Abstract

School voucher programs have become a prominent aspect of the education policy landscape in the United States. The DC Opportunity Scholarship Program is the only federally funded voucher program in the United States. Since 2004 it has offered publicly funded private school vouchers to nearly four thousand students to attend any of seventy-three different private schools in Washington, DC. An official experimental evaluation of the program, sponsored by the federal government's Institute of Education Sciences, found that the students who were awarded Opportunity Scholarships graduated from high school at a rate 12 percentage points higher than the students in the randomized control group. This article estimates the benefit/cost ratio of the DC Opportunity Scholarship Program, primarily by considering the increased graduation rate that it induced and the estimated positive economic returns to increased educational attainment. We find a benefit to cost ratio of 2.62, or $2.62 in benefits for every dollar spent on the program.

1.  Introduction

School vouchers “provide government resources to parents to enable them to enroll their children in independent private schools of their choosing” (Wolf 2008, p. 635). In the United States, vouchers for elementary and secondary schooling take two distinct forms: directly funded voucher programs and tax-credit scholarship programs. Under directly funded voucher programs, the government draws from general revenues to give parents checks that they can only redeem for tuition and fees at a participating private school. Tax-credit scholarship programs give businesses and individuals a reduction in their tax liability to compensate them for contributions to nonprofit organizations that then provide scholarships for students to attend private schools. In effect, government forgoes revenues that it otherwise would have collected, and those resources are directed to student scholarships. The school voucher programs in the United States are similar in concept to the government funding of private schools practiced in many countries of Europe, Canada, Australia, and Chile, except that constitutional provisions in the United States require that the funds flow from the government to parents and only then to private schools.1

A total of twenty school voucher and tax-credit scholarship programs operated in twelve U.S. states and the District of Columbia (DC) in 2010–11 (Alliance for School Choice 2011). In the spring of 2011, state and local policy makers enacted five new programs: 1) Douglas County, Colorado2, 2) the state of Indiana, 3) the state of Ohio, 4) the state of Oklahoma, and 5) Racine, Wisconsin. Lawmakers also significantly expanded voucher or tax-credit scholarship initiatives in Florida (two programs), Georgia, Indiana, Iowa, Louisiana, Ohio, and Utah as well as the District of Columbia and the cities of Cleveland and Milwaukee. School voucher programs are spreading briskly across the United States. Do they represent a wise investment of taxpayer resources?

To address that critical question, we draw on the results from an evaluation of the voucher pilot program in Washington, DC (Wolf et al. 2010). In the spring of 2004, the U.S. Congress passed and President George W. Bush signed into law a new program to provide low-income students in DC with vouchers worth up to $7,500 to attend any of seventy-three different DC private schools participating in the program. Through 2009 a total of 3,738 students had been offered Opportunity Scholarships. The overwhelming majority of students offered these scholarships (3,512) were not attending private schools at the time they applied to the program. Those “previous public school” students are the focus of this analysis. Although the Opportunity Scholarship Program (OSP) was the subject of partisan controversy for several years, with most Democrats opposing the program and most Republicans supporting it, at the insistence of Republican Speaker of the House John Boehner the OSP was reauthorized and expanded in April 2011 as part of the final agreement of the fiscal year 2011 U.S. budget. The OSP remains one of the most visible and storied school voucher programs in the United States.

This article proceeds as follows. Section 2 discusses the theory of school choice and why a school voucher program such as the OSP may or may not be expected to yield benefits in excess of its costs. Section 3 examines the program's implementation and the experimental evaluation that produced the educational attainment impact that informs the benefit/cost analysis, which appears in section 4 and comprises much of the article's content. Section 5 presents the results of sensitivity tests applied to the benefit/cost calculations. Section 6 concludes by discussing the implications of the benefit/cost analysis as well as important cautions about the results.

2.  Theory

Two bodies of theory inform this study: school choice and human capital theory. School choice programs have long been proposed as an alternative to the assignment of students to public schools based on their residential address (e.g., Paine 1791; Mill [1859] 1962). Economist Milton Friedman (1955) famously argued that government should subsidize the cost of K–12 education but that parents should choose which schools, public or private, educate their children. Friedman proposed that government vouchers be given to parents as the implementation mechanism to realize his vision of publicly financed but potentially privately delivered elementary and secondary education in the United States. Fellow economist Derek Neal (1997) found that Catholic schools in particular were more effective than neighborhood public schools in promoting educational attainment in the form of high school graduation, college enrollment, and college graduation. Neal speculated that Catholic schools might promote higher levels of self-discipline and a tolerance for delayed gratification in their charges—attributes that lead young people to persist in their pursuit of educational degrees.

The claims of Friedman and others that school vouchers would result in better outcomes for students have been strongly contested. Henry Levin (1998), for example, argues that many parents will choose private schools based on nonacademic criteria; we therefore might not expect their children to benefit academically from vouchers. Jeffrey Henig (1994) expresses concerns that the quality of school information given to school choice participants may be inadequate for them to be informed consumers of educational services. Stewart and Wolf (2011) argue that new school choosers in urban environments in particular require a substantial array of informational supports and guidance in order to shop for schools with confidence. All these challenges could in theory undermine expectations that school vouchers will tend to boost student outcomes.

The second area of theory that informs this study surrounds the returns to human capital. Measuring the return to education is an essential element of human capital theory. Individuals must make decisions about how to best invest in their own capital, and society has a vested interest in developing the capital of its citizens. Education is the primary mechanism that individuals and society have chosen to develop human capital. The optimal education system is one that spends the least amount of money to achieve the desired amount of human capital development.

From the seminal work of Jacob Mincer (1974) to the recent work by Nobel Laureate James Heckman (Heckman and Krueger 2003), economists have worked to estimate the rate of return on investment in education. While initially rates of return to individual years of education were estimated, later work on “sheepskin effects”—the bump that a diploma gives to earnings and other post-education outcomes (Hungerford and Solon 1987; Jaeger and Page 1996)—increased the estimated value of both a high school and a college diploma. In a review of the literature, Ferrer and Riddell (2001) concluded that the sheepskin effect of a high school diploma was from 9 percent to 13 percent of the present value of lifetime earnings.

Using data on the cost of education and estimation strategies for returns to education, the benefit/cost ratio of a particular education program can be estimated. Costs are relatively straightforward as a compilation of the dollars sent to schools to pay for the education of students. Benefits are more complicated to determine, as attempting to track the effects of education across all relevant indicators is difficult. Intuitively it makes sense that those with higher amounts of education would earn more, since higher paying jobs on average demand higher levels of education. This would, as Rouse (2005) points out, increase the opportunity cost of illegal behavior, decreasing the likelihood that a more educated person would commit a crime. Earning more and having the more diversified opportunity set that increased education provides would also decrease the likelihood that an individual would need to participate in government assistance programs. Finally, it would stand to reason that increased income and the lifestyle afforded by that increased income would lead people to live longer, healthier lives, as more personal resources would increase access to higher-quality health care, food, and recreation. We now consider the extent to which such benefits were realized by participants in the DC Opportunity Scholarship Program.

3.  Evaluating the Effectiveness of the OSP

The federal legislation that created the OSP also mandated that the pilot program be evaluated using “the strongest possible research design for determining the effectiveness of the program.”3 Working under the supervision of the U.S. Department of Education's Institute of Education Sciences, the independent research team tasked with studying the OSP implemented a randomized controlled trial (RCT) or experimental analysis. Most of the eligible applicants to the program faced a lottery to determine whether they would be awarded a voucher. Over time, and to gauge the program's impact, the outcomes for the treatment group of students offered vouchers were compared with those for the control group of students not offered vouchers.

The first cohort of students subject to a scholarship lottery applied to the program in the spring of 2004 and was randomly assigned to either the treatment or the control group in July of that year (Wolf et al. 2005). A total of 492 public school students in that initial cohort were entering the oversubscribed grades 6–12 and therefore participated in scholarship lotteries.4 The lottery assigned 299 cohort 1 students to receive a scholarship offer and serve in the treatment group and relegated the remaining 193 students to not receive an offer and serve in the control group.

The second cohort of students in spring of 2005 overwhelmed the program and left it oversubscribed at all grade levels. The 1,816 eligible public school applicants in cohort 2 all faced scholarship lotteries. A total of 1,088 were assigned to receive a scholarship offer and thus serve in the treatment group, and 728 students were randomly assigned to the control group. To increase the study power of the analysis, all the public school students in the two cohorts who had faced scholarship lotteries were combined into a single “impact sample” of 2,308 students, of which 1,387 were in the treatment group and 921 were in the control group. The program enrolled additional cohorts of students in 2006–2009 to replace Opportunity Scholarship recipients who declined to use their scholarship, moved away from DC, reverted to public school, or graduated. These subsequent cohorts of eligible applicants did not face scholarship lotteries and therefore were not added to the impact sample that was the subject of the experimental evaluation.

For purposes of the experimental analysis, the treatment was the mere offer of an Opportunity Scholarship. Thus the core of the experiment was an intent-to-treat analysis. Approximately 22 percent of scholarship recipients never used their voucher to switch to a private school. These scholarship decliners, or “no shows,” remained in the study as members of the treatment group, and their outcomes were averaged in with those of scholarship users in the treatment group to calculate the intent-to-treat programmatic impacts. The control group experienced business as usual, which in this case merely meant no Opportunity Scholarship offer. About 12 percent of control group members attended private schools without the assistance of an Opportunity Scholarship. These students presumably would have gone private even without the program, since they actually did so, and as such they represent an authentic component of the control group counterfactual. The private school attendees in the control group remained in the study as control group members, and their outcomes were averaged in with those from public school attendees in the control group for purposes of calculating programmatic impacts. In sum, assignment to treatment meant the offer of an Opportunity Scholarship and a 78 percent probability of subsequently attending a private school. Assignment to control meant no scholarship offer and a 12 percent probability of subsequently attending a private school. Therefore the analysis identified the impact of increasing a student's likelihood of attending a private school by about 66 percentage points.5

The students in the impact sample were tracked over the pilot's five-year period, from the spring of 2004 through the summer of 2009. A total of 500 of the 2,308 members of the impact sample were old enough to have graduated from high school by the summer of 2009. The parents of these older study participants were asked if their child had received a high school diploma. A total of 62.9 percent of treatment parents and 63.4 percent of control parents responded to the educational attainment survey. If parents answered yes to the question, their child was assigned a 1 for the indicator variable “received a high school diploma.” If parents answered no, their child was assigned a 0 for the variable. If parents replied that their child received a general equivalency degree (GED) in lieu of a high school diploma, their child was coded 0 for the high school diploma variable.6 Because less than 100 percent of the survey targets responded, the outcome data were weighted to rebalance the sample to reflect the equalization of the two groups achieved at baseline through randomization.7

The impact estimate used in this benefit/cost analysis was a regression-adjusted estimate of the program's experimental impact. Although the treatment and control groups were randomized at baseline and the outcome sample was rebalanced to preserve the equalization that randomization produced, the original evaluation obtained greater precision in the impact estimation by controlling for relevant baseline characteristics in a logit regression model of the form:
formula
1
where Xi is a vector of student and family characteristics measured at baseline and known to influence future academic achievement, including reading and math baseline test scores, age, grade, household income, number of children in household, number of months at current residence, number of days from 1 September until the actual date of testing, and indicator variables for having attended a school in need of improvement (SINI), gender, African American, in special education at baseline, mother employed full time, and mother employed part time.8 In this model, τ, the parameter of sole interest, represents the effect of the scholarship offer on a student's probability of having graduated from high school, conditional on Xi.9

The clearest finding from the government-sponsored evaluation of the OSP was that the program increased the likelihood of students graduating from high school. The subgroup of students in the evaluation who were old enough to have graduated from high school by 2009 did so at a rate of 82 percent if they had been offered a scholarship, which was 12 percentage points higher than the 70 percent graduation rate of the randomized control group. The impact of actually using a voucher was to increase the graduation rate from 70 percent to 91 percent (Wolf et al. 2010). Since graduating from high school, as opposed to dropping out, has been connected to a variety of positive life outcomes, this positive impact of the DC OSP on students’ educational attainment is the focus of this benefit/cost analysis.

The final impact report from the OSP evaluation stated, “There is no conclusive evidence that the OSP affected student achievement” (Wolf et al. 2010, p. xv). Although the lack of significant achievement gains might call into question the reliability of the significant attainment impacts reported, there are several reasons why both sets of findings might be accurate. First, there was suggestive evidence that the OSP had some positive impact on reading test scores. The third-year impact report from the study found that the voucher treatment increased student reading scores by an average of 4.5 scale scores or 0.13 standard deviation, an impact that was statistically significant beyond p < 0.01 (Wolf et al. 2009). For the fourth and final year analysis, 296 students “graded out” of the study and could no longer be tested (Wolf et al. 2010).10 Thus the sample of students for the test score analysis in the final year of the study, when no significant achievement impacts were reported, was 13 percent smaller than the test score analysis sample after three years, when significant reading impacts were observed. The positive impact of the OSP on student reading scores after four years would have been deemed statistically significant at a value of p < 0.06 (Wolf et al. 2010), but since the research team had committed to a significance level of p < 0.05, the result of the test was a finding of no significant difference (barely). It is possible that the finding of no significant OSP impact on reading scores in the final analysis is a type II error, induced by a smaller analytic sample and thus lower study power, coupled with a slightly lower reading impact estimate. The program's impact on math achievement was found not to be statistically significant in every year of the evaluation.

Second, the likelihood of high school graduation is probably a more malleable outcome for urban high school students than are standardized test scores. Students' test score performance in high school is heavily influenced by students' background factors and their previous schooling experiences. Moreover, in high school, social considerations and activities occupy an increasing proportion of a student's time and attention. As a result, students may very well forget as much testable knowledge as they learn in high school. Evidence from the DC OSP analysis confirms this hypothesis. On the vertically equated SAT 9 exam, students in the study averaged reading scores that were 10.625 scale score points higher per grade from kindergarten through eighth grade, ceteris paribus. From ninth grade through twelfth grade they gained an average of just 0.005 scale score points in reading per grade, ceteris paribus— essentially no net change in test score performance.11

Although urban high schools may have little leverage over students’ test score achievement during their four-year secondary school career, they may have substantial influence over their likelihood of graduating from high school. High school experiences of an inspiring teacher, an attentive counselor, engaging extracurricular activities, or overall nurturing school environment may be dispositive regarding attaining a high school degree even if such factors have little noticeable effect on test scores for high school students. Although no empirical research yet exists to test the hypothesis, private high schools might increase the likelihood of high school graduation primarily or even exclusively through nurturing character traits of grit, determination, and personal responsibility in students. Those habits, though generally supportive of academic success, in the short term are likely to have greater influence on how far students go than on how much they know. In sum, there are legitimate reasons not to be surprised or alarmed that the students exposed to the OSP treatment demonstrated much larger and clearer attainment impacts than achievement impacts.

Given that the attainment impacts of the OSP appear to be real, does the gain in educational attainment caused by the voucher initiative justify the amount of government resources expended on the program? Since the enactment or expansion of school voucher programs is under consideration in dozens of states and localities around the United States and in various countries worldwide, the answer to that question will inform important contemporaneous debates about the economics of education.

4.  Estimating the Benefits and Costs of Education in the District of Columbia

The first requirement of a successful benefit/cost analysis is the proper alignment of the elements in the numerator and denominator of the benefit/cost ratio. We use the intent-to-treat (ITT) instead of the impact-on-the-treated (IOT) estimate of the OSP's positive impact as the driver of benefits for two reasons. First, the ITT estimate is more conservative in that it includes the outcomes for treatment students who never used their scholarships on the treatment side of the comparison and the outcomes for control group students who self-financed private schooling on the control side. Second, the ITT is the estimate of the impact of awarding an Opportunity Scholarship to a public school student in Washington, DC, and we know that exactly 3,512 public school students were awarded scholarships during the five-year pilot. Thus the numerator of the ratio will be the value of the benefits realized by the 3,512 public school students offered Opportunity Scholarships ensuing from the fact that the scholarship offer increased their likelihood of graduating from high school by 12 percentage points, on average. Calculating the proper denominator for the benefit/cost ratio will be somewhat more complicated. One reason is that the students awarded scholarships who never used them did not require any extra government spending on their behalf. As a result, we omit treatment never-users from the calculation of aggregate program costs. The second complication is that education finance is a complex web of factors, especially in Washington, DC.

The District of Columbia is notorious for the high cost and low performance of its public schools. In an interview with the Washington Post, Secretary of Education Arne Duncan went so far as to say that “DC has had more money than God for a long time, but the outcomes are still disastrous.”12 Estimates of the amount of money spent on public education tend to range widely, so it is an important first step to determine the actual cost of educating a DC student. Because of the comparative nature of this report, in attempting to compare costs from the DC public schools (DCPS) to the DC OSP, it is important to develop a fair comparison group. Since the private schools that receive voucher students are not required to provide supplementary special education or English language learner assistance, it is inaccurate to compare what DCPS spend on these students with what voucher-receiving schools do. We therefore undertook a line-by-line analysis of the DC budget to attempt to parse out dollars that would go to educate students in DCPS compared with DC OSP students (table 1). We also excluded all DCPS capital expenditures from the comparison, as the private schools participating in the OSP all had existing buildings and therefore presumably had to cover only operational costs with the scholarship revenue. The comparable annual cost for DCPS to educate a DC OSP student is $14,939.

Table 1.
Calculating a Comparable DCPS Annual Cost
Total DCPS Students68,681
IEP students −8,450 
ELL students −4,370 
Charter school students −25,729 
Total students less IEP, ELL, and charter 30,132 
Total expenditures on remaining students $450,150,325 
DCPS per pupil cost for regular students $14,939 
Total DCPS Students68,681
IEP students −8,450 
ELL students −4,370 
Charter school students −25,729 
Total students less IEP, ELL, and charter 30,132 
Total expenditures on remaining students $450,150,325 
DCPS per pupil cost for regular students $14,939 

Notes: Total $ excludes special education, vocational ed, gifted and talented, early childhood, ESL, facilities and capital expenses, after school/summer school programs, transportation, Safe and Drug Free Schools initiatives, athletics, co/extracurricular activities, and scholarships. For more details see table A.1.

The calculation of the cost of DC OSP students was simpler to perform. The government appropriation for the five years of the program's pilot was $14 million per year. Thus the total cost of the program over the time of the study was $70 million. If the 78 percent of the 3,512 students offered scholarships that used them had instead received their education in the DCPS over that same time period, the cost to the government would have been $204,604,544. Therefore the operation of the OSP from its launch in 2004 through 2009 had the potential to save the government an estimated $134,604,544 (table 2). A critique could be raised that private schools subsidize the cost of educating students above and beyond the $7,500 value of the voucher. However, it is important to note that we are evaluating the costs and benefits of this particular program as it was implemented in Washington, DC. These are the costs to the taxpayers; if schools decide to contribute above and beyond, the taxpayers benefit from not having to bear that cost. We do acknowledge the implications of private school internal subsidies for the external validity of the study in our cautionary notes below.

Table 2.
Comparing DCPS and DC OSP Costs
OSP Costs
Total appropriation $70,000,000 
DCPS costs  
Comparable cost per pupil $14,939 
Number of students (3,512) * usage rate (78%) 2,739 
Users * 5 years of usage 13,696 
Total cost if voucher students in DCPS $204,604,544 
Total savings $134,604,544 
OSP Costs
Total appropriation $70,000,000 
DCPS costs  
Comparable cost per pupil $14,939 
Number of students (3,512) * usage rate (78%) 2,739 
Users * 5 years of usage 13,696 
Total cost if voucher students in DCPS $204,604,544 
Total savings $134,604,544 

The fact that the operation of the OSP generated a net savings of government resources normally would complicate any benefit/cost analysis of the program, since the cost element of the ratio would be negative (a net fiscal gain), meaning the ratio of benefits to that cost would be infinite. Fortunately for us (but not for U.S. taxpayers), although the cost structure of the OSP could have delivered net fiscal benefits to the government, local public officials entered into an agreement with DCPS to hold the public school district financially harmless when students left its schools through the voucher program. In other words, the public schools continued to receive about $15,000 per student in revenues even for OSP students that they were no longer required to educate. As a result, the cost of the OSP of $70,000,000 over five years of program operation represented a real additional cost to taxpayers. Did the subsequent benefits of the program justify that cost?

Estimating the Benefits of High School Graduation

This article uses the work of several economists to estimate the return on a high school diploma. Rouse's 2005 work “The Labor Market Consequences of an Inadequate Education” is used to estimate increased income and tax revenue due to a high school diploma. Muennig's 2005 work “The Economic Value of Health Gains Associated with Education Intervention” is used to estimate health benefits and their economic repercussions. Lochner and Moretti's 2004 work “The Effect of Education on Crime: Evidence from Prison Inmates, Arrests, and Self-Reports” is used to estimate decreased crime and its benefits.13 At times we divide the benefits into both public benefits (the positive externalities realized by society due to increased education) and private benefits (the positive internal benefits of increased education). However, in calculating the overall benefits of the program, we combine the two, as the overall purpose of education is both public and private.

An Important Caveat

All the authors cited took care to include in their analyses a caveat on estimating the effects of high school graduation on life outcomes. When using survey data and controlling for various characteristics, there is an inherent endogeneity issue in separating out the effect of a high school diploma. Most likely, attainment of a high school diploma is related to inherent characteristics of individuals such as innate motivation or intelligence. These inherent characteristics are most likely also correlated with positive later-life outcomes. Therefore we should take caution in inferring causation of positive-lifetime outcomes from a high school diploma, as most likely there were additional factors apart from the diploma that led to that success. As noted below, only the Lochner and Moretti (2004) analysis of the positive effect of graduation on the likelihood of committing crimes used a method to correct for endogeneity in the relationship between graduation and innate characteristics also related to positive life outcomes.

Importantly, the endogenous connection between innate characteristics and the likelihood of high school graduation was not a concern regarding the higher graduation rates of the treatment group compared with the control group in the OSP experiment. That is because the random assignment of a large number of eligible applicants ensured that the treatment and control groups are expected to be approximately similar in all relevant respects, including innate characteristics such as motivation and intelligence. The endogeneity problem may, however, have biased the estimates of the positive payoff to graduation that we apply to the higher OSP graduation rates, since student motivation and intelligence may be at least partly responsible for those outcomes net of any benefit exclusively due to educational attainment.

Income and Tax Revenue

Individuals with high school diplomas make more money and pay more taxes than their non-diploma-receiving counterparts. Rouse (2005) used data from the 2003 and 2004 March Current Population Survey (CPS) (a sample of over 300,000 individual cases) to track employment patterns. She found that only 53 percent of high school dropouts were employed, while 69 percent of individuals with a high school diploma had a job. For those high school dropouts that were employed, they earned on average $12,000 less per year than their counterparts with a high school diploma. As Rouse explains, “These lower earnings are a function of dropouts earning lower wages, and working fewer hours and weeks per year” (p. 17).

Taken over the course of a working life, a high school diploma has a net present value of $260,000 in increased earnings (after taxes). Rouse used the CPS data set to construct age-earnings profiles of the individuals in the CPS and estimated their lifetime earnings from those profiles. These estimations rest on some serious assumptions (e.g., that these profiles will remain consistent), but Rouse uses various combinations of productivity growth rates and discount rates (see table 7 for an adaptation of the table from her paper) to estimate lifetime earnings. She settles on a discount rate of 3.5 percent and a productivity growth rate of 1.5 percent for a total present value calculation from high school graduation of $260,000.

People who earn more money also pay more money in taxes. Using TAXSIM calculation software, Rouse's previous estimate of increased income, and assumptions matching those above, Rouse estimates a total payment of an extra $60,000 in state and federal taxes over the working life of a high school graduate. That estimate of the positive tax externalities of graduation is conservative, however, since it does not factor in additional wealth (e.g., property) or consumption (e.g., sales) taxes that individuals are more likely to pay if they graduated from high school.

Health

Young adults who graduate from high school live longer, healthier lives. Muennig (2005) cites Wong et al. (2002) as stating that high school graduates live 9.2 years longer than high school dropouts. Those students who fail to graduate, according to Muennig, are more likely to suffer from diseases correlated with poverty (such as diabetes) and are less likely to have access to the highest quality health care.

Muennig used a data set of medical expenditure information to estimate the effect of a high school diploma on health later in life. He drew on the 2002 Medical Expenditure Panel Survey that sampled over 40,000 individuals and used his findings to estimate the difference in quality-adjusted life years (QALY), a common metric in health economics. A QALY is calculated by multiplying the quality of health (from 1 being perfect health to 0 being deceased) of an individual by the additional years expected to result from a particular intervention. One QALY would be an additional year lived in perfect health. Muennig, using health economics literature as a guide, set the present value of a QALY at $80,000.14

Muennig found that a high school diploma resulted in about 1.1 additional QALY lived,15 at a value of $85,000. This calculation rests on a reasonable discount rate assumption of 3.5 percent. If an 8 percent rate of discount is used, the predicted lifetime savings associated with 1.1 QALY falls to $36,000.

Crime

Students who graduate from high school are less likely to commit crimes than are high school dropouts. Intuitively this makes sense, as the increased income and health previously ascribed to high school graduates would increase the opportunity cost of crime. Legitimate work would earn increasing wages, making the low-wage but high-risk life of crime less appealing. Lochner and Moretti (2004) further argue that increased education changes the psychic costs of crime, since those who are more educated tend to exhibit different preferences, such as increased patience or risk aversion, that make crime less palatable.

Lochner and Moretti directly addressed endogeneity problems inherent in estimating the effect of high school graduation by using changes in compulsory education as instruments for their instrumental variables (IV) analysis. They argued that changes in compulsory education laws affect levels of educational attainment but are uncorrelated with individual ability to complete high school. The authors then paired this instrument with data from the FBI Uniform Crime Report and data on criminal activity from the National Longitudinal Survey of Youth to estimate the effect of attainment level on future crime and incarceration. They also used a series of ordinary least squares (OLS) estimates as robustness checks.

Using these estimation techniques, Lochner and Moretti (2004, p. 23) found that “high school graduation reduces the probability [of incarceration] by 3−4 percentage points among white men aged 22−28 and 8−9 percentage points among black men.” This reduction in crime saves money for those who will not become criminals, those who will not be the victims of these crimes, and those who will not have to pay for incarcerating the perpetrators. The authors state that average victim costs for crimes range from $2.94 million for murder to $370 for larceny/theft. The average incarceration costs for crimes range from $845,455 for murder to $44 for larceny/theft. By decreasing the incidence of these crimes, both victim and incarceration costs decrease. In total, the authors estimate that “the social benefit per additional male graduate amounts to around $1,170 to $2,100” (p. 26) per year of the person's life.

Comparing Benefits to Costs

If students received only a $7,500 voucher to fund their education, the DC Opportunity Scholarship Program would have an infinite benefit/cost ratio, as this is less than the amount spent to finance their education in the public school alternative. However, because local officials decided to hold the District of Columbia public schools financially harmless for the loss of students to the school voucher program, the OSP constituted an additional expenditure by taxpayers. The following calculations are based simply on the $70 million appropriated for and spent on the program.

The DC Opportunity Scholarship Program increased graduation rates for those students offered vouchers. The graduation rate in the treatment group was 82 percent, compared with the 70 percent graduation rate in the control group. Since 3,512 students were offered scholarships during the program pilot, the higher graduation rate of 12 percent is forecast to produce 421 additional graduates. This number can then be multiplied by the estimates of average lifetime savings for each marginal graduate based on the studies reviewed above (table 3).

Table 3.
Benefits
Private benefit
Increased graduation rate 0.12 
Increased number of graduates 421 
Increased lifetime earnings $262,519 
Total benefit $110,636,007 
Increased health/quality of life $85,000 
Total benefit $35,822,400 
Total benefit (private) $146,458,407 
Public benefit Public benefit  
Decrease in cost of crime $28,030 
Total benefit $11,812,963 
Increased tax revenue $59,210 
Total benefit $24,953,462 
Total benefit (public) $36,766,425 
Total benefit (public and private) $183,224,832 
Private benefit
Increased graduation rate 0.12 
Increased number of graduates 421 
Increased lifetime earnings $262,519 
Total benefit $110,636,007 
Increased health/quality of life $85,000 
Total benefit $35,822,400 
Total benefit (private) $146,458,407 
Public benefit Public benefit  
Decrease in cost of crime $28,030 
Total benefit $11,812,963 
Increased tax revenue $59,210 
Total benefit $24,953,462 
Total benefit (public) $36,766,425 
Total benefit (public and private) $183,224,832 

The simplest calculations of benefits in table 3 come from the work of the economists who directly calculated the net present value of a high school diploma. Rouse (2005) valued a high school diploma at $262,519 of income and $59,210 in tax revenue and Muennig (2005) valued a high school diploma at $85,000 in health benefits. Wong et al. (2002) estimated that those with a high school diploma but no exposure to college lose 3.6 expected life years before seventy-five, versus 18.8 potential life years lost for high school dropouts. Reductions in crime, therefore, were calculated by multiplying the $1,170 in benefits per year of each additional graduate by the 53.4 additional years a high school graduate on average is expected to live past their eighteenth birthday (discounted at 3.5 percent).

Through these straightforward calculations we estimate that the operation of the DC Opportunity Scholarship Program during its pilot phase of 2004–9 generated benefits totaling $183,224,832 due to the higher high school graduation rates delivered by the program. Of this total, $146,458,407 is estimated to be realized by the individuals participating in the OSP, while $36,754,092 is estimated to be due to positive externalities realized by society writ large. Dividing the total benefit of $183,212,499 by the total expenditure of public funds ($70,000,000) yields a benefit/cost ratio of 2.62, or $2.62 of benefit for every dollar expended (table 4).

Table 4.
Benefit/Cost Ratio
Benefit-to-cost ratio
Total public benefit $36,766,425 
Total cost $70,000,000 
Public benefit/cost ratio 0.53 
Total private benefit $146,458,407 
Total cost $70,000,000 
Private benefit/cost ratio 2.09 
Total benefit $183,224,832 
Total cost $70,000,000 
Total benefit/cost ratio 2.62 
Benefit-to-cost ratio
Total public benefit $36,766,425 
Total cost $70,000,000 
Public benefit/cost ratio 0.53 
Total private benefit $146,458,407 
Total cost $70,000,000 
Private benefit/cost ratio 2.09 
Total benefit $183,224,832 
Total cost $70,000,000 
Total benefit/cost ratio 2.62 

5.  Sensitivity Analyses

Given the nature of the techniques used to determine both the graduation impacts of the DC OSP and the monetary value of a high school diploma, we performed a series of sensitivity tests to provide upper and lower bounds for the benefits of the program. The first sensitivity test uses the upper and lower bounds of the 95 percent confidence interval to provide high- and low-end estimates of the numbers of graduates attributed to the program (table 5). At the low end of the confidence interval, a 3 percent increase in graduation is found, resulting in 105 graduates who would not have completed high school absent the intervention. At the high end, a 21 percent increase is found, yielding 738 additional graduates.

Table 5.
Graduate Estimates by Confidence Interval
Graduation Rate Estimation
 Lower Bound Regression Estimate Upper Bound 
 0.03 0.12 0.21 
Estimated number of graduates 105.36 421.44 737.52 
Graduation Rate Estimation
 Lower Bound Regression Estimate Upper Bound 
 0.03 0.12 0.21 
Estimated number of graduates 105.36 421.44 737.52 

Upper and lower bounds of the program's fiscal benefit can then be determined by combining these extreme ends of the plausible range of positive high school graduation impacts with the estimates of the monetary value of a high school diploma (table 6). At the lower bound (the 3 percent increase), the benefit estimate is reduced to $45,806,209. At the upper bound (the 21 percent increase), the benefit estimate is increased to $320,643,458. Holding the cost ($70,000,000) constant offers benefit/cost ratios ranging from 0.65 at the lower bound (giving 65 cents of benefit for every dollar expended) to 4.58 at the upper bound ($4.58 of benefit for every dollar expended).

Table 6.
Benefit Estimates by Confidence Interval
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $262,519 $27,659,002 $110,636,007 $193,613,013 
Health/quality of life $85,000 $8,955,600 $35,822,400 $62,689,200 
Total private $347,519 $36,614,602 $146,458,407 $256,302,213 
Crime $28,030 $2,953,241 $11,812,963 $20,672,686 
Taxes $59,210 $6,238,366 $24,953,462 $43,668,559 
Total public $87,240 $9,191,607 $36,766,425 $64,341,245 
Total benefit  $45,806,209 $183,224,832 $320,643,458 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  0.65 2.62 4.58 
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $262,519 $27,659,002 $110,636,007 $193,613,013 
Health/quality of life $85,000 $8,955,600 $35,822,400 $62,689,200 
Total private $347,519 $36,614,602 $146,458,407 $256,302,213 
Crime $28,030 $2,953,241 $11,812,963 $20,672,686 
Taxes $59,210 $6,238,366 $24,953,462 $43,668,559 
Total public $87,240 $9,191,607 $36,766,425 $64,341,245 
Total benefit  $45,806,209 $183,224,832 $320,643,458 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  0.65 2.62 4.58 

The individual estimates of the benefits of a high school diploma can also be subject to sensitivity analysis. In the cited articles, two of the authors presented information on the discount and productivity growth calculations they made in order to arrive at their valuation of a high school diploma. Rouse (2005, 2007) provided three productivity growth rates and three discount rates for both earnings and tax estimates (adapted into table 7). Although she settled on a 1.5 percent productivity growth rate and a 3.5 percent discount rate, the estimates for earnings ranged from $121,074 for a 0 percent productivity growth rate and a 6 percent discount rate to $294,024 for a 2 percent productivity growth rate and a 3.5 percent discount rate for earnings. With respect to increased tax revenue, estimates ranged from $25,311 to $66,946, respectively.

Table 7.
Income and Tax Estimates with Varying Productivity Growth and Discount Ratesa
Productivity Growth
Discount Rate 1.5 
Earnings    
3.5 $190,230 $262,519 $294,024 
$172,559 $235,703 $263,084 
$121,074 $159,045 $175,175 
Taxes    
3.5 $41,683 $59,210 $66,946 
$37,456 $52,669 $59,349 
$25,311 $34,242 $38,080 
Productivity Growth
Discount Rate 1.5 
Earnings    
3.5 $190,230 $262,519 $294,024 
$172,559 $235,703 $263,084 
$121,074 $159,045 $175,175 
Taxes    
3.5 $41,683 $59,210 $66,946 
$37,456 $52,669 $59,349 
$25,311 $34,242 $38,080 

Adapted from Rouse 2005, table 3. Reprinted with permission.

Similarly, Muennig (2005, 2007) provided both a “conservative” estimate of the monetized health benefits of a high school diploma and what he termed an “average” estimate. The conservative figure (used in the above benefit/cost calculations) puts the value of a QALY at $80,000 and the value of the 1.1 additional QALYs a high school diploma brings at $85,000. Muennig states that using average and acceptable calculations sets the value of a QALY much higher, at $166,000, and thus the 1.1 additional QALYs of a high school graduate at $182,600. The calculations using the 3.5 percent discount rate that he felt most appropriate (table 8) were used to estimate the average estimate of a QALY for both 0 percent and 8 percent discount rates, ranging from $195,000 for the conservative estimate and $418,906 for the average estimate at a 0 percent discount rate to $36,000 for the conservative estimate and $77,336 for the average estimate at an 8 percent discount rate.

Table 8.
Health Benefit Estimates by Discount Ratesa
Discount RateConservative EstimateAverage Estimates
0 $195,000 $418,906 
3.5 $85,000 $182,600 
8 $36,000 $77,336 
Discount RateConservative EstimateAverage Estimates
0 $195,000 $418,906 
3.5 $85,000 $182,600 
8 $36,000 $77,336 

Adapted from data presented in Muennig 2005, 2007. Reprinted with permission.

These additional figures give us the ability to reestimate lower (table 9) and upper (table 10) estimates of benefit/costs with the lowest possible figures for earnings, taxes, and health benefits, and the highest possible figures ranging from lowest possible benefit/cost ratio of 0.36 (36 cents for every dollar expended) to the highest possible benefit/cost ratio of 7.82 ($7.82 for every dollar expended).

Table 9.
Lowest Estimates of Cost/Benefit Ratios
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $121,074 $12,756,357 $51,025,427 $89,294,496 
Health/quality of life $36,000 $3,792,960 $15,171,840 $26,550,720 
Total private $345,000 $16,549,317 $66,197,267 $115,845,216 
Crime $28,030 $5,399,974 $21,599,894 $37,799,815 
Taxes $25,311 $2,666,767 $10,667,068 $18,667,369 
Total public $124,463 $8,827,485 $35,309,940 $78,655,562 
Total benefit  $25,268,802 $101,507,207 $194,500,778 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  0.36 1.45 2.78 
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $121,074 $12,756,357 $51,025,427 $89,294,496 
Health/quality of life $36,000 $3,792,960 $15,171,840 $26,550,720 
Total private $345,000 $16,549,317 $66,197,267 $115,845,216 
Crime $28,030 $5,399,974 $21,599,894 $37,799,815 
Taxes $25,311 $2,666,767 $10,667,068 $18,667,369 
Total public $124,463 $8,827,485 $35,309,940 $78,655,562 
Total benefit  $25,268,802 $101,507,207 $194,500,778 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  0.36 1.45 2.78 
Table 10.
Highest Estimates of Cost/Benefit Ratios
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $294,024 $25,412,494 $101,649,977 $177,887,460 
Health/quality of life $481,906 $41,651,136 $166,604,542 $291,557,949 
Total private $345,000 $67,063,630 $268,254,520 $469,445,409 
Crime $50,310 $5,300,662 $21,202,646 $37,104,631 
Taxes $66,946 $5,786,143 $23,144,571 $40,502,999 
Total public $117,256 $11,086,805 $44,347,217 $77,607,630 
Total benefit  $78,150,435 $312,601,737 $547,053,039 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  1.12 4.47 7.82 
Benefits by Estimates
  Lower Bound Regression Estimate Upper Bound 
Earnings $294,024 $25,412,494 $101,649,977 $177,887,460 
Health/quality of life $481,906 $41,651,136 $166,604,542 $291,557,949 
Total private $345,000 $67,063,630 $268,254,520 $469,445,409 
Crime $50,310 $5,300,662 $21,202,646 $37,104,631 
Taxes $66,946 $5,786,143 $23,144,571 $40,502,999 
Total public $117,256 $11,086,805 $44,347,217 $77,607,630 
Total benefit  $78,150,435 $312,601,737 $547,053,039 
Total cost  $70,000,000 $70,000,000 $70,000,000 
Benefit/cost  1.12 4.47 7.82 

6.  Cautions and Conclusions

There are good reasons why we throw parties for young adults who graduate from high school. In the United States, graduating from high school, as opposed to dropping out, significantly raises the employment likelihood, earning potential, and various other important quality-of-life metrics for individuals that in turn generate substantial positive externalities for society. Because the DC Opportunity Scholarship Program demonstrated a clear positive impact on high school graduation rates, it delivered an important benefit that can be compared with its public costs. Drawing from previous work of well-known labor economists, we find that expending the additional dollars to fund the OSP over six years generated benefits that exceeded the program's costs, even counting the $70 million price tag as an additional cost over and above what was already provided to educate these students in traditional public schools. If, as is supposed to be the case in voucher programs, the voucher amount actually replaced the higher amount of $14,939 needed to pay for the education of these students in public school, the cost savings to society would have been even greater.

Admittedly, the estimations used to monetize the value of a high school diploma rely on some serious assumptions and are likely not the most reliable figures. The theory underpinning this analysis, however, is clear and compelling. Young adults with greater levels of education and the greater skills associated with higher educational attainment are more attractive on the job market and can thus demand a higher salary. This grants them access to better nutrition and health care, decreases the likelihood that they would ever need to take advantage of public assistance, and raises the opportunity cost of crime, greatly decreasing the likelihood of its occurrence. In short, increasing the number of high school graduates benefits both the individuals graduating and the taxpayers that funded their education and that will pay for social services if students fail to make use of it.

Some cautions are in order. First, the fact that student achievement was little affected by the OSP while attainment increased markedly, while not entirely surprising, has implications for our analysis. Do students generate personal and societal benefits from graduating high school primarily because graduates are smarter than nongraduates or primarily because graduates have a credential that nongraduates lack? If the former, then the actual return on investment of the OSP might prove to be lower than we estimate here. If the latter, then our results should hold, even though student test score achievement did not appear to increase much due to the OSP. This matter of why high school graduates do better than nongraduates is underexplored in the human capital literature, though some studies do indicate that the sheepskin effect of the diploma credential per se is responsible for the benefits of graduation (Hungerford and Solon 1987; Jaeger and Page 1996). Recent nonexperimental empirical research on the attainment effects of the Milwaukee Parental Choice Program found that Milwaukee voucher students graduated from high school at rates 5−7 percentage points higher than a matched sample of public school students and also enrolled and persisted in four-year colleges at the same higher rates (Cowen et al. 2013). Although a related achievement analysis concluded that there was little evidence that the Milwaukee voucher students scored higher on standardized tests as a result of the program (Witte et al. 2012), the fact that Milwaukee voucher students were more likely than similar peers to enroll in and progress through college indicates that private schooling through the voucher program somehow better prepared them for future academic success. Those results generally support the integrity of our analysis here.

Second, only 202 of the 3,512 students offered scholarships by the program during the pilot were in the experimental study sample and also old enough to have graduated from high school before the study ended. Although the treatment students in the attainment analytic sample were broadly representative of all students offered Opportunity Scholarships on observable characteristics (except for age, grade level, and application cohort), it is possible that they were exceptional in less obvious ways. The fact that they were among the first two cohorts of applicants to the program and were eager to switch from a public to a private school late in their educational career may mean that the treatment and control students whose attainment results informed the estimation of the program's attainment impact possessed unobservable characteristics—such as a sense that private schooling would be a good fit specifically for them—that contributed to the large attainment gains. Strictly speaking, such a situation would not bias the experimental impact estimations, as random assignment ensures that unobservable characteristics are equal in expectation across the treatment and control groups, but it could generate a bias when the attainment results are generalized to the entire group of OSP participants, as we had to do here to implement our benefit/cost analysis. The attainment gains from the OSP, and thus the average return on the program investment that we estimate here, might not have been quite so large for all OSP students compared with the students who were first out of the chute.

Finally, the reauthorized DC voucher program and future programs in other places may generate different benefit/cost payoffs than we have calculated for the OSP pilot. The estimated return on investment of 262 percent achieved by the DC OSP during its early years may not be as large in the future—for example, if the elite private high schools that initially participated did so merely to demonstrate their effectiveness and later decided to reduce or eliminate their internal subsidies of voucher student costs. Under such a scenario, the average attainment benefit realized by DC voucher students would decline, because fewer seats would be available in the best private schools, and thus the return on investment would decrease. On the other hand, the full-scale Milwaukee voucher program appears to be generating attainment effects that are about half the size of the ITT effects of the DC OSP pilot (Cowen et al. 2013), but the government actually saves money due to the program's operation. State money follows the child in Wisconsin, and the voucher maximum is less than the state per pupil expenditure in Milwaukee public schools, yielding an annual fiscal program benefit of over $50 million (Costrell 2010) on top of the attainment benefit.

This is all to say that benefit/cost analyses are decidedly influenced by context, and our assessment of the relative return yielded by the DC OSP is no different. Still, throughout the analysis we deliberately adopted approaches that were highly conservative so that, even in the face of circumstances we could not control and assumptions that had to be made, our estimation would be as likely to underestimate as to overestimate the actual return on the OSP investment. An experimental evaluation of the DC OSP demonstrated clearly that the federal school voucher program increased the high school graduation rate of program participants. Our analysis shows that the benefits estimated to be realized as a result of that higher graduation rate are probably between double and triple the documented program costs. In the field of the economics of education, as in the field of investment, anything with a demonstrable yield of over 100 percent is an attractive investment. The five-year pilot of the DC Opportunity Scholarship Program generated impacts on student educational attainment that are forecasted to pay a net return of 262 percent over the course of the lives of the low-income DC students who participated.

Notes

1 

Zelman v. Simmons-Harris. 2002. 536 U.S. 639.

2 

This program is currently being challenged in court.

3 

DC Parental Choice Incentive Act of 2003. 2004. U.S. Statutes at Large 118: 126–42. Available http://constitution.org/uslaw/sal/118_statutes_at_large.pdf. Accessed 15 August 2012.

4 

Student applicants were organized into “grade bands” of K−5, 6–8, and 9–12 that were matched up with the known number of slots available in participating private schools within those bands. Since the number of available slots exceeded the number of applicants entering K−5 in 2004, the cohort 1 eligible applicants in those elementary grades automatically received voucher awards. The cohort 1 eligible applicants entering grades 6–8 and 9–12 in 2004 faced different lottery scholarship probabilities depending on the degree of oversubscription. Students applying to the program from schools designated in need of improvement (SINI) under No Child Left Behind were advantaged in the lottery by having a higher lottery scholarship probability than non-SINI applicants. To implement this somewhat complicated lottery, students were randomized within various strata based on their grade band and SINI status. For purposes of the analysis, sample weights were used to eliminate any compositional differences across the treatment and control groups introduced into the data as a result of the differential lottery probabilities across the strata.

5 

It was essential to keep all lottery winners in the treatment group and all lottery losers in the control group, regardless of whether they subsequently attended a private school, because the treatment students who actually used their scholarships and the control students who actually remained in public schools were distinctive subgroups of their respective lottery groups. Any comparison of the outcomes from only treatment scholarship users to only control group public school students would be biased by the fact that relatively advantaged members of the treatment group actually used their scholarships, while relatively disadvantaged members of the control group remained in public school. Preserving the complete treatment and control groups as determined by the random lotteries, as was done here, avoids any such bias.

6 

An extensive empirical literature, summarized by Chaplin (2002), concludes that a GED does not produce increases in lifetime earnings and other quality-of-life indicators commensurate with the increases from obtaining an actual high school diploma.

7 

The weight was set as the inverse of the probability of response given key baseline characteristics identified through a logit estimation predicting survey response. A supplemental analysis confirmed that the nonresponse weights succeeded in rebalancing the sample to reflect baseline equivalence (Wolf et al. 2010).

8 

Some missing baseline data were imputed by fitting stepwise models to each covariate using all the available baseline covariates as potential predictors.

9 

The observations in the study were potentially nested in two ways. Some families contributed multiple children to the study sample, and some students in the sample attended the same school as other students in the sample. To prevent this nested structure of the data from producing spatial autocorrelation among the error terms, the observations were first clustered by family to produce the main effect estimate and then clustered by school to produce a robustness check on that main effect estimate. Clustering observations does not affect the estimates of the programmatic impact, although they can affect the level of statistical significance of those estimates by changing the size of the impact estimates’ standard errors. In this case, the p-value of the estimate of the program's impact on the likelihood of high school graduation increased from p < 0.01 when observations were clustered by family to p < 0.02 when observations were clustered by school (Wolf et al. 2010).

10 

The research design required that students be administered the test that was one grade higher than the test they had taken the previous year. Once students had taken the twelfth-grade test they could not be further tested, since the SAT 9 does not have a thirteenth-grade test, and they dropped out of the test score impact sample.

11 

Students’ standardized test scores also might stagnate in high school because they are learning relatively more specialized or esoteric content that is not adequately captured by grade-level norm-referenced tests.

13 

In earlier drafts of this article, we included estimates of the decrease in public assistance seen with an increase in high school graduation rates. Specifically, Waldfogel, Garfinkel, and Kelly's 2005 work was used to estimate the benefits from reductions in the use of public assistance due to high school graduation. This work, however, based its estimates on a subset of young adults (single mothers) that was too specific to generalize to our analysis here.

14 

There is some debate about the value of QALYs. For a more complete discussion see Weinstein (2008).

15 

This finding is curious (and is not explained by the author), given that a high school graduate lives on average 9.2 years longer and accrues only 1.1 QALYs, implying that those additional years are lived in poor health. Some possible explanations could be that the type of work in which high school diploma earners participate is still extremely physically demanding and, while it provides better benefits that prolong life, is at the end of a difficult life of hard work.

Acknowledgments

The authors would like to thank Dr. Robert Costrell, participants at the Association for Education Finance and Policy's 36th Annual Conference in Seattle, Washington, in March 2011, and participants in the Workshop on the Economics of Education II at the University of Barcelona in September 2011 who reviewed preliminary drafts of this article. The usual caveats apply.

REFERENCES

Alliance for School Choice
.
2011
.
School choice yearbook 2010–2011
.
Washington, DC
:
Alliance for School Choice
.
Chaplin
,
Duncan
.
2002
.
Tassels on the cheap.
Education Next
2
(
3
):
24
29
.
Costrell
,
Robert M.
2010
. The fiscal impact of the Milwaukee Parental Choice Program: 2010–2011 update and policy options. SCDP Milwaukee Evaluation Report No. 22, University of Arkansas.
Cowen
,
Joshua M.
,
David J.
Fleming
,
John F.
Witte
,
Patrick J.
Wolf
, and
Brian
Kisida
.
2013
,
in press
.
School vouchers and student attainment: Evidence from a state-mandated study of Milwaukee's Parental Choice Program.
Policy Studies Journal.
Ferrer
,
Ana
, and
W. Craig
Riddell
.
2001
. Sheepskin effects and the returns to education. Available http://qed.econ.queensu.ca/pub/jdi/deutsch/edu_conf/Ferrer.pdf. Accessed 25 September 2011.
Friedman
,
Milton
.
1955
.
The role of government in education
. In
Economics and the Public Interest
, edited by
Robert A. Solo
, pp.
123
44
.
New Brunswick, NJ
:
Rutgers University Press
.
Heckman
,
James J.
, and
Alan B.
Krueger
.
2003
.
Inequality in America: What role for human capital policies?
Cambridge, MA
:
MIT Press
.
Henig
,
Jeffrey R.
1994
.
Rethinking school choice: Limits of the market metaphor
.
Princeton, NJ
:
Princeton University Press
.
Hungerford
,
Thomas
, and
Gary
Solon
.
1987
.
Sheepskin effects in the return to education.
Review of Economics and Statistics
69
(
1
):
175
77
.
Jaeger
,
David A.
, and
Marianne E.
Page
.
1996
.
Degrees matter: New evidence on sheepskin effects in the returns to education.
Review of Economics and Statistics
78
(
4
):
733
40
.
Levin
,
Henry M.
1998
.
Educational vouchers: Effectiveness, choice, and costs
.
Journal of Policy Analysis and Management
17
(
3
):
373
92
.
Lochner
,
Lance
, and
Enrico
Moretti
.
2004
.
The effect of education on crime: Evidence from prison inmates, arrests, and self-reports.
American Economic Review
94
(
1
):
155
89
.
Mill
,
John Stuart
. [1859]
1962
.
On liberty. In
Utilitarianism, On liberty, Essay on Bentham
,
edited by Mary Warnock
, pp.
126
250
.
New York
:
New American Library
.
Mincer
,
Jacob
.
1974
.
Schooling, experience, and earnings
.
New York
:
NBER Press
.
Muennig
,
Peter
.
2005
. The economic value of health gains associated with education intervention. Paper presented at the Social Costs of Inadequate Education Symposium, Teachers College, Columbia University, October.
Muennig
,
Peter
.
2007
.
Consequences in health status and costs
. In
The price we pay: Economic and social consequences of inadequate education
, edited by
Clive R. Belfield and Henry M. Levin
, pp.
125
41
.
Washington, DC
:
Brookings Institution Press
.
Neal
,
Derek
.
1997
.
The effects of Catholic secondary schooling on educational attainment.
Journal of Labor Economics
15
(
1
):
98
123
.
Paine
,
Thomas
.
1791
.
Rights of man
.
London
:
J.S. Jordan
.
Rouse
,
Cecilia Elena
.
2005
. The labor market consequences of an inadequate education. Paper presented at the Social Costs of Inadequate Education Symposium, Teachers College, Columbia University, October.
Rouse
,
Cecilia Elena
.
2007
.
Consequences for the labor market
. In
The price we pay: Economic and social consequences of inadequate education
, edited by
Clive R. Belfield and Henry M. Levin
, pp.
99
124
.
Washington, DC
:
Brookings Institution Press
.
Stewart
,
Thomas
, and
Patrick J.
Wolf
.
2011
.
The evolution of parental school choice
. In
Customized schooling
, edited by
Frederick M. Hess and Bruno Manno
, pp.
91
106
.
Cambridge, MA
:
Harvard Education Press
.
Turque
,
Bill
, and
Maria
Glod
.
2009
. Stimulus to help retool education, Duncan says. Washington Post, 5 March.
Waldfogel
,
Jane
,
Irwin
Garfinkel
, and
Brendan
Kelly
.
2005
. Public assistance program: How much could be saved with improved education? Paper presented at the Social Costs of Inadequate Education Symposium, Teachers College, Columbia University, October.
Weinstein
,
M. C.
2008
.
How much are Americans willing to pay for a quality-adjusted life year?
Medical Care
46
(
4
):
349
56
.
Witte
,
John F.
,
Deven
Carlson
,
Joshua M.
Cowen
,
David J.
Fleming
, and
Patrick J.
Wolf
.
2012
. The MPCP longitudinal educational growth study final year report. SCDP Milwaukee Evaluation Report No. 29, University of Arkansas.
Wolf
,
Patrick
,
Babette
Gutmann
,
Nada
Eissa
,
Michael
Puma
, and
Marsha
Silverberg
.
2005
.
Evaluation of the DC Opportunity Scholarship Program: First year report on participation.
U.S. Department of Education, Institute for Education Sciences, National Center for Education Evaluation and Regional Assistance.
Washington, DC
:
U.S. Government Printing Office
.
Wolf
,
Patrick
,
Babette
Gutmann
,
Michael
Puma
,
Brian
Kisida
,
Lou
Rizzo
,
Nada
Eissa
, and
Matthew
Carr
.
2010
.
Evaluation of the DC Opportunity Scholarship Program: Final report.
U.S. Department of Education, Institute of Education Sciences, National Center for Education Evaluation and Regional Assistance.
Washington, DC
:
U.S. Government Printing Office
.
Wolf
,
Patrick
,
Babette
Gutmann
,
Michael
Puma
,
Brian
Kisida
,
Lou
Rizzo
, and
Nada O.
Eissa
.
2009
.
Evaluation of the DC Opportunity Scholarship Program: Impacts after three years.
U.S. Department of Education, Institute of Education Sciences, National Center for Education Evaluation and Regional Assistance.
Washington, DC
:
U.S. Government Printing Office
.
Wolf
,
Patrick J.
2008
.
Vouchers
. In
The international encyclopedia of education
, edited by
Gary McCulloch and David Crook
, pp.
635
36
.
London
:
Routledge
.
Wong
,
Mitchell D.
,
Martin F.
Shapiro
,
W. John
Boscardin
, and
Susan L.
Ettner
.
2002
.
Contribution of major diseases to disparities in mortality.
New England Journal of Medicine
347
(
20
):
1585
92
.

Appendix A

Table A.1.
Calculating the Comparable per Pupil Cost of a DCPS Student (2008)
Instructional programsGeneral Education$238,157,000
 Substitutes $3,376,000 
 Textbook Program $7,195,000 
 Library and Media $223,000 
 Tech and Support $1,222,000 
Instructional support services Curriculum Development and Implementation $11,500,000 
 Professional Development $957,000 
 Local Grants Administration $11,115,000 
 Parental Engagement $3,000,000 
Title II programs Administration LEA Programs $691,000 
 Professional Development Program $5,189,000 
 Prof. Development Literacy and Numeracy $1,301,000 
 Prof. Development Mentoring Program $2,000,000 
 Prof. Development Schools Program $3,067,000 
Title IV grants Administration LEA Programs $1,016,000 
Title V grants Administration LEA Programs $15,000 
 Innovative Education LEA Programs $377,000 
Title I SEA set-aside School Improvement Grants $1,000,000 
Title I part B Reading First $2,016,000 
Title II part D Instructional Tech/LEA Programs $476,000 
Title I grant LEA Administration $2,284,000 
 Homeless Children Program $250,000 
 Parental Involvement Reserve $329,000 
 Neglected and Delinquent Youth Reserve $475,000 
 School Improvement Program Reserve $3,093,000 
 Other Title 1 Services Reserve $6,801,000 
 Professional Development Reserve $1,621,000 
 Supplemental Services Reserve $4,639,000 
 Educational Program−School $16,935,000 
Student support services Student Services $546,000 
 Health Services $1,028,000 
 Transitory Services $510,000 
 Student Affairs $85,000 
Noninstructional support services Custodial $31,509,000 
 Food Services $26,340,000 
 Security Services $12,916,000 
 Public Utilities $42,628,000 
Other states function Assessment and Accountability $1,545,000 
 Grants Administration $281,000 
Office of the deputy mayor for education Total Budget $2,442,325 
 Total $450,150,325 
 Total Students 68,681 
 IEP Students 8,450 
 ELL Students 4,370 
 Charter School Students 25,729 
 Total Less IEP, ELL, and Charter 30,132 
 DCPS per Pupil Cost $14,939 
Instructional programsGeneral Education$238,157,000
 Substitutes $3,376,000 
 Textbook Program $7,195,000 
 Library and Media $223,000 
 Tech and Support $1,222,000 
Instructional support services Curriculum Development and Implementation $11,500,000 
 Professional Development $957,000 
 Local Grants Administration $11,115,000 
 Parental Engagement $3,000,000 
Title II programs Administration LEA Programs $691,000 
 Professional Development Program $5,189,000 
 Prof. Development Literacy and Numeracy $1,301,000 
 Prof. Development Mentoring Program $2,000,000 
 Prof. Development Schools Program $3,067,000 
Title IV grants Administration LEA Programs $1,016,000 
Title V grants Administration LEA Programs $15,000 
 Innovative Education LEA Programs $377,000 
Title I SEA set-aside School Improvement Grants $1,000,000 
Title I part B Reading First $2,016,000 
Title II part D Instructional Tech/LEA Programs $476,000 
Title I grant LEA Administration $2,284,000 
 Homeless Children Program $250,000 
 Parental Involvement Reserve $329,000 
 Neglected and Delinquent Youth Reserve $475,000 
 School Improvement Program Reserve $3,093,000 
 Other Title 1 Services Reserve $6,801,000 
 Professional Development Reserve $1,621,000 
 Supplemental Services Reserve $4,639,000 
 Educational Program−School $16,935,000 
Student support services Student Services $546,000 
 Health Services $1,028,000 
 Transitory Services $510,000 
 Student Affairs $85,000 
Noninstructional support services Custodial $31,509,000 
 Food Services $26,340,000 
 Security Services $12,916,000 
 Public Utilities $42,628,000 
Other states function Assessment and Accountability $1,545,000 
 Grants Administration $281,000 
Office of the deputy mayor for education Total Budget $2,442,325 
 Total $450,150,325 
 Total Students 68,681 
 IEP Students 8,450 
 ELL Students 4,370 
 Charter School Students 25,729 
 Total Less IEP, ELL, and Charter 30,132 
 DCPS per Pupil Cost $14,939 

Note: From FY 2009 Proposed Budget and Financial Plan, 9 June 2008, Government of the District of Columbia, available at www.cfo.dc.gov/cfo/cwp/view,a,1321,q,589949,cfoNav,%7C33210%7C.asp.

Appendix B: Data Sources

Details regarding the experimental protocol and data sources are available in Wolf et al. 2010,  Appendix A.

A restricted use file of the experimental data is being created by the National Center for Education Statistics.