Chapter 10

Keywords

Racial achievement gap; Charter schools; Racial inequality

1 Introduction

Racial inequality is an American tradition. Relative to whites, blacks earn twenty-four percent less, live five fewer years, and are six times more likely to be incarcerated on a given day. Hispanics earn twenty-five percent less than whites and are three times more likely to be incarcerated.² At the end of the 1990s, there were one-third more black men under the jurisdiction of the corrections system than there were enrolled in colleges or universities (Ziedenberg and Schiraldi, 2002). While the majority of barometers of economic and social progress have increased substantially since the passing of the civil rights act, large disparities between racial groups have been and continue to be an everyday part of American life.

Understanding the causes of current racial inequality is a subject of intense debate. A wide variety of explanations—which range from genetics (Jensen, 1973; Rushton, 1995) to personal and institutional discrimination (Darity and Mason, 1998; Pager, 2007; Krieger and Sidney, 1996) to the cultural backwardness of minority groups (Reuter, 1945; Shukla, 1971)—have been put forth. Renowned sociologist William Julius Wilson argues that a potent interaction between poverty and racial discrimination can explain current disparities (Wilson, 2010).

Decomposing the share of inequality attributable to these explanations is exceedingly difficult, as experiments (field, quasi-, or natural) or other means of credible identification are rarely available.³ Even in cases where experiments are used (i.e., audit studies), it is unclear precisely what is being measured (Heckman, 1998). The lack of success in convincingly identifying root causes of racial inequality has often reduced the debate to a competition of “name that residual”—arbitrarily assigning identity to unexplained differences between racial groups in economic outcomes after accounting for a set of confounding factors. The residuals are often interpreted as “discrimination,” “culture,” “genetics,” and so on. Gaining a better understanding of the root causes of racial inequality is of tremendous importance for social policy, and the purpose of this chapter.

This chapter contains three themes. First, relative to the 20th century, the significance of discrimination as an explanation for racial inequality across economic and social indicators has declined. Racial differences in social and economic outcomes are greatly reduced when one accounts for educational achievement; therefore, the new challenge is to understand the obstacles undermining the achievement of black and Hispanic children in primary and secondary school. Second, analyzing ten large datasets that include children ranging in age from eight months old to seventeen years old, we demonstrate that the racial achievement gap is remarkably robust across time, samples, and particular assessments used. The gap does not exist in the first year of life, but black students fall behind quickly thereafter and observables cannot explain differences between racial groups after kindergarten.

Third, we provide a brief history of efforts to close the achievement gap. There are several programs—various early childhood interventions, more flexibility and stricter accountability for schools, data-driven instruction, smaller class sizes, certain student incentives, and bonuses for effective teachers to teach in high-need schools, which have a positive return on investment, but they cannot close the achievement gap in isolation.⁴ More promising are results from a handful of high-performing charter schools, which combine many of the investments above in a comprehensive model and provide a powerful “existence proof”—demonstrating that a few simple investments can dramatically increase the achievement of even the poorest minority students.

An important set of questions is: (1) whether one can boil the success of these charter schools down to a form that can be taken to scale in traditional public schools; (2) whether we can create a competitive market in which only high-quality schools can thrive; and (3) whether alternative reforms can be developed to eliminate achievement gaps. Closing the racial achievement gap has the potential to substantially reduce or eliminate many of the social ills that have plagued minority communities for centuries.

2 The declining significance of discrimination

One of the most important developments in the study of racial inequality has been the quantification of the importance of pre-market skills in explaining differences in labor market outcomes between blacks and whites (Neal and Johnson, 1996; O’Neill, 1990). Using the National Longitudinal Survey of Youth 1979 (NLSY79), a nationally representative sample of 12,686 individuals aged 14 to 22 in 1979, Neal and Johnson (1996) find that educational achievement among 15- to 18-year-olds explains all of the black-white gap in wages among young women and 70% of the gap among men. Accounting for pre-market skills also eliminates the Hispanic-white gap. Important critiques such as racial bias in the achievement measure (Darity and Mason, 1998; Jencks, 1998), labor market dropouts, or the potential that forward-looking minorities underinvest in human capital because they anticipate discrimination in the market cannot explain the stark results.⁵

We begin by replicating the seminal work of Neal and Johnson (1996) and extending their work in four directions. First, the most recent cohort of NLSY79 is between 42 and 44 years old (15 years older than in the original analysis), which provides a better representation of the lifetime gap. Second, we perform a similar analysis with the National Longitudinal Survey of Youth 1997 cohort (NLSY97). Third, we extend the set of outcomes to include unemployment, incarceration, and measures of physical health. Fourth, we investigate the importance of pre-market skills among graduates of thirty-four elite colleges and universities in the College and Beyond database, 1976 cohort.

To understand the importance of academic achievement in explaining life outcomes, we follow the lead of Neal and Johnson (1996) and estimate least squares models of the form:

(1)

where i indexes individuals, X_i denotes a set of control variables, and R_i is a full set of racial identifiers.

Table 1 presents racial disparities in wage and unemployment for men and women, separately.⁶ The odd-numbered columns present racial differences on our set of outcomes controlling only for age. The even-numbered columns add controls for the Armed Forces Qualifying Test (AFQT)—a measure of educational achievement that has been shown to be racially unbiased (Wigdor and Green, 1991)—and its square. Black men earn 39.4% less than white men; black women earn 13.1% less than white women. Accounting for educational achievement drastically reduces these inequalities—39.4% to 10.9% for black men and 13.1% lower than whites to 12.7% higher for black women.⁷ An eleven percent difference between white and black men with similar educational achievement is a large and important number, but a small fraction of the original gap. Hispanic men earn 14.8% less than whites in the raw data—62% less than the raw black-white gap—which reduces to 3.9% more than whites when we account for AFQT. The latter is not statistically significant. Hispanic women earn six percent less than white women (not significant) without accounting for achievement. Adding controls for AFQT, Hispanic women earn sixteen percent more than comparable white women and these differences are statistically significant.

Labor force participation follows a similar pattern. Black men are more than twice as likely to be unemployed in the raw data and thirty percent more likely after controlling for AFQT. For women, these differences are 3.8 and 2.9 times more likely, respectively. Hispanic-white differences in unemployment with and without controlling for AFQT are strikingly similar to black-white gaps.

Table 2 replicates Table 1 using the NLSY97.⁸ The NLSY97 includes 8984 youths between the ages of 12 and 16 at the beginning of 1997; these individuals are 21 to 27 years old in 2006–2007, the most recent years for which wage measures are available. In this sample, black men earn 17.9% less than white men and black women earn 15.3% less than white women. When we account for educational achievement, racial differences in wages measured in the NLSY97 are strikingly similar to those measured in NLSY79— 10.9% for black men and 4.4% for black women. The raw gaps, however, are much smaller in the NLSY97, which could be due either to the younger age of the workers and a steeper trajectory for white males (Farber and Gibbons, 1996) or to real gains made by blacks in recent years. After adjusting for age, Hispanic men earn 6.5% less than white men and Hispanic women earn 5.7% less than white women, but accounting for AFQT eliminates the Hispanic-white gap for both men and women.

Black men in the NLSY97 are almost three times as likely to be unemployed, which reduces to twice as likely when we account for educational achievement. Black women are roughly two and a halftimes more likely to be unemployed than white women, but controlling for AFQT reduces this gap to seventy-five percent more likely. Hispanic men are twenty-five percent more likely to be unemployed in the raw data, but when we control for AFQT, this difference is eliminated. Hispanic women are fifty percent more likely than white women to be unemployed and this too is eliminated by controlling for AFQT. Similar to the NLSY79, controlling for AFQT has less of an impact on racial differences in unemployment than on wages.

Table 3 employs a Neal and Johnson specification on two social outcomes: incarceration and physical health. The NLSY79 asks the “type of residence” in which the respondent is living during each administration of the survey, which allows us to construct a measure of whether the individual was ever incarcerated when the survey was administered across all years of the sample.⁹ The NLSY97 asks individuals if they have been sentenced to jail, an adult corrections institution, or a juvenile corrections institution in the past year for each yearly follow-up survey of participants. In 2006, the NLSY79 included a 12-Item Short Form Health Survey (SF-12) for all individuals over age 40. The SF-12 consists of twelve self-reported health questions ranging from whether the respondent’s health limits him from climbing several flights of stairs to how often the respondent has felt calm and peaceful in the past four weeks. The responses to these questions are combined to create physical and mental component summary scores.

Adjusting for age, black males are about three and a half times and Hispanics are about two and a half times more likely to have ever been incarcerated when surveyed.¹⁰ Controlling for AFQT, this is reduced to about eighty percent more likely for blacks and fifty percent more likely for Hispanics. Again, the racial differences in incarceration after controlling for achievement is a large and important number that deserves considerable attention in current discussions of racial inequality in the United States. Yet, the importance of educational achievement in the teenage years in explaining racial differences is no less striking.

The final two columns of Table 3 display estimates from similar regression equations for the SF-12 physical health measure, which has been standardized to have a mean of zero and standard deviation of one for ease of interpretation. Without accounting for achievement, there is a black-white disparity of 0.15 standard deviations in self-reported physical health for men and 0.23 standard deviations for women. For Hispanics, the differences are −0.140 for men and 0.030 for women. Accounting for educational achievement eliminates the gap for men and cuts the gap in half for black women [−0.111 (0.076)]. The remaining difference for black women is not statistically significant. Hispanic women report better health than white women with or without accounting for AFQT.

Extending Neal and Johnson (1996) further, we turn our attention to the College and Beyond (C&B) Database, which contains data on 93,660 full-time students who entered thirty-four elite colleges and universities in the fall of 1951, 1976, or 1989. We focus on the cohort from 1976.¹¹ The C&B data contain information drawn from students’ applications and transcripts, Scholastic Aptitude Test (SAT) and the American College Test (ACT) scores, standardized college admissions exams that are designed to assess a student’s readiness for college, as well as information on family demographics and socioeconomic status in their teenage years.¹² The C&B database also includes responses to a survey administered in 1995 or 1996 to all three cohorts that provides detailed information on post-college labor market outcomes. Wage data were collected when the respondents were approximately 38 years old, and reported as a series of ranges. We assigned individuals the midpoint value of their reported income range as their annual income.¹³ The response rate to the 1996 survey was approximately 80%. Table A.3 contains summary statistics used in our analysis.

Table 4 presents racial disparities in income for men and women from the 1976 cohort of the C&B Database.¹⁴ The odd-numbered columns present raw racial differences. The even-numbered columns add controls for performance on the SAT and its square.¹⁵ Black men from this sample earn 27.3% less than white men, but when we account for educational achievement, the gap shrinks to 15.2%. Black women earn more than white women by 18.6%, which increases to an advantage of 28.6% when accounting for SAT scores. There are no differences in income between Hispanics and whites with or without accounting for achievement.

In developing countries, eradicating poverty takes a large and diverse set of strategies: battling disease, fighting corruption, building schools, providing clean water, and so on (Schultz and Strauss, 2008). In the United States, important progress toward racial equality can be made if one ensures that black and white children obtain the same skills. This is an enormous improvement over the battles for basic access and equality that were fought in the 20th century, but we must now work to close the racial achievement gaps in education—high-quality education is the new civil rights battleground.¹⁶

3 Basic facts about racial differences in achievement before kids enter school

We begin our exploration of the racial achievement gap with data on mental function in the first year of life. This approach has two virtues. First, nine months is one of the earliest ages at which one can reliably test cognitive achievement in infants. Second, data on the first year of life provide us with a rare opportunity to potentially understand whether genetics is an important factor in explaining racial differences later in life.¹⁷

There are only two datasets that both are nationally representative and contain assessments of mental function before the first year of life. The first is the US Collaborative Perinatal Project (CPP) (Bayley, 1965), which includes over 31,000 women who gave birth in twelve medical centers between 1959 and 1965. The second dataset is the Early Childhood Longitudinal Study, Birth Cohort (ECLS-B), a nationally representative sample with measures of mental functioning (a shortened version of the Bayley Scale of Infant Development) for over 10,000 children aged one and under. Summary statistics for the variables we use in our core specifications are displayed by race in Table A.4 (CPP) and Table A.5 (ECLS-B).

Figures 1 and 2 plot the density of mental test scores by race at various ages in the ECLS-B and CPP data sets, respectively.¹⁸ In Fig. 1, the test score distributions on the Bayley Scale at age nine months for children of different races are visually indistinguishable. By age two, the white distribution has demonstrably shifted to the right. At age four, the cognitive score is separated into two components: literacy (which measures early language and literacy skills) and math (which measures early mathematics skills and math readiness). Gaps in literacy are similar to disparities at age two; early math skills differences are more pronounced. Figure 2 shows a similar pattern using the CPP data. At age eight months, all races look similar. By age four, whites are far ahead of blacks and Hispanics and these differences continue to grow over time. Figures 1 and 2 make one of the key points of this section: the commonly observed racial achievement gap only emerges after the first year of life.

To get a better sense of the magnitude (and standard errors) of the change from nine months to seven years old, we estimate least squares models of the following form:

(2)

where i indexes individuals, a indexes age in years, and R_i corresponds to the racial group to which an individual belongs. The vector X_i captures a wide range of possible control variables including demographics, home and prenatal environment; ε; a is an error term. The variables in the ECLS-B and CPP datasets are similar, but with some important differences.¹⁹ In the ECLS-B dataset, demographic variables include the gender of the child, the age of the child at the time of assessment (in months), and the region of the country in which the child lives. Home environment variables include a single socioeconomic status measure (by quintile), the mother’s age, the number of siblings, and the family structure (child lives with: “two biological parents,” “one biological parent,” and so on). There is also a “parent as teacher” variable included in the home environment variables. The “parent as teacher” score is coded based on interviewer observations of parent-child interactions in a structured problem-solving environment and is based on the Nursing Child Assessment Teaching Scale (NCATS). Our set of prenatal environment controls include: the birthweight of the child (in 1000-gram ranges), the amount premature that the child was born (in 7-day ranges), and a set of dummy variables representing whether the child was a single birth, a twin, or one in a birth of three or more.

In the CPP dataset, demographic variables include the age of the child at the time of assessment (in months) and the gender of the child. Our set of home environment variables provides rich proxies of the environment in which children were reared. The set of home variables includes: parental education (both mother’s and father’s, which have been transformed to dichotomous variables ranging from “high school dropout” to “college degree or more”), parental occupation (a set of mutually exclusive and collectively exhaustive dummy variables: “no occupation,” “professional occupation,” or “non-professional occupation”), household income during the first three months of pregnancy (in $500 ranges), mother’s age, number of siblings, and each mother’s reaction to and interactions with the child, which are assessed by the interviewer (we indicate whether a mother is indifferent, accepting, attentive, over-caring, or if she behaves in another manner). The set of prenatal environment controls for the CPP is the same as the set of prenatal environment controls in the ECLS-B dataset. Also included in the analysis of both datasets is interviewer fixed effects, which adjust for any mean differences in scoring of the test across interviewers.²⁰ It is important to stress that a causal interpretation of the coefficients on the covariates is likely to be inappropriate; we view these particular variables as proxies for a broader set of environmental and behavioral factors.

The coefficients on the race variables across the first three waves of ECLS-B and CPP datasets are presented in Table 5. The omitted race category is non-Hispanic white, so the other race coefficients are relative to that omitted group. Each column reflects a different regression and potentially a different dataset. The odd-numbered columns have no controls. The even-numbered columns control for interviewer fixed effects, age at which the test was administered, the gender of the child, region, socioeconomic status, variables to proxy for a child’s home environment (family structure, mother’s age, number of siblings, and parent-as-teacher measure) and prenatal condition (birth weight, premature birth, and multiple births).²¹ Even-numbered columns for CPP data omit region and the parent-as-teacher measure, which are unique to ECLS-B.²²

In infancy, blacks lag whites by 0.077 (0.031) standard deviations in the raw ECLS-B data. Hispanics and Asians also slightly trail whites by 0.025 (0.029) and 0.027 (0.040), respectively. Adding our set of controls eliminates these trivial differences. The patterns in the CPP data are strikingly similar. Yet, raw gaps of almost 0.4 standard deviations between blacks and whites are present on the test of mental function in the ECLS-B at age two. Even after including extensive controls, a black-white gap of 0.219 (0.036) standard deviations remains. Hispanics look similar to blacks. Asians lag whites by a smaller margin than blacks or Hispanics in the raw data but after including controls they are the worst-performing ethnic group. By age four, a large test score gap has emerged for blacks and Hispanics in both datasets—but especially in the CPP. In the raw CPP data, blacks lag whites by almost 0.8 standard deviations and Hispanics fare even worse. The inclusion of controls reduces the gap to roughly 0.3 standard deviations for blacks and 0.5 standard deviations for Hispanics. In the ECLS-B, black math scores trail white scores by 0.337 (0.032) in the raw data and trail by 0.130 (0.036) with controls. Black-white differences in literacy are −0.195 (0.031) without controls and 0.020 (0.035) with controls. The identical estimates for Hispanics are −0.311 (0.029) and −0.174 (0.034) in math; −0.293 (0.028) and −0.103 (0.033) in literacy. Asians are the highest-performing ethnic group in both subjects on the age four tests. Racial disparities at age seven, available only in CPP, are generally similar to those at age four.

There are at least three possible explanations for the emergence of racial differences with age. The first is that the skills tested in one-year-olds are not the same as those required of older children, and there are innate racial differences only in the skills that are acquired later. For instance, an infant scores high if she babbles expressively or looks around to find the source of the noise when a bell rings, while older children are tested directly on verbal skills and puzzle-solving ability. Despite these clear differences in the particular tasks undertaken, the outcomes of these early and subsequent tests are correlated by about 0.30, suggesting that they are, to some degree, measuring a persistent aspect of a child’s ability.²³ Also relevant is the fact that the Bayley Scales of Infant Development (BSID) score is nearly as highly correlated with measures of parental IQ as childhood aptitude tests.

Racial differences in rates of development are a second possible explanation for the patterns in our data. If black infants mature earlier than whites, then black performance on early tests may be artificially inflated relative to their long-term levels. On the other hand, if blacks are less likely to be cognitively stimulated at home or more likely to be reared in environments that Shonkoff (2006) would label as characterized by “toxic stress,” disruptions in brain development may occur, which may significantly retard cognitive growth.

A third possible explanation for the emerging pattern of racial gaps is that the relative importance of genes and environmental factors in determining test outcomes varies over time. In contrast to the first two explanations mentioned above, under this interpretation, the measured differences in test scores are real, and the challenge is to construct a model that can explain the racial divergence in test scores with age.

To better understand the third explanation, Fryer and Levitt (forthcoming) provide two statistical models that are consistent with the data presented above. Here we provide a briefoverview of the models and their predictions.

The first parameter of interest is the correlation between test scores early on and later in life. Fryer and Levitt (forthcoming) assign a value of 0.30 to that correlation. The measured correlation between test scores early and late in life and parental test scores is also necessary for the analysis. Based on prior research (e.g., Yeates et al., 1983), we take these two correlations as 0.36 and 0.39, respectively.²⁴ The estimated black-white test score gap at young ages is taken as 0.077 based on our findings in ECLS-B, compared to a gap of 0.78 at later ages based on our findings in CPP.

The primary puzzle raised by our results is the following: how does one explain small racial gaps on the BSID test scores administered at ages 8 to 12 months and large racial gaps in tests of mental ability later in life, despite the fact that these two test scores are reasonably highly correlated with one another (ρ = 0.3), and both test scores are similarly correlated with parental test scores (ρ ≥ 0.3)?

4 Interventions to foster human capital before children enter school

In the past five decades there have been many attempts to close the racial achievement gap before kids enter school.³³ Table 6 provides an overview of twenty well-known programs, the ages they serve, and their treatment effects (in the cases in which they have been credibly evaluated).

Perhaps the most famous early intervention program for children involved 64 students in Ypsilanti, Michigan, who attended the Perry Preschool program in 1962. The program consisted of a 2.5-hour daily preschool program and weekly home visits by teachers, and targeted children from disadvantaged socioeconomic backgrounds with IQ scores in the range of 70–85. An active learning curriculum—High/Scope—was used in the preschool program in order to support both the cognitive and non-cognitive development of the children over the course of two years beginning when the children were three years old. Schweinhart et al. (1993) find that students in the Perry Preschool program had higher test scores between the ages of 5 and 27, 21% less grade retention or special services required, 21% higher graduation rates, and halfthe number of lifetime arrests in comparison to children in the control group. Considering the financial benefits that are associated with the positive outcomes of the Perry Preschool, Heckman et al. (2009) estimated that the rate of return on the program is between 7 and 10%, passing a cost-benefit analysis.

Another important intervention, which was initiated three years after the Perry Preschool program is Head Start. Head Start is a preschool program funded by federal matching grants that is designed to serve 3- to 5-year-old children living at or below the federal poverty level.³⁴ The program varies across states in terms of the scope of services provided, with some centers providing full-day programs and others only half-day. In 2007, Head Start served over 900,000 children at an average annual cost of about $7300 per child.

Evaluations of Head Start have often been difficult to perform due to the non-random nature of enrollment in the program. Currie and Thomas (1995) use a national sample of children and compare children who attended a Head Start program with siblings who did not attend Head Start, based on the assumption that examining effects within the family unit will reduce selection bias. They find that those children who attended Head Start scored higher on preschool vocabulary tests but that for black students, these gains were lost by age ten. Using the same analysis method with updated data, Garces et al. (2002) find several positive outcomes associated with Head Start attendance. They conclude that there is a positive effect from Head Start on the probability of attending college and—for whites—the probability of graduating from high school. For black children, Head Start led to a lower likelihood of being arrested or charged with a crime later in life.

Puma et al. (2005), in response to the 1998 reauthorization of Head Start, conduct an evaluation using randomized admission into Head Start.³⁵ The impact of being offered admission into Head Start for three and four year olds is 0.10 to 0.34 standard deviations in the areas of early language and literacy. For 3-year-olds, there were also small positive effects in the social-emotional domain (0.13 to 0.18 standard deviations) and on overall health status (0.12 standard deviations). Yet, by the time the children who received Head Start services have completed first grade, almost all of the positive impact on initial school readiness has faded. The only remaining impacts in the cognitive domain are a 0.08 standard deviation increase in oral comprehension for 3-year-old participants and a 0.09 standard deviation increase in receptive vocabulary for the 4-year-old cohort (Puma et al., 2010).³⁶

A third, and categorically different, program is the Nurse Family Partnership. Through this program, low-income first-time mothers receive home visits from a registered nurse beginning early in the pregnancy that continue until the child is two years old—a total of fifty visits over the first two years. The program aims to encourage preventive health practices, reduce risky health behaviors, foster positive parenting practices, and improve the economic self-sufficiency of the family. In a study of the program in Denver in 1994–95, Olds et al. (2002) find that those children whose mothers had received home visits from nurses (but not those who received home visits from paraprofessionals) were less likely to display language delays and had superior mental development at age two. In a long-term evaluation of the program, Olds et al. (1998) find that children born to women who received nurse home visits during their pregnancy between 1978 and 1980 have fewer juvenile arrests, convictions, and violations of probation by age fifteen than those whose mothers did not receive treatment.

Other early childhood interventions—many based on the early success of the Perry Preschool, Head Start, and the Nurse Family Partnership—include the Abecedarian Project, the Early Training Project, the Infant Health and Development Program, the Milwaukee Project, and Tulsa’s universal pre-kindergarten program. The Abecedarian Project provided full-time, high-quality center-based childcare services for four cohorts of children from low-income families from infancy through age five between 1971 and 1977. Campbell and Ramey (1994) find that at age twelve, those children who were randomly assigned to the project scored 5 points higher on the Wechsler Intelligence Scale and 5–7 points higher on various subscales of the Woodcock-Johnson Psycho-Educational Battery achievement test. The Early Training Project provided children from low-income homes with summertime experiences and weekly home visits during the three summers before entering first grade in an attempt to improve the children’s school readiness. Gray and Klaus (1970) report that children who received these intervention services maintained higher Stanford-Binet IQ scores (2–5 points) at the end of fourth grade. The Infant Health and Development Program specifically targeted families with low birthweight, preterm infants and provided them with weekly home visits during the child’s first year and biweekly visits through age three, as well as enhanced early childhood educational care and bimonthly parent group meetings. Brooks-Gunn et al. (1992) report that this program had positive effects on language development at the end of first grade, with participant children scoring 0.09 standard deviations higher on receptive vocabulary and 0.08 standard deviations higher on oral comprehension. The Milwaukee Project targeted newborns born to women with IQs lower than 80; mothers received education, vocational rehabilitation, and child care training while their children received high-quality educational programming and three balanced meals daily at “infant stimulation centers” for seven hours a day, five days a week until the children were six years old. Garber (1988) finds that this program resulted in an increase of 23 points on the Stanford-Binet IQ test at age six for treatment children compared to control children.

Unlike the other programs described, Tulsa’s preschool program is open to all 4-year-old children. It is a basic preschool program that has high standards for teacher qualification (a college degree and early childhood certification are both required) and a comparatively high rate of penetration (63% of eligible children are served). Gormley et al. (2005) use a birthday cutoff regression discontinuity design to evaluate the program and find that participation improves scores on the Woodcock-Johnson achievement test significantly (from 0.38 to 0.79 standard deviations).

Beyond these highly effective programs, Table 6 demonstrates that there is large variance in the effectiveness of well-known early childhood programs. The Parents as Teachers Program, for instance, shows mixed and generally insignificant effects on initial measures of cognitive development (Wagner and Clayton, 1999). In an evaluation of the Houston Parent-Child Development Centers, Andrews et al. (1982) find no significant impact on children’s cognitive skills at age one and mixed impacts on cognitive development at age two. Even so, the typical early childhood intervention passes a simple cost-benefit analysis.³⁷

There are two potentially important caveats going forward. First, most of the programs are built on the insights gained from Perry and Head Start, yet what we know about infant development in the past five decades has increased dramatically. For example, psychologists used to assume that there was a relatively equal degree of early attachment across children but they now acknowledge that there is a great deal of variance in the stability of early attachment (Thompson, 2000). Tying new programs to the lessons learned from previously successful programs while incorporating new insights from biology and developmental psychology is both the challenge and opportunity going forward.

Second, and more important for our purposes here, even the most successful early interventions cannot close the achievement gap in isolation. If we truly want to eliminate the racial achievement gap, early interventions may or may not be necessary but the evidence forces one to conclude that they are not sufficient.

5 The racial achievement gap in kindergarten through 12th grade

As we have seen, children begin life on equal footing, but important differences emerge by age two and their paths quickly diverge. In this section, we describe basic facts about the racial achievement gap from the time children enter kindergarten to the time they exit high school. Horace Mann famously argued that schools were “the great equalizer,” designed to eliminate differences between children that are present when they enter school because of different background characteristics. As this section will show, if anything, schools currently tend to exacerbate group differences.

6 The racial achievement gap in high school

We conclude our descriptive analysis of the racial achievement gap with high school-aged students using the National Education Longitudinal Survey (NELS).⁴⁶ The NELS consists of a nationally representative group of students who were in eighth grade in 1988 when the baseline survey and achievement test data were collected. Students were resurveyed in 1990 at the end of their tenth grade year and again in 1992 at the anticipated end of their high school career. All three waves consist of data from a student questionnaire, achievement tests, a school principal questionnaire, and teacher questionnaires; 1990 and 1992 follow-ups also include a dropout questionnaire, the baseline and 1992 follow-up also surveyed parents, and the 1992 follow-up contains student transcript information. NELS contains 24,599 students, in 2963 schools and 5351 math, science, English, and history classrooms initially surveyed in the baseline year. Eighty-two percent of these students completed a survey in each of the first three rounds.

The primary outcomes in the NELS data are four exams: math, reading comprehension, science, and social studies (history/citizenship/government). In the base year (eighth grade), all students took the same set of tests, but in order to avoid problematic “ceiling” and “floor” effects in the follow-up testing (tenth and twelfth grades for most participants) students were given test forms tailored to their performance in the previous test administration. There were two reading test forms and three math test forms; science and social studies tests remained the same for all students. Test scores were determined using Item Response Theory (IRT) scoring, which allowed the difficulty of the test taken by each student to be taken into account in order to estimate the score a student would have achieved for any arbitrary set of test items. Table A.12 provides descriptive statistics.

Table 16 provides estimates of the racial achievement gap in high school across four subjects. For each grade, we estimate four empirical models. We begin with raw racial differences, which are displayed in the first column under each grade. Then, we add controls for race, gender, age (linear, quadratic, and cubic terms), family income, and dummies for parents’ levels of education. The third empirical model includes school fixed effects and the fourth includes teacher fixed effects. The raw black-white gap in eighth grade math is 0.754 (0.025) standard deviations. Adding controls reduces the gap to 0.526 (0.021), and adding school fixed effects reduces the gap further to 0.400 (0.021), which is similar to the eighth grade disparities reported in ECLS. Including teacher fixed effects reduces the gap to 0.343 (0.031) standard deviations. In 10th and 12th grade, black-white disparities range from 0.734 (0.038) in the raw data to 0.288 (0.060) with teacher fixed effects in 10th grade, and 0.778 (0.045) to 0.581 (0.089) in 12th grade. Hispanics follow a similar trend, but the achievement gaps are nearly 40% smaller. In the raw data, Asians are the highest-performing ethnic group in eighth through twelfth grades. Including teacher fixed effects, however, complicates the story. Asians are 0.127 standard deviations ahead of whites in eighth grade. This gap diminishes over time and, by twelfth grade, Asian students trail whites when they have the same teachers.

Panels B, C, and D of Table 16, which estimate racial achievement gaps in English, history, and science, respectively, all show magnitudes and trends similar to those documented above in math. Averaging across subjects, the black-white gap in eighth grade is roughly 0.7 standard deviations. An identical calculation for Hispanics yields a gap of just under 0.6 standard deviations. Asians are ahead in math and on par with whites in all other subjects. In twelfth grade, black students significantly trail whites in science and math (0.911 (0.041) and 0.778 (0.045) standard deviations, respectively) and slightly less so in history and English. Hispanics and Asians demonstrate patterns in twelfth grade that are very similar to their patterns in eighth grade.

To close our analytic pipeline from nine months old to high school graduation, we investigate racial differences in high school graduation or GED acquisition within five years of their freshman year in high school [not shown in tabular form]. In the raw data, blacks are twice as likely as whites to not graduate from high school or receive a GED within five years of entering high school. Accounting for math and reading achievement scores in eighth grade explains all of the racial gap in graduation rates. Hispanics are 2.2 times more likely not to graduate and these differences are reduced to thirty percent more likely after including eighth grade achievement.

We learn four points from NELS. First, achievement gaps continue their slow divergence in the high school years. Second, gaps are as large in science and history as they are in subjects that are tested more often, such as math and reading. Third, similarly as in the preceding analysis, a substantial racial achievement gap exists after accounting for teacher fixed effects. Fourth, the well-documented disparities in graduation rates can be explained by eighth grade test scores. The last result is particularly striking.

7 Interventions to foster human capital in school-aged

CHILDREN

In an effort to increase achievement and narrow differences between racial groups, school districts have become laboratories of innovative reforms, including smaller schools and classrooms (Nye, 1995; Krueger, 1999), mandatory summer school (Jacob and Lefgren, 2004), merit pay for principals, teachers, and students (Podgursky and Springer, 2007; Fryer, 2010), after-school programs (Lauer et al., 2006), budget, curricula, and assessment reorganization (Borman et al., 2007), policies to lower the barrier to teaching via alternative paths to accreditation (Decker et al., 2004; Kane et al., 2008), single-sex education (Shapka and Keating, 2003), data-driven instruction (Datnow et al., 2008), ending social promotion (Greene and Winters, 2006), mayoral/state control of schools (Wong and Shen, 2002, 2005; Henig and Rich, 2004), instructional coaching (Knight, 2009), local school councils (Easton et al., 1993), reallocating per-pupil spending (Mar-low, 2000; Guryan, 2001), providing more culturally sensitive curricula (Protheroe and Barsdate, 1991; Thernstrom, 1992; Banks, 2001, 2006), renovated and more technologically savvy classrooms (Rouse and Krueger, 2004; Goolsbee and Guryan, 2006), professional development for teachers and other key staff (Boyd et al., 2008; Rockoff, 2008), and getting parents to be more involved (Domina, 2005).

The evidence on the efficacy of these investments is mixed. Despite their intuitive appeal, school choice, summer remediation programs, and certain mentoring programs show no effect on achievement (Krueger and Zhu, 2002; Walker and Vilella-Velez, 1992; Bernstein et al., 2009). Financial incentives for students, smaller class sizes, and bonuses for teachers in hard-to-staff schools show small to modest gains that pass a cost-benefit analysis (Fryer, 2010; Schanzenbach, 2007; Jacob and Ludwig, 2008). It is imperative to note: these programs have not been able to substantially reduce the achievement gap even in the most reform-minded school systems.

Even more aggressive strategies that place disadvantaged students in better schools through busing (Angrist and Lang, 2004) or significantly alter the neighborhoods in which they live (Jacob, 2004; Kling et al., 2007; Sanbonmatsu et al., 2006; Turney et al., 2006) have left the racial achievement gap essentially unchanged.

Table 17 describes seventeen additional interventions designed to increase achievement in public schools.⁴⁷ The first column lists the program name, the second column reports the grades treated, and the third column provides a brief description of each intervention. The final two columns provide information on the magnitude of the reported effect and a reference. The bulk of the evidence finds little to no effect of these interventions. Three programs seem to break this mold: Mastery Learning, Success for All, and self-affirmation essay writing. Mastery learning is a group-based, teacher-paced instructional model that is based on the idea that students must attain a level of mastery on a particular objective before moving on to a new objective. Guskey and Gates (1985) perform a meta-analysis of thirty-five studies on this instructional strategy and find that the average achievement effect size from mastery learning programs was 0.78 standard deviations. The effect sizes from within individual studies, however, ranged from 0.02 to 1.70 and varied significantly depending on the age of the students and the subject tested (Guskey and Gates, 1985).

Success for All is a school-level elementary school intervention that focuses on improving literacy outcomes for all students in order to improve overall student achievement that is currently used in 1200 schools across the country (Borman et al., 2007). The program is designed to identify and address deficiencies in reading skills at a young age using a variety of instruction strategies, ranging from cooperative learning to data-driven instruction. Borman et al. (2007) use a cluster randomized trial design to evaluate the impacts of the Success for All model on student achievement. Forty-one schools from eleven states volunteered and were randomly assigned to either the treatment or control groups. Borman et al. (2007) find that Success for All increased student achievement by 0.36 standard deviations on phonemic awareness, 0.24 standard deviations on word identification, and 0.21 standard deviations on passage comprehension.

The self-affirmation essay writing intervention was intended specifically to improve the academic achievement of minorities by reducing the impact of stereotype threat. Seventh grade students were randomly assigned to either a treatment or control group. Both groups were given structured writing assignments three to five times over the course of two school years, but the treatment group was instructed to write about their personal values and why they were important, while the control group was given neutral essay topics. Cohen et al. (2009) find that for black students, this intervention increased GPA by 0.24 points and that the impact was even greater for low-achieving black students (0.41 GPA points). They also find that the program reduced the probability of being placed in remedial classes or being retained in a grade for low-achieving black students. It is unclear what the general equilibrium effects of such psychological interventions are.

Despite trillions spent, there is not one urban school district that has ever closed the racial achievement gap. Figures 5 and 6 show the achievement gap in percentage of students proficient for their grade level across eleven major US cities who participate in the National Assessment of Educational Progress (NAEP)—a nationally representative set of assessments administered every two years to fourth, eighth, and twelfth graders that covers various subject areas, including mathematics and reading.⁴⁸

In every city there are large racial differences. In the Trial Urban District Assessment, among fourth graders, 43.2% of whites, 12% of blacks, and 16% of Hispanics are proficient in reading. In math, these numbers are 50.9, 14, and 20.9, respectively. Similarly, among eighth graders, 40.4% of whites, 10.6% of blacks, and 13.2% of Hispanics score proficient in reading. Math scores exhibit similarly marked racial differences. Washington, DC, has the largest achievement gap of participating cities in NAEP; there is a roughly seventy percent difference between blacks and whites on both subjects and both grade levels. At the other end of the spectrum, Cleveland has the smallest achievement gap—less than seventeen percentage points separate racial groups. Unfortunately, Cleveland’s success in closing the achievement gap is mainly due to the dismal performance of whites in the school district and not due to increased performance of black students. Remarkably, there is very little variance in the achievement of minority students across NAEP districts. There is not one school district in NAEP in which more than twenty-one percent of black students are proficient in reading or math.

The lack of progress has fed into a long-standing and rancorous debate among scholars, policymakers, and practitioners as to whether schools alone can close the achievement gap, or whether the issues children bring to school as a result of being reared in poverty are too much for even the best educators to overcome. Proponents of the school-centered approach refer to anecdotes of excellence in particular schools or examples of other countries where poor children in superior schools outperform average Americans (Chenoweth, 2007). Advocates of the community-focused approach argue that teachers and school administrators are dealing with issues that actually originate outside the classroom, citing research that shows racial and socioeconomic achievement gaps are formed before children ever enter school (Fryer and Levitt, 2004, 2006) and that one-third to one-half of the gap can be explained by family-environment indicators (Phillips et al., 1998a,b; Fryer and Levitt, 2004).⁴⁹ In this scenario, combating poverty and related social ills directly and having more constructive out-of-school time may lead to better and more focused instruction in school. Indeed, Coleman et al. (1966), in their famous report on equality of educational opportunity, argue that schools alone cannot solve the problem of chronic underachievement in urban schools.

The Harlem Children’s Zone (HCZ)—a 97-block area in central Harlem, New York, that combines reform-minded charter schools with a web of community services designed to ensure the social environment outside of school is positive and supportive for children from birth to college graduation—provides an extremely rare opportunity to understand whether communities, schools, or a combination of the two are the main drivers of student achievement.

Dobbie and Fryer (2009) use two separate statistical strategies to estimate the causal impact of attending the charter schools in the HCZ. First, they exploit the fact that HCZ charter schools are required to select students by lottery when the number of applicants exceeds the number of available slots for admission. In this scenario, the treatment group is composed of students who are lottery winners and the control group consists of students who are lottery losers. The second identification strategy explored in Dobbie and Fryer (2009) uses the interaction between a student’s home address and her cohort year as an instrumental variable. This approach takes advantage of two important features of the HCZ charter schools: (1) anyone is eligible to enroll in HCZ’s schools, but only students living inside the Zone are actively recruited by HCZ staff; and (2) there are cohorts of children that are ineligible due to the timing of the schools’ opening and their age. Both statistical approaches lead to the same result: HCZ charter schools are effective at increasing the achievement of the poorest minority children.

Figure 7A and B provide a visual representation of the basic results from Dobbie and Fryer (2009). Figure 7A plots yearly, raw, mean state math test scores, from fourth to eighth grade, for four subgroups: lottery winners, lottery losers, white students in New York City public schools and black students in New York City public schools. Lottery winners are comprised of students who either won the lottery or who had a sibling who was already enrolled in the HCZ Promise Academy. Lottery losers are individuals who lost the lottery and did not have a sibling already enrolled. These represent reduced form estimates.

In fourth and fifth grade, before they enter the middle school, math test scores for lottery winners, losers, and the typical black student in New York City are virtually identical, and roughly 0.75 standard deviations behind the typical white student.⁵⁰ Lottery winners have a modest increase in sixth grade, followed by a more substantial increase in seventh grade and even larger gains by their eighth-grade year.

The “Treatment-on-Treated” (TOT) estimate, which is the effect of actually attending the HCZ charter school, is depicted in Panel B of Fig. 7. The TOT results follow a similar pattern, showing remarkable convergence between children in the middle school and the average white student in New York City. After three years of “treatment,” HCZ Promise Academy students have nearly closed the achievement gap in math—they are behind their white counterparts by 0.121 standard deviations (p-value = 0.113). If one adjusts for gender and free lunch, the typical eighth grader enrolled in the HCZ middle school outscores the typical white eighth grader in New York City public schools by 0.087 standard deviations, though the difference is not statistically significant (p-value = 0.238).

Figure 8A plots yearly state ELA test scores, from fourth to eighth grade. Treatment and control designations are identical to those in Fig. 7A. In fourth and fifth grades, before they enter the middle school, ELA scores for lottery winners, losers, and the typical black student in NYC are not statistically different, and are roughly 0.65 standard deviations behind the typical white student.⁵¹ Lottery winners and losers have very similar ELA scores from fourth through seventh grade. In eighth grade, HCZ charter students distance themselves from the control group. These results are statistically meaningful, but much less so than the math results. The TOT estimate, depicted in Panel B of Fig. 8, follows an identical pattern with marginally larger differences between enrolled middle-school students and the control group. Adjusting for gender and free lunch pushes the results in the expected direction.⁵²

8 Conclusion

In 1908, W.E.B Dubois famously noted that “the problem of the 20th century is the problem of the color line.” America has undergone drastic changes in 102 years. The problem of the 21st century is the problem of the skill gap. As this chapter attempts to make clear, eliminating the racial skill gap will likely have important impacts on income inequality, unemployment, incarceration, health, and other important social and economic indices. The problem, to date, is that we do not know how to close the achievement gap.

Yet, there is room for considerable optimism. A key difference between what we know now and what we knew even two years ago lies in a series of ”existence proofs” in which poor black and Hispanic students score on par with more affluent white students. That is, we now know that with some combination of investments, high achievement is possible for all students. That is an important step forward. Ofcourse, there are many questions as to how one can use these examples to direct interventions that have the potential to close the achievement gap writ large.⁵⁶ An economist’s solution might be to create a market for gap-closing schools with high-powered incentives for entrepreneurs to enter. The government’s role would not be to facilitate the daily workings of the schools; it would simply fund those schools that close the achievement gap and withhold funds from those that do not. The non-gap-closing schools would go out of business and would be replaced by others that are more capable. In a rough sense, this is what is happening in Louisiana post-Hurricane Katrina, what cities such as Boston claim to do, and what reform-minded school leaders such as Chancellor Joel Klein in New York City have been trying accomplish within the constraints of the public system.

A second, potentially more politically expedient, way forward is to try and understand what makes some schools productive and others not. Hoxby and Murarka (2009) and Abdulkadiroglu et al. (2009) show that there is substantial variance in the treatment effect of charter schools—even though all are free from most constraints of the public system and the vast majority do not have staffs under collective bargaining agreements. Investigating this variance and its causes could reveal important clues about measures that could be taken to close the racial achievement gap.

Independent of how we get there, closing the racial achievement gap is the most important civil rights battle of the twenty-first century.

Appendix Data description

References

1. Abdulkadiroglu Atila, Angrist Joshua, Dynarski Susan, Kane Thomas J, Pathak Parag. Accountability and flexibility in public schools: evidence from Boston’s charters and pilots Working paper no 15549. Cambridge, MA: NBER; 2009.

2. Administration for Children and Families. Preliminary Findings from the Early Head Start Prekindergarten Followup. Washington, DC: US Department of Health and Human Services Report; 2006.

3. Andrews Susan Ring, Blumenthal Janet Berstein, Johnson Dale L, et al. The skills of mothering: a study of parent child development centers. Monographs of the Society for Research in Child Development. 1982;47(6):1–83.

4. Angrist Joshua D, Lang Kevin. Does school integration generate peer effects? Evidence from Boston’s Metco program. The American Economic Review. 2004;94(5):1613–1634.

5. Angrist Joshua D, Dynarski Susan M, Kane Thomas J, Pathak Parag A, Walters Christopher R. Who benefits from KIPP? Working paper no 15740. Cambridge, MA: NBER; 2010.

6. Banks James A. Approaches to multicultural curriculum reform. In: Banks James A, Banks Cherry AM, eds. Multicultural Education: Issues and Perspectives. New York: John Wiley & Sons, Inc., 2001; fourth ed.

7. Banks James A. Cultural Diversity and Education: Foundations, Curriculum, and Teaching. Boston, MA: Pearson Education, Inc., 2006.

8. Barton Paul E. Parsing the achievement gap: baselines for tracking progress. Princeton, NJ: Policy Information Report, Educational Testing Service Policy Information Report; 2003.

9. Bayley Nancy. Comparisons of mental and motor test scores for ages 1 to 15 months by sex, birth order, race, geographical location, and education of parents. Child Development. 1965;36:379–411.

10. Becker Douglas F, Forsyth Robert A. Gender differences in mathematics problem solving and science: a longitudinal analysis. International Journal of Educational Research. 1994;21(4):407–416.

11. Bernstein Lawrence, Dun Rappaport Catherine, Olsho Lauren, Hunt Dana, Levin Magorie, et al. Impact evaluation of the US Department of Education’s Student Mentoring Program: final report. Institute of Education Sciences, Washington, DC: US Department of Education; 2009.

12. Bethel, James, Green, James L., Kalton, Graham, Nord, Christine, 2004. Early childhood longitudinal study, birth cohort (ECLS-B), sampling. Volume 2 of the ECLS-B Methodology Report for the 9-Month Data Collection, 2001–02, US Department of Education, NCES, Washington, DC.

13. Bloom, Dan, Gardenhire-Crooks, Alissa, Mandsager, Conrad, 2009. Reengaging high school dropouts: early results of the National Guard Youth Challenge Program evaluation, MDRC Report, New York.

14. Borman Geoffrey D, Slavin Robert E, Cheung Alan CK, Chamberlain Anne M, Madden Nancy A, Chambers Bette. Final reading outcomes of the national randomized field trial of Success for all. American Educational Research Journal. 2007;44(3):701–731.

15. Bornstein Marc H, Sigman Marian D. Continuity in mental development from infancy. Child Development. 1986;57(2):251–274.

16. Boyd Donald, Grossman Pamela, Lankford Hamilton, Loeb Susanna, Wyckoff James. Teacher preparation and student achievement Working paper no 14314. Cambridge, MA: NBER; 2008.

17. Brooks-Gunn Jeanne, Liaw Fong-ruey, Klebanov Pamela Kato. Effects of early intervention on cognitive function of low birth weight preterm infants. Journal of Pediatrics. 1992;120(3):350–359.

18. Campbell Frances A, Ramey Craig T. Cognitive and school outcomes for high-risk African-American students at Middle Adolescence: positive effects of early intervention. American Educational Research Journal. 1994;32(4):743–772.

19. Campbell Jay R, Hombo Catherine M, Mazzeo John. NAEP 1999 trends in academic progress: three decades of student performance. NCES, Washington, DC: US Department of Education; 2000.

20. Carneiro Pedro, Heckman James. Human capital policy Working paper no 9495. Cambridge, MA: NBER; 2003.

21. Chenoweth Karin. It’s Being Done: Academic Success in Unexpected Schools. Cambridge, MA: Harvard University Press; 2007.

22. Cohen Geoffrey L, Garcia Julio, Purdie-Vaughns Valerie, Apfel Nancy, Brzutoski Patricia. Recursive processes in self-affirmation: intervening to close the minority achievement gap. Science. 2009;324(5925):400–403.

23. Coleman James S, Campbell Ernest Q, Hobson Carol J, et al. Equality of educational opportunity. Washington, DC: US Department of Health, Education, and Welfare, Office of Education; 1966.

24. Congressional Record, No. 11, p. H417 (daily ed. Jan. 27, 2010) (statement of The President).

25. Cook Thomas D, Habib Farah-Naaz, Phillips Meredith, Settersten Richard A, Shagle Shobha C, Degirmencioglu Serdar M. Comer’s School Development Program in Prince George’s county, Maryland: a theory-based evaluation. American Educational Research Journal. 1999;36(3):543–597.

26. Corrin William, Somers Marie-Andree, Kemple James J, Nelson Elizabeth, Sepanik Susan, et al. The enhanced reading opportunities study: findings from the second year of implementation. Washington, DC: US Department of Education, Institute of Education Sciences; 2009.

27. Currie Janet, Thomas Duncan. Does Head Start make a difference? American Economic Review. 1995;85(3):341–364.

28. Curto Vilsa E, Fryer Roland G, Howard Meghan L. It may not take a village: increasing achievement among the poor. Harvard University: Unpublished paper; 2010.

29. Darity Jr. William A, Mason Patrick L. Evidence on discrimination in employment: codes of color, codes of gender. Journal of Economic Perspectives. 1998;12(2):63–90.

30. Datnow Amanda, Park Vicki, Kennedy Brianna. Acting on data: how urban high schools use data to improve instruction. USC Rossier School of Education, Los Angeles: Center on Educational Governance; 2008.

31. Decker Paul, Mayer Daniel, Glazerman Steven. The effects of teach for America on students: findings from a national evaluation. Princeton, NJ: Mathematica Policy Research, Inc., Report; 2004.

32. Dee Thomas. Conditional cash penalties in education: evidence from the learnfare experiment Working paper no 15126. Cambridge, MA: NBER; 2009.

33. Dickens William T, Flynn James R. Heritability estimates versus large environmental effects: the IQ paradox resolved. Psychological Review. 2001;108(2):346–369.

34. Dickens William T, Flynn James R. Black Americans reduce the racial IQ gap: evidence from standardization samples. Psychological Science. 2006;17(10):913–920.

35. Dobbie Will, Fryer Jr Roland G. Are high quality schools enough to close the achievement gap? Evidence from a social experiment in Harlem Working paper no 15473. Cambridge, MA: NBER; 2009.

36. Domina Thurston. Leveling the home advantage: assessing the effectiveness of parental involvement in elementary school. Sociology of Education. 2005;78(3):233–249.

37. Easton John Q, Flinspach Susan Leigh, O’Connor Carla, Paul Mark, Qualls Jesse, Ryan Susan P. Local school council governance: the third year of Chicago school reform. Chicago, IL: Chicago Panel on Public School Policy and Finance; 1993.

38. Farber Henry S, Gibbons Robert. Learning and wage dynamics. Quarterly Journal of Economics. 1996;111(4):1007–1047.

39. Franzini L, Ribble JC, Keddie AM. Understanding the Hispanic paradox. Ethnicity and Disease. 2001;11:496–518.

40. Fryer Roland G, Levitt Steven D. Understanding the black-white test score gap in the first two years of school. Review of Economics and Statistics. 2004;86(2):447–464.

41. Fryer Roland G, Levitt Steven D. The black-white test score gap through third grade. American Law and Economics Review. 2006;8(2):249–281.

42. Fryer Roland G, Levitt Steven D. An empirical analysis of the gender gap in mathematics. American Economic Journal: Applied Economics. 2010;2(2):210–240.

43. Fryer, Roland G., Levitt, Steven D., Testing for racial differences in the mental ability of young children. American Economic Review (forthcoming).

44. Fryer Roland G. Financial incentives and student achievement: evidence from randomized trials. Harvard University: Unpublished paper; 2010.

45. Garber Howard L. The Milwaukee Project: preventing mental retardation in children at risk. Washington, DC: National Institute of Handicapped Research Report; 1988.

46. Garces Eliana, Thomas Duncan, Currie Janet. Longer-term effects of Head Start. American Economic Review. 2002;92(4):999–1012.

47. Garet Michael S, Cronen Stephanie, Eaton Marian, et al. The impact of two professional development interventions on early reading instruction and achievement. Washington, DC: US Department of Education, Institute of Education Sciences; 2008.

48. Goolsbee Austan, Guryan Jonathan. The impact of Internet subsidies in public schools. Review of Economics and Statistics. 2006;88(2):336–347.

49. Gormley Jr. William T, Gayer Ted, Phillips Deborah, Dawson Brittany. The effects of universal Pre-K on cognitive development. Developmental Psychology. 2005;41(6):872–884.

50. Gray Susan W, Klaus Rupert A. The early training project: a seventh-year report. Child Development. 1970;41:909–924.

51. Greene Jay P, Winters Marcus A. Getting ahead by staying behind: an evaluation of Florida’s program to end social promotion. Education Next. 2006;6(2):65–69.

52. Guryan Jonathan. Does money matter? Regression-discontinuity estimates from education finance reform in Massachusetts Working paper no 8269. Cambridge, MA: NBER; 2001.

53. Guskey Thomas R, Gates Sally L. A synthesis of research on group-based mastery learning programs. Chicago, IL: American Educational Research Association Presentation; 1985.

54. Hanushek Eric A, Kain John, Rivkin Steven, Branch Gregory. Charter school quality and parental decision making with school choice Working paper no 11252. Cambridge, MA: NBER; 2005.

55. Hart Betty, Risley Todd R. Meaningful Differences in the Everyday Experience of Young American Children. Baltimore, MD: Brookes; 1995.

56. Hawkins J, David Kosterman, Rick Catalano, Hill Richard F, Karl Abbott G, Robert D. Effects of social development intervention in childhood fifteen years later. Archives of Pediatrics & Adolescent Medicine. 2008;162(12):1133–1141.

57. Heckman James J, Moon Seong Hyeok, Pinto Rodrigo, Savelyev Peter A, Yavitz Adam. The rate of return to the High/Scope Perry Preschool program Working paper no 15471. Cambridge, MA: NBER; 2009.

58. Heckman James J. Detecting discrimination. Journal of Economic Perspectives. 1998;12(2):101–116.

59. Heckman James J. Policies to foster human capital Working paper no 7288. Cambridge, MA: NBER; 1999.

60. Henig Jeffrey R, Rich Wilbur C. Mayors in the Middle: Politics, Race, and Mayoral Control of Urban Schools. Princeton, NJ: Princeton University Press; 2004.

61. Hoxby Caroline M, Murarka Sonali. Charter schools in New York City: who enrolls and how they affect their students’ achievement Working paper no 14852. Cambridge, MA: NBER; 2009.

62. Hoxby Caroline M, Rockoff Jonah E. The impact of charter schools on student achievement. Harvard University: Unpublished paper; 2004.

63. Jacob Brian A, Lefgren Lars. Remedial education and student achievement: a regression- discontinuity analysis. Review of Economics and Statistics. 2004;86(1):226–244.

64. Jacob Brian A, Ludwig Jens. Improving educational outcomes for poor children Working paper no 14550. Cambridge, MA: NBER; 2008.

65. Jacob Brian A. Public housing, housing vouchers, and student achievement: evidence from public housing demolitions in Chicago. American Economic Review. 2004;94(1):233–258.

66. Jacob Brian A. Accountability, incentives and behavior: the impact of high-stakes testing in the Chicago public schools. Journal of Public Economics. 2005;89:761–796.

67. James-Burdumy Susanne, Mansfield Wendy, Deke John, et al. Effectiveness of Selected Reading Comprehension Interventions: Impacts on a First Cohort of Fifth-Grade Students. Institute of Education Sciences, Washington, DC: US Department of Education; 2009.

68. Jencks Christopher. Racial bias in testing. In: Jencks Christopher, Phillips Meredith, eds. The Black-White Test Score Gap. Washington, DC: The Brookings Institution Press; 1998;55–85.

69. Jensen Arthur R. Educability and Group Differences. New York: The Free Press; 1973.

70. Jensen Arthur R. Genetic and behavioral effects of nonrandom mating. In: Noble Clyde E, ed. Human Variation: Biogenetics of Age, Race, and Sex. New York: Academic Press; 1978.

71. Jensen Arthur R. The G Factor: The Science of Mental Ability. Westport, CT: Praeger; 1998.

72. Kane Thomas J, Rockoff Jonah E, Staiger Douglas O. What does certification tell us about teacher effectiveness? Evidence from New York City Working paper no 12155. Cambridge, MA: NBER; 2008.

73. Kemple James J. Career academies: long-term impacts on labor market outcomes, educational attainment, and transitions to adulthood. New York: MDRC Report; 2008.

74. Kemple James J, Herlihy Corinne M, Smith Thomas J. Making progress toward graduation: evidence from the talent development high school model. New York: MDRC Report; 2005.

75. Klebanov Pamelo Kato. Does neighborhood and family poverty affect mothers’ parenting, mental health, and social Support? Journal of Marriage and Family. 1994;56(2):441–455.

76. Kling Jeffrey R, Liebman Jeffrey B, Katz Lawrence F. Experimental analysis of neighborhood effects. Econometrica. 2007;75(1):83–119.

77. Knight Jim (Ed). Coaching: Approaches and Perspectives. Thousand Oaks, CA: Corwin Press; 2009.

78. Krieger Nancy, Sidney Stephen. Racial discrimination and blood pressure: the CARDIA study of young black and white adults. American Journal of Public Health. 1996;86(10):1370–1378.

79. Krueger Alan B. Experimental estimates of education production functions. Quarterly Journal of Economics. 1999;114(2):497–532.

80. Krueger Alan B, Whitmore Diane. Would smaller classes help close the black white achievement gap? Working paper no 451. Princeton University: Industrial Relations Section; 2001.

81. Krueger Alan B, Zhu Pei. Another look at the New York City school voucher experiment Working paper no 9418. Cambridge, MA: NBER; 2002.

82. Lally J Ronald, Mangione Peter L, Honig Alice S. The Syracuse University Family Development Research Program: Long-Range Impact of an Early Intervention with Low-Income Children and their Families. San Francisco, CA: Center for Child & Family Studies, Far West Laboratory for Educational Research & Development; 1987.

83. Lang Kevin, Manove Michael. Education and labor-market discrimination Working paper no 12257. Cambridge, MA: NBER; 2006.

84. Lauer Patricia A, Akiba Motoko, Wilkerson Stephanie B, Apthorp Helen S, Snow David, Martin-Glenn Mya L. Out-of-school-time programs: a meta-analysis of effects for at-risk students. Review of Educational Research. 2006;76(2):275–313.

85. Levin Jessica, Quinn Meredith. Missed opportunities: how we keep high-quality teachers out of urban classrooms. The New Teacher Project: Unpublished paper; 2003.

86. Lewis Michael, McGurk Harry. Evaluation of infant intelligence. Science. 1972;178(December 15):1174–1177.

87. List John A. The behavioralist meets the market: measuring social preferences and reputation effects in actual transactions Working paper no 11616. Cambridge, MA: NBER; 2005.

88. Lochner Lance, Moretti Enrico. The effect of education on crime: evidence from prison inmates, arrests, and self-reports. American Economic Review. 2004;94(1):155–189.

89. Marlow Michael L. Spending, school structure, and public education quality: evidence from California. Economics of Education Review. 2000;19(1):89–106.

90. McCall Robert B, Carriger Michael S. A meta-analysis of infant habituation and recognition memory performance as predictors of later IQ. Child Development. 1993;64(1):57–79.

91. Morrow-Howell Nancy, Jonson-Reid Melissa, McCrary Stacey, Lee YungSoo, Spitznagel Ed. Evaluation of experience corps: student reading outcomes. St. Louis, MO: Unpublished paper, Center for Social Development, George Warren Brown School of Social Work, Washington University; 2009.

92. Neal Derek A, Johnson William R. The role of premarket factors in black-white wage differences. Journal of Political Economy. 1996;104(5):869–895.

93. Neal Derek. Why has black-white skill convergence stopped? Working paper no. NBER, Cambridge, MA: 11090; 2005.

94. Neisser Ulric, Boodoo Gwyneth, Bouchard Jr. Thomas J, et al. Intelligence: knowns and unknowns. American Psychologist. 1996;51(2):77–101.

95. Nelson Charles A. The neurobiological bases of early intervention. In: Shonkoff Jack P, Meisels Samuel J, eds. Handbook of Early Childhood Intervention. New York: Cambridge University Press; 2000.

96. Nisbett Richard E. Race, genetics, and IQ. In: Jencks Christopher, Phillips Meredith, eds. The Black-White Test Score Gap. Washington, DC: The Brookings Institution Press; 1998;86–102.

97. Niswander KR, Gordon M. The Women and their Pregnancies: The Collaborative Perinatal Study of the National Institute of Neurological Diseases and Stroke. Washington, DC: US Government Print Office; 1972.

98. Nord Christine, Andreassen Carol, Branden Laura, et al. Early Childhood Longitudinal Study, Birth cohort (ECLS-B), user’s manual for the ECLS-B nine-month public-use data file and electronic code book. NCES, Washington, DC: US Department of Education; 2005.

99. Nye KE. The effect of school size and the interaction of school size and class type on selective student achievement measures in Tennessee elementary schools. University of Tennessee, Knoxville, TN: Unpublished doctoral dissertation; 1995.

100. Olds David, Henderson Charles R, Cole Robert, et al. Long-term effects of nurse home visitation on children’s criminal and antisocial behavior. Journal of the American Medical Association. 1998;280(14):1238–1244.

101. Olds David L, Robinson JoAnn, O’Brien Ruth, et al. Home visiting by paraprofessionals and by nurses: a randomized, controlled trial. Pediatrics. 2002;110(3):486–496.

102. O’Neill June. The role of human capital in earnings differences between black and white men. Journal of Economic Perspectives. 1990;4(4):25–45.

103. r Devah. The use of field experiments for studies of employment discrimination: contributions, critiques, and directions for the future. Annals of the American Academy of Political and Social Science. 2007;609(1):104–133.

104. Phillips Meredith, Brooks-Gunn Jeanne, Duncan Greg J, Klebanov Pamela, Crane Jonathan. Family background, parenting practices, and the black-white test score gap. In: Jencks Christopher, Phillips Meredith, eds. The Black-White Test Score Gap. Washington, DC: The Brookings Institution Press; 1998a;103–147.

105. Phillips Meredith, Crouse James, Ralph John. Does the black-white test score gap widen after children enter school? In: Jencks Christopher, Phillips Meredith, eds. The Black-White Test Score Gap. Washington, DC: The Brookings Institution Press; 1998b;229–272.

106. Plomin Robert, DeFries John C, McClearn Gerald E, McGuffin Peter. Behavioral Genetics. New York: Worth; 2000.

107. Podgursky Michael J, Springer Matthew G. Teacher performance pay: a review. Journal of Policy Analysis and Management. 2007;26(4):909–949.

108. Protheroe Nancy J, Barsdate Kelly J. Culturally Sensitive Instruction and Student Learning. Arlington, VA: Educational Research Center; 1991.

109. Puma Michael, Bell Stephen, Cook Ronna, Heid Camilla, Lopez Michael, et al. Head Start Impact Study: First Year Findings. Washington, DC: US Department of Health and Human Services; 2005.

110. Puma Michael, Bell Stephen, Cook Ronna, Heid Camilla, et al. Head Start Impact Study: Final Report. Washington, DC: US Department of Health and Human Services; 2010.

111. Reuter EB. Racial theory. American Journal of Sociology. 1945;50(6):452–461.

112. Rock Donald A, Stenner Jackson. Assessment issues in the testing of children at school entry. The Future of Children. 2005;15(1):15–34.

113. Rockoff Jonah E. The impact of individual teachers on student achievement: evidence from panel data. American Economic Review. 2004;94(2):247–252.

114. Rockoff Jonah E. Does mentoring reduce turnover and improve skills of new employees? Evidence from teachers in New York City Working paper no 13868. Cambridge, MA: NBER; 2008.

115. Rouse Cecilia E, Krueger Alan B. Putting computerized instruction to the test: a randomized evaluation of a ‘scientifically based’ reading program. Economics of Education Review. 2004;23(4):323–338.

116. Rushton J Philippe, Jensen Arthur. Thirty years of research on race differences in cognitive ability. Psychology, Public Policy, and Law. 2005;11(2):235–294.

117. Rushton J Philippe. Race and crime: international data for 1989–1990. Psychological Reports. 1995;76(1):307–312.

118. Sanbonmatsu Lisa, Kling Jeffrey R, Duncan Greg J, Brooks-Gunn Jeanne. Neighborhoods and academic achievement: results from the moving to opportunity experiment. The Journal of Human Resources. 2006;41(4):649–691.

119. Schanzenbach Diane Whitmore. What have researchers learned from project STAR? Brookings Papers on Education Policy 2006/07: 2007;205–228.

120. Schultz T Paul, Strauss John. Handbook of Development Economics, vol 4. Amsterdam, New York: North-Holland; 2008.

121. Schweinhart Lawrence J, Barnes Helen V, Weikart David P. Significant benefits: the High/Scope Perry Preschool study through age 27. Ypsilanti, MI: High Scope Press; 1993.

122. Shapka Jennifer D, Keating Daniel P. Effects of a girls-only curriculum during adolescence: performance, persistence, and engagement in mathematics and science. American Educational Research Journal. 2003;40(4):929–960.

123. Shonkoff Jack P. A promising opportunity for developmental and behavioral pediatrics at the interface of neuroscience, psychology, and social policy: remarks on receiving the 2005 C. Anderson Aldrich Award Pediatrics. 2006;118:2187–2191.

124. Shukla S. Priorities in educational policy. Economic and Political Weekly. 1971;6(30–32):1649–1651 1653–1654.

125. Taggart Robert. Quantum Opportunity Program Opportunities. Philadelphia, PA: Industrialization Center of America; 1995.

126. Thernstrom Abigail. The drive for racially inclusive schools. Annals of the American Academy of Political and Social Science. 1992;523:131–143.

127. Thompson Ross A. The legacy of early achievements. Child Development. 2000;71(1):145–152.

128. Turney Kristin, Edin Kathryn, Clampet-Lundquist Susan, Kling Jeffrey R, Duncan Greg J. Neighborhood effects on barriers to employment: results from a randomized housing mobility experiment in Baltimore. Brookings-Wharton Papers on Urban Affairs. 2006;2006:137–187.

129. Wagner Mary M, Clayton Serena L. The parents as teachers program: results from two demonstrations. The Future of Children. 1999;9(1):91–115.

130. Walker Gary, Vilella-Velez Frances. Anatomy of a Demonstration: The Summer Training and Education Program (STEP) from Pilot through Replication and Postprogram Impacts. Philadelphia, PA: Public/Private Ventures; 1992.

131. Wigdor Alexandra K, Green Bert F. Performance Assessment for the Workplace, vol 1. Washington, DC: National Academies Press; 1991.

132. Wilson William Julius. More than Just Race: Being Black and Poor in the Inner City (Issues of Our Time). New York: W.W. Norton & Company; 2010.

133. Wong Kenneth L, Shen Francis X. Do school district takeovers work? Assessing the effectiveness of city and state takeovers as school reform strategy. State Education Standard. 2002;3(2):19–23.

134. Wong Kenneth L, Shen Francis X. When mayors lead urban schools: assessing the effects of takeover. In: Howell William G, ed. Beseiged: School Boards and the Future of Education Politics. Washington, DC: The Brookings Institution Press; 2005;81–101.

135. Yeates Keith Owen, MacPhee David, Campbell Frances A, Ramey Craig T. Maternal IQ and home environment as determinants of early childhood intellectual competence: a developmental analysis. Developmental Psychology. 1983;19(5):731–739.

136. Ziedenberg Jason, Schiraldi Vincent. Cellblocks or classrooms? The funding of higher education and corrections and its impact on African American men. Justice Policy Institute: Unpublished paper; 2002.