The Perry Preschool Project was an intensive preschool program for low-income at-risk children for 3- to 5-year-olds in Ypsilanti, Michigan. Between 1962 and 1967, it was evaluated using a randomized control trial that allocated 123 children into a treatment and control group. While initial results suggested improvements in test scores, these effects faded after 10 years. However, later analysis showed improvements in earnings and other young adult outcomes.
Hendren and Sprung-Keyser (2020) form an MVPF estimate of the Perry Pre-School experiment from the 1960s using the results from the cost-benefit analysis provided in Heckman et al. (2010). To do so, they use the estimates of the impacts on earnings and other outcomes provided in Heckman et al. (2010). The program found large gains to the treatment group, although the small sample sizes lead to MVPFs that contain wide confidence intervals.
MVPF = 43.6
The upfront costs of Perry Pre-School were $17,759 in 2006 USD. Long-run reductions in transfer payments and increases in tax revenue offset roughly 92% of these upfront costs. Heckman et al. (2010) estimate significant earnings increases from ages 19-40, and an increase in earnings of 26% at age 40. Hendren and Sprung-Keyser (2020) combine their estimated earnings effects with a forecast to age 65 this into a lifetime earnings impact of $70,535. Using a state and federal combined tax rate of 12.9%, this implies an increase in tax revenue of $9,607. Heckman et al. (2010) also estimate that the policy led to a reduction of payments on welfare programs of $3,941. In addition, there are also induced costs of college attendance and vocational training whose incidence falls on the government. Heckman et al. (2010)’s estimates actually imply a fall in such costs, saving the government $2,805. This suggests $16,353 is repaid to the government, implying a net cost of $1,406 (95% CI of [-9,235, 12,126]). Roughly 92% of the upfront spending is repaid to the government. The upper end of the confidence interval for the net cost ($12,126) suggests 32% is recouped.
To harmonize across policies, these baseline cost measures exclude other cost components that are included in the analysis of Heckman et al. (2010). In particular, Hendren and Sprung-Keyser (2020) do not include their estimated $4,287 marginal cost increases from K-12 education (costs which arise because of increased special education or GED attainment), nor do they include any costs of crime. Heckman et al. (2010) find reductions in crime, which they estimate reduced the cost of policing by $19,894 and correctional costs by $8,281. On net, including these additional cost reductions implies a net cost of $-22,483 (95% CI of [-48,668, 3,899]). Incorporating the cost savings from crime yields a point estimate that suggests the policy pays for itself.
Hendren and Sprung-Keyser (2020) measure WTP as the change in post-tax-and-transfer earnings, which is $60,928, plus the estimated reduction in private outlays on college tuition and vocational training of $393. This generates a willingness to pay of $61,321. The confidence interval ranges from $21,121 to $101,224.
The WTP of $61,321 implies an MVPF of 61,321 / 1,406 = 43.6. Hendren and Sprung-Keyser (2020) estimate a 95% CI of [1.83, \infty]. This suggests one cannot rule out MVPFs of around 2 or infinity.
Hendren and Sprung-Keyser (2020) also consider alternative specifications. Including the additional government cost reductions from the criminal justice system generates an MVPF of \infty (95% CI of [5.42, \infty]). Further including WTP externalities from reduced crime victimization further increases the MVPF to \infty with a 95% CI of [20.65, \infty].
Heckman, James J., Seong Hyeok Moon, Rodrigo Pinto, Peter A. Savelyev and Adam Yavitz (2010). “The Rate of Return to the HighScope Perry Preschool Program.” Journal of Public Economics, 94(1-2), 114-128. https://doi.org/10.1016/j.jpubeco.2009.11.001
Hendren, Nathaniel and Ben Sprung-Keyser (2020). “A Unified Welfare Analysis of Government Policies.” The Quarterly Journal of Economics, 135(3): 1209–1318. DOI: https://doi.org/10.1093/qje/qjaa006