and the share of circumstance variable). As relative income mo- overall income inequality that is attributed to bility increases, more of the income inequality inequality of opportunity. Thus, despite impor- in the children’s generation income will be tant conceptual differences between the two within circumstance types rather than be- concepts, intergenerational mobility has a

: 6 opportunity, mobility, and increased inequality strong empirical association with equality of tional income persistence (and so less inter- opportunity. generational income mobility) than the United States. Key Facts: Intergenerational Mobility in the United States U.S. Trends Sawhill and Reeves provide in this issue an The trend in intergenerational income mobil- overview of current patterns of relative inter- ity in the United States over recent decades generational income mobility, and Timothy and especially its implications for changes in Smeeding, also in this issue, examines which equality of opportunity are somewhat contro- subgroups of the population appear to face versial. The data requirements for analyzing particularly severe obstacles to economic ad- changes over time in intergenerational mobil- vancement. (For comprehensive reviews of the ity are daunting, resulting in a paucity of stud- research literature on intergenerational mobil- ies on this topic. Using census data on adults ity, see Solon 1999; Black and Devereux 2011; matched to synthetic parents in the previous and Jäntti and Jenkins 2015.) generation, Daniel Aaronson and Bhashkar Drawing on both their research and that of Mazumder (2008) find evidence that intergen- others, Sawhill and Reeves document that in erational income mobility decreased in the the United States the position of one’s parents late twentieth century. Deirdre Bloome and in the income distribution strongly predicts Bruce Western (2011) document a similar de- one’s own place in the income distribution in cline in mobility based on data from two co- adulthood. This is especially true for those horts in the National Longitudinal Survey, and born to families at the top and bottom of the David Levine and Mazumder (2007) use data income distribution: one set of estimates sug- on brothers to come to a similar conclusion. gests that 60 percent of those born into a In contrast, both Chul-In Lee and Gary Solon bottom-quintile family will themselves be in (2009) and Raj Chetty, Nathaniel Hendren, Pat- the bottom two quintiles of the income distri- rick Kline, Emmanuel Saez, and Nicholas bution at age forty, and that 56 percent of those Turner (2014), using two different sources of born into a top-quintile family will be in the intergenerational longitudinal data, find little top two quintiles of the distribution at age evidence of changes over time in intergenera- forty. Other researchers have estimated quali- tional income mobility, at least for recent adult tatively similar relationships. For example, us- cohorts. ing a different data source, Susan Urahn and As discussed earlier, changes in intergen- her colleagues (2012) find that 70 percent of erational mobility may be only loosely con- those growing up in families in the bottom two nected to changes in inequality of opportunity. quintiles of the income distribution will also If we consider parental income a circumstance be in the bottom two quintiles as adults. variable, then any measure of inequality of op- Contrary to popular impression, measured portunity would increase as inequality of par- intergenerational income mobility in the ents’ income increases. Increasing income United States tends to be less than that in inequality and largely unchanged intergenera- many other advanced economies. Jo Blanden tional income mobility would combine to pro- (2013) provides a recent compilation of esti- duce a trend of increasing inequality of oppor- mates of intergenerational mobility across tunity. This appears to have been the pattern countries. Although sampling variance makes in late twentieth-century America. it difficult to draw definitive conclusions, point estimates of the intergenerational income elas- Population Subgroups ticity (one measure of intergenerational in- Particular subgroups of the population appear come persistence) are greater for the United to be especially vulnerable to lack of mobility. States than for Australia and several countries Examination of differences in mobility across in western Europe. Of the twelve countries for subgroups is of interest in its own right and which Blanden is able to identify comparable can also give us insight into the mechanisms estimates, only Brazil has greater intergenera- that may underlie barriers to opportunity.

: introduction 7

Smeeding reports in this issue that among peo- raises the question of the impact inequality of ple born into bottom-quintile families, blacks, outcomes has on inequality of opportunity. In children of never-married mothers, and chil- simple terms, the outcomes-affect-opportunity dren of parents lacking a high school diploma hypothesis is that as the overall distribution of are especially likely to be in the bottom quin- outcomes becomes more unequal it reduces tile of the income distribution in adulthood. low- income children’s access to education and Mazumder (2011) cites earlier studies and of- to other opportunities to accumulate human fers as well new evidence that blacks experi- capital and move up the income ladder, and it ence both lower rates of upward mobility and increases high-income children’s access to en- higher rates of downward mobility than whites richment beyond schooling, in turn enhancing do. Chetty, Hendren, Kline, and Saez (2014a) the ability of advantaged children to stay at the present estimates showing that intergenera- top. Both these changes tie individuals’ eco- tional mobility varies greatly across geographic nomic prospects more tightly to their parents’ areas of the United States. The variability in economic success. rates of intergenerational mobility raises the Several research papers explore this ques- question of what underlies differences in mo- tion, focusing on how unequal outcomes lead bility. To what extent are characteristics such to unequal opportunity or how unequal out- as race and geographic location causal factors comes reduce mobility, which is interpreted as affecting mobility, and to what extent are they indicating unequal opportunity. Miles Corak instead proxies for other less easily measured explores a variety of mechanisms that link in- characteristics, such as limited access to qual- come inequality, equality of opportunity, and ity schooling, that are the more direct causal intergenerational mobility.2 Introducing his factors? We return to this question in our dis- analysis, he notes that “an emerging body of cussion of mechanisms. evidence suggests that more inequality of in- Membership in the subgroups discussed comes in the present is likely to make family can be considered beyond one’s control. In the background play a stronger role in the adult context of the measures of inequality of oppor- outcomes of young people, with their own hard tunity discussed earlier in this essay, it is not work playing a commensurately weaker role” differences in intergenerational mobility (2013a, 79). Corak first establishes the empiri- across circumstance types that are most rele- cal regularity—labeled the Great Gatsby curve vant, but instead the extent to which inequality by Alan Krueger—that countries with greater of outcomes is explained by circumstance inequality of incomes at a point in time also types. Inequality between circumstance types “tend to be countries in which a greater frac- defined by factors beyond an individual’s con- tion of economic advantage and disadvantage trol, such as race, parents’ education, family is passed on between parents and their chil- structure, or place where one lives while grow- dren” (80). ing up, forms the basis of the ex ante measures To understand the causal links, Corak then of inequality of opportunity. That intergenera- investigates the various channels through tional mobility is especially low among specific which parents’ income can influence their chil- subgroups that are already overrepresented in dren’s accumulation of human capital and the bottom quintile of the income distribution their adult outcomes, influences that he notes suggests that barriers to economic opportunity are mediated by the different balance struck are especially severe for these groups. between family, labor market, and public policy in determining outcomes across countries. D icynam s: Inequality of For example, high returns to education not Outcomes and Inequality of only make the income distribution more un- Opportunity equal and thereby provide rich families with The close connection between intergenera- relatively more resources to invest in their chil- tional mobility and equality of opportunity dren, but also increase the incentive for the

2. In his Journal of Economic Perspectives article of that title.

: 8 opportunity, mobility, and increased inequality rich to make such investments. Corak argues tribution of outcomes. They use these findings that parents with high incomes create advan- to argue that inequality should be evaluated tages for their children both through monetary not only from the point of view of its direct investments (better schools, enrichment expe- impact on growth or other aspects of the econ- riences) and by passing along nonmonetary omy but also in terms of fairness. They note advantages—behavior, motivation, aspira- that inequality reflecting circumstances be- tions, and connections. One example of non- yond the individual’s control is widely viewed monetary advantage is the guidance and cul- as unfair. Their paper focuses on an ex ante ture supportive of college attendance. measure of inequality of opportunity used by Corak also discusses public policies that others in the literature (including Ferreira et can either exacerbate or blunt inequality of al. 2014; Marrero and Rodriguez 2013), which outcomes, such as public provision of early quantifies the extent of inequality between childhood education: he notes that public pol- groups of people defined in terms of circum- icies outside of education, such as in health stances beyond their control (see Bradbury and care and fiscal (tax and transfer) policy, can Triest later in this issue). also intervene or not between parental income Brunori and his colleagues (2013) examine and children’s outcomes. He argues that pub- the cross-sectional correlations between the lic policies in the United States, including even inequality of opportunity measure and other public K–12 schooling, are particularly tilted country characteristics, including per capita toward the advantaged. He also notes that pub- output, inequality of outcomes, and intergen- lic provision of health care, as in most other erational mobility. Like Corak, they find a pos- developed nations, helps level the playing field, itive relationship between inequality of op- leading to more preventive care for those with portunity and income inequality. They also low incomes and hence to fewer negative note a positive correlation between this health shocks that could have longer-term con- between-group inequality of opportunity sequences” (2013a, 97). measure and the standard intergenerational Corak concludes by pointing out that “in- mobility measure (the intergenerational elas- equality lowers mobility because it shapes ticity of income) as well as the intergenera- opportunity. It heightens the income conse- tional correlation of education, even when the quences of innate differences between individ- measures come from different papers and are uals; it also changes opportunities, incentives, based on different data sources. They con- and institutions that form, develop, and trans- clude by saying mit characteristics and skills valued in the labor market; and it shifts the balance of power inequality of opportunity is the missing link so that some groups are in a position to struc- between the concepts of income inequality ture policies or otherwise support their chil- and social mobility. If higher inequality dren’s achievement independent of talent” makes intergenerational mobility more diffi- (2013a, 98). Regarding policies to address in- cult, it is likely because opportunities for eco- equality of opportunity, he reminds us of Ro- nomic advancement are more unequally dis- emer’s argument that policy should offset only tributed among children. Conversely, the way those aspects of differential success that relate lower mobility may contribute to the persis- to circumstances, and argues that different na- tence of income inequality is through mak- tions may well make different judgment calls ing opportunity sets very different among the regarding which circumstances are appropri- children of the rich and the children of the ate to offset. poor. (2013, 17) In a sense, Brunori, Ferreira, and Peragine (2013) begin their analysis here, citing behav- Pablo Mitnik, Erin Cumberworth, and Da- ioral economics experiments indicating that vid Grusky “eke out as much evidence on people do distinguish between factors over [whether opportunities to get ahead are grow- which individuals have control and those they ing more unequal] as the available data will do not, when evaluating the fairness of the dis- allow” (2013, 1). They focus on measuring the

: introduction 9 trend in intergenerational social-class mobility examining how intergenerational mobility var- and find evidence of recent rigidification in the ies with inequality of outcomes in the geo- U.S. class structure.3 The negative trend in class graphic area where people grew up. Using in- mobility is especially pronounced among dividual data from two longitudinal surveys, younger cohorts, for whom the rise in inequal- she regresses children’s adult incomes on their ity would have had maximum influence during parents’ incomes interacted with inequality childhood, and is focused on professional and observed in their state of residence when the managerial parents, who have increasingly children were growing up.4 She finds that “the been successful in passing along status to their best available data cannot confirm the hypoth- children. esis that inequality and mobility are system- Along similar lines, Smeeding (2013) argues atically linked in the United States” (22). Given that existing empirical work on U.S. intergen- that she has improved considerably on the pre- erational mobility cannot tell us much about cision of earlier estimates, if a relationship ex- the impact of rising inequality of outcomes on ists, she argues that it must be quite small. She mobility because the recent and current young also notes that the estimated relationship may adults whose mobility can be examined were reflect countervailing trends; for example born and mostly grew up before inequality wid- (much as Corak noted), inequality-associated ened considerably beginning in the 1980s. We higher barriers to college completion among should look for evidence of growing divergence the poor may be partly offset by the increased in success at early life stages between children incentives for those at the bottom of the in- raised in rich and poor families, Smeeding come or wealth distribution to attempt a col- maintains. He notes that most developed na- lege degree. tions show differences in school readiness as- We next present new empirical evidence in sociated with parental socioeconomic status favor of the Great Gatsby curve, and show that (SES) and that some nations’ education institu- this relationship holds up, though in some- tions succeed in reducing these differences what attenuated fashion, when other factors somewhat, but most do not. Lane Kenworthy are controlled for. Our analysis begins where (2012) points to specific early-lif e indicators Corak’s does, by establishing the empirical re- that have worsened, citing Sean Reardon’s lationship, in our case within the United States. (2011) work on growing school performance Figures 1 and 2 plot inequality of income and gaps between high- and low-SES children and intergenerational income mobility across com- Martha Bailey and Susan Dynarski’s (2011) re- muting zone (CZ) areas.5 The vertical axes are search showing rising SES gaps in college com- transformed versions of Chetty, Hendren, pletion. Kline, and Saez relative mobility and absolute Bloome (2015) provides a comprehensive mobility measures (2014b).6 The horizontal and careful recent addition to this literature by axes are a measure of inequality of income in

3. They define social class mobility in occupational terms, professionals/managers at the top and unskilled workers at the bottom.

4. In addition to simple interaction terms, she allows the intergenerational coefficient to vary with inequality through the use of state fixed effects and random coefficients estimates. She examines inequality when the children were teens and, alternatively, when they were around age four.

5. Commuting zones generally coincide with metropolitan areas in urban locations, but also include combina- tions of counties in rural areas, so as to exhaust the territory of the United States.

6. Chetty et al.’s measure of relative (im)mobility is the elasticity of child income rank at age thirty to thirty-one with respect to parent rank when the child was in his or her teens. We have inverted that measure (by subtracting from one, and also multiplied the result by 100) so that it is larger—more positive—where intergenerational mobility is higher; we refer to the result as “adjusted relative mobility” to call attention to this transformation. Chetty et al.’s absolute mobility measure indicates the expected adult rank of a child whose parents were at the 25th income percentile when the child was growing up.

: 10 opportunity, mobility, and increased inequality

Figure 1. Relative Mobility and Inequality of Parental Income

70 Relative Mobility 60

50 .2 .4 .6 .8 Parental Gini Fitted values Relative mobility

Source: Authors’ calculations based on Chetty, Hendren, Kline, and Saez 2014b. RM = 79.07*** –28.28***Gini (***p<0.001) R2 = 0.12

Figure 2. Absolute Mobility and Inequality of Parental Income

Absolute Mobility 40

.2 .4 .6 .8 Parental Gini Fitted values Absolute mobility

Source: Authors’ calculations based on Chetty, Hendren, Kline, and Saez 2014b. AM = 60.93*** –41.56***Gini (***p<0.001) R2 = 0.33 the parental generation, the Gini coefficient.7 that places with greater inequality of income As in Corak’s depiction, both figures indicate also display less mobility, both relative and ab-

7. The Gini coefficient may take on values ranging from 0, indicating that all incomes are equal, to 1, indicating maximal inequality.

: introduction 11

Figure 3. Relative Mobility and Size of the Middle Class While Growing Up

70 Relative Mobility 60

.2 .4 .5 .6 .7 Proportion Middle Class Fitted values Relative mobility

Source: Authors’ calculations based on Chetty, Hendren, Kline, and Saez 2014b. AM = 16.95*** + 49.08***middle class (***p<0.001) R2 = 0.20

Figure 4. Absolute Mobility and Size of the Middle Class While Growing Up

50 Absolute Mobility 40

.3 .4 .5 .6 .7 Proportion Middle Class Fitted values Absolute mobility

Source: Authors’ calculations based on Chetty, Hendren, Kline, and Saez 2014b. AM = 16.95*** + 49.08***middle class (***p<0.001) R2 = 0.46 solute. Figures 3 and 4 use the size of the mid- class is inversely related to inequality and fairly dle class (proportion of CZ parents whose strongly related to mobility—places with a incomes are between the 25th and 75th percen- larger middle class display more mobility. tiles of the national income distribution) as the One of the issues raised by those who have indicator of inequality. The size of the middle challenged the import of Corak’s empirical re-

: 12 opportunity, mobility, and increased inequality lationship relates to timing: critics argue that confounding factors, although it is mostly sta- the inequality measure should refer to the pe- tistically significantly different from zero. We riod when the children whose mobility is mea- find that demographic characteristics have the sured were growing up. This is exactly what expected relationships with mobility: past im- these scatter plots refer to—the inequality of migration, more highly educated residents, parental income, by location, when individuals and fewer children living in single-parent fam- whose outcomes as thirty-y ear-olds are mea- ilies are positively associated with mobility. sured were in their mid-teens, living with their Lower commuting times are associated with parents. Nonetheless, scatter plots are simple higher absolute (but not relative) mobility, con- correlations and not evidence of causation.8 sistent with the hypothesis that spatial segre- An additional consideration is whether the gation (proxied here, as in Chetty, Hendren, relationship between inequality of family cir- Kline, and Saez by commuting times) makes cumstances while growing up and intergener- upward mobility more difficult for low-income ational mobility holds up when potentially residents. confounding factors are controlled for. For ex- The reduction in the magnitude of the ample, it might be that areas with a high pro- Great Gatsby curve relationship when other portion of single-parent households have a factors are controlled for should not be sur- high degree of inequality of outcomes and also prising, because some of the demographic generate a lower rate of upward mobility. If controls can be seen as reflecting the mecha- this is true, and if we do not statistically con- nisms through which inequality among par- trol for single-parent households, then we ents is likely to be passed along to their chil- might mistakenly attribute low upward mobil- dren. These mechanisms include support ity to general inequality, rather than to the (financial as well as via encouragement and prevalence of single-parent households. Using expectations) by more-educated parents for multiple regression analysis, we control for a their children to persist in school, the inherent variety of potentially confounding factors, in- time constraints that reduce the parental at- cluding a measure of past immigration (per- tention available to children of single parents centage of foreign-born residents), the mix of compared with their two-parent counterparts, educational attainment among adults age and the neighborhood segregation mecha- twenty- fiv e and older, the proportion of house- nisms discussed later. holds with children headed by a single mother, Even as increased inequality of outcomes the proportion of workers with average com- among parents appears to be associated with muting times of less than fifteen minutes, a reduction in their children’s income mobil- labor force participation rates of men and ity, a likely further impact of increased paren- women, and economic variables including per tal inequality is to increase inequality of op- capita income, growth in per capita income in portunity given any level of mobility. The the prior decade, and population size. Details reason for this is simply that the parent gen- and complete regression results are presented eration’s outcomes essentially constitute the in the appendix. circumstances (in the Roemer sense) of the The simple correlation represented by the children’s generation while growing up. In- Great Gatsby curve persists even in the pres- equality of opportunity will increase with a ence of demographic controls and the use of a widening of the distribution of circumstances parental inequality measure that predates the even if the relationship between circumstances period during which the adult children’s mo- and outcomes (intergenerational mobility) bility is measured. The estimated association does not change. As Bloome notes, although between inequality of circumstances while rising inequality of (parental) outcomes may growing up and economic mobility is consid- not cause a reduction in mobility, “the eco- erably smaller when controlling for potentially nomic consequences of growing up rich or

8. Chetty, Hendren, Kline, and Saez (2014a) report the same look at the Great Gatsby curve across the 709 CZs, except via regression coefficients (their table V) rather than scatter plots.

: introduction 13 poor have risen, simply because the distance then affects final educational attainment, between the rich and poor has increased” which affects income in adulthood. (2015, 29).9 Furthermore, the ongoing rise in Circumstances, behavior, institutions, and inequality of outcomes in the United States policies may interact in complex ways to create (and other nations) heightens the need for fur- or hinder economic opportunity. Joseph Fish- ther research to understand the mechanisms kin (2014, 2016) notes that opportunities for that underlie the observed relationships. economic mobility may be limited by bottlenecks one must pass through to reach the next Me chanisms stage in the pursuit of a goal. Analyzing in- Factors that enhance or diminish opportuni- equality of opportunity as due to a set of bottle- ties for economic advancement are present necks that are difficult to negotiate or circum- throughout the life cycle. Circumstances be- vent is useful in thinking about the obstacles yond the control of children affect their social to economic opportunity at different stages of and cognitive development and their opportu- the life cycle and in thinking about the effects nities for education and training from concep- of institutions and policies on economic op- tion through early adulthood. Institutions and portunity. In the sections that follow, we survey policies interact with circumstances and indi- the main findings of the research literature on vidual behavior in ways that may either hinder the effects of bottlenecks and barriers to op- or promote opportunities for education and portunity facing individuals as they progress advancement and also affect other outcomes, through life. such as conviction for criminal activity. In adulthood, exogenous circumstances continue E arly Childhood Influences to impinge on opportunities, both directly and on Mobility and Economic through their effect on investments and out- Opportunity comes earlier in life. A large and growing research literature docu- The Social Genome Model (SGM) developed ments the important role of early childhood by the Brookings Institution (Sawhill and development, when many of the barriers to Reeves in this issue; Winship and Owen 2013) economic mobility first arise, in shaping op- models the mechanisms through which oppor- portunities and outcomes later in life. Com- tunities for economic mobility cascade over prehensive reviews of this literature are pro- the life cycle and is used by researchers to vided in Douglas Almond and Janet Currie study the factors affecting intergenerational (2011a) and Currie and Maya Rossin-Slater mobility. The model simulates the progression (2015). Here, citing only a sampling of the most of individuals through the stages of life, start- relevant studies, we provide a more limited re- ing by specifying circumstances at birth. Initial view of the main themes and findings emerg- circumstances—such as family structure, ma- ing from this research. This topic is especially ternal education, and birth weight—directly af- pertinent to understanding the mechanisms fect the probability of successful outcomes through which inequality of opportunity throughout childhood and adolescence and, arises. Factors associated with early childhood through these early stage outcomes, have indi- development are circumstances beyond the rect effects on outcomes later in life. Income ability of young children to control, and so in adulthood is influenced by the cumulative clearly represent inequality of opportunity. effects of outcomes at earlier ages. For exam- Moreover, early childhood factors may have ple, circumstances at birth affect school read- an especially important influence on the op- iness, which in turn affects academic and so- portunities open to children in adolescence cial development in middle childhood, which and adulthood. In a series of papers, Flavio

9. Chetty, Hendren, Kline, Saez, and Turner (2014) make a similar comment: “children entering the labor market today have the same chances of moving up in the income distribution (relative to their parents) as children born in the 1970s. However, because inequality has risen, the consequences of the ‘birth lottery’—the parents to whom a child is born—are larger today than in the past” (2014, online abstract [emphasis added]).

: 14 opportunity, mobility, and increased inequality

Cunha and James Heckman (2007, 2008) and that prenatal and neonatal health is affected later with Susanne Schennach (2010) develop a by environmental factors such as maternal nu- model of human capital development and pro- trition, maternal health, and exposure to pol- vide empirical evidence showing that early lution. Neonatal health, in turn, affects child childhood development increases an individu- development and outcomes later in life. Some al’s capacity to benefit from investment at later of the environmental factors associated with stages of the life cycle. In another study, Cunha poor neonatal health are borne disproportion- and his colleagues (2006) survey earlier re- ately by relatively low-income women, and search on skill development over the life cycle. therefore this is a mechanism by which in- A key feature of human capital investment that equality of outcomes in the parents’ genera- underlies the importance of early childhood tion leads to inequality of opportunity among investments is the tendency for skills acquired children. at one stage of life to augment the skills ac- A surprisingly robust relationship exists be- quired later in life, which some refer to as dy- tween indicators of prenatal health and out- namic complementarity (see Cunha and Heck- comes later in life. Birth weight is an indicator man 2007). Early-lif e skills not only persist, but of prenatal health that has received much at- also increase children’s capacity to develop fur- tention, in part because it is very widely mea- ther. For example, basic language acquisition sured. It is well documented that low birth in infancy allows a young child to progress in weight is associated with reduced wages in further cognitive development. A closely re- adulthood (see, among others, Currie 2011). Re- lated phenomenon is the ability of invest- search supports the hypothesis that this rela- ments at one stage of life to increase the pro- tionship is causal, rather than being due sim- ductivity of investments at later stages, or ply to the correlation between birth weight and self-productivity (see Cunha and Heckman other unobserved factors affecting outcomes. 2007). For example, the efficacy of investment For example, Sandra Black, Paul Devereux, and in elementary and secondary school education Kjell Salvanes (2007) examine a sample of twins programs may be greater for a child who has to control for other factors related to a child’s attended a high-quality preschool than for the circumstances and find that the estimated same child had she or he not attended pre- long- term effects of low birth weight hold up school. In the words of Cunha and Heckman, to the controls for these factors. Much of the these two features of human capital invest- effect of birth weight on economic outcomes ment “produce multiplier effects which are the appears to operate through its effect on cogni- mechanisms through which skills beget skills tive development. David Figlio and his col- and abilities beget abilities” (2007, 6). leagues (2014) estimate a positive relationship Investments made early in childhood will between birth weight and standardized test generally have larger multiplier effects than scores in a sample of twins (to control for un- those made later in life for the simple reason observed factors). Interestingly, they also find that early childhood investment can affect that family background is more strongly asso- more stages of subsequent development. Con- ciated with test performance than neonatal versely, bottlenecks interfering with develop- health is; low-bi rth-w eight children with highly ment and opportunity early in life are likely to educated parents outscore high-birth-weight be especially damaging to children’s future op- children with relatively poorly educated par- portunities and outcomes. ents. This suggests that, in principle, the disadvantages associated with low birth weight Prenatal and Neonatal Factors can be offset by other factors. However, in prac- The earliest bottlenecks interfering with eco- tice, low birth weight and poor health in child- nomic opportunity and mobility occur in utero. hood tend to be associated with being born to Currie (2011) and Almond and Currie (2011a, low-income parents (Case, Lubotsky, and Pax- 2011b) synthesize the evidence on the role of son 2002; Currie and Moretti 2007). Thus, in prenatal health factors in health and economic the absence of public intervention, compensa- outcomes later in life. Evidence is substantial tory measures may not be available to low-

: introduction 15 income families and this mechanism may con- tant gains in later life outcomes. In the United tribute to low rates of intergenerational income States, the Special Supplemental Program for mobility. Women, Infants, and Children (WIC) is the The inequality of outcomes associated with public program most directly targeted at im- prenatal health represents inequality of oppor- proving maternal and early childhood nutri- tunity whether or not a causal relationship tion. Currie and Rossin-Sl ater survey research runs from prenatal health to outcomes. An al- on the effects of WIC, and conclude that alternative possibility is that prenatal health though identifying effects of the program on serves as a proxy for unobserved early life cir- child health outcomes is difficult, “recent work cumstances. Nevertheless, to design effective that carefully attempts to identify the causal programs and policies to reduce inequality of effects of WIC nevertheless points to positive opportunity associated with early life circum- and relatively large effects of the program” stances, the causal relationships between early (Currie and Rossin-Slater 2015, 222). life circumstances and outcomes need to be The Supplemental Nutrition Assistance Pro- understood. gram (SNAP), formerly known as the Food Researchers have used natural experiments Stamp Program (FSP), provides means-tested to identify causal links from prenatal health to payments to households to be used for food outcomes later in life. For example, Almond expenditures, although many analysts have and his colleagues (2010) examine the long- concluded that SNAP/FSP is effectively an in- term effects of being subjected in utero to the come support program. Almond, Hilary 1959 to 1961 Chinese famine and find substan- Hoynes, and Diane Schanzenbach (2011) use tial reductions in literacy and employment for geographic variation in the initial rollout of the this birth cohort in the 2000 Chinese census. FSP to identify causal effects of potential pre- Similarly, in other research, Almond (2006) ex- natal participation in the FSP on birth weight amines long-term effects of prenatal exposure and find that the program increased birth to the 1918 influenza pandemic and finds the weights, especially at the low end of the birth- effects to be increased risk of adult disability weight distribution. and reduced educational attainment and earn- Given the success of the FSP in improving ings. Adam Isen, Maya Rossin-Sl ater, and Reed birth weight, one would expect other policies Walker (2014) use geographic variation in that boost family income to also result in im- changes in air pollution concentration follow- proved neonatal health outcomes and eventual ing implementation of the Clean Air Act of improvements in adult economic outcomes. 1970 to identify the effect of exposure to air The Earned Income Tax Credit (EITC) has be- pollution in the year of birth. They find that come a significant source of income for fami- greater exposure to pollution leads to lower lies with low earnings. Hoynes, Douglas Miller, earnings in adulthood, driven mostly by de- and David Simon (2012) use variation in EITC creased labor force participation. generosity due to tax reforms to identify the Overall, it is now well established that causal effect of EITC payments on birth out- health in early childhood affects outcomes comes and find that an increase in EITC in- later in life. Although health risks are corre- come is associated with a reduction in the in- lated with other factors such as low family incidence of low birth weight and an increase in come that may also be associated with barriers average birth weight. Using a similar identifi- to development and opportunity, the role of cation strategy, Gordon Dahl and Lance Loch- early child health in affecting outcomes is a ner (2012) estimate that an increase in EITC mechanism distinct from the more general income causes a substantial increase in chil- contribution of other associated factors. dren’s standardized test scores.

Compensatory Programs and Policies Early Childhood Education Programs that improve maternal and early life The policy instrument with arguably the great- health and nutrition have the potential to im- est potential to increase equality of opportu- prove child health outcomes, with concomi- nity in early childhood is the provision of high-

: 16 opportunity, mobility, and increased inequality quality early childhood education (ECE). high- quality compensatory preschool educa- Katherine Magnuson and Greg Duncan pro- tion targeted to economically disadvantaged vide in this issue an overview of the evidence children. The evidence on the efficacy of ECE on the overall effectiveness of ECE and an anal- in improving school readiness is generally fa- ysis of the extent to which expanded ECE can vorable, Magnuson and Duncan write, though promote equality of opportunity. variation is considerable in the magnitude of The way in which ECE combines with chil- the effects across programs and test score dren’s classroom readiness to promote learn- gains associated with preschool attendance ing and development, and its implications for fade as students age. how ECE might be most effectively targeted, Of perhaps greater relevance to the topic of has provoked some controversy. Magnuson inequality of opportunity, several research and Duncan note that developmentalists tend studies have found favorable long-term im- to hold a view of the role of ECE that differs pacts of preschool attendance. One of the most from the “skills-beget-skills” model of human studied preschool programs is Head Start, a capital investment described earlier. That large federally funded U.S. program targeted at model predicts that the productivity of ECE economically disadvantaged children. Using will be greatest for children who start pre- comparisons with siblings who did not attend school equipped with the cognitive and socio- Head Start to control for confounding factors, emotional skills needed to take full advantage Eliana Garces, Duncan Thomas, and Currie of preschool learning opportunities. Develop- (2002) find that Head Start attendance results mentalists instead view the productivity as in increased earnings and educational attain- driven by how well an ECE program matches ment among whites and decreased probability the developmental needs of the child in ques- of being booked or charged with a crime tion. Children who enter preschool with rela- among blacks. David Deming (2009) also uses tively low skill levels due to factors associated siblings who did not attend Head Start as con- with economic disadvantage may especially trols and finds that Head Start attendance re- benefit from high-quality preschool programs sults in a substantial increase in a summary designed to compensate for their economic index of young adult outcomes, including high disadvantage. In this example, the ECE pro- school graduation, college attendance, and re- gram would substitute for skill development ductions in crime and teen parenthood. Strik- outside school rather than largely complement skills developed before entering the program. ingly, Head Start participation closes one-third Of course, it is possible that the skills-beget- of the gap in this index between children with skills model might hold over broader phases median and bottom-quartile family income. of the life cycle even if the developmentalists’ Two high-quality preschool programs targeted view is more accurate with regard to early at disadvantaged children, the Perry Preschool childhood education. Furthermore, compensa- and the Abecedarian Project, randomly as- tory programs at young ages, such as with pre- signed potential participants into enrollment school programs, would be especially impor- and control groups and have been extensively tant if the skills-beget-skills phenomenon studied. Research has documented substantial holds for human capital investments at later positive effects on a variety of long-term out- ages. comes for both programs. Magnuson and Duncan document in this Magnuson and Duncan document in their issue that measures of children’s prekindergar- article in this issue that participation in ECE ten skills differ greatly by socioeconomic sta- programs is higher among children in top- tus. These differences are especially pro- income-quintile families than among their nounced for math and reading skills, and counterparts from lower in the income distri- smaller for attention skills. The large differ- bution. This suggests that expanding preschool ences in pre-ki ndergarten skills across socio- enrollment among children from relatively economic strata point to the possibility of re- low- income families is likely to be an effective ducing inequality of opportunity by expanding way to reduce inequality of opportunity.

: introduction 17

Eutd ca ion and Inequality of from zero) between per-pupil district spending Opportunity and district per capita income in the represen- The widespread American prescription to ame- tative states of Massachusetts and Illinois from liorate rising inequality is to advance educa- 1900 through 1980, controlling for per-pupil tion. However, if the education system rein- property valuation and selected demographic forces existing differences rather than leveling variables. Several studies examine school fi- the playing field, we cannot expect more edu- nance reforms and show that as spending be- cation to cure inequality. This section reviews comes more equal across districts, so does stu- evidence on the degree to which the education dent performance (Hoxby 2001; Card and system in the United States mitigates or ampli- Payne 2002; Chaudhary 2009; Jackson, John- fies the cumulative effects on children entering son, and Persico 2014). These studies provide school of current and past advantages and dis- evidence that when funding depends at least advantages. in part on local parental resources, both spending and test scores will be higher (and dropout Primary and Secondary Education: rates lower) in districts where parents are rela- Kindergarten Through Twelfth Grade tively better off. School finance reforms in As discussed, substantial differences in school many states have reduced dependence on local readiness exist between children of poor and tax bases, but they have not eliminated the re- rich parents, reflecting a wide range of influ- lationship between school spending and local ences, including prenatal health, child care wealth. arrangements, and number of books in a Furthermore, even within local school dis- child’s home. Sawhill and Reeves report in tricts, individual schools typically have neigh- this issue that 66 percent of children born borhood catchment areas, so income seg more advantaged begin school with accept- regation among neighborhoods within a able prereading and math skills and generally community may translate into lower-quality school-appropriate behavior, whereas only 46 education for poor children either in the pres- percent of less advantaged children do.10 A key ence of peer effects or because within-district question is whether primary schools, once school- le vel per-pupil spending is not adjusted children come under their care, level the play- to compensate for differences in the costs of ing field and reduce these disparities. Most re- educating children from advantaged versus search findings suggest that they do not, for a disadvantaged backgrounds. Kendra Bischoff variety of reasons. and Sean Reardon (2014) document significant One part of the answer has to do with the growth in the last forty years in neighborhood fact that most primary and secondary educa- segregation of families by socioeconomic sta- tion is provided as a local government service tus. Similarly, Joseph Altonji and Richard Mans in the United States; that is, to the extent that field (2011) find that neighborhoods and high rich and poor families live in different com- schools have become increasingly segregated munities, their children will go to different by socioeconomic status, even as racial segre- schools. And to the extent that K–12 education gation has decreased. Annette Lareau and Kim- relies on local financing, those schools’ avail- berly Goyette note the prevalence of neighbor- able resources are likely to be correlated with hood schools, examine the links between parental income. Even without local financing, choice of residential location and school, and if peers influence the quality of education, liv- argue that differential access to information ing in a poorer community may negatively af- (or access to different information sources) fect the quality of education a child receives. and institutional/structural and financial con- Caroline Hoxby (1998) shows a positive rela- straints imply that richer families “may be tionship (statistically significantly different more easily and freely able to enact their ideal

10. Sawhill and Reeves distinguish disadvantaged and advantaged not just on the basis of parental income or socioeconomic status. The Social Genome Model they use in their analysis defines advantaged as being born at normal birth weight to a nonpoor married mother with at least a high school diploma.

: 18 opportunity, mobility, and increased inequality preferences,” while poorer families face more strengthened.12 Reardon reports that the in- trade- offs and operate within a more limited come achievement gap is large when children choice set (2014, xiv–xv). Although school enter school and does not appear to change choice programs are often seen as a way to re- appreciably during K–12 school attendance. duce socioeconomic or racial segregation of To the degree that children benefit from in- schools, Lareau and Goyette argue that they heriting innate abilities from their parents as “may not always reduce inequalities in school well as from exposure to parental attitudes and quality across families from different social the advantages that money can buy, these re- backgrounds, but instead may reproduce or lationships between parental income and even exacerbate them” (xv). school achievement overstate the influence of Another link in the chain is provided by opportunity. However, even controlling for Bruce Sacerdote (2011), who cites studies that math achievement, low-SES children dispro- find significant effects of peer ability on stu- portionately fail to complete high school: dents’ academic achievement, though some among children with top-quartile eighth grade disagree about the magnitude of these effects. math scores in 1988, 10.7 percent of low-SES William Duncombe, Phuong Nguyen-Hoa ng, children had not completed high school by and John Yinger (2015) document substantial 2000, whereas “rounds to 0” percent of high- additional costs associated with educating dis- SES children had dropped out; for those in the advantaged students. Bruce Baker (2009) sum- middle two quartiles of math scores, 12.4 per- marizes a number of studies that examine cost cent of low-SES children and 0.6 percent of differentials related to student characteristics high-SES children had not completed high among schools within districts and adds to school twelve years later (Fox, Connolly, and that literature, emphasizing that equal per- Snyder 2005, table 21). pupil spending does not provide equal educa- Duncan and Murnane describe later in this tional opportunity when student-body compo- issue the mechanisms by which these growing sition, such as poverty incidence, varies across educational gaps have been developing: some schools. operate through the family and some via Even without unequal school districts or un- schools. They point to differences that arise in equal schools within districts, children with early childhood before children begin formal rich parents benefit from greater enrichment schooling, differential enrichment expendi- expenditures than children with poor parents tures, and parental time investments before do. These benefits include music and art les- and throughout the school-age years that am- sons, books and toys, trips, and tutoring (Kau plify the advantages of high-income children, shal, Magnuson, and Waldfogel 2011; Duncan plus increasing segregation by income of U.S. and Murnane, this issue). schools and the associated concentration in Tallying the impact of a full range of fac- lower-income districts of children whose tors, Sean Reardon documents a growing behavioral problems negatively affect class- school achievement gap between low and high mates’ ability to learn. Sawhill and Reeves re- socioeconomic status students over the last port, also in this issue, that 66 percent of more- fifty years (Reardon 2011).11 He finds, in addi- advantaged children graduate high school with tion, that the relationship between parents’ a GPA of 2.5 or better and without having been educational attainment and their children’s convicted of a crime or having become a par- achievement has been relatively stable even ent, but only 37 percent of less-advantaged as the relationship with parental income has children do so. The Social Genome Model in-

11. Reardon combines results from many previous cross-sectional studies and measures the test-score gap between children with parents at the 90th and at the 10th income percentiles; test score differences are measured in standard deviation units.

12. In recent years, Reardon notes, the income achievement gap has approached the size of the parental- education achievement gap, but parental education remains somewhat more important.

: introduction 19 volves gross flows in both directions at each such as rigor of curriculum and disciplinary transition. However, the success rate for those practices (2013, 714). born advantaged (66 percent) is the same at adolescence and in early childhood (being Postsecondary Education school-ready when they begin school), whereas Martha Bailey and Susan Dynarski (2011) help the success rate of those born disadvantaged elaborate the issue of postsecondary education (46 percent in early childhood) deteriorates be- gaps and trends; they use panel data to com- tween these two transitions (37 percent at the pare higher education enrollment, persistence, end of adolescence), which suggests that the and completion for high-income and low- K–12 schooling years do not erase—and may income members of the National Longitudinal worsen—the disparities that children present Survey of Youth (NLSY) 1979 cohort (born be- when entering school. tween 1961 and 1964) and the NLSY 1997 cohort Completing high school—or not—and the (born between 1979 and 1982). They find that quality of high school education are critical college-entry rates were higher for the later co- determinants, together with parental educa- hort, regardless of income, but enrollment of tional attainment and parental income, of the students who grew up in the richest quartile of next stage in the lives of youth. Some enter families rose faster than enrollment of those the workforce, some go on to additional train- in the poorest quartile. A 40 percentage-point ing, and others continue their education in gap in college enrollment of students born in more academic settings, beginning two- or the early 1960s between poorest-quartile and four- y ear college programs. Many of the fac- richest- quartile students expanded to a 51 tors that play a role in high school graduation point gap for the later cohort; similarly, the ear- and performance have independent effects lier cohort’s 31 point gap in college completion also on college enrollment and completion. between rich and poor grew to a 45 point gap Returning to the issue of socioeconomic seg- for the later cohort.13 A 2015 paper by Greg Dun- regation between districts and between can, Kenneth Lee, Ariel Kalil, and Kathleen schools, Gregory Palardy, for example, finds Ziol- Guest uses similar NLSY data and also in- that “socioeconomic segregation has a strong cludes data from the Panel Study of Income association with high school graduation and Dynamics (PSID) on same-age cohorts to esti- college enrollment. Controlling for an array mate educational gaps between children grow- of student and school factors, students who ing up in low-income versus high-income fam- attend high socioeconomic composition ilies. Like Bailey and Dynarski (2011), they find (SEC) schools are 68 percent more likely to en- that educational attainment was measurably roll at a four-year college than students who higher for later cohorts than for earlier ones at attend low SEC schools. . . . The results indi- all income levels, but the gap between high- cate the association between SEC and attain- and low-income students between the two co- ment is due more to [socioeconomic-based] horts grew larger.14 Their data, like Bailey and peer influences” than to other school effects Dynarski’s, show further increases over time that reflect resource disparities and practices, (between cohorts) in the fraction of students

13. Bailey and Dynarski (2011) find that it is especially among women that the income gap in college entry and college completion increased between the two cohorts; women with high-income parents saw the steepest increases in both college entry and completion between the two cohorts.

14. Duncan and colleagues’ data show that students who grew up in the poorest income quintile in the late 1970s completed 11.7 to 12.0 years of school, on average, compared with 13.9 to 14.0 years for richest-quintile students in the early cohort; later-cohort students from low-income families completed from 12.1 to 12.3 years versus the 15.0 years for the richest-quintile students, implying that the gap between rich and poor expanded from between 2.0 and 2.3 years to between 2.7 and 2.9 years. The year ranges reported in this sentence represent data from the NLSY79 and NLSY97 compared with corresponding cohorts in the PSID (fourteen- to sixteen-years-old in the late 1970s and fourteen- to sixteen-year-olds in the late 1990s); it is remarkable how close these educational attainment estimates from the two data sources are.

: 20 opportunity, mobility, and increased inequality of all income levels who complete college, and ysis, Belley and Lochner conclude that al- expanding gaps between those who grew up in though “ability is equally important for both richest- quintile and poorest-quintile families.15 cohorts, family income plays a substantially Although these average attainment data re- more important role in determining college at- flect all the influences on and behaviors of low- tendance for the NLSY97 youth” than for the income versus high-income students and earlier cohort (14). hence overstate the opportunity disparities, data including eighth grade math test scores Obstacles to Postsecondary Education on a cross-sectional snapshot of students half- Considerable research has examined the na- way between the late 1970s and late 1990s co- ture of the obstacles facing low-income stu- horts indicate that a substantial part of the dents in pursuing higher education. Some completion gap reflects inequality of opportu- concern the attitudes, expectations, and aspi- nity. Mary Anne Fox, Brooke Connolly, and rations regarding postsecondary education Thomas Snyder (2005) report that among low- that surround rich versus poor children among income students with top-quartile eighth- parents, peers, and teachers during their ear- grade mathematics scores in 1988, 74 percent lier years of schooling; some reflect the quality continued their education beyond high school of the education (or the quality of the creden- and 29 percent completed a bachelor’s degree tial) that students receive in high school. Be- or higher by 2000.16 By contrast, virtually all cause many low-income parents did not go high- SES high-scorers continued in school be- beyond high school and some high schools yond high school graduation and 74 percent serving concentrated-po verty populations do obtained a bachelor’s degree or higher by 2000. not have college-focused guidance depart- High- SES students with low eighth-grade test ments, lack of information about the benefits scores (scores in the bottom quartile) contin- of degree attainment or the process of applying ued beyond high school and graduated from to postsecondary schools can hinder low- college at higher rates (83 percent and 30 per- income students. Furthermore, high and ris- cent, respectively) than low-SES children with ing college costs can create substantial per- the highest test scores did. Thus, even attaining ceived and actual financial barriers, even in the top- quartile 8th grade math test scores did not presence of need-responsiv e financial aid pro- earn low-income children as much access to grams. higher education as their rich classmates with Brian Jacob and Tamara Linkow (2011) pres- the poorest test scores enjoyed. Philippe Belley ent data showing a strong link between tenth and Lochner (2007) report higher gaps in col- graders’ expectations about attending college lege attendance between students with low and their parents’ educational attainment or and high family income for a cohort born in income level.18 Although SES-associ ated gaps 1979–1982 (NLSY97) than for a cohort born in in expectations decreased between the early 1961–1964 (NLSY 79).17 In their regression anal- 1970s and mid-2000s, tenth graders in the mid-

15. Only 2 to 6 percent of students who grew up in the poorest income quintile in the late 1970s completed college, on average, versus 36 to 38 percent for richest-quintile students in the early cohort; 8 to 9 percent of low- income later-cohort children completed college versus 54 to 59 percent for the richest quintile. Thus, the college-completion gap between rich and poor students increased from 32 to 34 percentage points for the late 1970s cohort to 45 to 51 points for the later cohort: (Duncan et al. 2015, table 1).

16. Continuing beyond high school graduation includes some college, received certificate or license, received associate’s degree, and received bachelor’s degree or higher.

17. They use the Armed Forces Qualifying Test as the measure of ability and group students into quartiles; they also group family incomes into quartiles. College attendance by the age of twenty-one is measured because of the youth of the NLSY97 sample.

18. The intergenerational transmission of educational attainment is a widely researched topic (for a review of recent literature, see Black and Devereux 2011).

: introduction 21

2000s with college-educated parents were still taking can differ depending on college- roughly 10 percentage points (girls) to 16 per- attendance intentions, which are partly a func- centage points (boys) more likely than tenth tion of income even controlling for ability, as graders whose parents did not complete col- noted in the previous paragraph. Students in lege to expect to obtain a bachelor’s them- high- poverty schools take fewer high-le vel selves. The college-expectations gap between math and science courses than students in tenth graders with parents in the richest and low- poverty schools: National Center for Edu- poorest quintiles of family income remained cation Statistics data show that 80 percent of high in the early 2000s at 40 percentage points 2009 high school graduates from low-po verty (girls) to 48 percentage points (boys). These au- schools completed algebra II–trigonometry thors also show that expectations “have an im- and 23 percent took calculus, while only 71 per- portant influence on college enrollment and cent and 7 percent, respectively, of graduates persistence” (159). Data for high school seniors from high-poverty schools did.19 Similarly, 40 similarly show that students whose parents percent of graduates from low-po verty high were more highly educated or had higher SES schools completed biology, chemistry, and expected to obtain more education themselves physics, compared to only 23 percent of gradu- (Aud, KewalRamani, and Frohlich 2011, table ates from high-poverty schools. 52; Aud et al. 2012, table A-35- 1). Joshua Good- Advanced placement (AP) course taking also man (2008) looks at college-going intentions differs between schools. Brett Lane and Phom- of Massachusetts 2003 and 2004 public high daen Souvanna (2013) report much lower AP school graduates and similarly finds substan- participation rates and lower overall success tial differences between low- and high-income rates in high-need schools than in low-need students; these differences shrink but do not schools in the absence of the Mass Math + Sci- disappear when he controls for skills (test ence Initiative, a program implemented in scores). He finds specifically that “low income some schools in Massachusetts to increase students in the middle and upper parts of the low-income students’ participation and suc- skill distribution appear the most constrained, cess in AP courses.20 Schools that participated particularly with respect to four-y ear public in the program saw immediate and dramatic colleges” (Goodman 2008, 5). increases in AP participation and success rates One issue in college-going for low-income and have sustained those gains over the four- students is the academic preparation they re- plus years that the program has run.21 The suc- ceive in high school. Course offerings some- cess of this program implies that barriers un- times differ between low- and high-income related to their abilities prevent students at schools and districts, and a student’s course low- income schools from getting ahead in this

19. Low-poverty schools are those in which 0 to 25 percent of students qualify for free or reduced-pric e lunches, while 76 to 100 percent of students receive free or reduced-pric e lunches in high-p overty schools. See Aud et al. 2012, Appendix A, table A-31-1, pp. 234–35.

20. Their data show participation rates (the number of AP exams taken divided by the number of juniors and seniors in the school) of 21 percent for the 2011–2012 school year at low-ne ed schools not participating in the program as compared with 13 percent in high-need non-participating schools; low need means fewer than 35 percent of students qualify for free or reduced-pric e lunch while high need indicates the school has more than 50 percent of its students eligible for free or reduced-price lunch or has been classified as level 3 or 4 in terms of school accountability status. The overall success rate (number of exams scoring 3 or better divided by the number of juniors and seniors) was 14 percent for low-ne ed schools and less than 6 percent in high-need non- participating schools. Authors’ calculations based on Lane and Souvanna 2013, table 3, p. 11.

21. Students in participating schools were more than 2.5 times as likely to take an AP course as students in otherwise similar comparison group schools, and overall success rates were also about twice as high as those in comparison schools. The simple success rate—test scores of 3 or better per test—is lower for participating schools, as would be expected when broadening the pool of test-takers.

: 22 opportunity, mobility, and increased inequality way. Taking AP courses and completing them Aid (FAFSA) along with estimates of aid (and successfully is said to increase college applica- tuition) at nearby colleges. They find the com- tions, enrollment, and persistence. bined treatment increased significantly the The availability of college counselors varies proportion of students who had completed considerably between high- and low-income two years of college three years after receiving high schools, reducing the chances that low- services, but find no effect of aid and tuition income children receive useful advice about information without filling out the FAFSA. college- going. Christopher Avery, Jessica How- Some low-income individuals, however, lack ell, and Lindsay Page document less availabil- both the necessary information and the finan- ity of college counseling for low-income high cial resources to pay for college. Although net school students, which they attribute largely to tuition at public and private nonprofit four- “inadequate school finances, insufficient coun- year institutions is typically lower for low- selor training programs, and a lack of clarity income students than for their high-income about how school counselors should allocate counterparts because grants and scholarships their time” (2014, 1). They conclude, “Armed are higher, net tuition still represents a consid- with less information about colleges than their erably higher proportion of income for low- higher- income peers, students from modest income students or their parents.22 As a result, backgrounds may be at greater risk of selecting many students and families borrow to meet a postsecondary alternative that is not a good college costs. Belley and Lochner, as noted ear- fit” (2). Hoxby and Avery document that many lier, document a substantial increase in the ef- low- income, high-ac hieving high school stu- fects of family income on college attendance dents fail to apply to high-quality selective col- between the NLSY79 and NLSY97 cohorts. They leges that would be a better academic fit than develop a model that includes credit con- the much less selective colleges they typically straints and conclude, “Overall, it is likely that attend. They note that these students lack ac- borrowing constraints have become more cess to appropriate college counseling because stringent over the past few decades and that they “come from districts too small to support this is at least partially responsible for the in- selective public high schools, are not in a crit- crease in college attendance gaps by family in- ical mass of fellow high achievers, and are un- come” (2007, 32). Janet Yellen shows in this is- likely to encounter a teacher who attended a sue much higher student loan debt burdens selective college” (2013, 1). for families in the lower half of the wealth dis- Recent studies and experiments have shed tribution than for richer families and reports light on additional information problems that that these disparities increased between 1995 impede low-income students in the college- and 2013. going process. Carrell and Sacerdote report on All in, children of affluent parents graduate an inexpensive intervention involving cash bo- from college at substantially higher rates than nuses and coaching for high school seniors children of low-income parents, and the gap identified by guidance counselors and find persists even when controlling for ability in the large effects on college enrollment and persis- form of test scores. A variety of mechanisms tence for young women but no effect for men; serve to prepare poor children less well for col- in addition, “offering cash bonuses alone with- lege in addition to making it more difficult for out mentoring has no effect” (2013, 1). Eric Bet- them to attend and persist through graduation tinger and his colleagues (2012) offered low- even with equal preparation. The conse- income adults who received tax-preparation quences of these parental-income gaps in help simultaneous assistance filling out the schooling are enormous, as educational attain- lengthy Free Application for Federal Student ment is a key determinant of labor market suc-

22. For 2011–2012 data on average net price of tuition at public and private nonprofit four-year institutions by income group (net price nets out all grants and scholarships but does not take account of loan aid), see the 2014 study by Grace Kena and her colleagues (indicator 35, figures 2 and 4).

: introduction 23 cess. Four-year college graduates have higher with any educational disparities. Here, we pro- labor force participation rates, lower unem- vide an overview of research on the role of bar- ployment rates (that is, higher employment riers to opportunity in the labor market, and rates given participation), and higher pay for the role of the labor market in amplifying the full- time, full-year employment than high effects of barriers encountered earlier in life. school graduates or individuals with some col- Some parents can provide their children lege or AA degrees. As discussed in the next with especially useful employment informa- section, Jo Blanden and her colleagues (2014) tion and networks when they seek a job. Re- document a dominant role for educational at- search indicates that friends and family are an tainment as a factor contributing to the corre- important source of referrals or information lation between an adult’s earnings and his par- on job openings during job search (Holzer ents’ family income when he was growing up. 1988; Ionnides and Loury 2004). Explicit nepo- Thus, unequal education is both an effect and tism would amplify the effects of more benign a cause of unequal opportunity. information disparities. Moreover, some parents may be able to pay for (or support their Inequality of Opportunity in the children during) work internships and other Labor Market forms of work enrichment or work experience It seems likely that much of the intergenera- beyond education that enhance job prospects, tional transmission of labor market success is providing support that poorer parents or accomplished via educational attainment, poorer young labor market entrants cannot both because education is an important deter- afford. In addition, Miles Corak (2013a) and minant of labor earnings and because research others have written about intergenerational shows strong intergenerational correlations in transmission of attitudes, values, preferences, educational attainment. This section investi- aspirations, and soft skills that can enhance or gates the degree to which labor market oppor- hurt workers’ labor market prospects. tunities may differ among people who grow up An extensive literature examines intergen- in different circumstances beyond disparities erational earnings elasticities or correlations, in education. That is, to what extent do the especially between fathers and sons (for recent children of low-income parents see lower earn- reviews, see Solon 1999, 2002; Bjorklund and ings as adults than children of high-income Jäntti 2009; Black and Devereux 2011; on daugh- parents, even controlling for educational at- ters, see Chadwick and Solon 2002; for inter- tainment? Given the disparities in educational generational elasticity estimates on daughters, attainment documented earlier and elsewhere see Solon 1999, table 6). This literature does in this issue, any disparities added on in the not generally control for education because it labor market will compound the degree of in- seeks to quantify the full relationship, includ- equality of opportunity and inequality of out- ing the contribution of education. Nonethe- comes. less, some authors do shed light on the net-of- What mechanisms could cause labor mar- education question. For example, in his Labor ket opportunities to be distributed unequally Handbook article, Solon (1999) lays out a sim- among young workers with similar education? plified version of the theoretical model posited Children of low-income parents may be less by Gary Becker and Nigel Tomes (1979), in healthy, have different noncognitive skills and which a child’s earnings in adulthood reflect attitudes (including, for example, lower expec- parental investments in his/her human capital tations of labor market success) and have in- (education) as well as his/her endowment of ferior access to personal networks, connec- earnings capacity and market luck (a stochas- tions, and internships that are instrumental in tic element). That endowment, in turn, is de- the job search and advancement processes, re- termined “by the reputation and ‘connections’ sulting in access to less prestigious and less of their families, the contribution to the ability, remunerative positions. In addition, labor race, and other characteristics of children from market institutions and processes may widen the genetic constitutions of their families, and or reduce the degree of inequality associated the learning, skills, goals, and other ‘family

: 24 opportunity, mobility, and increased inequality commodities’ acquired through belonging to in which the labor market works and how a particular family culture” (Becker and Tomes ‘good jobs’ are obtained, and the income in- quoted in Solon 1999, 1764). equalities between parents.” (2013b, 114). Solon (1999) goes on to note several “cru- Another aspect of interaction among the cial” aspects of the intergenerational transmis- processes determining outcomes is that the in- sion of earnings status implied by the simple fluence of educational attainment on earnings model, including that intergenerational trans- depends on the rate of return to education in mission occurs through a multitude of pro- the labor market; that rate of return—or edu- cesses and that parental income is not the only cational wage premium—is a labor market intergenerational influence on child’s earn- characteristic. Nations or eras with greater dis- ings. Thus, to go beyond education, as we want parities in pay levels according to educational to do here, we need to think about children’s attainment will, other things equal, have endowments and about how the various pro- higher intergenerational earnings elasticities cesses determining earnings interact. Regard- (hence, lower mobility) because any level of in- ing that interaction, a child’s endowment is tergenerational correlation in education trans- correlated with the parental endowment; in ad- lates into greater differences in earnings and, dition, some elements of the child’s endow- hence, higher correlation of parent and child ment may help determine the degree to which earnings. parental investments translate into educa- Jo Blanden, Paul Gregg, and Lindsey Mac- tional attainment and the degree to which they millan group the key factors or mechanisms in have a direct impact on earnings, even after intergenerational earnings persistence— controlling for education. “those variables that are related to family in- Solon concludes that the intergenerational comes and that have a return in the [labour] earnings elasticity, a measure of the relation- market”—into four categories: noncognitive ship between parents’ earnings and that of skills, cognitive ability, early labor market ex- their children in adulthood, for U.S. men is periences, and educational attainment (2007, “somewhere around 0.4” (1999, 1795) and notes 1).23 Although it is impossible to separate these that this estimate is higher than similar esti- influences (because, for example, cognitive mates for Canada, Finland, and Sweden; an ability helps to determine educational attain- elasticity of zero would indicate no relation- ment as well as later earnings), it is still in- ship between the earnings of children and par- structive to delve into them one by one to learn ents, and an elasticity of one would indicate a what we can infer about labor market influ- near- exact correspondence. In a later paper, ences net of education; indeed, Blanden, Solon (2004) also includes the influence of gov- Gregg, and Macmillan argue that “many of the ernment investments in human capital (public associations operate in a sequential way” (4). financing of education), which can be redis- In their initial examination of intergenera- tributive or not. That is, to the degree that pub- tional persistence for a 1970 British cohort, licly funded education is focused on poorer they find the intergenerational elasticity of children or public funds are inversely related earnings to be 0.32.24 They then decompose to parental income, the intergenerational earn- that elasticity by examining the relationships ings elasticity will be lower. Corak notes that among the factors and family background / pa- “we can expect the intergenerational elasticity rental income and success-re wards in the labor [of earnings] to differ across countries for rea- market. They find that “better off children have sons associated with the costs and returns of better noncognitive traits and perform better investing in a child’s human capital, the way in all cognitive tests . . . achieve more at all lev-

23. The title of Blanden, Gregg, and Macmillan’s paper mentions only three, but in the text they explore early labor market experience as well.

24. The intergenerational elasticity is derived from the estimated relationship between child earnings at age thirty and parental family income.

: introduction 25 els of education and have greater [labour] mar- measures of occupation at ages thirty and ket attachment in their teens and 20s” (8). Fur- thirty- four. This study retains the pathways of thermore, the cognitive variables are generally education and early labor market attachment. more strongly associated with parental income Overall, they find higher mobility in the British than the noncognitive traits. After analyzing sample than in the U.S. sample. They report how these factors are related to earnings at age that “the linkage between parental income thirty, they use both sets of regression results and offspring earnings [in the United States] to decompose the estimated earnings elasticity is largely accounted for by the offspring- into components explained by each of the fac- education pathway, whereas in Great Britain, tors and an unexplained component, which offspring occupation plays a much stronger amounts to 46 percent of the elasticity. In this role. The difference in the strength of the edu- analysis, they find educational attainment to cation pathway is due to relative differences in be the most important factor, accounting for the returns to education in the two countries 31 percent of the total estimated elasticity.25 rather than to relative differences in the influ- Early labor market attachment comes in a dis- ence of parental income on educational attain- tant second among the explanatory factors, ex- ments” (Blandon et al. 2014, 442). Note that la- plaining 9 percent of the elasticity; noncogni- bor market returns to education reflect the tive and cognitive factors explain only 6 to 7 operation of the labor market rather than the percent, largely because their influence ap- heritability of education. Quantifying the con- pears to work mostly through education. If we tributions, their data indicate that education attribute the entire earnings elasticity exclud- accounts for 26 percent of the persistence in ing the education component to what goes on the United States and 12 percent in Great Brit- in the labor market, we obtain an upper-bound ain, even when controlling for occupation. The estimate of the importance of variations in la- other pathways between parental income and bor market opportunity of close to 70 percent. an offspring’s earnings—early marriage, early The direct influence of early labor market ex- labor force attachment, and marital status and perience, together with the effects of cognitive health at age thirty—account for only 6 percent and noncognitive skills not operating via edu- of persistence in both the United States and cation, yields a lower-bound estimate of 23 per- Great Britain.26 Occupational choices account cent for this British cohort of sons. They un- for 24 percent in the United States and 34 per- dertake a similar decomposition to understand cent in Great Britain. Interestingly, a simula- the sources of a drop in mobility observed be- tion exercise indicates that education’s contri- tween 1958 and 1970 cohorts and find that a bution to the overall U.S. elasticity would be strengthening relationship between parental roughly cut in half if the returns to education income and both educational attainment and in the United States were at the lower, British early labor market attachment accounts for level. most of the change. Other relevant literature examines the A 2014 paper by Blanden, Robert Haveman, transmission of employers between fathers Smeeding, and Kathryn Wilson compares in- and sons (see, for example, Bingley, Corak, and tergenerational transmission for men in the Westergard-Nielsen 2011; Corak and Piraino United States and Great Britain using decom- 2011). The transmission of employers is part of positions that lack the measures of noncogni- a broader mechanism regarding parental pro- tive traits and cognitive ability (because they vision of both information and social networks are not available in the U.S. Panel Study of In- that can enhance the labor market prospects come Dynamics), and add in early marriage, of their offspring; parents may also invest in marital status and health at age thirty, and firm-specific types of human capital. Miles

25. They measure educational attainment at and after age sixteen, as number of O-levels at age sixteen, number of A-levels, staying in school after sixteen, earning a degree, and staying in school after eighteen.

26. Missing variables “explain” 4 percent of U.S. persistence and less than 1 percent in Great Britain.

: 26 opportunity, mobility, and increased inequality

Corak and Patrizio Piraino (2011) also note that are said to provide important job experience transmission of employers may reflect fathers’ and connections to young graduates and some- possible direct role in the hiring process, in- times lead to offers of paid employment. But cluding nepotism. They measure inherited em- low- income labor market entrants, with or ployers in two ways: whether the son has ever without college debt, often cannot afford to worked for an employer that ever employed work without pay and therefore lose out on their father; whether the son’s main employer these opportunities. Some court decisions at age thirty-three was also the father’s main have limited employers’ exploitation of young employer when the son was in his teens. They workers and some colleges have raised funds find that about 40 percent of a cohort of Cana- to provide scholarships to provide living sti- dian young men meet the first criterion, largely pends to low-income students who want to reflecting early jobs (in the teen years and early serve as interns, but these remedies are un- twenties), and that 6 percent to 9 percent have likely to have made a serious dent in the prev- the same main employer in adulthood. Al- alence of the practice and its disparate im- though Corak and Piraino are unable to infer pact.27 As Ross Eisenbrey of the Economic causality, their findings are consistent with all Policy Institute observes, “It’s hard to quantify of these hypothesized mechanisms. Intergen- the impact of this phenomenon [internships] erational transmission of employers is higher on the decline in economic mobility, but I sus- when the father has self-employment income pect it has been substantial and will continue and is at or near the top of the fathers’ earn- to grow until the Department of Labor cracks ings distribution. They also find that transmis- down on what is, in many cases, illegal exploi- sion of employers contributes to nonlinearities tation” (2012).28 in intergenerational earnings elasticities: high Existing research finds an important role elasticities in the middle and at the upper tail for parental income and other family-re lated of the fathers’ distribution reflect the pattern factors in determining labor market earnings, of those who inherit an employer from their even beyond parental influence on educational father. Similar research using Canadian and attainment. These effects occur via transmis- Danish data yields similar findings (Bingley, sion (by both nature and nurture) of attitudes, Corak, and Westergard-Nie lsen 2011). Regard- skills, preferences, and social networks, and ing the correlation with father’s earnings level, even nepotism. Depending on one’s interpreta- the study notes that “mobility out of the bot- tion of the unexplained portion of intergen- tom has little to do with inheriting an em- erational earnings elasticities, noneducation ployer from the father, while the preservation factors account for one-quarter to three- of high income status is distinctly related to quarters of U.S. earnings transmission from this tendency” (1). parents to sons.29 These numbers are higher One example of recent U.S. labor market in- when taking account that part of the education stitutions that may be contributing to unequal portion reflects returns to educational attain- opportunity, even among those with a college ment determined in the labor market; indeed, education, is internships. Unpaid internships much of the difference in intergenerational

27. Data are generally lacking on the prevalence of unpaid internships. In promoting a 2011 event to discuss internships in the U.S. labor market, and the book Intern Nation by Ross Perlin, the Economic Policy Institute noted, “Internships have become a principal point of entry for young people seeking white-collar careers, and it is estimated that half of all college students will do an internship before graduating. Between 1 and 2 million people overall will work as interns this year in the United States, saving firms $600 million dollars.”

28. Two key criteria in determining the legality of an internship at a for-pr ofit private sector employer is that “the internship experience is for the benefit of the intern,” “the intern does not displace regular employees, but works under close supervision of existing staff,” and “the employer . . . derives no immediate advantage from the activities of the intern” (U.S. Department of Labor 2010).

29. These figures are based on reported U.S. estimates (Blanden et al. 2014, table 6, column 1).

: introduction 27 transmission between the United States and also reduce inequality of opportunity in the Great Britain represents differences between labor market. the two nations in education returns. Do these findings suggest policy interven- T he Role of Geography in tions that might reduce parental influence in Inequality of Opportunity the labor market? Although rules exist to limit Children have no choice over the geographic outright nepotism—at least in public employ- area where they are born and raised, so geog- ment, many interventions aimed directly at re- raphy is clearly a dimension of circumstance ducing parental influence would be seen as im- that should be considered in evaluating the ex- pinging on parents’ autonomy in raising their tent of inequality of opportunity. Research on children. However, policies that enhance op- the role of geography in economic opportunity, portunities for all children to succeed, as dis- which we survey in this section, shows that cussed earlier would have two positive impacts economic opportunity varies substantially on the labor market: More disadvantaged chil- across geographic areas, though the mecha- dren would attain higher education, thereby nisms underlying this relationship are not yet directly improving their labor market out- fully understood. comes, and the resulting increase in the supply Branko Milanovic (2015) summarizes the de- of educated workers would moderate the high gree of inequality of opportunity associated returns to education that still prevail in U.S. with geography globally by decomposing total labor markets and contribute, as documented, global income inequality into inequality of op- to the heritability of earnings. portunity (between-country inequality) and re- Additional policies in the labor market sidual inequality. He finds that inequality of could further equalize opportunities. Jared opportunity constitutes a “huge but decreas- Bernstein (2014), for example, argues that after ing” share of overall inequality between 1988 education, the next most important policy ac- and 2008, amounting to almost 70 percent of tions governments must undertake are to level total interpersonal inequality in 2008. He re- the playing field for workers who seek to form ports that a measure of intercountry inequal- or join unions, and to increase the minimum ity, the mean log deviation of income, was 0.68 wage to counter the lack of bargaining power in 2008, down from 1998 and 1988, and that the of many in the workforce. Raising the mini- global interpersonal mean log deviation was mum wage and promoting unions is likely to 0.98 in 2008, also lower than in 1998 and 1988 reduce the labor market returns to education (see the top panel of table 1).30 (and occupation) relative to less-reg ulated Milanovic provides a metaphor for his ana- market outcomes. Furthermore, labor regula- lytical framework: he suggests seeing the world tions and policies that nudge firms toward income distribution as a long pole and each adopting human resource practices that result country’s distribution as being represented by in opportunities for workers to learn by doing a plaque on that pole. An individual’s income and to acquire occupation- and firm-specific lies within the range covered by the plaque rep- human capital (perhaps by creating career lad- resenting his or her home country, and that ders in positions with relatively modest educa- home country substantially circumscribes the tional requirements) may reduce the associa- person’s economic prospects. The plaque rep- tion between parental income and early labor resents two “circumstances beyond individual market attachment and advance the interests control: level of development of one’s country of employers by reducing worker turnover. As of residence, proxied by its GDP per capita or noted, action to limit the prevalence and ex- average number of years of education, and in- ploitative aspects of internships or to provide equality of distribution within that country” more equitable access to internships would (2015, 456). Milanovic uses the term circum-

30. The mean log deviation measure is often referred to as Theil L. Milanovic’s dataset includes 118 countries in 2008, representing 92 percent of the world population. This inter-country inequality measure depends only on average income in each nation.

: 28 opportunity, mobility, and increased inequality

Table 1. Geographic Inequality of Opportunity Inequality is measured as mean log deviation (Theil L) of income

Between-Area Inter-Areab Percentage of Geography, Income Unit, and Year Total Inequalitya (Geographic) Total

Global, across countries (Milanovic) 1988 1.070 0.862 80.6 1998 1.035 0.764 73.8 2008 0.983 0.677 68.9

United States households 1988 0.401 1998 0.488 1999 0.476 2008 0.541

U.S. commuting zones, cohort familiesc 1996–2000 0.492 0.031 6.3

U.S. counties, cohort familiesc 1996–2000 0.492 0.055 11.2

U.S. census tracts, households 1999 0.081 17.0

Source: Authors’ compilation based on Milanovic 2015 and U.S. Census Bureau 2014a; authors’ calculations based on Chetty, Hendren, Kline, and Saez 2014b and U.S. Census Bureau 2014b. aTotal measures inequality between individuals, households, or cohort families. For cohort families, total is inequality across centiles of the distribution, not individual families. bInter-area measures inequality between areas, assuming area mean income applies to all units in an area. cCohort families are tax filing units that claimed children born in 1980–1982 on tax returns in 1996– 2000. stances in the Roemerian sense, as elements of level or for which only national governments “fate, decided at birth” (as he puts it) or public can be held responsible” (2015, 452–53). Geo- goods reflecting the country in which they re- graphic aspects of inequality of opportunity side. To improve her lot, an individual has within the United States are the focus of the three options: she can rely on hard work or remainder of this section, but Milanovic’s luck to rise within her country’s distribution; global measures (for comparison) and his pole she can hope for her country to experience metaphor also prove useful in what follows. strong growth and have its plaque move up the Looking within the United States, overall in- pole; or she can migrate to a higher-income terhousehold income inequality is, of course, country. considerably smaller than Milanovic reports Milanovic notes that “the topic of inequality among global individuals and across nations.31 of opportunity is traditionally studied at the The Census Bureau reports a mean log devia- national level” at least in part because of “the tion of household income equal to 0.54 in 2008, unstated view that equality of opportunity is up from 1998 and 1988 (see second panel of something that ought to hold at the national table 1). Data on family income for parents re-

31. If the U.S. data were those that Milanovic included for the United States in his global measures, they would necessarily show less inequality within the United States than globally, given that the United States is one nation among many.

: introduction 29 ported by Chetty, Hendren, Kline, and Saez the area “plaques” are wider and have much (2014b) show a mean log deviation across cen- greater overlap. Moreover, the measures of geo- tile means of 0.49 during the 1996 to 2000 pe- graphic location used in the decompositions riod.32 With these data, we can also follow Mi- are for current location, not where people grew lanovic and decompose total inequality into up. Given the prevalence of geographic migra- the part associated solely with geography and tion, especially between neighborhoods, one the residual. This exercise applied to commut- should not necessarily interpret the interarea ing zones as the geographic unit within the component as measuring inequality of oppor- United States yields a between-area mean log tunity, although it may indicate inequality of deviation of 0.03 during the 1996–2000 period, circumstances for the generation currently which amounts to only 6 percent of total in- growing up. equality (measured across centiles for the same tax-filing population of parents). Repeat- Inequality of Opportunity Across U.S. ing the same exercise across counties, the Commuting Zones and Counties smaller geographic building blocks of com- Chetty, Hendren, Kline, and Saez (2014a) ex- muting zones, yields a mean log deviation of plore the geography of inequality of opportu- 0.06, or 11 percent of total inequality. nity in the United States, using commuting Given the prevalence of economic segrega- zones as their geographic units; they analyze tion in residential location, one might expect data on 709 commuting zones covering virtu- that a higher share of inequality would be ex- ally all U.S. territory and population.33 Using plained by a more-detailed geography. That is, measures of relative and absolute intergenera- in fact, the case: if we go down to the neighbor- tional mobility, they document wide variations hood level, we find somewhat greater interarea in mobility across commuting zones within disparities, with a mean log deviation of 1999 the United States. Moreover, they note that rel- household incomes across census tracts na- ative mobility patterns across CZs are highly tionwide of 0.08, or 17 percent of the 0.48 na- correlated with absolute mobility.34 They pro- tionwide 1999 mean log deviation of household vide a much-cited heat-map of absolute mobil- incomes published by the census (see table 1). ity variations across CZs nationwide, showing Although this represents greater interarea dis- broad regional patterns of mobility as well as parities than the CZ or county figures, it is still variations within regions and differences be- small compared with Milanovic’s intercountry tween rural and urban areas. Even though vari- differences. Differences in mean incomes ation within regions is substantial, the map across commuting zones, across counties, or shows concentrations of low mobility across even across census tracts do not explain much the southern United States. This is quite con- of the total inequality of incomes in the nation; sistent with the finding of Gustavo Marrero within the United States, the length of the and Juan Rodriguez (2013) that southern U.S. “pole” in Milanovic’s metaphor is shorter and states exhibit high inequality of opportunity

32. This mean log deviation figure is remarkably close to the U.S. Census Bureau figure cited above for all households in 1998. This is surprising, given that 0.49 from Chetty, Hendren, Kline, and Saez is an underestimate of total household inequality for at least two reasons. First, the inequality of incomes across individual households or filing units should be higher than inequality across centile means of households or filing units. (Milanovic’s analysis, however, also used centiles for each country and computed the (weighted) mean log deviation among those values.) Probably more important, the Chetty data are tax-filing-unit incomes of a subset of the population that is undoubtedly more homogeneous than the set of all tax filers; their 1996 to 2000 sample is parents with children born between 1980 and 1982.

33. The 709 CZs with data contained 99.96 percent of the U.S. population in 2000.

34. Their measure of relative mobility is the slope coefficient from the child-rank on parent-rank regression, which indicates the difference in expected ranks between children in the richest versus poorest families. Their primary measure of absolute mobility is the mean percentile income rank of children whose parents were at the 25th percentile of the national income distribution.

: 30 opportunity, mobility, and increased inequality when measured across circumstance groups low mobility in CZs characterized by high defined by race and parental education (see es- prevalence of single-parent households is not pecially figures 2a, 2b, and table 1). simply a compositional result; children of What mechanisms might be at work to married-couple families show lower mobility, cause such substantial differences in opportu- on average, in CZs with high fractions of single nity—and hence differences in intergenera- parents (2014a). tional mobility—among commuting zone ar- A third strong correlate of mobility across eas? Chetty, Hendren, Kline, and Saez explore CZs is the quality of local schools. We dis- a number of covariates of mobility at the CZ cussed earlier the importance of K–12 school level. These explorations do not attempt to quality and the U.S. pattern of local public establish causation, but they do suggest school provision in generating opportunity. some geography-related forces that may be Chetty, Hendren, Kline, and Saez find mea- at work. Discussing such factors may help us sures of social capital strongly related to mo- sort out circumstances, as Milanovic uses the bility outcomes across commuting zones, cit- term (based on Roemer), that are mediated by ing earlier studies that establish the importance physical nearness versus those that are not, of social networks and engagement in commu- even though they may exhibit geographic pat- nity organizations in determining social and terns. economic outcomes (2014a). They measure so- One of Chetty, Hendren, Kline, and Saez’s cial capital with indicators of violent crime five important covariates of income mobility rates, religiosity, and a social capital index con- across CZs is the degree of income inequality structed by other researchers, which they ag- in the CZ (2014a). We explored this factor in gregate to the CZ level. To the degree that so- discussing the Great Gatsby curve—the rela- cial capital has a geographic aspect, it is tionship between inequality of outcomes and presumably at a considerably smaller geo- subsequent inequality of opportunity or inter- graphic unit than the commuting zone. For ex- generational mobility. ample, Robert Putnam’s work focuses on the The most important covariates of income community and neighborhood level, investi- mobility across CZs in the Chetty, Hendren, gating people’s interaction, trust, and cooper- Kline, and Saez analysis relate to family struc- ation with neighbors and their social peers ture, specifically the fraction of families with (2000, 2015, 2016). His discussion comments children that are headed by single mothers accompanying this issue examine the interac- (2014a).35 CZs in which a substantial fraction tion between propinquity (geographic near- of children are living in single-parent families ness) and social networks in various contexts. display lower absolute mobility. Analysts offer It is among the traditional working classes that many links, both causal and associational, be- neighborhood ties have tended to be most im- tween single motherhood and low family in- portant, and researchers have documented come that explain why more children in CZs greater success for those living in neighbor- with a high proportion of single mothers start hoods with what the literature calls collective at the bottom (of the national income distri efficacy. Putnam notes that the nongeographic bution)—and hence are likely to be nearer networks characteristic of higher-income par- the bottom as adults. In addition, hypotheses ents and children appear to provide access to about why, given parental income, children in broader opportunity. single-parent households may have less oppor- The last of Chetty, Hendren, Kline, and tunity to advance, focus on factors such as pa- Saez’s five most important covariates is the de- rental education and associated aspirations gree of geographic segregation—by income or for children, as well as parental time available by race—within a CZ (2014a). Measures of seg- to devote to interacting with children. Chetty, regation likely capture shared versus separate Hendren, Kline, and Saez note, however, that experiences of children raised in rich and poor

35. In the family structure category, they also examine the fractions of CZ adults who are divorced and who are married, and they find similar results, but not as strong as for single mothers

: introduction 31 families within a CZ and their exposure to fam- require controlling for (subtracting out) geo- ilies in the other group. Partly because spatial graphic disparities that are not based on pro- residential segregation is often associated with pinquity.37 separate schools and other public institutions from parks to libraries, it is likely to limit fa- Inequality of Opportunity vorable peer effects and positive economic role Across Neighborhoods models for low-income children in addition to Inequality of opportunity across neighbor- reducing the funding and quality of the public hoods potentially involves different mecha- services to which they have access. Like social nisms than inequality of opportunity associ- capital measures, these segregation indicators ated with coarser measures of geography. point to smaller geographies as important loci Although causality has been difficult to pin in which opportunity takes shape. down, the association between the character- In a 2015 paper, Chetty and Hendren explore istics of the neighborhoods children grow up another path for improving one’s situation in and the economic outcomes of those chil- highlighted by Milanovic—moving to a new dren as adults is clear. area (geographic mobility) to achieve upward William Julius Wilson (1987) is credited with mobility. The authors find that children gain being among the first to hypothesize that the positive outcomes of destination counties neighborhood environment has an important in proportion to how young they were when role in shaping opportunity. In particular, he their parents moved there from a county with focused on the potentially negative conse- poorer outcomes.36 Unlike Milanovic’s sce- quences of growing up in inner-city areas of nario, however, Chetty and Hendren do not concentrated poverty after both manufactur- find that the positive effects of moving extend ing jobs and the black middle class had largely to adults (twenty-four and older at the time of moved out. Socially isolated, with few positive the move). They identify causal effects of resi- (employed) role models and few jobs available dence location (county) when growing up and locally, children growing up in severely disad- estimate these effects for every county nation- vantaged neighborhoods were likely to be wide, interpreting them as neighborhood ex- scarred in terms of educational attainment posure effects. They also explore county char- and eventual employment. acteristics associated with better outcomes, as Patrick Sharkey’s paper in this issue sum- indicated by more positive causal effects and marizes key research on the role of neighbor- find better outcomes for children who grow up hoods in enhancing or limiting access to op- in counties with “less concentrated poverty, portunity. As noted, many of the mechanisms less income inequality, better schools, a larger at work at the neighborhood level relate to spa- share of two-parent families, and lower crime tial segregation, which limits exposure of rates” (Chetty and Hendren 2015, 1). members of different income groups, racial Covariates such as family structure or groups, cultural groups, or other dimension of school quality, though they have geographic segregation to each other. This can in turn per- patterns, are not inherently related to physical petuate the disadvantages suffered by one gen- adjacency in the way that social capital, crime eration, passing them on to their children. rates, and segregation are, because the latter These arguments are based on the idea that depend at least in part on neighborhood prox- neighborhoods are more than just the combi- imity. Thus, isolating geography’s role might nation of individuals who live there and in-

36. “Better” destination counties are those with more positive outcomes for children who live there throughout childhood compared with outcomes for children who live in the origin county throughout childhood. The outcome on which they focus is income as a young adult (age twenty-four or twenty-six), but they find similar “exposure” effects on college attendance, teenage birth rates, and marriage rates.

37. Julia Burdick-Will and her colleagues (2011), for example, report the conclusion of Dobbie and Fryer that test scores are unresponsive to changes in neighborhood environments in the absence of school-quality changes.

: 32 opportunity, mobility, and increased inequality stead also depend on peer effects, role models, influence, and also find that crime rates and social networks, and the like being important exposure to violence are negatively related to aspects of access to opportunity. children’s test scores.38 They note, however, Sharkey notes in this issue that observa- that this evidence is mostly circumstantial, tional studies find strong correlations be- based on large differences in levels of violent tween child neighborhood conditions and crime and neighborhood disadvantage be- adult economic outcomes and that the conse- tween the MTO cities of Baltimore and Chi- quences in terms of academic performance cago (where test score improvements did occur and educational attainment of growing up in among the MTO treatment group) and the a disadvantaged environment appear to be cu- other three MTO cities of Boston, New York, mulative; that is, outcomes worsen with length and Los Angeles. of exposure. Sharkey also summarizes his Chetty, Hendren, and Lawrence Katz (2015) joint work with Bryan Graham (2013), which bring new data to bear and successfully recon- examines the links between spatial segrega- cile the all-cities experimental MTO results tion and economic mobility, noting that the with observational studies, finding substantial tight connection between family economic neighborhood “exposure” effects on adult out- status and neighborhood economic status comes (earnings and college-going) of children that segregation creates will increase the whose families were offered the MTO voucher transmission of family economic status across requiring they move to a low-po verty neighbor- generations. Their paper confirms a relation- hood when they were young (younger than ship between mobility and spatial economic thirteen). They combine Internal Revenue Ser- segregation using three data sets, but notes vice data with MTO data to look at recent that the association does not provide evidence (young adult) outcomes (through 2012) and dif- of causation. ferentiate by age of the child when the family As Julia Burdick-Will and her colleagues was randomly assigned to one of the MTO (2011) point out, many nonexperimental stud- treatments or to control group status. They ies find substantial effects of neighborhood on find negative effects, sometimes significantly children’s life chances—findings that admit- different from zero for children who were older tedly include some bias from selection ef- when the MTO moves occurred, a finding that, fects—yet the Moving to Opportunity (MTO) they argue, may reflect disruption effects; experiment found no discernable neighbor- these would be offset for younger children by hood effects. Burdick-Will and her colleagues positive exposure effects proportional to the attempt to reconcile experimental, quasi- length of exposure to lower-poverty neighbor- experimental, and observational studies of hoods. neighborhood influence, and come out in the In summing up existing research on the role middle, concluding that some neighborhood of neighborhoods in access to opportunity, circumstances do matter for children’s out- Sharkey (this issue) notes the wide variation in comes. In particular, they argue that what opportunity and economic mobility across seems to matter is whether children live in the geographic areas in the United States. Investi- most economically distressed or dangerous gating sources of that variation, the research neighborhoods. They find little support for ei- mostly identifies correlations but has not suc- ther neighborhood differences in school qual- ceeded in establishing causal relationships or ity or racial composition playing a key role in even pinned down key mechanisms; as a re- children’s differential school performance out- sult, he notes, “as a whole, however, the re- comes. By contrast, they see concentrated search explaining geographic variation in eco- neighborhood disadvantage as an important nomic mobility remains at a very early stage.”

38. They measure “concentrated neighborhood disadvantage” following Sampson, Sharkey, and Raudenbush (2008), as a weighted average of neighborhood poverty, percentage of residents who are black, percentage of adults who are unemployed, percentage of households with a female head, percentage of residents on welfare, and percentage of residents under age eighteen.

: introduction 33

T he Relationship Between are unevenly measured or possibly because Inequality of Opportunity and such effects are not detectable or do not oper- Economic Growth ate across nations with different cultural and Our focus has been on the mechanisms institutional contexts. By contrast, Marrero through which barriers to opportunity arise and Rodriguez (2013) analyze inequality of op- and how they affect economic outcomes for in- portunity and growth across U.S. states and dividuals, along with potential policy reme- time and use eight circumstance groups, de- dies. This raises the question of what effect fined on the basis of four categories of parental policies that reduce barriers to opportunity educational attainment, cross-classified by two would likely have on aggregate economic per- racial groups. They find significant and nega- formance. If removing barriers to opportunity tive effects of inequality of opportunity on promotes economic growth, these spillovers growth, effects that persist through various ro- reduce the overall cost of such policies. bustness checks. They argue that “returns to The literature relating inequality of oppor- effort may encourage people to invest in edu- tunity and aggregate economic performance is cation and to exert an effort, while inequality extremely limited and most of it was developed of opportunity may not favor human and phys- out of the considerably more extensive litera- ical capital accumulation in the more talented ture relating overall inequality (that is, inequal- individuals” (120). ity of outcomes) and growth. Hypotheses re- Bradbury and Triest measure inequality or garding the effects of overall inequality on equality of opportunity in a different way in growth lay out plausible effects with both pos- this issue; they use indicators of relative and itive and negative signs and the empirical lit- absolute intergenerational mobility at the erature is correspondingly inconclusive. By commuting-zone level of geography within the contrast, theory suggests that inequality of op- United States that were developed and pub- portunity will be a drag on economic growth, lished by Chetty, Hendren, Kline, and Saez because individuals who lack opportunity will (2014a, 2014b). Intergenerational mobility mea- not be able to produce to their full potential sures reflect how equal or unequal opportunity and thus some capital will not be put to its is by indicating how tied a child’s adult income most productive use. is to the (parental-income) circumstances in which he or she grew up. Bradbury and Triest Evidence on Effects of Unequal find positive effects of intergenerational mobil- Opportunity on Growth ity, especially absolute mobility, on income Two papers that directly investigate the influ- growth between 2000 and 2013 in a cross- ence of inequality of opportunity on growth section of commuting zones, indicating that decompose total inequality into two parts, in- inequality of opportunity, as proxied by low equality of opportunity and a residual inequal- intergenerational mobility, acts as a drag on ity labeled inequality of effort, and include the local- area growth. two parts in a growth regression in place of a A fourth paper takes an entirely different total inequality measure (Ferreira et al. 2014; approach, investigating the addition to output Marrero and Rodriguez 2013). They measure associated with the reduction over the last sev- inequality of opportunity as the between- eral decades in the inequality of access to high- group dispersion in outcomes, where groups level occupations suffered earlier by women are defined in terms of circumstances individ- and blacks compared with white men in the uals face that are not within their control, as United States. Chang-Tai Hsieh and his col- in Roemer’s framework, summarized earlier. leagues (2013) do not directly estimate the ef- Francisco Ferreira and his colleagues (2014) fect of inequality of opportunity on growth. use a set of circumstance variables observed Instead, they mark the stark differences in oc- only selectively for individuals in an interna- cupations in the United States in 1960 between tional panel analysis. They fail to find negative white men, on one hand, and white women, effects of inequality of opportunity on growth, black women, and black men, on the other. perhaps because their circumstance groups They argue that, because these differences

: 34 opportunity, mobility, and increased inequality were so great, they cannot possibly be random system. As noted earlier, a rapidly growing em- or the result of unequal talent—white men, for pirical literature indicates that early childhood example, accounted for 94 percent of U.S. phy- interventions targeted at disadvantaged popu- sicians and 96 percent of lawyers in 1960—and lations have high social rates of return; this instead largely reflect unequal opportunity. topic is also addressed by Magnuson and Dun- They then estimate the addition to U.S. output can in their paper in this issue. Health, nutri- made by the degree to which those differences tion, and preschool programs help to lessen shrank in the ensuing years and attribute that the effect of poverty in early childhood on long- contribution—a remarkable 15 to 20 percent of term outcomes. Payoffs would also be positive U.S. economic growth—to equalization of op- for policies to lessen the link between the qual- portunity via great reductions in the barriers ity of elementary and secondary schooling and women and blacks face in their access to parental income or wealth, as discussed earlier. skilled professions, encouraging members of The Duncan and Murnane paper in this issue these formerly severely disadvantaged groups focuses on specific policies to raise K–12 edu- to invest in their own human capital and gain cational quality for disadvantaged children. the ability to contribute more fully to the na- Beyond high school, policies are also tional output. needed to break the link between family eco- These various approaches to measuring the nomic status and college attendance. Children impact of inequality of opportunity on growth of low-income parents are much less likely point to negative effects, at least in the United than their high-income counterparts to enroll States. (The international panel analysis by in and graduate from four-y ear college pro- Ferreira and his colleagues [2014] failed to find grams, even conditional on standardized test negative effects.) As noted earlier, this is not scores. Better-targeted financial aid programs, surprising, given that theory consistently sug- greater outreach to disadvantaged students, gests negative effects to the degree that un- and more widespread and effective compen equal opportunity prevents individuals from satory programs to guide disadvantaged stu- performing to their full potential and prevents dents through college have the potential to capital from being invested in the most high- reduce the effect of parents’ income on post- value projects. Although not unexpected, these secondary schooling investment. findings imply that the economic effects of un- Other policies are needed to provide more equal opportunity are large enough to be mea- opportunities for people to get back on track surable at the macro level of output or eco- after suffering the effects of being born disad- nomic growth and hence that policymakers vantaged. Children from disadvantaged back- have an additional impetus to reduce inequal- grounds are more likely to fall off standard ac- ity of opportunity. ademic and career tracks, and do not enjoy the same degree of insurance more affluent fami- Policies to Improve both Economic lies offer. Enhanced opportunities for adult Opportunity and Aggregate Performance education and degree completion, and pro- The empirical evidence reviewed suggests that grams to improve labor market and educa- removing barriers to economic opportunity tional opportunities for people with criminal may also improve aggregate economic perfor- records would attenuate some of the obstacles mance. Certain policies may have the potential facing those who have slipped off track. to achieve both of these objectives. Bernstein (2014) points toward policies that Sm um ary have positive output effects via enhancing op- The research reviewed offers grounds for both portunities without negative side-effects from optimism and pessimism regarding prospects reducing incentives for others to achieve or in- for addressing inequality of opportunity in the vest. His first policy recommendation is that United States. Barriers to economic opportu- governments should enhance their invest- nity are pervasive, and the growth in inequality ments aimed at helping disadvantaged chil- of outcomes has both increased the stakes as- dren overcome barriers in the U.S. educational sociated with confronting barriers and in-

: introduction 35 creased the difficulty in overcoming these ob- will exists to enact policies of sufficient scale stacles. Absent substantial increases in the and scope to address the problem. scope and scale of policy interventions, inequality of opportunity is likely to persist, A ppendix along with stagnation or deterioration of eco- This appendix reports the results of the mul- nomic mobility. On the other hand, research tiple regression analysis discussed earlier. The has made great strides in identifying and un- regression results underlie our analysis of derstanding the mechanisms relating barriers whether the Great Gatsby curve relationship to opportunity to economic outcomes. Al- between intergenerational mobility and in- though detailed research and analysis are re- equality of outcomes for the older generation quired to evaluate the success of specific poli- holds up when one controls for other factors cies in breaking down barriers to opportunity that arguably might also affect intergenera- and improving labor market outcomes, such tional mobility. Descriptive statistics (variable policies clearly exist and some have already means, etc.) for the variables used in the re- been implemented. In considering whether gressions are presented in table A1. the future will bring improvements in eco- Tables A2 and A3 report regression results nomic opportunity and mobility, the most in a cross-section of commuting zones, with problematic question and thus the largest area relative and absolute mobility as the depen- of uncertainty may be in whether the political dent variables, measures of inequality as

Table A1. Summary Statistics

Mean SD

Absolute mobility 43.94 5.681 Relative mobility 67.49 6.479 Gini (inequality) of parental income 0.410 0.0792 Parental middle class 0.550 0.0786 Top 1 percent income share 10.84 5.049 Per capita income, 1980 8.538 1.777 Per capita income, 1990 15.88 2.989 Per capita income growth, 1970–1980 149.2 26.42 Per capita income growth, 1980–1990 87.68 20.83 Foreign born, 1980 0.0252 0.0318 Foreign born, 1990 0.0275 0.0391 Workers with commute < fifteen minutes, 1980 0.508 0.142 Workers with commute < fifteen minutes, 1990 0.489 0.139 Households with kids headed by single mom, 1980 0.0491 0.0168 Households with kids headed by single mom, 1990 0.0582 0.0194 Less than high school, 1980 0.381 0.110 Less than high school, 1990 0.287 0.0907 More than high school, 1980 0.399 0.0809 More than high school, 1990 0.441 0.0912 Male labor force participation rate, 1980 72.27 6.050 Male labor force participation rate, 1990 70.62 5.787 Female labor force participation rate, 1980 46.07 5.966 Female labor force participation rate, 1990 52.87 6.206 Logarithm of population, 1980 11.57 1.406 Logarithm of population, 1990 11.60 1.454 Observations 709

Source: Authors’ calculations based on U.S. Census Bureau 2014c, U.S. Bureau of Eco- nomic Analysis 2014, and Chetty, Hendren, Kline, and Saez 2014b.

: 36 opportunity, mobility, and increased inequality

Table A2. Mobility Regressions, 1990 Demographics

Absolute Mobility Relative Mobility

(1) (2) (3) (4) (5) (6)

Gini (inequality) of parental –8.082*** –16.057*** –2.806 –16.735** income (2.148) (3.780) (3.133) (5.576) Parents middle class 16.581*** 11.903*** 12.154** 7.905+ (2.910) (3.090) (4.276) (4.559) Top 1 percent income share 0.143*** 0.214*** (0.041) (0.060) Per capita income growth, 0.009 0.007 0.011+ 0.007 0.008 0.010 1980–1990 (0.007) (0.006) (0.006) (0.010) (0.009) (0.009) Per capita income, 1990 –0.089 –0.036 –0.074 –0.460*** –0.409*** –0.481*** (0.063) (0.064) (0.065) (0.092) (0.093) (0.095) Foreign born, 1990 17.775*** 20.535*** 21.656*** 53.898*** 56.814*** 57.243*** (3.576) (3.590) (3.569) (5.215) (5.275) (5.264) Workers with commute 12.199*** 10.821*** 11.080*** 3.997 2.584 3.515 < fifteen minutes, 1990 (1.727) (1.736) (1.733) (2.519) (2.551) (2.557) Households with kids headed –122.617*** –96.319*** –89.355*** –137.107*** –114.768*** –105.832*** by single mom, 1990 (7.036) (8.945) (9.010) (10.262) (13.146) (13.291) Less than high school, 1990 –2.648 3.380 2.564 –8.714+ –3.584 –5.165 (3.616) (3.797) (3.768) (5.275) (5.581) (5.558) More than high school, 1990 7.000* 10.147** 10.493** 14.188** 17.475*** 16.639*** (3.188) (3.234) (3.227) (4.649) (4.752) (4.759) Male labor force participation 0.221*** 0.226*** 0.215*** 0.001 –0.002 –0.004 rate, 1990 (0.032) (0.031) (0.031) (0.046) (0.046) (0.046) Female labor force participation –0.250*** –0.249*** –0.267*** –0.009 –0.011 –0.032 rate, 1990 (0.034) (0.033) (0.033) (0.049) (0.049) (0.049) Logarithm of population, 1990 0.149 –0.023 0.119 –0.453+ –0.571* –0.402 (0.165) (0.165) (0.166) (0.240) (0.242) (0.245) Constant 41.513*** 26.052*** 33.224*** 83.439*** 72.856*** 79.590*** (3.635) (4.271) (4.558) (5.302) (6.277) (6.723)

Regional fixed effects Yes Yes Yes Yes Yes Yes

Observations 709 709 709 709 709 709 R2 0.781 0.787 0.792 0.642 0.646 0.652

Source: Authors’ calculations based on U.S. Census Bureau 2014c, U.S. Bureau of Economic Analysis 2014, and Chetty, Hendren, Kline, and Saez 2014b. +p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001

explanatory variables, and various other from 1980, just before they were born (1980 to commuting-zone characteristics as exogenous 1982). Columns 1 through 3 (absolute mobility and predetermined control variables that help as dependent variable) and 4 through 6 (rela- explain mobility. These regressions build on tive mobility) of tables A2 and A3 include alter- those reported elsewhere in this issue (Brad- native measures of inequality; specifically, col- bury and Triest, table 4). Table A2 uses 1990 umns 1 and 4 include the Gini coefficient to measures of those characteristics—when the measure inequality (as in figure 1), columns 2 children whose mobility is observed were and 5 the proportion middle class (as in figure about age ten, and table A3 uses measures 2), and columns 3 and 6 include the Gini, the

: introduction 37

Table A3. Mobility Regressions, 1980 Demographics

Absolute Mobility Relative Mobility

(1) (2) (3) (4) (5) (6)

Gini (inequality) of parental –5.855** –14.972*** –3.459 –16.073** income (1.845) (3.444) (2.930) (5.487) Parents middle class 11.530*** 7.000** 14.966*** 11.141** (2.438) (2.694) (3.867) (4.293) Top 1 percent income share 0.153*** 0.218*** (0.036) (0.058) Per capita income growth, 0.014** 0.015*** 0.017*** 0.025*** 0.027*** 0.028*** 1970–1980 (0.005) (0.005) (0.004) (0.007) (0.007) (0.007) Per capita income, 1980 –0.397*** –0.339** –0.420*** –0.890*** –0.820*** –0.917*** (0.105) (0.104) (0.104) (0.166) (0.165) (0.166) Foreign born, 1980 28.319*** 32.071*** 32.006*** 67.457*** 73.760*** 73.178*** (3.672) (3.787) (3.744) (5.832) (6.009) (5.965) Workers with commute < fifteen 11.413*** 10.540*** 11.188*** 3.251 1.760 2.818 minutes, 1980 (1.586) (1.591) (1.581) (2.518) (2.524) (2.518) Households with kids headed –162.948*** –140.580*** –133.013*** –154.203*** –121.291*** –109.104*** by single mom, 1980 (7.775) (9.547) (9.648) (12.348) (15.146) (15.372) Less than high school, 1980 –3.853 –1.263 –2.034 –12.532** –8.713* –9.649* (2.740) (2.794) (2.763) (4.352) (4.433) (4.403) More than high school, 1980 9.468** 10.542*** 10.278** 16.325** 18.828*** 17.659*** (3.159) (3.143) (3.122) (5.017) (4.987) (4.974) Male labor force participation 0.278*** 0.279*** 0.272*** 0.103** 0.097* 0.09* rate, 1980 (0.025) (0.024) (0.024) (0.039) (0.039) (0.039) Female labor force participation –0.245*** –0.248*** –0.262*** –0.077+ –0.079+ –0.101* rate, 1980 (0.026) (0.026) (0.026) (0.042) (0.041) (0.041) Logarithm of population, 1980 0.526*** 0.378* 0.519*** –0.204 –0.368 –0.228 (0.149) (0.149) (0.151) (0.236) (0.236) (0.240) Constant 31.422*** 21.541*** 27.994*** 74.786*** 63.204*** 68.971*** (3.455) (3.792) (4.095) (5.488) (6.017) (6.524)

Regional fixed effects Yes Yes Yes Yes Yes Yes

Observations 709 709 709 709 709 709 R2 0.815 0.818 0.823 0.641 0.648 0.655

Source: Authors’ calculations based on U.S. Census Bureau 2014c, U.S. Bureau of Economic Analysis 2014, and Chetty, Hendren, Kline, and Saez 2014b. +p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001

proportion middle class, and the income share rates for men and women, the proportion of of the top 1 percent in the CZ. workers with average commuting times of less The control variables include a rough mea- than fifteen minutes, and economic variables sure of past immigration (percentage of for- including per capita income, growth in per eign-born residents), the mix of educational capita income in the prior decade, and popula- attainments among adults ages twenty-fiv e tion size. The regressions also include fixed ef- and older in the CZ, the proportion of single- fects for census divisions. mother households, labor force participation The strong negative coefficient on the Gini

: 38 opportunity, mobility, and increased inequality coefficient in the scatter plot lines persists into NBER-EASE vol. 19. Chicago: University of Chi- column 1 in the two tables (with 1990 or 1980 cago Press. controls) for absolute mobility but is not sta- Almond, Douglas, Hilary W. Hoynes, and Diane tistically significant for relative mobility (col- Whitmore Schanzenbach. 2011. “Inside the War umn 4). In both cases, the slope coefficient is on Poverty: The Impact of Food Stamps on Birth much smaller when controls are added; in the Outcomes.” Review of Economics and Statistics relative mobility case, the estimate is so small 93(2): 387–403. that it is dwarfed by the standard error. How- Altonji, Joseph G., and Richard K. Mansfield. 2011. ever, columns 2 and 5 indicate that even for “The Role of Family, School, and Community relative mobility, the size of the (parental) mid- Characteristics in Inequality in Education and dle class has a statistically significant positive Labor-Market Outcomes.” In Whither Opportu- relationship with children’s mobility. And both nity? Rising Inequality, Schools, and Children’s estimated relationships show up in columns 3 Life Chances, edited by Greg J. Duncan and and 6, where the Gini and proportion middle Richard J. Murnane. New York: Russell Sage class obtain coefficient estimates significantly Foundation. different from zero, as does the size of the in- Aud, Susan, William Hussar, Frank Johnson, Grace come share of the richest 1 percent of families Kena, Erin Roth, Eileen Manning, Xiaolei Wang, in the commuting zone. The latter coefficient and Jijun Zhang. 2012. The Condition of Educa- is opposite in sign to that of the Gini, even tion 2012. NCES 2012–045. Washington: U.S. though they are both indicators of inequality; Department of Education. this sign reversal for the top 1 percent income Aud, Susan, Angelina KewalRamani, and Lauren share suggests that while inequality in the Frohlich. 2011. America’s Youth: Transitions to lower parts of the CZ income distribution con- Adulthood. NCES 2012–026. Washington: U.S. strains mobility, inequality at the very top does Department of Education. not. The use of 1980 measures (table A3) versus Avery, Christopher, Jessica S. Howell, and Lindsay 1990 (table A2) for the control variables makes Page. 2014. “A Review of the Role of College little difference to the coefficients on the in- Counseling, Coaching, and Mentoring on Stu- equality measures. dents’ Postsecondary Outcomes.” College Board Research, Research Brief (October). New York: R eferences The College Board. Accessed January 3, 2016. Aaronson, Daniel, and Bhashkar Mazumder. 2008. http://research.collegeboard.org/sites/default “Intergenerational Economic Mobility in the /files/publications/2015/1/college-board United States, 1940 to 2000.” Journal of Human -research-brief-role-college-counseling-coach Resources 43(1): 139–72. ing-mentoring-postsecondary-outcomes.pdf. Almond, Douglas. 2006. “Is the 1918 Influenza Pan- Baker, Bruce D. 2009. “Within-Dist rict Resource demic Over? Long-Term Effects of in Utero Influ- Allocation and the Marginal Costs of Provid- enza Exposure in the Post-1940 U.S. Population.” ing Equal Educational Opportunity: Evidence Journal of Political Economy 114(4): 672–712. from Texas and Ohio.” Education Policy Analy- Almond, Douglas, and Janet Currie. 2011a. “Human sis Archives 17(3): 1–31. Accessed January 3, Capital Development Before Age Five.” In Hand- 2016. http://files.eric.ed.gov/fulltext/EJ83 book of Labor Economics, vol. 4B, edited by Da- 5083.pdf. vid Card and Orley Ashenfelter. Philadelphia, Pa.: Bailey, Martha J., and Susan M. Dynarski. 2011. “In- Elsevier. equality in Postsecondary Education.” In Whither ———. 2011b. “Killing Me Softly: The Fetal Origins Opportunity? Rising Inequality, Schools, and Chil- Hypothesis.” Journal of Economic Perspectives dren’s Life Chances, edited by Greg J. Duncan 25(3): 153–72. and Richard J. Murnane. New York: Russell Sage Almond, Douglas, Lena Edlund, Hongbin Li, and Jun- Foundation. sen Zhang. 2010. “Long-Term Effects of Early- Becker, Gary, and Nigel Tomes. 1979. “An Equilib- Life Development: Evidence from the 1959 to rium Theory of the Distribution of Income and In- 1961 China Famine.” In The Economic Conse- tergenerational Mobility.” Journal of Political quences of Demographic Change in East Asia, Economy 87(6): 1153–89.

: introduction 39

Belley, Philippe, and Lance Lochner. 2007. “The bility in the United States and Great Britain: A Changing Role of Family Income and Ability in Comparative Study of Parent-C hild Pathways.” Determining Educational Achievement.” NBER Review of Income and Wealth 60(3) (September): working paper 13527. Cambridge, Mass.: Na- 425–49. tional Bureau of Economic Research. Bloome, Deirdre. 2015. “Income Inequality and Inter- Bernstein, Jared. 2014. “The Impact of Inequality on generational Income Mobility in the United Growth.” Washington, D.C.: Center for American States.” Social Forces 93(3): 1047–80. Progress. Bloome, Dierdre, and Bruce Western. 2011. “Cohort Bettinger, Eric P., Bridget Terry Long, Philip Oreo- Change and Racial Differences in Educational poulos, and Lisa Sanbonmatsu. 2012. “The Role and Income Mobility.” Social Forces 90(2): 375– of Application Assistance and Information in Col- 95. lege Decisions: Results from the H&R Block Brunori, Paolo, Francisco H.G. Ferreira, and Vito Per- FAFSA Experiment.” Quarterly Journal of Eco- agine. 2013. “Inequality of Opportunity, Income nomics 127(3): 1205–42. Inequality and Economic Mobility: Some Interna- Bingley, Paul, Miles Corak, and Niels Westergard- tional Comparisons.” Policy Research working Nielsen. 2011. “The Intergenerational Transmis- paper no. 6304. Washington, D.C.: The World sion of Employers in Canada and Denmark.” IZA Bank. discussion paper series no. 5593 (March). Bonn: Burdick-Will, Julia, Jens Ludwig, Stephen W. Institute for the Study of Labor. Raudenbush, Robert J. Sampson, Lisa Sanbon- Bischoff, Kendra, and Sean F. Reardon. 2014. “Resi- matsu, and Patrick Sharkey. 2011. “Converging dential Segregation by Income 1970–2009.” In Evidence on Neighborhood Effects on Children’s Diversity and Disparities: America Enters a New Test Scores: An Experimental, Quasi- Century, edited by John Logan. New York: Russell Experimental, and Observational Comparison.” In Sage Foundation. Whither Opportunity? Rising Inequality, Schools, Bjorklund, Anders, and Markus Jäntti. 2009. “Inter- and Children’s Life Chances, edited by Greg J. generational Income Mobilty and the Role of Duncan and Richard J. Murnane. New York: Rus- Family Background.” In The Oxford Handbook of sell Sage Foundation. Economic Inequality, edited by Wiemer Salverda, Card, David, and A. Abigail Payne. 2002. “School Fi- Brian Nolan, and Timothy M. Smeeding. Oxford: nance Reform, the Distribution of School Spend- Oxford University Press. ing, and the Distribution of Student Test Scores.” Black, Sandra E., and Paul J. Devereux. 2011. “Recent Journal of Public Economics 83(1): 49–82. Developments in Intergenerational Mobility.” In Carrell, Scott, and Bruce Sacerdote. 2013. “Late In- Handbook of Labor Economics, edited by David terventions Matter Too: The Case of College Card and Orley Ashenfelter. Philadelphia, Pa.: El- Coaching in New Hampshire.” Cambridge, Mass.: sevier. National Bureau of Economic Research. Black, Sandra E., Paul J. Devereux, and Kjell G. Sal- Case, Anne, Darren Lubotsky, and Christina Paxson. vanes. 2007. “From the Cradle to the Labor Mar- 2002. “Economic Status and Health in Child- ket? The Effect of Birth Weight on Adult Out- hood: The Origins of the Gradient.” American comes.” Quarterly Journal of Economics 122(1): Economic Review 92(5): 1308–34. 409–39. Chadwick, Laura, and Gary Solon. 2002. “Intergen- Blanden, Jo. 2013. “Cross-Count ry Rankings in Inter- erational Income Mobility Among Daughters.” generational Mobility: A Comparison of Ap- American Economic Review 92(1)(March): 335– proaches from Economics and Sociology.” Jour- 44. nal of Economic Surveys 27(1): 38–73. Chaudhary, Latika. 2009. “Education Inputs, Student Blanden, Jo, Paul Gregg, and Lindsey Macmillan. Performance and School Finance Reform in 2007. “Accounting for Intergenerational Income Michigan.” Economics of Education Review 28(1): Persistence: Noncognitive Skills, Ability, and Ed- 90–98. ucation.” Economic Journal 117(519) (March): Chetty, Raj, and Nathaniel Hendren. 2015. “The Im- C43–60. pacts of Neighborhoods on Intergenerational Mo- Blanden, Jo, Robert Haveman, Timothy Smeeding, bility: Childhood Exposure Effects and County- and Kathryn Wilson. 2014. “Intergenerational Mo- Level Estimates.” Cambridge, Mass.: Harvard

: 40 opportunity, mobility, and increased inequality

University. Accessed January 3, 2016. http:// Cunha, Flavio, James J. Heckman, and Susanne M. scholar.harvard.edu/files/hendren/files/nbhds_ Schennach. 2010. “Estimating the Technology of paper.pdf?m=1430722623. Cognitive and Noncognitive Skill Formation.” Chetty, Raj, Nathaniel Hendren, and Lawrence F. Econometrica 78(3): 883–931. Katz. 2015. “The Effects of Exposure to Better Currie, Janet. 2011. “Inequality at Birth: Some Neighborhoods on Children: New Evidence from Causes and Consequences.” American Economic the Moving to Opportunity Experiment.” NBER Review: Papers and Proceedings101(3): 1–22. working paper no. 21156. Cambridge, Mass.: Na- Currie, Janet, and Enrico Moretti. 2007. “Biology as tional Bureau of Economic Research.. Destiny? Short-and Long-R un Determinants of Chetty, Raj, Nathaniel Hendren, Patrick Kline, and Intergenerational Transmission of Birth Weight.” Emmanuel Saez. 2014a. “Where Is the Land of Journal of Labor Economics 25(2): 231–63. Opportunity? The Geography of Intergenera- Currie, Janet, and Maya Rossin-Slater. 2015. “Early- tional Mobility in the United States.” Quarterly Life Origins of Life-Cy cle Well-Being: Research Journal of Economics 129(4): 1553–623. and Policy Implications.” Journal of Policy Analy- ———. 2014b. “Data from Chetty, Hendren, Kline, and sis and Management 34(1): 208–42. Saez (2014): Descriptive Statistics by County Dahl, Gordon B., and Lance Lochner. 2012. “The Im- and Commuting Zone.” Data downloaded from pact of Family Income on Child Achievement: the Equality of Opportunity Project. Accessed Evidence from the Earned Income Tax Credit.” February 2, 2016. http://equality-of-opportunity American Economic Review 102(5): 1927–56. .org/index.php/data. Deming, David. 2009. “Early Childhood Intervention Chetty, Raj, Nathaniel Hendren, Patrick Kline, Em- and Life-Cycle Skill Development: Evidence from manuel Saez, and Nicholas Turner. 2014. “Is the Head Start.” American Economic Journal: Applied United States Still a Land of Opportunity? Re- Economics 1(3): 111–34. cent Trends in Intergenerational Mobility.” Ameri- Duncan, Greg. J., Kenneth T. H. Lee, Ariel Kalil, and can Economic Review: Papers and Proceedings Kathleen M. Ziol-Guest . 2015. “Parent Income- 104(5): 141–47. Based Gaps in Schooling, Earnings and Family Corak, Miles. 2013a. “Inequality from Generation to Income: Cross-Cohort Trends in the NLSYs and Generation: The United States in Comparison.” the PSID.” Accessed February 2, 2016. http:// In The Economics of Inequality, Poverty, and Dis- sites.uci.edu/gduncan/files/2013/06/RSF crimination in the 21st Century, edited by Robert -Memo-Duncan-et-al-010815.pdf. S. Rycroft. Santa Barbara, Calif.: ABC-CLIO . Duncombe, William, Phuong Nguyen-H oang, and ———. 2013b. “Income Inequality, Equality of Oppor- John Yinger. 2015. “Measurement of Cost Differ- tunity, and Intergenerational Mobility” Journal of entials” In Handbook of Research in Education Economic Perspectives 27(3): 79–102. and Policy, edited by Helen F. Ladd and Margaret Corak, Miles, and Patrizio Piraino. 2011. “The Inter- E. Goertz. New York: Routledge. generational Transmission of Employers.” Journal Economic Policy Institute. 2011. “Unpaid and Ex- of Labor Economics 29(1): 37–68. ploited? Examining Interns in the U.S. Labor Cunha, Flavio, and James Heckman. 2007. “The Market.” EPI forum and discussion, May 18. Technology of Skill Formation.” American Eco- Washington, D.C.: Economic Policy Institute. Ac- nomic Review Papers and Proceedings 97(2): 31– cessed December 27, 2015. http://www.epi.org 47. /event/unpaid-and-exploited-examining-interns ———. 2008. “Formulating, Identifying and Estimat- -in-the-u-s-labor-market. ing the Technology of Cognitive and Noncogni- Eisenbrey, Ross. 2012. “Unpaid Internships Hurt Mo- tive Skill Formation.” Journal of Human Re- bility.” Economic Policy Institute blog post, Janu- sources 43(4): 738–82. ary 12. Accessed April 7, 2016. http://www.epi Cunha, Flavio, James J. Heckman, Lance Lochner, .org/blog/unpaid-internships-economic and Dimitriy V. Masterov. 2006. “Interpreting the -mobility/. Evidence on Life Cycle Skill Formation.” In Hand- Ferreira, Francisco H. G., Christoph Lakner, Maria book of the Economics of Education, edited by Ana Lugo, and Berk Ozler. 2014. “Inequality of Eric Hanushek and Finis Welch, vol. 1. Philadel- Opportunity and Economic Growth: A Cross- phia, Pa.: Elsevier. Country Analysis.” Policy Research working

: introduction 41

paper no. 6915. Washington, D.C.: The World Hoynes, Hilary W., Douglas L. Miller, and David Si- Bank. mon. 2012. “Income, the Earned Income Tax Figlio, David, Jonathan Guryan, Krzysztof Karbownik, Credit, and Infant Health.” NBER working paper and Jeffrey Roth. 2014. “The Effects of Poor Neo- no. 18206. Cambridge, Mass.: National Bureau of natal Health on Children’s Cognitive Economic Research. Development.”American Economic Review Hsieh, Chang-Tai, Eric Hurst, Charles I. Jones, and 104(12): 3921–55. Peter J. Klenow. 2013. “The Allocation of Talent Fishkin, Joseph. 2014. Bottlenecks: A New Theory of and U.S. Economic Growth.” NBER working pa- Equal Opportunity. Oxford: Oxford University per no. 18693. Cambridge, Mass.: National Bu- Press. reau of Economic Research. ———. 2016. “From Subgroups to Bottlenecks: New Ionnides, Yannis M., and Linda Datcher Loury. 2004. Directions for the Empirical Study of Intergener- “Job Information Networks, Neighborhood Ef- ational Mobility: Comments on ‘Multiple Barriers fects, and Inequality.” Journal of Economic Litera- to Economic Opportunity in the United States.’” ture 42(4): 1056–93. In Additional Materials. RSF: The Russell Sage Isen, Adam, Maya Rossin-Slater, and Reed Walker. Journal of the Social Sciences, volume 2, issue 2. 2014. “Every Breath You Take—Every Dollar You’ll Available at: http://www.rsfjournal.org. Make: The Long-Term Consequences of the Fleurbaey, Marc, and Vito Peragine. 2013. “Ex Ante Clean Air Act of 1970.” NBER working paper no. Versus Ex Post Equality of Opportunity.” Eco- 19858. Cambridge, Mass.: National Bureau of nomica vol. 80(317): 118–30. Economic Research. Fox, Mary Anne, Brooke A. Connolly, and Thomas D. Jackson, C. Kirabo, Rucker Johnson, and Claudia Snyder. 2005. Youth Indicators 2005: Trends in Persico. 2014. “The Effect of School Finance Re- the Well-Being of American Youth. NCES 2005– forms on the Distribution of Spending, Academic 050. Washington: U.S. Department of Education. Achievement, and Adult Outcomes.” NBER work- Garces, Eliana, Duncan Thomas, and Janet Currie. ing paper no. 20118. Cambridge, Mass.: National 2002. “Longer-Term Effects of Head Start.” Bureau of Economic Research. American Economic Review 92(4): 999–1012. Jacob, Brian A., and Tamara Wilder Linkow. 2011. Goodman, Joshua. 2008. “Skills, Schools, and Credit “Educational Expectations and Attainment.” In Constraints: Evidence from Massachusetts.” De- Whither Opportunity? Rising Inequality, Schools, partment of Economics discussion paper series and Children’s Life Chances, edited by Greg J. no. 0809–03. New York: Columbia University. Duncan and Richard J. Murnane. New York: Rus- Graham, Bryan, and Patrick Sharkey. 2013. “Mobility and the Metropolis: The Relationship between In- sell Sage Foundation. equality in Urban Communities and Economic Jäntti, Markus, and Stephen P. Jenkins. 2015. “In- Mobility.” Washington, D.C.: The Pew Charitable come Mobility.” In Handbook of Income Distribu- Trusts. tion, edited by Anthony B. Atkinson and François Holzer, Harry J. 1988. “Search Method Use by Unem- Bourguignon, vol. 2A. Philadelphia, Pa.: Elsevier. ployed Youth” Journal of Labor Economics 6(1): Kaushal, Neeraj, Katherine Magnuson, and Jane 1–20. Waldfogel. 2011. “How Is Family Income Related Hoxby, Caroline M. 1998. “How Much Does Spend- to Investments in Children’s Learning?” In ing Depend on Family Income? The Historical Whither Opportunity? Rising Inequality, Schools, Origins of the Current School Finance Dilemma.” and Children’s Life Chances, edited by Greg J. Papers and Proceedings of the American Eco- Duncan and Richard J. Murnane. New York: Rus- nomic Association 88(2): 309–14. sell Sage Foundation. ———. 2001. “All School Finance Equalizations Are Kena, Grace., Susan Aud, Frank Johnson, Xiaolei Not Created Equal.” Quarterly Journal of Econom- Wang, Jujun Zhang, Amy Rathbun, Sidney ics 116 (November): 1189–231. Wilkinson-Flicker, and Paul Kristapovich. 2014. Hoxby, Caroline M., and Christopher Avery. 2013. The Condition of Education 2014. NCES 2014– “‘The Missing “One-Off s’: The Hidden Supply of 083. Washington: U.S. Department of Education. High-Achieving, Low Income Students.” Brook- Kenworthy, Lane. 2012. “It’s Hard to Make It in ings Papers on Economic Activity (Spring): 1–65. America: How the United States Stopped Being

: 42 opportunity, mobility, and increased inequality

the Land of Opportunity.” Foreign Affairs 91(6): sion paper no. 6672. Bonn: Institute for the 97–109. Study of Labor. Lane, Brett, and Phomdaen Souvanna. 2013. “Mass Reardon, Sean F. 2011. “The Widening Academic Math + Science Initiative 2013 High School Im- Achievement Gap Between the Rich and the pact Study.” Boston, Mass.: Institute for Strategic Poor: New Evidence and Possible Explanations.” Leadership and Learning. In Whither Opportunity? Rising Inequality, Lareau, Annette, and Kimberly Goyette. 2014. Pref- Schools, and Children’s Life Chances, edited by ace. In Choosing Homes, Choosing Schools. New Greg J. Duncan and Richard J. Murnane. New York: Russell Sage Foundation. York: Russell Sage Foundation. Lee, Chul-In, and Gary Solon. 2009. “Trends in In- Roemer, John E. 1993. “A Pragmatic Theory of Re- tergenerational Income Mobility.” Review of Eco- sponsibility for the Egalitarian Planner.” Philoso- nomics and Statistics 91(4): 766–72. phy and Public Affairs 22(2): 144–66. Levine, David E., and Bhashkar Mazumder. 2007. ———. 1998. Theories of Distributive Justice. Cam- “The Growing Importance of Family: Evidence bridge, Mass.: Harvard University Press. from Brothers’ Earnings.” Industrial Relations Roemer, John E., and Alain Trannoy. Forthcoming. 46(1): 7–21. “Equality of Opportunity: Theory and Measure- Marrero, Gustavo A., and Juan G. Rodriguez. 2013. ment.” Journal of Economic Literature. “Inequality of Opportunity and Growth.” Journal Sacerdote, Bruce. 2011. “Peer Effects in Education: of Development Economics 104 (September): How Might They Work, How Big Are They and 107–22. How Much Do We Know Thus Far?” In Handbook Mazumder, Bhashkar. 2011. “Black-White Differences of the Economics of Education, vol. 3, edited by in Intergenerational Economic Mobility in the Eric A Hanushek, Stephen J. Machin, and Ludger United States.” Working paper no. 2011-10. Chi- Woessmann. Philadelphia, Pa.: Elsevier. cago: Federal Reserve Bank of Chicago. Sampson, Robert J., Patrick Sharkey, and Stephen Milanovic, Branko. 2015. “Global Inequality of Op- Raudenbush. 2008. “Durable Effects of Concen- portunity: How Much of Our Income Is Deter- trated Disadvantage on Verbal Ability Among mined by Where We Live?” Review of Economics African-American Children.” Proceedings of and Statistics 97(2): 452–60. the National Academy of Sciences 105(3): 845– Mitnik, Pablo A., Erin Cumberworth, and David B. 53. Grusky. 2013. “Social Mobility in a High Inequal- Smeeding, Timothy M. 2013. “On the Relationship ity Regime.” Stanford Center for the Study of Between Income Inequality and Intergenerational Poverty and Inequality. Stanford, Calif: Stanford Mobility.” GINI discussion paper no. 89. Madison: University. University of Wisconsin. Accessed April 7, 2016. Palardy, Gregory J. 2013. “High School Socioeco- http://gini-research.org/system/uploads/589 nomic Segregation and Student Attainment.” /original/89.pdf?1385131374. American Education Research Journal 50(4): 714– Solon, Gary. 1999. “Intergenerational Mobility in the 54. Labor Market.” In Handbook of Labor Economics, Putnam, Robert D. 2000. Bowling Alone: The Col- edited by Orley Ashenfelter and David Card. Phil- lapse and Revival of American Community. New adelphia, Pa.: Elsevier. York: Simon & Schuster. ———. 2002. “Cross-Country Difference in Intergen- ———. 2015. Our Kids: The American Dream in Crisis. erational Earnings Mobility.” Journal of Economic New York: Simon & Schuster. Perspectives 16(3): 59–66. ———. 2016. “Social Networks and Geography: Com- ———. 2004. “A Model of Intergenerational Mobility ments on ‘How Does Geography Affect Equality Variation over Time and Place.” In Generational of Opportunity.’” In Additional Materials. RSF: The Income Mobility in North America and Europe, Russell Sage Journal of the Social Sciences, vol- edited by Miles Corak. Cambridge: Cambridge ume 2, issue 2. Available at: http://www.rsf University Press. journal.org. U.S. Bureau of Economic Analysis. 2014. “Regional Ramos, Xavier, and Dirk Van de Gaer. 2012. “Empiri- Data: GDP and Personal Income,” tables CA1–3 cal Approaches to Inequality of Opportunity: (Personal income summary), CA4 (Personal in- Principles, Measures, and Evidence.” IZA discus- come and employment summary), CA25 (Total

: introduction 43

full- time and part-time employment by SIC in- sion 2.0. Minneapolis, Minn.: University of Minne- dustry), and CA25N (Total full-time and part- sota. Accessed August 27, 2014. https://nhgis time employment by NAICS industry). Washing- .org/. ton: U.S. Government Printing Office. Accessed U.S. Department of Labor, Wage and Hour Division. September 17, 2014. http://bea.gov/itable 2010. “Fact Sheet #71: Internship Programs un- /iTable.cfm?ReqID=70&step=1#reqid=70&step=1 der the Fair Labor Standards Act.” Washington: &isuri=1. U.S. Government Printing Office. Accessed De- U.S. Census Bureau. 2014a. “Selected Measures of cember 2, 2015. http://www.dol.gov/whd/regs Household Income Dispersion: 1967 to 2014.” /compliance/whdfs71.pdf. Washington: U.S. Government Printing Office. Urahn, Susan K., Erin Currier, Dana Elliott, Lauren Accessed December 15, 2014. https://www Wechsler, Denise Wilson, and Daniel Colbert. .census.gov/hhes/www/income/data/historical 2012. Pursuing the American Dream: Economic /inequality/Table%20IE-1.pdf. Mobility Across Generations. Washington, D.C.: ———. 2014b. Census tract data accessed via Minne- Pew Charitable Trusts. sota Population Center. National Historical Geo- graphic Information System: Version 2.0. Minne- Wilson, William Julius. 1987. The Truly Disadvan- apolis, Minn.: University of Minnesota. Accessed taged: The Inner City, the Underclass, and Public December 19, 2014. https://nhgis.org/. Policy. Chicago: University of Chicago Press. ———. 2014c. Data from decennial censuses ac- Winship, Scott, and Stephanie Owen. 2013. “Guide cessed via Minnesota Population Center. National to the Brookings Social Genome Model.” Wash- Historical Geographic Information System: Ver- ington, D.C.: Brookings Institution.