As I understand it, the Perry Preschool Project is the gold standard of early childhood interventions. One of the main criticisms of deploying the results of the Perry Preschool Project to justify an expansion of pre-K programs is that it would be extremely difficult to replicate the intensity of the Perry Preschool Project at reasonable cost, given the fact that it relied on high-quality personnel and deep relationships with families that might prove impracticable at scale.
And so I was struck by the findings of a 2006 Committee on Economic Development report on the evidence from early childhood intervention demonstration projects:
The Perry Preschool program is estimated to generate nearly $230,000 in benefits per student, much of which is attributable to avoiding the tangible and intangible costs of crime. The long-term follow-ups of these targeted model programs suggest that every dollar invested will return about $4 to $16, with the public recouping one-half to three-quarters of the investment. [Emphasis added]
This theme recurs throughout the report, e.g.:
Considering just the payback to state budgets from implementing preschool programs (i.e., excluding the considerable increase in participants’ earnings), stateprekindergarten programs are estimated to return$1.18 to $2.25 for every dollar states invest. The benefits are largely attributable to near- and long-term savings in crime and education. Schools save money when students arrive better prepared to learn, and teachers are more satisfied with their jobs.
So does the idea that marginal improvements in high school completion will yield significant dividends:
Improvements in education can produce savings in health, crime, and social welfare and help both federal and state governments balance their budgets. For example, improving high school dropout rates couldsave as much as $11 billion annually in welfare, food stamp, and housing assistance. Boosting the high school completion rate of adult men by just 1 percent would save up to $1.4 billion a year in crime costs. Furthermore, providing a single cohort of high school dropouts with one more year of education would generate nearly $42 billion in health care savings, enough to provide every child with a seat in a pre-kindergarten classroom.
To return to crime for a moment:
Crime Outcomes. High-quality preschool is aneffective crime-deterrent program, providing anopportunity to prevent criminal behavior before it begins rather than relying on later rehabilitation. Students who attend high-quality preschool programs are less likely to be arrested as juveniles, with the effects persisting into adulthood. Even when involved in crime, preschool students are less likely to become violent, hardened criminals. As a result, students who attend preschool are also less likely to be sentenced to prison or jail and serve fewer months if incarcerated.
High-quality preschool programs may also make children’s home environments safer and reduce the likelihood that the children will become victims ofcrime. For instance, the incidence of child abuse orneglect among children in the Chicago Child-Parentprogram was nearly one-half that of similar childrenwho did not participate. The cost savings associated with reduced criminal behavior among preschool students is large. The savings from crime in the Chicago Child-Parent Centers program are estimated at $6,000 per student, while the savings in the Perry Preschool program, which include the intangible costs of crime, are estimated to be about $47,000 per student. [Emphasis added]
Investments in preschool programs are strongly justified by the favorable returns from high-quality programs targeted toward disadvantaged children. Benefit/cost analyses for the Carolina Abecedarian, Chicago Child-Parent Centers, and Perry Preschool programs, and a meta-analysis reviewing more than 50 programs, suggest they generate $2 to $16 in benefits for every dollar invested (see Table 3). Viewed another way, the large-scale Chicago Child-Parent Centers program generated more than $40,000 in benefits per student enrolled, and other more intensive programs generated even larger benefits.* The largest net benefits per student accrue from reducing crime among boys, and boosting the earnings of girls. Those programs that demonstrate the largest effects on crime have the largest benefits accruing directly to taxpayers.†
The footnotes are particularly revealing:
The extraordinary net benefits from the Perry Preschool program are, in part, a result of including the savings from reducing the victim cost of crime. When only including the direct cost savings from reductions in criminal activity, the returns from Perry Preschool are more inline with the other programs. [Lynn A. Karoly and James H. Bigelow, The Economics of Investing in Universal Preschool Education inCalifornia (Santa Monica, CA: RAND Corporation, 2005)]
Of the public returns on investment in the Perry Preschool study, the majority of savings (88 percent) came from reductions in crime. Education savings account for 4 percent of savings. Increased taxes from higher earnings account for 7 percent of the savings, and the remaining 1 percent come from savings in welfare payments. Of the public return in the Chicago Child-Parent Centers program, reductions in crime and its associated costs also provided the majority of the benefits (52 percent), while increased tax revenues from earnings accounted for 28 percent of benefits, and savings on school remediation accounted for the remaining 18 percent. Unlike the other model programs, the majority of benefits in the Abecedarian program (more than 80 percent) accrue to the individuals. Because the Abecedarian program did not significantly reduce criminal activity, the public benefits were limited to reductions in K-12 spending, better health, and lower welfare payments. [Clive R. Belfield, Milagros Nores, Steve Barnett, and Lawrence Schweinhart, “The High/Scope Perry Preschool Program: Cost-BenefitAnalysis Using Data from the Age-40 Follow-Up,” Journal of Human Resources, vol. 41 issue 1, pp. 184-186; Arthur J. Reynolds, Judy A.Temple, Dylan L. Robertson, and Emily A. Mann, “Age 21 Cost-Benefit Analysis of the Title I Chicago Child-Parent Centers,” Educational Evaluation and Policy Analysis, vol. 24, no. 4 (Winter 2002), pp. 267-303; Leonard Masse and W. Steven Barnett, “A Benefit-Cost Analysis ofthe Abecedarian Early Childhood Intervention” (New Brunswick, NJ: NIEER, 2002)] [Emphasis added]
So what kind of intangible benefits of crime control were scholars measuring in assessing the Perry Preschool study? We get an indication of this during a brief discussion of studies that did not take into account these intangible benefits:
Relying on more conservative findings (from the Chicago Child-Parent Centers program), a publicly funded, part-day, part-year preschool program for all California four-year-olds is estimated to generatenearly $7,000 in net present-value benefits for every child enrolled. About $2.7 billion in net present-value benefits would be generated from one year of student participation, providing an annual return on investment of 10 percent over 60 years. These benefits are conservative because they do not include the intangible costs of crime, such as victim suffering, nor potential health and intergenerational benefits. Incorporating the savings from just the intangible costs of crime would increase the benefits to California by 50 percent, and boost the rate of return to 13 percent. [Emphasis added]
Victim suffering is of course very important, but measuring the value of counterfactual victim suffering is presumably a difficult proposition.
The CED offers an assessment of potential economy-wide benefits:
Considering just poor children under the optimistic assumption that they attend a very high-quality, Perry-type preschool and experience similar benefits, an initial investment of $19 billion is expected to generate $31 billion in net budgetary savings (in 2004 dollars) by 2030, and to nearly double to $61 billion in savings by 2050. Earnings increases are expected to increase gross domestic product (GDP) by nearly one-half percent, while crime savings would total more than $150 billion by 2050. However, the large benefits associated with Perry Preschool suggest these are upper-bound effects. It is reasonable to expect that a large-scale program similar to the Chicago Child-Parent Centers program could reduce these benefits to one-fifth of the effects modeled from the Perry Preschool program. [Emphasis added]
Crime control remains front and center:
Between 40 and 50 percent of the fiscal benefits that states receive by implementing widely accessible preschool programs are likely to come from later savings in the criminal justice system. The criminal justice savings alone are expected to pay for one-half to more than four-fifths of the cost to expand preschool to all students. The crime effects are particularly important because the United States now spends about $167 billion a year on crime, or $586 for every person in the United States. Crime exacts a large toll on society because of both the direct costs of policing, prosecuting, and incarcerating criminals, and the financial and emotional burden on crime victims. Given the high cost of crime, both in prevention and victimization costs, even a small reduction in criminal activity can have a large effect on state and federal crime expenditures. [Emphasis added]
A few quick thoughts on the CED’s findings:
It seems as though the main vehicle through which well-designed, intensive pre-K programs deliver benefits is through reducing future crime rates and marginally improving high school completion rates. So one assumes that we should compare well-designed, intensive pre-K programs against other crime control strategies and other strategies designed to improve high school completion rates, e.g.:
(1) Increasing the number of police officers, decreasing incarceration rates, and embracing “hot-spot policing,” strategies that John Tierney has identified as potential contributing factors in New York city’s crime decline;
(2) reforming the regulation of narcotics and alcohol with an eye towards reducing violent crime;
(3) embracing organizational innovation in K-12 schools, i.e., reforming work rules, pursuing blended learning models that can be customized to student needs, etc., to increase high school completion rates;
(4) and reducing childhood exposure to lead.
There are versions of (1), (2), and (3) that could generate immediate savings while also generating substantial savings in both the tangible and the intangible costs of crime. (4), in contrast, involves a big upfront investment, but its advocates promise substantial long-term benefits. Though I can’t say for sure that these approaches would be more cost-effective than scaling up the highly-intensive Perry Preschool Project or the less effective Chicago Child-Parent Centers, it seems to be the kind of thing we ought to consider before taking the plunge.
Why? Partly because the reason these interventions generated such substantial savings is that the U.S. has high crime rates. If these crime rates were much lower to start with, the savings would be commensurately lower. Another issue to keep in mind, a friend reminds me, is that there is a difference between a partial equilibrium and a general equilibrium — a small program might help the treatment group avoid a life of crime, but a larger program might not have the same effect because the small program might have simply shifted which people were engaged in criminal activity. This would particularly make sense if the underlying drivers of criminal activity, e.g., the lack of adequate police presence, remained in place.
Finally, I recommend reading Gabriel Rossman’s 2009 post on publication bias:
One of the things I try to stress to my grad students is all the ways that stats can go wrong. Case in point is publication bias (the tendency of scientists to abandon, and journals to reject, work that is not statistically significant). The really weird thing about publication bias is that it means that the p-value means different things depending on where you read it. When you run numbers in Stata and it tells you ”p<.05″ it more or less means what you think it does (i.e., probability of seeing these results if the null were true). However, when you publish that same result and I read it I should interpret the p-value more conservatively.
The reason is that when you get a null finding you tend to give up and if you are so tenacious as to submit it, the peer reviewers tend to reject it. So if you think of the literature as a sample, the null findings tend to be censored out. This wouldn’t necessarily be a problem except that the literature does not also censor false positives. We would expect there to be rather a lot of false positives since the conventional alpha means that about 1 in 20 analyses of noise data would appear significant. [Emphasis added]
It is possible that this kind of publication bias has had no effect on the literature on early childhood interventions, i.e., there have been virtually no null findings, and false positives have not been used to make the case for more investment in early childhood interventions. It is also possible that this is not the case. I should stress that I’m not an expert on this literature, which is why I’ve relied heavily on the CED report.