Presentation Transcript

1. CBA of interventions in education: the benefit side

2. References “Class Size and Earnings?” by S. Peltzman, as described in Public Finance, by Harvey Rosen “The Economics of Investing in Universal Preschool Education in California,” by Lynn A. Karoly, James H. Bigelow “A Cost Benefit Analysis of Preschool Education in Turkey,” by MehmetKaytaz
3. Criteria in CBA Present value criteria: A project is admissible only if its net present value is positive When two projects are mutually exclusive, the preferred project is the one with the higher net present value PV=B0-C0 + (B1-C1)/(1+r) + (B2-C2)/(1+r)2 + …
4. Criteria in CBA Internal rate of return (IRR) The discount rate that makes the present value of a project just equal to zero, i.e. the  that solves the equation B0-C0 + (B1-C1)/(1+ ) + (B2-C2)/(1+ )2 + … = 0 A project is admissible if its IRR is larger than the discount rate Benefit-cost ratio: the ratio of the present value of benefits to the present value of costs A project is admissible if its benefit-cost ratio is greater than 1
5. Criteria in CBA Note: IRR and benefit-cost ratio are not necessarily useful in comparing projects. Large projects with a small IRR may have larger net benefits than small projects. Classification of an item as negative benefit rather than costs changes the benefit cost ratio, though nothing fundamentally has changed. Net present value criterion is the most reliable guide
6. Estimating benefits (in general) Step 1 Obtain a relationship between proposed intervention and outcomes (e.g., the effect of reduced class size on educational attainment) Step 2 Value outcomes using market prices when available and appropriate (reflecting social cost cost) E.g., increase in wages as result of higher educational attainment
7. Estimating benefits (in general) Market prices may not reflect resource costs, in that case obtain “shadow prices” (=social marginal cost) -e.g., monopoly, unemployment, externalities, taxes Other methods to infer benefits: Value of life; value of time based on cost for life and time-saving devices, compensating differentials between low and high risk occupations Hedonic pricing (regression analysis to infer value of e.g., noise reduction from a regression of house prices on home and neighborhood characteristics) Travel time (indicates demand for recreational activities) Step 1 and step 2 can sometimes be combined
8. Estimating benefits (in general) Step 3 Convert benefits into certainty equivalent amounts in case variance in benefits is important relative to individual incomes Calculating present discounted value of benefits
9. Short example: Impact of reducing class-size [Peltzman, 1997] Do children in relatively small classes have higher earnings as adults, everything else equal? Step 1 & 2 Study by Card and Krueger (1996) estimates that a 10 percent reduction in class size is associated with future annual earnings increases of 0.4 to 1.1%: mid-point 0.75% Applied to 1994 median earnings for male workers 25 and older ($30,000), this amounts to $225 higher earnings per year
10. Short example: Impact of reducing class-size [Peltzman, 1997] Step 3 Peltzman assumes individuals go to work right after school and work for 50 years PV of benefits = 225/(1+r)13 + 225/(1+r) 14 + … 225/(1+r) 62 Notes Shortcuts (e.g. use of male earnings; ignoring trend in earnings) Analysis ignores external effects, such as reduced crime, better informed choices in elections, etc.

11. Karoly and Bigelow: The Economics of Investing in Universal Preschool Education in California

12. Step 1--Methodological issues What is the effect of preschool attendance on children’s outcomes compared with what would be observed for the same children had they not attended preschool, holding everything else constant? Experimental studies: a study population is randomly assigned either to participate in a particular preschool program—the treatment group—or to not participate in the preschool program—the control group. Get an unbiased estimate of the treatment effect except if there is non-random attrition from the study population. May be unethical to exclude the control group
13. Step 1--Methodological issues Quasi-experimental studies: control for aspects that affect outcomes, such as parental income and education -however, there are unobservable differences that affect outcomes for the two groups, such as parental motivation. If these differences are correlated with who is selected to receive education, we have bias owing to self-selection. This bias is mitigated if manage to have high enrollment rate in a given area. Need long-term follow-up with outcomes measured in late adolescence or early adulthood.
14. Step 1--Methodological issues Karoly and Bigelow use evaluations based on 4 programs The Early Training Project and Perry Preschool Project demonstration programs, operating in one site experimental evaluations based on relatively small samples. The Chicago Child-Parent Centers (CPC) and Head Start programs larger-scaleprograms, in the former case operating in multiple sites in Chicago and in the latter case operating nationwide. quasi-experimental designs based on longitudinal survey data All four programs include long-term follow-up so that outcomes are measured in late adolescence or early adulthood.
15. Step 1--Methodological issues Caroly and Bigelow take the results from the CPC as their base for the effect of early childhood education for disadvantaged children -the CPC program is large scale and similar in important respects to what we assume for a high-quality universal preschool program in California Note that the results from the CPC will overestimate effects for advantaged children in the short-term (e.g., in terms of school readiness or early test scores), various studies show positive but weaker benefits In the long-term, one quasi-experimental study of longer-term benefits of untargeted preschool found that participating children were no better off in terms of high school or college completion, earnings, or criminal justice system involvement than those not going to any preschool.
16. Step 1--Evaluation of Chicago-child parent centers (CPC) Large-scale publicly funded preschool program operating in Chicago public schools serving low-income children. The program began in 1967 in four sites serving children ages 3 to 5 prior to kindergarten entry, using Title I funding. In 1998, 25 centers served over 4,000 preschoolagedchildren in high poverty neighborhoods at an average annual cost per child of $4,400.
17. Step 1—Evaluation of Chicago-child parent centers (CPC) a structured half-day program (three hours per day) during the school year with diverse learning experiences designed to prepare children for school through the promotion of language arts and math. The lead classroom teachers all have bachelor’s degrees and certification in early childhood education, and they participate in regular professional development opportunities. The staff-child ratio is 2:17 (with a lead teacher and an aide), and the teachers are relatively well paid as public school employees. the CPC preschool program provides health screening, speech therapy services, and meals, as well as outreach activities including home visiting. A high priority is also placed on parent involvement, from volunteering in the classroom to participating in school events and field trips. These comprehensive services mirror those emphasized in the Head Start program.
18. Step 1--Evaluation of Chicago-child parent centers (CPC) Evidence of the effects of the Chicago CPC program comes from the Chicago Longitudinal Study (CLS), a prospective study of a cohort of 1,539 low-income minority children born in 1980 who attended kindergarten in one of 25 sites in Chicago in the 1985–1986 school year. Of this cohort, 989 children attended one or two years of preschool and kindergarten in one of 20 CPC programs, and a subset also continued to participate in affiliated elementary schools through age 9. The remaining 550 children did not attend a CPC preschool (and most did not attend an alternative preschool program, although about one fifth were enrolled in Head Start), but they did attend full-day kindergarten either at a CPC-affiliated school or one of five other randomly selected schools in the Chicago area not served by a CPC program.
19. Step 1--Evaluation of Chicago-child parent centers (CPC) The evaluation, which has followed children until age 20 or 21, shows that the CPC program had a range of statistically significant, meaningful effects: • Advantages in reading achievement scores as late as age 14 • Lower likelihood of retention in grade by age 15 • Reduced use of special education through age 18 • Lower incidence of child abuse and neglect from ages 4 to 17 • Lower likelihood of involvement in the juvenile justice system by age 18 • Greater likelihood of high school completion by age 20.
20. Specific program considered for California
21. Assumptions Low risk children will have benefits only 25% those of disadvantaged children, and in between risk children will have benefits of 50% of disadvantaged children. Children otherwise attending current public programs would realize some benefit (if not low risk) because we assume the new state program would be higher in quality than most current public programs. Finally, children otherwise attending private programs (third column) are assumed not to experience any gain in the quality of their preschool experience—hence no gain from switching to the public program. 25 percent of California children of preschool age are high risk, like the CPC program children; 55 percent are low risk; and 20 percent are in between (based on data on % of children living with a single parent, living in poverty, or having a foreign-born parent) 70% of pre-school age children will enroll in the public program
22. Assumptions
23. Step 1--Results
24. Step 2—Linking outcomes to monetary values Costs of repeating grades state and local government cost associated with each grade repeated is the average state-wide annual K–12 public education cost in California equal to $6,961 per child in school year 2002–2003. This savings is applied as of age 19 of the child on the assumption that any grades repeated will extend the time in school by an added year at the end of K–12 education.
25. Step 2—Linking outcomes to monetary values Cost of special education Data from the 1998–1999 school year indicated that the ratio of special education costs to general education costs in California was 1.38. Applying this ratio to the California general K–12 education costs in 2002–2003 noted above gives an annual cost of special education for the same year equal to $9,637. This distribution of savings over the government sectors is applied at age 12 of the child, the midpoint of the age interval when special education use is measured.
26. Step 2—Linking outcomes to monetary values Increased high school graduation rate and years of schooling (negative net benefit part) The additional years of schooling generate a cost in terms of California K–12 education. Use the same annual cost of K–12 education discussed above and apply it as an added cost to state and local government as of age 19 of the child Increased university attendance (negative net benefit part) Similar method to calculate costs
27. Step 2– Linking outcomes to monetary values Increased university attendance (negative net benefit part) The increased high school graduation rate can be projected to lead to an increased rate of college attendance for preschool children, compared with their nonpreschool counterparts. Indeed, 62 percent of the CPC treatment group had graduated from high school as of age 21, and 47 percent were attending college at the same age (Reynolds et al., 2002). Infer from data on subsequent education of high school graduates that graduation from high school is associated with an additional 1.5 years of postsecondary education, on average. For the 2002–2003 academic year, the average state and local cost of higher education in California was $6,678 per full-time-equivalent student. Student fees were an average of $780 for the same year, or about 10 percent of the combined public and private costs. These added state and local government costs, as well as participant costs, are applied for 1.5 additional years of schooling to the 11-percentage point high school graduation rate differential as of ages 19 and 20.
28. Step 2– Linking outcomes to monetary values Reduced child maltreatment Foster care in California is estimated to cost nearly $19,000 per year in 2003 dollars, while data for several other states on the cost of in-home care indicate those costs equal about $3,400 per year. Applying the proportions of cases that result in these two outcomes to these figures, and adding an estimate for administrative costs per substantiated case (about $1,000), gives a weighted cost per substantiated case of $9,349 for the California child welfare system in 2003. The reduction in child maltreatment also results in less harm to the victim (i.e., preschool program participants), which can be valued in terms of a reduction in tangible losses—e.g., those associated with medical care and mental health treatment. Estimates of these costs indicate that nearly $7,800 of losses in 2003 dollars are associated with a case of child abuse, while a case of child neglect costs about $1,200. Weighting by the proportion of California cases in these two categories (61 and 39 percent, respectively) gives an estimate of $5,231 in tangible victim costs for each case of abuse and neglect. The cost savings to the government from the reduction in child maltreatment is applied as of age 10.
29. Step 2– Linking outcomes to monetary values Fewer petitions to juvenile court This results in savings to the justice system at the state and local level in terms of costs for the police and court system, as well as for incarceration. We estimate the costs based on the distribution of juvenile court petitions in California as of 2003 and the costs associated with each of those outcomes. For example, about 22 percent of juvenile petitions lead to incarceration in a county facility, which cost an estimated $25,200 annually in California as of 2003. A sentence in the California Youth Authority is much more rare (about 1 percent of cases) but much more expensive ($49,200 per year in 2003). Probation, other dispositions, and dismissal are the other major outcomes. Weighting the cost for each outcome by the share of petitions with this disposition, plus adding an estimate of costs for arrests and adjudication, results in an estimate of $9,480 in justice system costs for each juvenile petition in 2003. This savings to state and local government is applied as of age 14.
30. Step 2—Linking outcomes to monetary values The reduction in juvenile crime also lowers the costs to victims of crime, which falls into the “rest of society” category. Here we focus on the tangible costs to crime victims, which include property loss, lost productivity, medical care, and mental health costs. Following the approach used by Aos et al. (2004), estimate victim tangible costs based on the distribution in California of juvenile petitions and tangible costs by type of crime, with a scaling factor to account for crimes that are not reported to police or that do not generate a court action. Using this method, estimate the tangible victim costs per juvenile court petition at $13,259 as of 2003. This savings is applied to the rest of society, also as of age 14.
31. Step 2—Linking outcomes to monetary values Reduction in crime by the preschool program participants after the observed 18 years old. This projection covers the most crime-prone years (i.e., up to age 44), often called the adult criminal career. Following the methodology applied by Karoly et al. (1998) and also applied by Reynolds et al. (2002), estimate that the net present value of the savings to the criminal justice system at age 19 due to the reduction in the adult criminal career is $3,536. (This figure, applied to the state and local government, is further discounted to the same age as all other cost and benefit terms.) Applying an estimate that indicates that the tangible victim costs for an adult criminal career are about 1.05 times the justice system costs gives an estimate of the former equal to $3,708 in 2003,again a present discounted value as of age 19 (also further discounted).
32. Step 2 – Linking outcomes to monetary values Earnings of high school graduates versus non-high school graduates use 2003 data on mean annual earnings for different levels of educational attainment, disaggregated by age group and race/ethnicity (white non-Hispanic, black, Hispanic, Asian, and other). These data are used to calculate the expected lifetime earnings differential (to age 65) between the average high school dropout and the average high school graduate (accounting for the distribution of additional schooling among high school graduates). This earnings differential is weighted by the racial/ethnic composition of the California cohort from birth to age 4 in 2000, thereby reflecting the demographics of the future California workforce. Allow real wage growth of 0.5 percent per year in the 2003 cross-sectional earnings profile due to productivity gains, and fringe benefits are assumed to equal 20 percent of cash earnings.
33. Step 2—Linking outcomes to monetary values Next project the differential in total compensation between high school dropouts and high school graduates each year between ages 18 and 65.
34. Step 2-- Linking outcomes to monetary values Value of Child Care The families of preschool participants receive the value of the child care provided under a preschool program. Follow Reynolds et al. (2002) in valuing the care received at the minimum wage. For California, the current minimum wage is $6.75 per hour. If apply this amount to a program providing care for 525 hours annually, we obtain a benefit to participants in 2003 dollars of $3,544 for each year of preschool.
35. Step 3 No certainty equivalents calculated here Calculate PDV as of age 3 of the participating child, using a 3 percent annual real discount rate.
36. Sensitivity Analysis On the downside: Assume no benefits will accrue to low- and medium-risk children. On the upside Assume children moving from other public preschools to the universal program would realize the same benefit gain as those moving from no preschool and low- and medium-risk children moving from private schools would also realize some benefits.
37. Sensitivity Analysis On the upside.. Intangible benefits associated with reduced pain and suffering, fear of crime, and so on. Raise the net present value benefits by about $3,400 per child compared to baseline. This is nearly a 36 percent increase for U.S. society. The benefit-cost ratio increases to nia (from 2.62) and 3.93 for the U.S. (from 3.15). Not included nor estimated: These include near-term labor force benefits for California businesses and longer-term benefits for the state in terms of economic growth and competitiveness and economic and social equality.
38. A Cost Benefit Analysis of PreschoolEducation in Turkey, by MehmetKaytaz Step 1. Link between intervention and outcomes Study of CigdemKagitcibasi, Diane Sunar, and SevdaBekman (2001, 2004) of Turkish Early Enrichment Project (TEEP) of MOCEP—an intervention with two different types of early intervention, child-focused (center-based) and mother-focused (home-based).
39. Key findings
40. “The two studies reported in this paper comprise the Turkish Early Enrichment Project (TEEP) spanning a period of 10 years. Both studies were conducted with mothers and children in low-income, low-education areas of Istanbul. Study 1 involved an examination over 4 years of the effects of two different types of early enrichment (intervention), child-focused (center-based) and mother-focused (home-based). Study 2 was a follow-up of Study 1, 7 years after the end of project intervention, Although both interventions produced superior cognitive skills and school adjustment at the end of the program, follow-up assessments in Study 2 revealed that parent-focused intervention had numerous sustained effects in terms of school attainment, higher primary school grades and vocabulary scores, more favorable attitudes towards school, and better family and social adjustment, while most effects of center-based intervention had dissipated (with the notable exception of negative effects of custodial, as opposed to educational, day care). It is concluded that home-based early enrichment through the mediation of the mother is a highly effective strategy with multiple positive outcomes in contexts of socioeconomic disadvantage.”
41. Step 2 Linking educational outcomes to wages
42. 2003 household data
43. Smoothed data
44. How to use this data? Concentrate on high-school graduates versus non-graduates? Concentrate on additional earnings per year of education? Use all information by educational achievement (wages by completed education level and last education level by use of program)
45. Kaytaz: Illustrative scenarios