9
Amer. J. Orthopsychiat. 63(4), October 1993 DELIVERY OF SERVICES BENEFIT-COST ANALYSIS OF PRESCHOOL EDUCATION: Findings From a 25-Year Follow-Up W. Steven Barnett, Ph.D. A 25-year study of the Perry Preschool program provides the basis for a com- prehensive benefit-cost analysis of the long-rerm efsects of a preschool education program on children growing up in poverty. Findings indicate that the preschool education program produced economic benefits to participants and to the general public that greatly exceeded the costs of the program. critical issue for public policy regard- A ing Head Start and other preschool education programs for children in poverty is the extent to which they produce long- term effects that have significant economic impact. Preschool programs that can facili- tate long-term improvement in children’s lives are properly viewed as investments, and public policy decisions can be guided (at least in part) by information on their economic costs and benefits. This paper pre- sents a benefit-cost analysis of one pre- school education program for poor children based on a 25-year follow-up study. The source of the data for this analysis is the Perry Preschool study (Berrueta-Clem- ent, Schweinhart, Barnett, Epstein, & Wei- kart, 1984). Begun in 1962, it is one of the strongest and best known longitudinal stud- ies of the effects of preschool education on poor children. An earlier benefit-cost analy- sis, based on the study’s results through age 19 (Barnett, -1985), has been criticized for its dependence on projections beyond that age. Data are now available fiom a follow-up at age 28, making it possible to conduct a new analysisthat is less reliant on projections. METHOD Subjects The initial sample in the Perry Preschool study consisted of 128 African-American children and their parents, all of low socio- economic status The children, born be- tween 1958 and 1962, entered the study in five waves in a single school attendance area in Ypsilanti, Michigan. Five children were lost to the study because they moved away shortly after study-entry or died dur- ing the preschool years, so that the sample for follow-up was composed of 123 child- ren from 100 families. Assignment to the preschool program and control groups was randomized, with two minor exceptions. First, younger sib- lings of children already in the study were assigned to the same group as the older sib- ling. Otherwise some children in the con- A revised version of a paper submitted to the Journal in April 1993. Author is at the Graduate School o f Education, Rutgers, State llniversiw of New Jersey, New Brunswick.. 0 1993 American Orthopsychiatric Association, Inc. 500

BENEFIT-COST ANALYSIS OF PRESCHOOL EDUCATION: Findings From a 25-Year Follow-Up

Embed Size (px)

Citation preview

Amer. J . Orthopsychiat. 63(4), October 1993

DELIVERY OF SERVICES

BENEFIT-COST ANALYSIS OF PRESCHOOL EDUCATION:

Findings From a 25-Year Follow-Up

W. Steven Barnett, Ph.D.

A 25-year study of the Perry Preschool program provides the basis for a com- prehensive benefit-cost analysis of the long-rerm efsects of a preschool education program on children growing up in poverty. Findings indicate that the preschool education program produced economic benefits to participants and to the general public that greatly exceeded the costs of the program.

critical issue for public policy regard- A ing Head Start and other preschool education programs for children in poverty is the extent to which they produce long- term effects that have significant economic impact. Preschool programs that can facili- tate long-term improvement in children’s lives are properly viewed as investments, and public policy decisions can be guided (at least in part) by information on their economic costs and benefits. This paper pre- sents a benefit-cost analysis of one pre- school education program for poor children based on a 25-year follow-up study.

The source of the data for this analysis is the Perry Preschool study (Berrueta-Clem- ent, Schweinhart, Barnett, Epstein, & Wei- kart, 1984). Begun in 1962, it is one of the strongest and best known longitudinal stud- ies of the effects of preschool education on poor children. An earlier benefit-cost analy- sis, based on the study’s results through age 19 (Barnett, -1985), has been criticized for its dependence on projections beyond that

age. Data are now available fiom a follow-up at age 28, making it possible to conduct a new analysis that is less reliant on projections.

METHOD Subjects

The initial sample in the Perry Preschool study consisted of 128 African-American children and their parents, all of low socio- economic status The children, born be- tween 1958 and 1962, entered the study in five waves in a single school attendance area in Ypsilanti, Michigan. Five children were lost to the study because they moved away shortly after study-entry or died dur- ing the preschool years, so that the sample for follow-up was composed of 123 child- ren from 100 families.

Assignment to the preschool program and control groups was randomized, with two minor exceptions. First, younger sib- lings of children already in the study were assigned to the same group as the older sib- ling. Otherwise some children in the con-

A revised version of a paper submitted to the Journal in April 1993. Author is at the Graduate School of Education, Rutgers, State llniversiw of New Jersey, New Brunswick..

0 1993 American Orthopsychiatric Association, Inc. 500

W. STEVEN BARNETT

trol group would have been exposed to the influence of older siblings who had attend- ed preschool and parents who had received home visits. Second, several children (only two can be identified with certainty) were shifted from the preschool to the control group because their mothers' employment presented problems for attending preschool or participating in the home visits. Howev- er, mothers' employment at entry was un- related to outcomes in the economic analy- sis and had no significant effects on the re- sults.

Participants' characteristics at study en- try are described by group in TABLE 1. The only significant difference between groups at entry is in maternal employment, which resulted from the previously mentioned shift of some children with working moth- ers from one group to another. This differ- ence was temporary; no difference was found between groups when maternal em- ployment was measured again at age 15. Attrition was low throughout the study, with a median of 4.9% and a mean of 8.7% across all times and measures. Attrition was not significantly different between groups and did not vary significantly with entry characteristics. At the age 28 follow- up, 1 17 of 121 participants still living com- pleted interviews, and official records data were reviewed for all 123.

501

Treatment The Perry Preschool program consisted

of daily 2%-hour classes on weekday mornings and weekly 90-minute teacher- conducted home visits with mother and child in the afternoons. The program was provided by the local public school and op- erated for 30 weeks each year from Octo- ber to May. Classroom staff were certified public school teachers, and the staff-child ratio was roughly one teacher to six chil- dren. The curriculum was not entirely con- stant over the years, but evolved from a fairly traditional nursery school to an ex- plicitly Piagetian approach. Nevertheless, there probably was greater homogeneity across the years in the Perry Preschool than in a cross-section of Head Start or public school prekindergarten programs (Frede & Barnett, 1992; House, Glass, McLean, & Walker, 1978; Stbek, Daniels, Galluzzo & Milburn, 1992). In any case, the effects of differences among well-implemented cur- ricula tend to be small compared to the generic effects of preschool education (Consortium for Longitudinal Studies, 1983; Schweinhart, Weikart, & Larner, 1986).

Although the program was carefully su- pervised and well funded, the degree to which it was insulated from the problems experienced by ordinary programs should

Table 1 BACKGROUND CHARACTERISTICS OF SUBJECTS BY GROUP AT STUDY ENTRY: MEANS (WITH STANDARD DEVIATIONS IN PARENTHESES) AND PERCENTAGES

CHARACTERISTIC Child's Stanford-Binet IQ Child's Age in Months Child's Sex: Male One-Parent Families Receiving Welfare Parental Employment

Both employed Father alone employed Mother alone employed Neither employed

Mother's Highest Grade Father's Highest Gradea Persons Per Room in Home

PRESCHOOLa 79.6 (5.9) 42.7 (6.9) 57% 45% 58%

9% 39% 4% 48% 9.5 (2.4) 8.4 (2.3) 5.2 (1.2)

NO-PRESCHOOLa 78.5 (6.9) 41.9 (5.9) 60% 49% 45%

11% 34% 22% 74% 9.4 (2.0) 8.8 (2.5) 5.2 (1.6)

pb .38 .44 .87 .76 .23 .02

.84

.58

.93

aPreschool N=58, No-preschool N=63, except for Father's Highest Grade, where Preschool and No-preschool N=34. bo-values based on t-test for means, chi-sauare for Dercentaaes.

502 BENEFIT-COST OF PRESCHOOL EDUCATION

not be exaggerated (Haskins, 1989). Aver- age attendance in the preschool was 69% the first year. Home visit appointments were not always kept. The preschool began in the auditorium of a community center and moved to an old school gymnasium in its second year. Teacher turnover required that ten teachers be hired to staff the four positions over five years. In these respects, the program experienced some of the same difficulties confronted by Head Start today (Granger & Man, 1990; Office of Inspector General, 1993).

Thirteen children in the preschool group entered the program at age four and at- tended for one year. The remaining 45 en- tered the program at age three and attended for two years. All study participants en- tered public kindergarten at age five. The small number of children attending for one year, greatly limits the study’s power to de- tect differences in effects between one and two years. No statistically significant dif- ferences in effects were found between one and two years, but differences of even 50% would not have been significant. Data from all waves were combined for analysis.

Measures The Perry Preschool study is an unusu-

ally comprehensive data set. Data on par- ticipants were collected at entrance to the study, annually through age 1 1, and at ages 14, 15, 19, and 28. In the early years, data collection focused on standardized tests of cognitive development, academic achieve- ment, and teacher ratings. Face-to-face in- terviews about activities and attitudes, as well as searches of official records, were conducted at ages 15, 19, and 28. Test ad- ministrators, interviewers, and public rec- ord searches were blind to participants’ group assignments. A complete description of the data set has been provided by Schweinhart, Bames, Weikart, Bamett, and Epstein (1993).

Economic Analysis From an economic perspective, the Perry

Preschool Program was an investment by

the larger society in the human capital (Becker, 1964) of children whose families and communities had relatively limited ca- pacities for investing in the development of young children. Benefit-cost analysis is a tool for evaluating the return on this invest- ment and investigating the distribution of returns between participants and the gen- eral public (who paid for the program). This procedure was carried out following standard principles for benefit-cost analy- sis (Thompson, 1980).

The design of the economic analysis was guided by human capital theory and previ- ous research on the economic returns to ed- ucation. Potential benefits were identified based on expected improvements in educa- tion and on economic research linking in- creases in the quality and the quantity of education with improvements in adult life outcomes (Behrman & Birchall, 1983; Have- man & Wove, 1984; Juster, 1975; Leslie & Brinkman, 1988; Mincer, 1974; Schultz, 1974). Eight categories of potential effects were identified for which the study pro- vided measures and for which it was feasi- ble to estimate monetary values: I) the preschool program’s cost, 2) child care provided by the program, 3) elementary and secondary education, 4) adult educa- tion, 5) higher education, 6 ) employment, 7) crime and delinquency, and 8) public welfare.

The eight categories listed above provide a comprehensive assessment of short- and long-term costs, but fail to include two ma- jor categories of potential benefits identi- fied in the literature-health and family life. The Perry study’s data on health were simply too limited to provide a basis for es- timating benefits, and it was not feasible to develop reasonable estimates of the mone- tary value of family life effects. The impli- cations of these and other limitations are discussed later.

Like every benefit-cost analysis, the foundation of this study was the estimation of program effects. To provide an indica- tion of the magnitude of the effects on

W. STEVEN BARNETT 503

Table 2 PRESCHOOL EFFECTS RELATED TO ECONOMIC BENEFITSa

OUTCOME VARIABLE PRESCHOOL (N) NO-PRESCHOOL Education Effects

California Achievement Test at Age 9 172.8 (54) 145.5

Classified Mentally Retardedb 15% (54) 35% California Achievement Test at Age 14 122.2 (49) 94.5

Graduated From High School 67% (58) 49% Employment Effects

Employed at Age 19 50% (58) 32% Monthly Earnings at Age 28 $1,219 (54) $766

Arrested by Age 19 31% (58) 51% Crime Effects

5 or More Arrests By Age 28 7% (58) 35%

Received Welfare by Age 28 59% (58) 80%

Welfare Effects Received Welfare at Age 19 18% (58) 32%

=All group differences statistically significant at .05 level. bAt least one veal in a classroom for “educablv mentally impaired“ children.

which benefits depend, estimates of effects in the major benefit categories are present- ed in TABLE 2. Cost and benefit figures were estimated from more detailed effects data in most cases (e.g., individual educational placements each year) and were adjusted to represent average costs for a population consisting of 50% males and 50% females (the policy-relevant percentages). Obvi- ously, there are complex patterns of differ- ences between men and women in their adult lives that affect the economic analysis, but substantial economic benefits were found for both sexes. Only the overall results are summarized here; a detailed description of gender differences in the patterns of costs and benefits is available elsewhere (Bar- nett, 1993).

Finally, when a benefit-cost analysis covers multiple years, adjustments must be made among dollar values of different time periods. Part of the adjustment removes the effects of inflation using a price de- flator that translates all dollars into their common purchasing power in a single year-in this case, 1992. The other part of the adjustment discounts dollars from different years to their “present value,” which is their value at the beginning of the preschool program (the most appro- priate point at which to evaluate the in- vestment).

Present value is calculated using a “real

discount rate,” essentially the interest rate over and above inflation that could have been earned had resources not been in- vested in the program. The further ahead in time a cost or benefit occurs, the less its present value. Following U.S. General Ac- counting Office (1992) recommendations, the results presented here are based on a real discount rate of 3%. Each dollar of benefits at age 28 has a present value of only 30 cents using a 3% discount rate.

RESULTS Estimates of the present value of costs

and benefits in each of the eight cate- gories are presented in TABLE 3. The first column presents totals for society as a whole; these are the relevant figures for determining social policy. The second and third columns show the distribution of costs and benefits between partici- pants and the general public. Each cost and benefit is addressed individually be- fore its summation is discussed. These discussions are necessarily brief; more detailed description of estimation proce- dures and results has been provided by Barnett (1 993).

Program cost. Based on a detailed de- scription of the Perry Preschool program and the resources it required, the cost per child was estimated to be $7,601 for one year and $14,415 for two years. The pres-

504 BENEFIT-COST OF PRESCHOOL EDUCATION

ent value of the weighted-average cost for all preschool participants was $12,356 per child. This weighted-average figure is the appropriate one for comparison with the present value of benefits that were esti- mated for the one-year and two-year partic- ipants combined. It also can be considered to approximate the results of a program in which roughly 20% of participants choose to attend one year of a two-year program.

Child care. The Perry Preschool pro- gram provided a small amount of child care, allowing parents to concentrate on other activities while their child was in at- tendance. This care was only incidental to the provision of the educational program, but it had economic value nevertheless. The value of care to parents was estimated at about $1 S O per hour (in 1992 dollars), roughly the average paid for care in a fam- ily day care home or for-profit center in 1990 (Willer et al., 1991). The estimated present value of the total child care benefit was $738.

Elementary and secondary education. By improving school success, the pre-

school program produced reductions in the costs of public schooling. Estimates of cost reductions were derived from individual school cost histories constructed for each of the 112 participants with complete school records. The most important influences on cost were placements in special education and number of years of schooling. The esti- mated present value of preschool’s effects on elementary and secondary education was $6,872 per child. This cost reduction was found despite a lower dropout rate for the those who attended preschool.

Adult education. By age 27, some of those who had not completed high school by age 19 had enrolled in adult education courses in order to complete their high school educations. Costs of adult education were estimated for each participant based on adult education records. The present value of the preschool program’s estimated effect on the costs of adult education was a reduction of $283 per person.

Post-secondary education. At age 27, those study participants embarked on high- er education had virtually completed it.

Table 3 PRESENT VALUE OF COSTS AND BENEFITS PER CHILD

RECIPIENTS OF COSTS AND BENEFITS PRESCHOOL

COST OR BENEFIT WHOLE SOCIETY PARTICIPANTS GENERAL PUBLIC Preschool Costa -$12.356 $0 -$ I 2,356

Child Care 738 738 0 K-12 Education 6,872 0 6,872 Collegeb -868 0 -868 Adult Education 283 0 203 EmploymentC 14,498 10,269 4,229 Crime 49,044 0 49,044 Welfare 219 -2,193 2,412

Benefit Subtotal $70,876 $8,814 $61,972 Projected Benefits

Earnings 15,833 11,215 4,618 Crime 21,337 0 21,337 Welfare 46 -460 506

Total Benefits $1 08,002 $19,569 $88,433 Net Present Value $95,646 $19,569 $76,077

Costs and cost increases appear as negative numbers. bSome small portion of college costs are likely to have been borne by the participants, but these could not be estimated from the available information. CThe benefits reported for employment include all costs paid by the employer to hire a participant. Allocation to participants and the general public assume that: aj the marginal tax rate is 25%. bj the value of fringe benefits received by the employee equals 10% of salary, and c) the value of other fringes paid by the employer (e.g., the employer’s share of social security) equals 10% of salary.

Measured Benefits

W. STEVEN BARNETT 505

Costs were estimated for each person who reported higher education based on the number of courses and credits for which they had enrolled and the type of institution attended (public two-year or four-year). Courses were included without regard to whether they were eligible for college credit. In this case, the preschool program was found to increase the costs of education by increasing participation in higher educa- tion. The present value of the estimated in- crease was $868 per person.

Employment compensation. The preschool program’s effects on earnings and fringe ben- efits were estimated in two stages. First, ef- fects on earnings and fringe benefits through age 27 were estimated from job and pay histories obtained in interviews at ages 19 and 28. Second, effects on earnings and fringe benefits from age 28 to age 65 were forecast based on educational attainment at age 27 and national data relating annual earnings of African-Americans to highest year of education by age and sex (adjusted by the percentage predicted to survive to each age). The present value of the total ef- fect of preschool on fringes and earnings was estimated to be $14,498 per person through age 27 and $15,833 per person be- yond age 27.

Crime and delinquency. Both interviews and official records of the police and courts indicated that the preschool program re- duced juvenile delinquency and crime. Rec- ords were searched at ages 19 and 28, with the last search providing one more year of data (for age 28) than was available for the other benefits. Effects of preschool experi- ence on victim costs and on criminal jus- tice system (CJS) costs were estimated in two stages-through age 28 and beyond age 28.

To estimate the effects through age 28, crime histories were constructed from ar- rest histories and national data on crimes per arrest by type of crime. These crime histories were combined with estimates of victim costs by type of crime. CJS costs were estimated from the CJS records and

national data on CJS costs per arrest by type of crime and the costs of incarceration and probation supervision. The preschool program was estimated to reduce the pre- sent value of crime costs (victim and CJS) by $49,044 per person through age 28.

Effects on crime costs beyond age 28 were projected based on arrests in each group up to age 28 and national data on ar- rests by age, sex, and type of crime. These data were used to forecast the percentage of arrests and crime costs at each age from 28 to 65. The present value of the pro- gram’s projected effect on crime costs be- yond age 28 was $2 1,337.

Wevare. The effects of preschool on the costs of welfare programs were estimated through age 27 based on welfare program participation histories constructed from so- cial services records and interview data. Projections of welfare program participa- tion beyond age 27 were based on an esti- mate of the exit rate for persons still receiv- ing welfare payments at age 27 (Ellwood, 1986).

The preschool program was estimated to reduce the present value of welfare pay- ments and Medicaid expenses by $2,193 per person through age 27 and $460 per person beyond age 27. However, these fig- ures are not cost savings to society as a whole, but reductions in transfer payments from taxpayers to recipients. Society as a whole gains only from the associated re- ductions in administrative costs, about 10% of the payment reductions-$2 19 per per- son through age 27 and $46 per person be- yond age 27.

Summation of Costs and Benefits From the first column in TABLE 3, it can

be seen that, for society as a whole, the pre- sent value of benefits far exceeded the pre- sent value of cost-$108,002 compared to $12,356. Net present value (benefits minus cost) exceeded $95,000. These results are remarkably robust. No projections at all are needed to establish that net present value was positive as benefits substantially ex-

506 BENEFIT-COST OF PRESCHOOL EDUCATION

ceeded cost based on results observed by age 27. Moreover, net present value re- mains large and positive when any one of the benefit estimates is excluded or if all benefit estimates are reduced by half. Fur- ther analyses (available from the author) show that net present value is positive at real discount rates exceeding 7% (arguably an upper limit) so that the result is not de- pendent on the choice of discount rate.

Distribution of Costs and Benefits The second and third columns of TABLE 3

disaggregate costs and benefits to allocate them to preschool program participants and the general public. Net present value is positive for participants and taxpayers. Even if the benefits to participants alone were weighed against the program’s costs, net present value would be positive. This holds even though a substantial part of par- ticipants’ income gains are offset by wel- fare payment reductions. Finally, net pre- sent value of the preschool program to the general public was large and positive by age 27.

Omitted Outcomes As noted earlier, benefits could not be

estimated in two categories. Although im- proved education is expected to increase longevity and decrease health care costs (Haveman & Wolfe, 1984), health benefits were not estimated. No official health rec- ords were obtained, no examinations were conducted, and the interviews revealed no differences in health-related behavior or health problems. Significantly more of the preschool group than the control group re- ported hospitalization in the previous year (30% vs 15%, N=117, Pearson x2p=.04), but interpretation of this finding is compli- cated by the lack of differences in self- reported health. Possibly, the preschool group had greater access to employment- related health insurance. Lack of access to medical care might pose fiture health prob- lems for the control group.

Education is known to have a wide range

of effects on family formation, child bear- ing, and family investments in children that affect the survival, health, development, and education of the next generation (Have- man & Wove, 1984). The Perry Preschool study provided relatively limited data about these outcomes, but some of the data collected indicate differences. For exam- ple, preschool appears to have significantly increased the percentage of women mar- ri5d at age 28 (40% vs 8%, N=47, Pearson x p=.03) and reduced the percentage who had an abortion (4% vs 28%, N=42 as five control women declined to answer this ques- tion, Fisher’s Exact Test One-Tail p=.04). These results indicate important family- related benefits, but it was not possible to develop reasonable estimates of their mon- etary value.

DISCUSSION Although benefit-cost analyses are not

available for most other government pro- grams, it is difficult to believe that large numbers of them would generate benefit- cost ratios in excess of 7:1, as did this preschool program. Clearly the preschool program accomplished its primary objec- tive, improving the quality of life for chil- dren born into poverty. Even the rough and incomplete measure provided by this anal- ysis establishes that the benefits to the chil- dren exceeded the costs. Thus, the pre- school program could be justified as a worthwhile venture based on the benefits to the children alone.

However, the program proved to be much more productive from a taxpayer perspec- tive than might have been expected. The vast majority of the benefits from the pre- school program-wer $88,000 per child- accrued to the general public. Most of the public benefits were due to reductions in crime, one of our most serious and intract- able social problems. Thus, the benefit-cost analysis establishes that preschool educa- tion can serve taxpayer self-interest, as well as the more altruistic interests of im- proving the lives of the poor and increasing

W. STEVEN BARNETT 507

equality of educational and economic op- portunity. Based on these findings, taxpay- ers have extremely strong incentives to pay for preschool education for children in pov- erty.

Of course, implications for public policy must be drawn with caution. No one should expect the results of this benefit-cost analy- sis to be exactly replicated across the na- tion or in any particular location, although other preschool studies have found the types and magnitudes of effects on educa- tion that underlie the economic benefits (Barnett, 1992). The precise results of pre- school education programs should be ex- pected to vary with characteristics of: a) the children and families served, b) the preschool program, and c) the child’s so- cial environment (Barnett, 1992; Bronfen- brenner, 1989). In addition, because of the omission of important benefits from the an- alysis, the benefit estimates may be con- sidered lower-bound figures. A full account- ing of the benefits of the Perry Preschool program would have indicated a greater re- turn.

One of the most vexing public policy is- sues is how much to spend on preschool education. The Perry Preschool program was much more expensive than the typical public school preschool or Head Start pro- gram today, in which costs are lower be- cause programs are limited to one year at age four and have larger class sizes, lower staff qualifications, and lower salaries. Such cost-cutting can reduce the quality of chil- dren’s preschool experiences (Bruner, 1980; Chafel, 1992; Howes, Phillips, & White- book, 1992; Ruopp et al., 1979). Given the magnitude of the benefits this could jeop- ardize, the low expenditure per child in most programs for disadvantaged children may be unduly risky. Thus, it might be more prudent to increase public funding to the level needed to produce the quantity and quality of services at levels provided by the Perry Preschool until research can provide better guidance on the effects of cost re- ductions.

REFERENCES Barnett, W.S. (1985). Benefit-cost analysis of the

Perry Preschool program and its policy implica- tions. Educational Evaluation and Policy Analy- sis, 7, 333-342.

Barnett, W.S. (1992). Benefits of compensatory pre- school education. Journal of Human Resources,

Barnett, W.S. (1993). Benefit-cost analysis. In L.J. Schweinhart, H.V. Barnes, & D.P. Weikart. Signif icant benefits: The HigWScope Perry Preschool Study through Age 27. (Monographs of the HigNScope Educational Research Foundation, 10). Ypsilanti, MI: HigWScope Press.

Becker, G.S. (1964). Human capital: A theoreti- cal and empirical analysis with special refer- ence to education. New York: Columbia Univer- sity Press.

Behrman, J.R. & Birdsall, N. (1983). The quality of schooling: Quantity alone is misleading. American Economic Review, 73,92846.

Berrueta-Clement, J.R., Schweinhart, L.J., Barnett, W.S., Epstein, AS. , & Weikart, D.P. (1984). Changed lives: The effects of the Perry Preschool program on youths through age 19. Ypsilanti, MI: HigWScope Press.

Bronfenbrenner, U. (1989). Ecological systems the- ory. In R. Vasta (Ed.), Six theories of child devel- opment (pp. 185-246). Greenwich, CT: JAI Press.

Bruner, 3. (1980). Under five in Britain. Ypsilanti, MI: High/Scope Press.

Chafel, J.A. (1992). Funding Head Start: What are the issues? American Journal of Orthopsychiatry,

Consortium for Longitudinal Studies (1983). As the twig is bent .... lasting effects of preschool pro- grams. Hillsdale, NJ: Erlbaum.

Ellwood, D. (1986). Targeting the would-be long- term recipient of AFDC: Who should be served? Princeton, NJ: Mathematica.

Frede, E.C., & Barnett, W.S. (1992). Developmen- tally appropriate public school preschool: A study of implementation of the HigWScope curriculum and its effects on disadvantaged children’s skills at first grade. Early Childhood Research Quarterly 7,483499.

Granger, R.C., & Marx, E. (1992). The policy impli- cations ofjob satisfaction ratings for recruiting and re- taining early childhood teachers. Child & Youth Care Forum, 21(4), 229-246.

Haskins, R. (1989). Beyond Metaphor: The efficacy of early childhood education. American Psycholo- gist, 44, 274-287.

Haveman, R.H., & Wolfe, B.L. (1984). Schooling and economic well-being: The role of nonmarket effects. Journal of Human Resources, 19, 377- 407.

House, E.R., Glass, G.V., McLean, L.D., & Walker, D.F. (1978). No simple answer: Critique of the “Follow Through” evaluation. Harvard Education Review 48, 128-160.

Howes, C., Phillips, D.A., & Whitebook, M. (1992). Thresholds of quality: Implications for the social development of children in center-based child care. Child Development, 63, 449460.

27(2), 279-312.

62,9-2 1.

BENEFIT-COST OF PRESCHOOL EDUCATION

Juster, F.T. (Ed.) (1975). Education, income, and hu- man behavior. New York: McGraw-Hill.

Leslie, L. & Brinkman, P. (1988). The economic value ofhigher education. New York: Macmillan.

Mincer, J. (1974). Schooling, experience, and earn- ings. New York: Columbia University Press.

Office of Inspector General (1993). Evaluating Head Start expansion through performance indicators. San Francisco: Office of Inspector General, De- partment of Health and Human Services.

Ruopp, R., Travers, J., Glantz, F., & Coelen, C. (1979). Children at the center: Final report of the National Day Care Study. Cambridge, MA: Abt Associates.

Schultz, T.W. (Ed.) (1974). Economics ofthefamily: Marriage, children, and human capital. Chicago: University o f Chicago Press.

Schweinhart, L.J., Barnes, H.V., & Weikart, D.P., with Barnett, W.S., & Epstein, A S . (1993). Significant beneJts: The HigWScope Perry Pre- school Study through Age 27. (Monographs o f the

HigWScope Educational Research Foundation, 10). Ypsilanti, MI: HigWScope Press.

Schweinhart, L.J., Weikart, D., & Lamer, M. (1986). Consequences of three preschool curriculum mod- els through age 15. Early Childhood Research Quarterly, I , 15-45.

Stipek, D., Daniels, D., Galluzzo, D., & Milbum, S. (1992). Characterizing early childhood education programs for poor and middle-class children. Early Childhood Research Quarterly, 2, 1-19.

Thompson, M. (1980). BeneJt-cost analysisfor pro- gram evaluation. Beverly Hills, CA: Sage.

U.S. General Accounting Office (1992). Early inter- vention: Federal investments like WIC can pro- duce savings (GAO/HRD-92- 18). Gaithersburg, MD: Author.

Willer, B., Hofferth, S.L., Kisker, E.E., Divine- Hawkins, P., Farquhar, E., & Glantz, F.B. (1991). The demand and supply of child care in 1990. Washington, DC: National Association for the Ed- ucation of Young Children.

For repnnts:W. Steven Barnett, Ph.D., Rutgers University Graduate School of Education, 10 Seminary PI., New Brunwick, NJ 08903