A Benefit Cost Analysis
Abecedarian Early Childhood Intervention
Leonard N. Masse
W. Steven Barnett
National Institute for Early Education Research
120 Albany St., Suite 500
New Brunswick, New Jersey 08901
A Benefit Cost Analysis of the Abecedarian Early Childhood Intervention
Leonard N. Masse
W. Steven Barnett
A commonly proposed approach to improving the educational success of children in
poverty is the provision of early childhood education programs. These interventions, most
notably Head Start, typically begin at age three or four and operate on a school calendar. Such
programs seem able to boost cognitive scores and school success, though some evidence
suggests that at least some of the effects fade out as children proceed in school (Barnett 1998).
A less common approach is the provision of full-day, year round, child care and preschool
services starting soon after birth. These programs can be considered more preventative in the
sense that services begin before any marked educational deficit can occur.
The Carolina Abecedarian Study is an experiment in the provision of intensive pre-school
services to children in low-income families from infancy to five years of age. The program
began in 1972, and research on program effects found that experimental group children
experienced durable gains in IQ, and achievement in mathematics and reading (Campbell and
Ramey 1995). Comparison of the findings for the Abecedarian preschool project to other
interventions suggests that effects may be more persistent if a program is preventative, intensive,
and starts very early in life (Ramey and Ramey 1998).
The increment to academic achievement and cognitive development experienced by the
Abecedarian children has been fairly well documented. A question that remains, however, is
whether or not expenditures on programs based on the Abecedarian preschool model represent
sound social investments. Simply put, are the benefits worth the costs when viewed in the light
of the many alternative uses of scarce public and private funds? This paper presents the findings
of a benefit-cost analysis of the Abecedarian preschool program. The primary data sources are
follow-up surveys and official school records through age 21.
The program followed an experimental design and originally involved 112 children, mostly
of African American descent, who were born between 1972 and 1977 and whose family
situations were believed to put the children at risk of retarded intellectual and social
development. A "High-Risk Index" was used to determine risk for retarded cognitive
development. The index was constructed based on factors such as household income, parental
education, school histories of family members, welfare payments, parental intelligence, and
parental occupations (Ramey and Campbell 1984). Selected background characteristics at
program entry were: maternal education of approximately 10 years, maternal IQ of 85, 25
percent of households with both parents, and 55 percent of households on Aid to Families with
Dependent Children - AFDC (Ramey and Campbell 1984; Campbell et al. 1998). Between 6
and 12 weeks of age children were randomly assigned to either a preschool program or a
control group. By 1978, 104 participants remained in the study and the follow-up at age 21
involved all 104 of these participants.
The preschool program was center-based with teacher/child ratios that ranged from 1:3 for
infants/toddlers to 1:6 for older children. The center was operated from 7:30 a.m. to 5:30 p.m.,
five days per week, and fifty weeks out of the year, with free transportation available. The
curriculum is called “Partners in Learning” and is discussed in Ramey and Ramey (1998). The
curriculum emphasized language development, but addressed the needs of children in all
developmental domains. Children at the center also received medical and nutritional services.
In order to avoid the confounding effects of these factors on intellectual development, the same
medical and nutritional services were provided to the children in the preschool control group.
The educational results of the program are summarized in Table 8.1. Early assessments
indicated substantial gains in intellectual development. Children in the preschool group
consistently outscored children in the control group on standard measures of intelligence (Ramey
and Campbell 1984). At age 8 participants were assessed and it was found that children in the
preschool group had IQ scores that were significantly higher than the scores of the control
group. Further, at 8 years of age children who had received the preschool intervention also
scored significantly higher on a set of achievement tests in mathematics and reading (Campbell
and Ramey 1995).
An additional assessment was conducted at age 12 and the results were similar to those
discussed above, indicating durable gains in intelligence and achievement (Campbell and Ramey
1994). An assessment at age 15 indicated that the effect on IQ tended to "fade" but that the
effects on reading and mathematics scores remained positive and significant (Campbell and
Ramey 1995). The most recent assessment at age 21 indicated similar effects with respect to
measures of intelligence and achievement. Importantly, the age 21 data demonstrated that the
experimental group children were much more likely to have attended a four year college than the
control group children (P=36%, C=13%, p=.01)1. In general, the results from all the
assessments supported the claim that the preschool intervention was effective in improving
measures of intelligence and achievement over the long term.
Comparisons of the two groups revealed benefits of the program beyond those discussed
above. Campbell and Ramey (1995) reported that preschool participants experienced lower
levels of grade retention and placements in special education classes. Clearly, these cost-savings
to school districts and families represent real economic benefits of the Abecedarian program.
Following the example set by the Perry Preschool Program, researchers examined the
relationship between program participation and the incidence of youth crime to an average age
of 21 and found no statistically significant differences between the groups (Clarke and Campbell
1998). The differences in the nature of community life experienced by the Perry families and the
Abecedarian families could account for the differing results. Although further examination of the
relationship between preschool participation and crime is possible, it does not appear likely that
1 All tests of significance are two-tailed.
crime reduction and cost-savings to victims will represent significant benefits in the Abecedarian
Researchers also investigated the impact of preschool availability on the lives of the
subsample of teenage mothers (under 18 years of age) who participated in the study (Campbell
et al. 1986). When children were approximately 54 months of age, it was found that teenage
mothers of preschool children were more likely to have graduated high school, to have received
post-secondary training, to be self-supporting, and less likely to have borne subsequent
children. It was also reported that mothers with children in the preschool group were generally
more likely to be employed and to obtain jobs with a classification of "skilled or semi-skilled"
(Ramey et al. 1983). To the extent that additional training, job experience, and education was
realized in increased earnings and/or decreased future reliance on social assistance, the above
effects on mothers represent a direct and quantifiable benefit of the program.
Economic Measures and Analysis
This study presents a benefit-cost analysis of the Carolina Abecedarian Preschool
Program. As informed by economic theory, our perspective is that education is both a
consumption good that confers immediate benefits and an investment good that confers personal
and social benefits well into the future (Becker 1964; Haveman and Wolfe 1984). Benefit-cost
analysis involves estimating the monetary values of streams of cost and benefits in order to
measure the program's net value as a social investment.
The benefit-cost analysis of the Abecedarian Project will follow the standard procedures
set forth by Thompson (1980) and Levin and McEwan (2001), and followed by Barnett in the
analysis of the Perry Preschool Program (Barnett 1996). The two core parts of a benefit-cost
analysis are a detailed estimation of program costs and the identification and estimation of
program benefits or effects. In this case, records provided by the program sponsor (Frank
Porter Graham Child Development Center - FPG) are the primary data sources used for
estimation of program costs and effects.
In this benefit-cost analysis, program costs are estimated for three different resource
"settings" in which the program might be offered. Program benefits are generated for 6
categories for which it was possible to obtain monetary estimates: 1) earnings and fringe benefits
of participants, 2) earnings and fringe benefits of future generations, 3) maternal employment
and earnings, 3) elementary and secondary education cost-savings, 4) improved health, 5)
higher education costs, and 6) welfare use. The effects of the program on crime and delinquency
appear to be negligible given earlier research in this area (Clarke and Campbell 1998).
As the analysis involves streams of cost and benefits over time, estimated benefits and
costs are converted into constant dollars (deflated) and discounted to the present using
appropriate rates of discount. The rate of discount reflects the opportunity cost of public
resources. A range of discount rates from zero to seven percent is employed in this analysis.
The analysis estimates the present value of benefits minus costs for each alternative rate of
discount. Additionally, estimates of the internal rate of return, the rate at which the project
benefits are equal to its costs, can be generated.
Table 8.2 presents estimates of the present value of program costs and benefits at
various rates of discount. Some of the benefits and costs accrue to the program participants
and some to the general public. The distribution of benefits and costs is important to the
political viability of an instrument of public policy. A relevant question is whether or not society
realizes returns in excess of public funds and resources that are dedicated to the program. As
we will point out below, the Abecedarian program does “pay for itself” at healthy rates of
discount when all benefits and costs are included in the analysis. However, Masse (2002)
estimates that taxpayers benefits alone (excluding benefits to participants) fall short of program
costs at a discount rate of 3 percent.2 This may be considered a relatively low price for an
effective targeted intervention that is consistent with social or governmental goals concerning
access to education, learning, and economic opportunity.
Resources employed for a representative sample of program years were identified by the
Frank Porter Graham Development Center (FPG). The resources, or program ingredients,
were broadly classified according to function (Levin and McEwan 2001). Categories included
labor resources (paid staff and volunteer workers), and non-labor resources (equipment,
supplies, facilities, etc.). The cost of reproducing the Abecedarian program according to its
resource requirements is clearly relevant for policy purposes. Resources are therefore valued
at the prices typically paid by two institutions that might provide such programs on a large scale:
2 Unless otherwise stated, all values are in 2002 dollars.
public schools and child care centers. This is in addition to estimating cost based on the actual
prices paid by FPG during the program’s operation.
Table 8.3 presents the yearly costs of providing the Abecedarian treatment by program
year in the three different cost settings. Average enrollment in the nursery was about 12 infants
and the staff/child ratio was 1:3. Average age at entry was 4.4 months. In program years 2 and
3 the average unit of instruction/care was 7 children for both age groups and the staff/child ratio
was 1:3.5. In program years 4 and 5 the average unit of instruction/care was 12 children for
both age groups and the staff/child ratio was 1:6.
The undiscounted total resource costs for the FPG and public school settings are clearly
greater than the costs for the child care setting. A few comments are in order. First, it is not
surprising that the cost of executing the program in the FPG and public school settings are
similar. FPG paid workers what they considered to be competitive public school salaries. The
difference in the two estimates is due, in part, to the lower cost of living and level of salaries in
North Carolina relative to the national average. Second, the relatively low cost of executing the
program in a child care setting is presented mostly as a benchmark. It is unlikely that the input
quality necessary to execute a high-quality program could be maintained at the prices and wages
paid in this setting. The extent to which cost savings, represented by a movement along the
resource continuum from the public preschool setting to the child care setting, can be discovered
while preserving benefits is important even if the program is found to lead to substantial net
benefits in the highest cost setting. Although movement away from a successful setting involves
risk of lost benefits, this would have to weighed against the probable cost-savings. A full
treatment of this issue is beyond the scope of the current work and is suggestive of an area for
Cost of Care - Control Children
The cost of the program, properly considered, is the additional cost of the Abecedarian
treatment over the cost of child care arrangements experienced by the control group. Both sets
of experiences involve a stream of costs and a stream of benefits. The measurement of benefits
is necessarily marginal (i.e. the difference between groups consists of benefits beyond the
benefits received by the control group) and the appropriate comparison is with the marginal cost
of the Abecedarian treatment.
Data on the child care experiences of the control children are somewhat limited. Data
were collected on the use of center-based child-care by age. The percentage rates of
participation are 18, 29, 67, 78 and 73 for the first five years of life. Compared to national and
regional data, these rates seem high, especially for years 3 to 5. Possibly, families that
volunteered for the study were exceptionally predisposed to use center- based care. There is
indication, however, that the community in which the experiment took place was one that was
unusually supportive of the care and education of young children (Burchinal, Lee and Ramey
1989). To the extent that higher quality center-based care was available to the control group,
this analysis may underestimate net marginal benefits if the Abecedarian program were provided
Meaningful cost estimates for the child care received by the control children require
estimates of participation rates and hours of care by type of care. Since the analysis seeks to
inform current public policy, the estimation procedure considers the nature of care as it currently
exists. Using data from the National Household Education Survey of 1995, estimates of the
number of hours a child was in center-based care, relative care, and non-relative care for each
of the five program years or age groups are obtained. One of the advantages of using the NHES
data is that it permits the estimation of the use of relative and non-relative care arrangements for
the control group children. These data are not available from FPG but are clearly important to
the calculations of the cost of care for non-treatment children.
The cost estimate for the care of the control children is based on the actual participation
rates of the control group children in center-based child care. In addition, the NHES data is
used to obtain estimates of participation rates and hours of care in non-center-based care
arrangements. The weekly cost of care for each program year is calculated by multiplying the
average number of hours of care by a weighted average (based on participation rates) of the
cost of care. Yearly costs are generated for the non-parental care arrangements of the control
group children. These estimates are used to calculate the marginal cost of the program.
The benefit cost analysis seeks to weigh the marginal benefits that accrue due to program
participation against the marginal costs that are incurred. The marginal cost of the program is
the difference between the cost of the intervention and the cost of the care experienced by the
control group children. The program provided an average of 40 hours of care per week. The
control group children also experienced care for the same 40 hours but a portion of the care
was parental. Since it is the difference in the quality and composition of the care during these 40
hours that leads to program benefits, then it is consistent to obtain a cost for the full 40 hours of
care experienced by the control group children. In order to accomplish this it is necessary to
obtain estimates for the parental component of care and to combine these with the estimates for
In order to estimate the cost of parental care a price needs to be assigned to an hour of
parental-provided care. Information is available on the prices paid to individuals for the care of
young children. The prices of non-relative and relative care are estimated at $2.12 and $1.34,
respectively (Hofferth et al. 1991). The price of relative care may be conservatively low, and
not reflect market prices, for a number of reasons. Individuals may provide care at a subsidized
rate for children of relatives either because of reciprocity agreements between family members
(exchange of services) or merely due to a sense of family responsibility. A relative may also
receive a lower than market wage to reflect the fact that he/she may receive a benefit from
participating in the care of a child to which there is some attachment. For these reasons, an
hour of parental provided care is valued at the price for non-relative care.
The cost of care for the control group children is subtracted from the cost of care for the
program group children to estimate a yearly net cost for the program at each age or program
year. The average marginal yearly costs for the program are $7565 at FPG, $8849 in a public
preschool setting and $2818 in a child care setting. Table 8.4 presents the present value of the
marginal costs under various rates of discount. As detailed above, the cost of implementing the
Abecedarian program in a public preschool is far more expensive than implementation in a child
care setting. Both options are presented as suitable endpoints for the analysis. Benefit-cost
analysis is an important component of a full program evaluation but it cannot provide answers to
all relevant policy questions. Measures that are cost-saving and quality-preserving are clearly
relevant as policy makers consider movement away from the public preschool model and to the
child care model. The benefit cost analysis, at the minimum, should provide information on the
magnitude of the required movement.
The average annual total cost of the Abecedarian Program is approximately $13,900. By
comparison, the annual amounts for Head Start and the Perry Preschool Program are
approximately $7000 and $9200, respectively ( Barnett 1996; USDHHS 2000). The
Abecedarian treatment is clearly more intensive than the other two and this is reflected in its
higher costs. This analysis is partially aimed at determining whether or not the higher costs of
the Abecedarian Program are associated with sufficient benefits to justify the intervention on
purely economic grounds.
At the federal level, the United States in 2001 spent approximately 16 billion dollars on the
early care and education of young children (Barnett and Masse in press). State and local
governments spent an additional 9 billion dollars and direct expenditures by families (not
accounting for parental-provided care) is estimated at 30 billion dollars (Barnett and Masse in
press). What would be the effect on funding levels of providing the Abecedarian program to all
poor children? How much additional funds would have to be allocated by government to early
care and education?
According to the United States Census Bureau (2002), there are approximately 19 million
children less than 5 years of age in the United States. Assuming that 20% of these children are
poor, then the target population for the Abecedarian program totals 3.8 million children. The
total annual cost of providing the Abecedarian program to poor children in the United States is
therefore approximately 53 billion dollars. This is greater than two times the level of current
federal and state expenditures for early childhood education and about equal to the level of total
current expenditures (including federal, state and household expenditures).
The costs of the program may seem prohibitively high for replication on a large scale.
Governments and policy makers may experience “sticker shock” at first but must bear in mind
that costs alone offer little guidance. The costs of a program must be compared against the
benefits that the program generates. Benefit-cost ratios that are greater than one for acceptable
rates of discount indicate that a program is worthy of consideration regardless of the absolute
level of program costs.
Earnings are forecasted on the basis of educational attainment based on the standard
method first presented in Miller and Hornseth (1967). Using cross-sectional data from the
Bureau of the Census, earnings estimates are obtained by age, race, and gender for various
categories of educational attainment (United States Bureau of the Census 1998). Each category
corresponds to an estimated stream of lifetime income. An individual’s estimated lifetime
income depends on educational attainment at age 21 and the probability of higher educational
attainment later in life.
FPG provided data on the educational levels of the Abecedarian control and program
group participants at age 21. The schooling levels of the participant fell into 8 categories (less
than 9 years, 9 to 11 years, GED enrollee, GED graduate, high school graduate with no college
credits, some college but no degree, enrolled in an AA program, and enrolled in a BA
program). In order to estimate future earnings it was assumed that educational status at age 21
did not necessarily represent an individual’s final educational status. It was therefore necessary
to calculate the expected value of an individual’s future stream of income. In order to
accomplish this, it was necessary to assign probabilities to each level of future educational
attainment (9 census categories) for each level of current educational status (8 study categories).
The conditional probabilities were based on the results of United States Department of
Education longitudinal studies that follow the educational advances of specific age cohorts
(Adelman 1999; Boesal et al. 1998; McCormick et al. 1999) and cross sectional data on high
school dropout and graduation rates (USDOE 1998). For each level of current education, the
expected value of future income is the sum of the products of the probability of obtaining each
level of higher education and the present value of the income stream associated with each
The procedure for estimating lifetime earnings therefore involved several steps. First,
earnings for ages 22-65 were estimated using cross-sectional data for the nine levels of future
educational attainment. Second, these earnings were multiplied by the probability that a
participant would survive to each age. Survival rates were estimated from data from the
National Center for Health Statistics (1998). Discounted lifetime earnings were then calculated
for each level of future educational attainment. The estimated probabilities for future educational
attainment were then employed to calculate the expected value of discounted lifetime earnings
given the level of educational attainment at age 21.
The simple use of cross-sectional data assumes that there is no productivity-induced
growth in real income over the lifetimes of participants. Government data show that the output
per hour of all persons employed grew at an average annual rate of 2.3 percent over the period
from 1948-1997. More recently, the average annual rate in gross domestic product per worker
hour was 1.2 percent over the period from 1979-1990 and was 1.3 percent over the period
from 1990-1997 (Bureau of Labor Statistics 2000). In this analysis, therefore, earnings are
adjusted assuming a 1.0 percent growth in real income.
The estimates for the program effects on lifetime compensation beyond age 21 are
presented in Table 8.5. Compensation includes base salary and fringe and employer-provided
benefits that are valued at 20% of base salary. In this benefit-cost analysis results are not
presented by gender. It is noted however that gender differences in program effects on
academic achievement and attainment seem to have translated into effects with respect to
earnings. The mechanism through which females participants are more likely than male
participants to realize a marginal effect on higher educational attainment and earnings is an area
of research that warrants further attention.
The program effect on lifetime compensation beyond age 21 is approximately $37,500
at a discount rate of 3 percent. Overall, lifetime compensation beyond age 21 is somewhat
conservatively estimated. The use of cross sectional data assumes that age-earnings profiles are
relatively stable over time. In particular, it assumes that the labor force participation rates of
men and women that gave rise to the current cross-sectional earnings data will prevail over the
working lives of our sample. However, the labor force participation rates of women have
shown a significant upward trend over the past 50 years for women of all ages (Fullerton 1999).
Using 1998 cross-sectional data, the labor force participation rates for women aged 35-44 is
estimated at 77 percent. However, Fullerton (1999) estimates that in 2015 this rate will had
increased to 82 percent. For women aged 45-54, the 1998 and 2025 estimates are 76 percent
and 82 percent, respectively. Similarly, Fullerton’s estimate of 59 percent for the LFP in 2025
of women aged 55-64 can be used as an estimate for the LFP of Abecedarian women when
they reach this age interval (approximately 2035). The estimate employed for this age group
using cross sectional data is 51 percent.
In 1998, 2015, 2025, and 2035, the Abecedarian participants (were) will be
approximately 23, 40, 50 and 60 years of age. The use of 1998 cross-sectional data, therefore,
seems to underestimate the labor force participation of program participants by approximately
5-6 percent at ages 40 and 50 and 8 percent at age 60. Therefore, projecting female earnings
based on cross-sectional data is conservative and leads to estimates that are below the actual
earnings that will be realized by program participants.
Earnings of Future Generations
In this section, the magnitude of benefits that accrue to the descendants of the
Abecedarian participants is explored. There are a number of clear mechanisms through which
benefits from the preschool program may be transmitted across generations. In theory, most
benefits that accrue to program participants are sources of benefits for the children of
participants. These include effects on academic achievement, educational attainment, earnings,
criminal behavior, welfare use, educational cost-savings, job-satisfaction and status, self-esteem,
pro-social behavior, household management, fertility and birth weight. As is the case with the
effects for the program participants, some of these effects are difficult to monetize and will
remain unmeasured. The overall ratio of program benefits to costs is conservatively estimated
for this reason.
There is a significant amount of evidence that supports the positive relationship between
parental education and income and the educational attainment and income of children (Birdsall
and Cochrane 1982; Wolfe and Behrman 1985; Singh 1992; Glewwe and Jacoby 1994; Leigh
1998). Measures of household income and family background are standard variables used in
estimating wage and earnings functions (Cohn and Geske 1990). Using cross sectional data,
Peters (1992) presents the conditional probabilities of a child’s income attainment given the
income attainment of the father or male head-of-household. In general, the probability that a
child’s income attainment is greater than or equal to that of the parent is greater than .50. Peters
(1992) also estimates an earnings function and finds that the elasticity of child income with
respect to the income of the father is approximately .26. Estimates from other studies range
from .07 to .44 (Atkinson 1981; Behrman and Taubman 1985; Solon 1992).
In order to estimate the program’s effect on the earnings of future generations, elasticities
estimates presented in Altonji and Dunn (1990) are employed. Using data from the National
Longitudinal Surveys of Labor Market Experience, Altonji and Dunn derive estimates of the
elasticity of child income with respect to the income of parent. In particular, they find that the
elasticity of the income of a son (daughter) with respect to the income of the father is equal to
.210 (.335). The elasticity of the income of a son (daughter) with respect to the income of the
mother is equal to .148 (.348).
In order to use these elasticities to estimate the earnings of future generations, it is first
assumed that they can be applied to undiscounted lifetime earnings. It is also assumed that the
program effect on generation one (G1) parental income can be considered an increment to
income relative to the base level achieved by the control group. Using the program effects by
gender, the percentage change in G1 income associated with each effect is calculated.
Employing the elasticities given above, the corresponding change in generation two income (G2)
associated with the calculated change in parental income is calculated. Once the program effect
in G2 income is calculated, the process can be repeated and effects calculated for future
generations in an iterative manner. For the purpose of this analysis, estimates for the combined
program effects on generations two through four are provided.
In Table 8.5 the discounted values for combined incomes of future generations are
presented. It is assumed, conservatively, that each participant (parent) has one child at age 25
and that the children will have earned income from age 22 to age 65. The overall effects are the
weighted average of the individual effects for males and females. We can see from Table 8.5
that the program effects on the earnings of future generations are not economically insignificant.
At an interest rate of 3%, these effects equal $5722 per participant, an amount equal to
approximately 16 percent of the per child marginal cost in the FPG setting.
Elementary and Secondary Education
The effects of the program on the elementary and secondary education costs of participants
were estimated. School histories were constructed for 99 of the study participants based on
data that was originally gathered from official school record data by FPG. For each participant,
a school placement was assigned to every year that a child was in school. The major distinction
was between special education placements and regular educational placements, with the former
being more resource intensive and, hence, more costly.
Costs were mapped onto the schooling histories in the following manner. A school year
that involved at least one special education category was assigned the yearly estimate for special
education. All other school years were assigned the cost estimate for regular education. The
estimates for the costs of regular education and special education are adjusted from data
presented in Parrish, O’Rielly, Duenas and Wolman (1997), which are based on data from the
national cost study conducted by Moore and colleagues (1988). According to Parrish and
colleagues, the average annual real rate of growth in per pupil special education costs over the
period from 1968 to 1986 was 4.1 percent. The corresponding rate for regular education was
2.1 percent (Parrish, et al 1997). Assuming that education costs grew at these rates over the
period from 1986 to 1999, revised national estimates for the costs of regular education and
special education are $7931 and $18341, respectively.
In Table 8.5 the program effects on educational costs are presented. At a discount rate
of 3% the cost reduction was equal to $8836, an amount equal to approximately 25 percent of
the per child marginal cost in the FPG setting. It was expected, however, that the savings from
reduced rates of grade retention and special education would be somewhat larger. Campbell
and Ramey (1995) reported that the rates of placement in special education by the end of grade
9 were 25% and 48% for the program group and control group, respectively (n=92). These
rates represent the percentage of children that had ever received special education services by
the end of grade 9. The current reexamination of the schooling data results in comparable
estimates of 31% and 49% for the program and control groups (n=99, p=.0672). The
difference between the two estimates is likely due to the change in the sample size over the
years as more complete schooling data became available.
In addition to the above measure, the percentage of total school years in special
education was calculated for the program and control groups. The program effect using this
measure was not nearly as pronounced. The estimates for the program and control groups
were found to be 12% and 18%, respectively (p=.0082). Since years in special education are
more directly related to cost than the former measure, the expected program effect is somewhat
Smoking and Health
Schooling is related to an individual’s ability to obtain and process information related to
matters of health (Grossman 1972; Grossman and Kaestner 1997). Higher-schooled
individuals can make more informed and better decisions regarding their personal health (ex.
they may have a healthier diet, visit the doctor more regularly, and be able to provide a higher
standard of personal health care than someone who is less informed). There may be a number
of mechanisms through which schooling increases the opportunity for an individual to lead a
healthier life. Education increases the ability to be an effective consumer of health care services
and producer of personal health. Education also increases earning power, the ability to
command wages, fringe benefits, vacation time, and the ability to avoid working conditions that
may be detrimental to personal health. Education also increases income that allows one to
purchase higher quality and quantity of health services and to establish living conditions that are
conducive to good health.
Another proposed mediating factor in the relationship between schooling and general
health is the degree to which an individual has concern and regard for the future. Someone who
is willing to invest in human capital demonstrates that he/she is willing to trade off a certain
degree of current consumption for returns that will mainly accrue in the future (Fuchs 1982,
1996). Such an individual may also, therefore, be more willing to engage in behaviors that
reflect a concern for future health. In this view, it is willingness to consider future events in
present decisions that is responsible for the investments in both education and personal health
(Becker and Mulligan 1994). Further, some have argued that schooling, concern for the future
(time preference), and cognitive ability, all independently affect the probability that an individual
will engage in healthy behavior (Sander 1998).
Sorting out the independent contributions of schooling, achievement, income, and time
preference on health-related behaviors is important but beyond the scope of this study. The
above information indicates that there are possible health-related benefits to an early childhood
intervention that had positive effects on the schooling, achievement, and income levels of
participants. Although the independent effects of all three are of some interest, it is much more
important in this case to realize that the program effects on these variables are likely responsible
for whatever differences are observed in the health-related behaviors of the program and
control groups. Isolating, measuring, and valuing possible health effects of the program
contributes to a more complete analysis of program benefits and costs.
The measurement of possible health effects in the current study is limited to effects related
to smoking. Data collected by FPG indicate that there are differences in the rates of smoking
between the program and control group children, although the rates seem high relative to
national average data. However, any program or policy that can reduce smoking rates has
the clear potential to generate significant economic benefits. The benefits include, but are not
limited to, improvements in health and longevity, and reductions in the cost of health care.
Sander (1998) found that cognitive ability, educational attainment, and time preference all
affected the probability that an individual smoked. Grossman and Kaestner (1997) found a
negative relationship between achievement scores in high school and the likelihood of smoking.
National data indicate that there are a number of strong associations between smoking and
certain demographic characteristics (USDHHS 1997). First, individuals with less than a high
school degree currently smoke at a rate of 47%, which is nearly four times the rate of 12% for
college graduates. Second, smoking is negatively associated with the level of income. At
household income levels that are less than 150 percent of the poverty level, the rate of adult
smoking is 38 percent. At household income levels that are 300 percent of the poverty level,
the corresponding rate is 25 percent. The national data also suggest that individuals that live in
households that do not include both biological parents are more likely to be smoking as adults.
It can be argued that these data indicate that situations that involve certain forms of stress raise
the possibility that an individual will smoke, in part, as a reaction to his/her situation in life
(USDHHS 1998). In this view, policies aimed at reducing stress from these sources for any
reason should consider as a benefit the possible effect of the policy or program on the rate of
smoking for the target group.
Data on smoking by Abecedarian participants come from a 1993 youth risk behavior
survey conducted by FPG. The rates of smoking for the control group and program group were
estimated at 55% and 39%, respectively (p=.106). The results are clearly suggestive, if not
strictly significant from a statistical point of view. In order to estimate the economic value of the
program’s effect on the rate of smoking, it is necessary to translate these rates into monetary
returns. For the purposes of this analysis, the effects on morbidity (illness) prior to death are
ignored and the focus is only on the value of differences in expected mortality between the two
groups. Ignoring the effect of smoking on illness prior to death simplifies the estimation
procedure at the cost of underestimating potential benefits by potentially significant amounts.
However, the effect on mortality may still contribute significantly to program benefits and this
suggests a future area for research. In any case, data on smoking behavior should be collected
in follow-ups of early intervention programs.
Cutler and associates use national data to estimate the life expectancy of individuals who
either are or had been a regular smoker by age 20 (Cutler et al. 2000). Being a non-smoker at
age 20 increases longevity by approximately 6.5 years. In order to value these additional years
of life, an economic estimate of the value of an additional year of life is needed. It is assumed
that additional years gained occur after the average age for life expectancy by gender. The
value of a life (L) is associated with the years from 70-76 for male non-smokers and for the
years 77-82 for female non-smokers. The estimates for L are then discounted to program
entry and the discounted values are multiplied by the average difference in smoking rates
between the two groups in order to obtain estimates of program effects.
In order to execute the above simple procedure the non-simple task of attaching a value
(L) to a year of life is necessary. There is a vast literature in the area of health economics that
corresponds to the economic value of an increase in mortality and/or a decrease in morbidity
(Oster, Colditz and Kelly 1984; Moore and Viscusi 1988; Manning et al. 1989; Miller,
Calhoun and Arthur 1990; Tolley, Kenkel and Fabien 1994; Adams and Young 1999). For
example, values are associated with decreases in government expenditures on Medicaid, an
individual's willingness to pay for reductions in health risks, income loss due to premature death,
and property loss or damage due to fire. Following the example of Cutler and associates
(2000) and Gruber and Zinman (2000), in their respective works for the National Bureau of
Economic Research, the range for the value of a year of life that emerges from this literature is
between $100,000 and $200,000 (1999 dollars).
Table 8.6 presents the estimates for the program effects on smoking and the economic
value of increased longevity assuming that a year of life has a value of $161,000 ($150,000 in
1999 dollars). The discounted values of increased longevity between males and females were
not significantly different and, therefore, average values were used to calculate program effects.
It is clear from Table 8.6 that the benefits from a reduction in the rate of smoking are not
insignificant. It is also clear from the estimates that the assumption that benefits accrue in the last
years of life results in a large reduction in benefits at higher rates of discount. At a discount rate
of 3 percent, the program effect on smoking is equal to approximately 50 percent of the per
child marginal cost in the FPG setting. At a discount rate of 7 percent, the program effect on
smoking is reduced to approximately 3 percent of the per child marginal cost. Although not
explicitly measured here, there are benefits from reductions in the consumption of cigarettes that
occur in the present. General health is improved and individuals can lead more active and
productive lives. There is also arguably a benefit in the reduction of the number of individuals
captive to a physical addiction that exceeds whatever benefits individuals may experience due to
the act of smoking. However, even ignoring these benefits and the substantial benefits from
reduction in the pain and suffering associated with illness, program effects on the rate of smoking
result in benefits that are not economically insignificant at lower rates of discount.
In some sense the effect on smoking is an unexpected result. The program had its main
goal of improving the cognitive ability of young children and increasing the probability of school
and workplace success later in life. However, a relative increase in cognitive ability, coupled
with significant differences in achievement and school-related experiences, can arguably result in
the program group children making relatively more productive choices about personal health.
The general nature of this finding may be limited by the fact that the smoking rates were much
higher for the Abecedarian participants relative to the national population. Measured benefits
for a different population of children will likely be less than those presented here. However,
given the great concern over youth smoking, and the strong relationship between youth and
adult smoking, the results here are particularly encouraging.
Maternal Productivity and Earnings
Benasich and colleagues (1992) reviewed the literature on early intervention and
maternal outcomes and summarized the results of a variety of programs that are child-centered
and/or parent-centered and reported outcomes for mothers. The review was restricted to
experimental and quasi-experimental studies on interventions that began before age 3, lasted for
at least 6 months, and were targeted at educationally disadvantaged families. Positive outcomes
for mothers are generally reported in four areas: education and employment, fertility, parent-
child interaction, and in the quality of the home environment.
Seven studies of child-centered preschool programs meet the criteria established by
Benasich and colleagues: the Abecedarian Project , the Birmingham Parent-Child Development
Center , the Infant Health and Development Project , the Milwaukee Project, the Perry
Preschool Project, the Teenage Parenting Project, and the Teenage Pregnancy Intervention
Program. The majority of these programs were center-based and provided care and education
on a full-time basis for a number of years.
The Abecedarian project reported that, relative to a control group, experimental group
mothers had higher levels of educational attainment and held higher-paying jobs when their
children were age 5 (Campbell and Ramey 1994). The current study reports that program
group mothers held an earnings advantage over the control group mothers at various times since
program entry. Pungello and colleagues (2000) report that program group mothers were more
likely to have a skilled versus unskilled job when the program child was 21 years of age. The
Birmingham Parent-Child Development Center project reports that program group mothers
were more likely to return to paid employment (Andrews et al. 1982). The Infant Health and
Development Project reports that program group mothers held at an employment advantage
over control group mothers when the target child was age 3 (Brooks-Gunn et al. 1994). The
Milwaukee Project reports that program mothers experienced more stable employment and
higher weekly earnings (Garber 1988). The Teenage Parenting Project and the Teenage
Pregnancy Intervention Program report that participating mothers were more likely to complete
high school (Field et al. 1982; Roosa and Vaughan 1983). The Perry Preschool Project
measured maternal outcomes on employment and education and found no significant effects
(Schweinhart et al. 1993). This is not surprising given that the Perry program operated on a
part-time, part-year basis. More so than the other programs mentioned, the Perry Preschool
Project did not have a major child care component and therefore did not substantially reduce
the necessary quantity of maternal-provided care.
Despite the fact that econometric studies on child care and maternal employment have
produced mixed results (Kimmell 1998), there is some experimental and quasi-experimental
evidence to support the position that quality child care results in benefits for mothers in
disadvantaged households. It is not clear that pure custodial care, with less attention to the
educational experiences of children, would result in similar benefits. The stability and general
quality of child care arrangements may be related to the ability of mothers to focus or
concentrate more consistently on matters related to work or employment (Vandell and Wolfe
2002). Mothers, feeling that their children are safe and well cared for, may be more willing and
able to participate effectively in the labor force and to reallocate time to employment activities
(Meyers 1993; Ross and Paulsell 1998; Vandell and Wolfe 2002). Furthermore, this
reallocation may not come fully at the expense of time spent caring for children. Bianchi (2000)
argues that working mothers trade off time spent working in the home, volunteering, sleeping,
and engaging in general leisure activities in an attempt to preserve time spent caring for children.
Overall, the use of quality and stable care arrangements may not significantly decrease the
amount of maternal-provided care a child receives and also permits mothers and employers to
establish relationships that are continuous, reliable, and productive.
In the current study, it appears that the provision of 5 years of high-quality, full-time
care and education increased the opportunities of mothers to obtain employment, training, and
other productivity-enhancing activities. These opportunities resulted in increased earnings for
the program group mothers relative to the earnings of the mothers in the control group. Self-
reported income data were available on maternal earnings at participant ages of 12, 15 and 21.
Corresponding maternal ages were approximately 32, 35 and 41. Based on these data, a
yearly program effect on maternal earnings of $3750 per child is estimated.
Table 8.7 presents the program effect on maternal earnings for various rates of discount.
The program effects through age 41 are estimated based on the actual earnings. The program
effects from ages 42 to 60 are extrapolations based on the effects through age 41 and assume,
conservatively, that there is no increase in the earnings differential between the two groups. Due
to a lack of earlier data on maternal earnings, the estimates are also conservative in that they
assume that the earnings differential does not occur until maternal age 26. Despite these
assumptions, the program effect on maternal earnings is quite substantial and is greater than the
per child marginal cost of the intervention in the FPG setting at a discount rate of 7 percent.
This analysis indicates that an important benefit of the program is the effect of fully
subsidized preschool on the labor market success of mothers. This issue, by itself, is important
to the child care debate because program effects on the household go beyond those that involve
the children receiving care. Maternal employment is clearly related to labor market experience,
training, and earnings, all of which promote self-sufficiency and an improved quality of life for all
members of the household. Society benefits as well from improvements in the productive
capacity of female workers and from a decrease in the need for social assistance. The results of
this analysis with respect to maternal employment, therefore, are encouraging and warrant
Cost of Higher Education
The program group participants have higher levels of educational attainment at age 21
than the control group participants. The higher levels of educational attainment reflect, among
other things, higher levels of academic achievement and are assumed to result in higher
productivity and individual earnings. However, the cost of attending institutions of higher
education must be taken into account. In this section the program effects on the costs of higher
education are estimated. Since the program group has a higher rate of attendance in higher
education by age 21, it is expected that the program effect due to cost of higher education will
The estimated program effects for the costs of higher education are included in Table
8.2. The effects are fairly significant in size due to the large differences in the educational
attainment of the program and control groups. The increase in cost due to higher education is
approximately $8128 at a 3% rate of discount. The effects due to the cost of higher education
decrease overall program effects and are therefore negative in value.
The cost of adult education, particularly the cost of a GED program is not employed in
this study for a number of reasons. First, the rates of participation at age 21 for the program
group and the control group are 11% and 15%, respectively, and the difference is not
statistically significant. Second, because of the higher rate of GED attendance for the control
group, the inclusion of a cost estimate for GED education would tend to increase overall
program effects. Lastly, the adult education cost estimate employed by Barnett (1996) in the
benefit-cost analysis of the Perry Preschool Program is $1710 per class in 1992 dollars
(Varden 1982). Using this figure, the undiscounted program effect of the cost of adult
education is calculated to be $93 per person for the Abecedarian Preschool Program. Given the
relatively small nature of this effect, and the reasons detailed above, the adult education program
effects are not included in the final analysis of benefits and costs.
Income-tested Programs at age 21
A reduction in welfare payments to program participants represents a transfer of money
to the general taxpayer and does not change total social benefits associated with the program.
The program effect is therefore the reduction in the cost of administration for a lower number of
AFDC-related cases. At age 21 the use of Aid to Families with Dependent Children (AFDC)
in both the program and control groups was restricted to females. The rates of AFDC use for
the program and control groups were 8% and 16%, respectively (p=.2340). Neither a two
tailed test of significance (p=.2340) or a one-tailed test (p=.1170) indicated a significant
difference between the two groups. Restricting attention to females only, the corresponding
rates were 17% and 32%. Again, neither a two-tailed test (p=.2224) or a one tailed test
(p=.1112) indicated a significant difference between the two groups.
Although the rates are not statistically different at age 21, the relative differences are
quite substantial. The rate of AFDC use for control group is approximately twice the rate for
the program group. The small sample size, especially when the attention is directed at females
only, limits the statistical power of the findings. In this section, estimates are therefore generated
for the cost-savings from the reduction in the rate of AFDC use that is associated with program
participation. As stated above, the magnitude of the effect is expected to be small relative to the
major benefits and costs of the program.
The Committee on Ways and Means (1998) reports the average value of income-tested
programs for a household that participated in the AFDC program in 1995. In particular, the
average value of AFDC was $3935 per participating household. Given that a household
participated in AFDC, the probability that the household participated in the Food Stamp
program was .83 and the average value received was $2306. Similarly, the probability that an
AFDC household received a housing subsidy was .29 and the average value of that subsidy was
$2650. The probability that an AFDC household received Supplemental Security Income
(SSI) was .15 and the average value was $5380. The probability that an AFDC household
received Medicaid benefits was .98 and the amount of these benefits was $2057. The
expected value, therefore, of all income-tested programs is calculated to be $9441 per
household in 1995 dollars. Converting to 2002 dollars the estimate becomes $10,715 per
In order to estimate program effects due to AFDC use, a number of assumptions are
employed. First, effects are estimated assuming that the average value of total assistance per
AFDC household is $10,715. Consistent with current welfare law that limits the term of AFDC
assistance but permits states to extend benefits for 20% of participating households because of
economic hardship, AFDC use is assumed to be for five years with a 20% probability of
continuation thereafter and no future reentry into the program (Committee on Ways and Means
1998). These assumptions likely underestimate the use of AFDC and income-tested programs
over the lives of the participants. The assumptions ignore households who could enter the
program for the first time after age 21. It also does not permit reentry of households after age
26 who are receiving AFDC at age 21. However, given the higher rate of entry into AFDC by
the control group at age 21, these assumptions will provide conservative estimates of program
In Table 8.2 the effects of program participation on the use of AFDC and other
income-tested programs are presented.. The overall program effect at a discount rate of 3% is
estimated to be $196 per participant. Again, the program effect is related to the savings in
administrative costs that are associated with a relatively lower number of AFDC cases. It is
clear from Table 8.2 that reductions in the use of AFDC and other income-tested programs due
to program participation results in a relatively minor benefit when measured in dollar terms. It is
noted, however, that unmeasured benefits include additional psychological or personal benefits
that may accrue to participants and their families resulting from a decreased reliance on social
The last two lines of Table 8.2 present the net present values of benefits and costs at three
different rates of discount for program replication in the FPG and public school settings. In both
cases the net present value is greater than zero for discount rates up to 7 percent. The same is
clearly true for replication in the lower cost child care setting. If we include all measured
benefits, then the internal rate of return for the Abecedarian intervention appears to be slightly
greater than 7 percent. The positive results are not highly sensitive to the presence or exclusion
of any one benefit. Excluding maternal earnings from ages 42-60 yields an internal rate of return
of between 5 and 7 percent in the FPG and the public school resource settings. Excluding
forecasted participant earnings and the earnings of future generations also results in an internal
rate of return between 5 and 7 percent. Excluding the estimates for smoking and health still
results in an internal rate of return greater than 7 percent. If we confine attention to the benefits
that accrue mainly to the children (participant earnings and smoking/health), then the rate of
return to the program is between 3 and 5 percent. Overall, the rate of return to the
Abecedarian project is no less than 3 percent and is likely higher than 7 percent.
The Abecedarian program results in healthy returns for the investment of public resources
targeted at a disadvantaged group. This occurs even when viewed in the light of significant
unmeasured benefits from improved education, such as the personal consumption value of
learning and educational experiences, increases in civic and pro-social behavior, increases in the
overall quality of life, and improvements in personal decision-making and household
management (Haveman and Wolfe 1984). In addition, Donohue (1999) argues that lower rates
of discount or estimated internal rates of return may actually be appropriate if government
programs help future generations avoid some irreversible damage. Market rates of return, and
hence market rates of discount, may not lead to appropriate decisions if markets alone cannot
bring about the desired program effects. If the goal is to increase the income and prospects of a
disadvantaged group, and there exists no other clear mechanism for doing so, it may make
sense to apply a lower rate of discount to projects that accomplish this goal. An alternative is to
recognize that a dollar of program benefit to a target group has more value than a dollar of
program cost to others. Favoring the disadvantaged group may help improve distributional
equity at the expense of efficiency in resource allocation. In this case, the effects of program
participation on the educational attainment, productivity, and earnings of at-risk children result in
an improvement in overall social equity. Change in equity remains, therefore, a potentially large
unmeasured benefit of the Abecedarian program.
It is unlikely that the results of the Abecedarian program can be replicated perfectly in
all settings and for all populations. The benefits that accrued to the participants were due to the
marginal differences in the quality of the care received by the program group children and the
quality of the care received by the control group children in the first five years of life. In the
cases where the care currently being received is of a higher quality, then the marginal effects will
not be as great. However, if attention is limited to the 20% of our nation's children that are
estimated to be living in poverty, then the results of the study are more directly applicable.
Replication in other settings will also affect the magnitude of specific benefits. It is possible
that the effects on the smoking behavior of participants may not be as great outside of North
Carolina and the southern region of the United States. However, this is relatively minor
measured benefit. More importantly, it is also very possible that the effects of the intervention
on criminal behavior may be more pronounced in higher-crime areas. We would, therefore, not
be surprised if the Abecedarian intervention resulted in greater program effects and returns than
estimated above if replicated on a large scale for at-risk children in areas where the quality of
care currently being received was relatively low.
The issue of the optimal form and intensity of a preschool program cannot be settled
with the encouraging results of the Abecedarian project. How many years of full-time quality
preschool and child care are needed to produce the results outlined above? As a matter of
research, more information is needed on the long-term results of programs that vary the amount
and form of care before this issue can be settled. As a matter of principle, all children should
receive quality care in the first five years of life. A concern for the lives of the children
considered to be most "at-risk" can, by itself, direct public resources to an intervention that will
provide quality experiences to children and, over the long run, result in benefits that exceed
Adams, K.E. and T.L. Young. 1999. “Costs of Smoking: A Focus on Maternal, Childhood,
and other Short-Run Costs.” Medical Care Research and Review, Volume 56, Number
1, pp. 3-29.
Adelman, C. 1999. Answers in the Tool Box: Academic Intensity, Attendance Patterns, and
Bachelor’s Degree Attainment. Washington, D.C.: United States Department of Education,
Office of Educational Research and Improvement.
Altonji, J.G. and T.A. Dunn. 1990. Family Background and Labor Market Outcomes. Final
Report to the United States Department of Labor, Washington, D.C.: Bureau of Labor
Andrews, S., J. Blumenthal, D. Johnson, A. Kahn, C. Fergusen, T. Lasater, P. Malone, and D.
Wallace. 1982. "The Skills of Mothering: A Study of Parent-Child Development Centers."
Monographs of the Society for Research in Child Development, Volume 46, Number 6,
Serial Number 198.
Atkinson, A.B. 1981. "On Intergenerational Income Mobility in Britain." Journal of Post-
Keynesian Economics, Volume 3, pp. 194-217.
Barnett, S. and L. Masse. in press. “Funding Issues for Early Care and Education in the United
States.” in Early Childhood Education and Care in the United States. edited by Debby
Cryer. Baltimore, Maryland: Brookes Publishing.
Barnett, W.S. 1996. Lives in the Balance: Age-27 Benefit-Cost Analysis of the High/Scope
Perry Preschool Program. Monographs of the High/Scope Educational Research
Foundation, Number 11, Ypsilanti, Michigan: High/Scope Press.
Barnett, W.S. 1998. "Long-Term Cognitive and Academic Effects of Early Childhood
Education on Children in Poverty." Preventive Medicine, Volume 27, pp. 204-207.
Becker, G.S. 1964. Human Capital. New York: Columbia University Press.
Becker, G.S. 1981. A Treatise on the Family. Massachusetts: Harvard University Press.
Becker. G.S. and C.B. Mulligan. 1994. "On the Endogenous Determination of Time
Preference." NORC Working Paper, Number 9, Volume 2, pp. 1-67.
Behrman, J. and P. Taubman. 1985. "Intergenerational Earnings Mobility in the United States:
Some Estimates and a Test of Becker's Intergenerational Endowments Model." Review of
Economics and Statistics, Volume 67, pp. 144-151.
Benasich, A.A., J. Brooks-Gunn, and B.C. Clewell. 1992. "How do Mothers Benefit from
Early Intervention Programs?" Journal of Applied Developmental Psychology, Volume 13,
Bianchi, S.M. 2000. "Maternal Employment and Time with Children: Dramatic Change or
Surprising Continuity?" Demography, Volume 37, Number 4, pp. 401-414.
Birdsall, N. and S.H. Cochrane. 1982. “Education and Parental Decision-Making: A Two-
Generation Approach.” in Education and Development. edited by L. Anderson and D.M.
Windham. Lexington, Massachusetts: DC Heath.
Boesal, D., N. Alsalam and T.M. Smith. 1998. Educational and Labor Market Performance of
GED Recipients. Washington, D.C.: United States Department of Education, Office of
Educational Research and Improvement.
Brooks-Gunn, J., M.C. McCormick, S. Shapiro, A.A. Benasich, and G.W. Black. 1994. "The
Effects of Early Education Intervention on Maternal Employment, Public Assistance, and
Health Insurance: The Infant Health and Development Program." American Journal of
Public Health, Volume 84, Number 6, pp. 924-931.
Burchinal, M.R., M.W. Lee and C.T. Ramey. 1989. "Type of Daycare and Preschool
Cognitive Performance of in Disadvantaged Children." Child Development, Volume 60,
Bureau of Labor Statistics. 2000. “Multifactor Productivity Trends, 1997 [WWW document].”
URL http://www.bls.gov/news.release/prod3.-nws.htm, visited March 3, 2000.
Campbell,F.A., B. Breitmayer and C.T. Ramey (1986). “Disadvantaged Single Teenage
Mothers and Their Children: Consequences of Free Educational Day Care.” Family
Relations, Volume 35, pp. 63-68.
Campbell, F.A., R. Helms, J.J. Sparling and C.T. Ramey. 1998. “Early Childhood Programs
and Success in School.” in Early Care and Education for Children in Poverty. edited by
W.S. Barnett and S.S. Boocock. New York: State University of New York Press.
Campbell,F.A. and C.T. Ramey. 1994. “Effects of Early Intervention on Intellectual and
Academic Achievement: A Follow-up Study of Children from Low-Income Families.”
Child Development, Volume 65, Number 2, pp. 684-689.
Campbell, F.A., and C.T. Ramey. 1995. "Cognitive and School Outcomes for High-Risk
African American Students at Middle Adolescence: Positive Effects of Early Intervention."
American Education Research Journal, Volume 32, Number 4,pp. 743-772.
Campbell, F.A., C.T. Ramey, E. Pungello, J. Sparling and S. Miller-Johnson. 2002. “Early
Childhood Education: Young Adult Outcomes from the Abecedarian Project.” Applied
Developmental Science, Volume 6, Number 1, pp. 42-57.
Clarke, S.H. and F.A. Campbell. 1998. “Can Intervention Early Prevent Crime Later? The
Abecedarian Project Compared with Other Programs.” Early Childhood Research
Quarterly, Volume 13, Number 2, pp. 319-343.
Cohn, E. and T.G. Geske. 1990. The Economics of Education. New York: Pergamon Press.
Committee on Ways and Means. 1998. 1998 Green Book: Background Material and Data on
Programs Within the Jurisdiction of the House Committee on Ways and Means.
Washington, D.C. : United States Government Printing Office.
Cutler, D.M, J. Gruber, R.S. Hartman, M.B. Landrum, J.P. Newhouse and M.B. Rosenthal.
2000. “The Economic Impacts of the Tobacco Settlement.” Cambridge, Massachusetts:
National Bureau of Economic Research, Working Paper 7760.
Donohue, J.J. 1999. “Why We Should Discount the Views of Those who Discount
Discounting.” Yale Law Journal, Volume 108, pp. 1901-1910.
Field, T., S. Widmayer, S. Stringer, and E. Ignatoff. 1982. "Effects of Parent Training on
Teenage Mothers and their Infants." Pediatrics, Volume 69, pp. 703-707.
Fuchs, V.R., 1982. "Time Preference and Health: An Exploratory Study." in Economic Aspects
of Health. edited by V.R. Fuchs. New York: National Bureau of Economic Research, pp.
Fuchs, V.R. 1996. "Economics, Values, and Health Care Reform." American Economic
Review, Volume 86, pp. 1-2.
Fullerton, H.N. 1999. “Labor Force Participation: 75 Years of Change, 1950-1998 and
1998-2025.” Monthly Labor Review, December.
Garber, H.L. 1988. The Milwaukee Project: Prevention of Mental Retardation in Children at
Risk. Washington, D.C.: American Association on Mental Retardation.
Glewwe, P. and H. Jacoby. 1994. “Student Achievement and Schooling Choice in Low Income
Countries: Evidence from Ghana.” Journal of Human Resources, Volume 23, Number 3,
Grossman, M. 1972. "On the Concept of Health Capital and the Demand for Health." Journal
of Political Economy, Volume 80, pp. 223-255.
Grossman, M., and R. Kaestner. 1997. "Effects of Education on Health." in The Social Benefits
of Education. edited by J. Behram and N. Stacey. Ann Arbor, Michigan: University of
Gruber, J. and J. Zinman. 2000. “Youth Smoking in the U.S.: Evidence and Implications.”
Cambridge, Massachusetts: National Bureau of Economic Research, Working Paper
Haveman, R. and B. Wolfe. 1984. "Schooling and Economic Well-Being: The Role of
Nonmarket Effects." Journal of Human Resources, Volume 19, Number 3, pp. 377-407.
Hofferth, S., A. Brayfield, S. Deich and P. Holcomb. 1991. National Child Care Survey, 1990.
Washington, D.C.: The Urban Institute Press.
Kimmel, J. 1998. "Child Care Costs as a Barrier to Employment for Single and Married
Mothers." The Review of Economics and Statistics, Volume 80, Number 2, pp. 287-299.
Leigh, J.P. 1998. “Parents’ Schooling and the Correlation between Education and Frailty.”
Economics of Education Review, Volume 17, Number 3, pp. 249-358.
Levin, H.M. and P.J. McEwan. 2001. Cost-Effectiveness Analysis – 2nd Edition. Thousand
Oaks, California: Sage Publications.
Masse, L.N. 2002. A Benefit-Cost Analysis of the Carolina Abecedarian Preschool Program.
Unpublished doctoral dissertation, Rutgers University.
McCormick, A.C., A. Nunez, V. Shah, S. Choy and P.R. Knepper. 1999. Life After College:
A Descriptive Summary of 1992-93 Bachelor’s Degree Recipients in 1997. United States
Department of Education, Office of Educational Research and Improvement.
Manning, W.G., E.B. Keeler, J.P. Newhouse, E.M. Sloss and J. Wasserman. 1989. “The
Taxes of Sin: Do Smokers and Drinkers Pay their Way?” Journal of the American
Medical Association, Volume 261 (March 17), pp. 1604-1609.
Meyers, M.K. 1993. Child Care in Jobs Employment and Training Programs: What Difference
Does Quality Make?" Journal of Marriage and the Family, Volume 55, pp. 767-783.
Miller, H.P. and R.A. Hornseth. 1967. Present Value of Estimated Lifetime Earnings.
Washington, D.C.: United States Department of Commerce, Bureau of the Census,
Technical Paper Number 16.
Miller, T.R., C. Calhoun and W.B. Arthur. 1990. Utility-Adjusted Impairment Years: A Low-
Cost Approach to Morbidity Valuation. Washington, D.C.: Federal Highway
Moore, M.T., E.W. Strange, M. Schwartz and M. Braddock. 1988. Patterns in Special
Education Service Delivery and Cost. Washington, D.C.: Decision Resources Corporation.
Moore, M.J. and W.K. Viscusi. 1988. “The Quantity-Adjusted Value of Life.” Economic
Inquiry, Volume 26, pp. 369-388.
National Center for Health Statistics. 1998. “United States Abridged Life Tables, 1996.”
National Vital Statistics Reports, Volume 47, Number 13, 1998. Washington, D.C.:
United States Department of Health and Human Services.
Oster, G., G.A. Colditz and N.L. Kelly. 1984. The Economic Costs of Smoking and Benefits
of Quitting. Lexington, Massachusetts: D.C. Heath and Company.
Parish, T.B., F. O'Reilly, I.E. Duenas and J.M. Wolman. 1997. State Special Education
Finance Systems, 1994-1995. Palo Alto, California: Center for Special Education
Finance, American Institutes for Research.
Peters, H.E. 1992. “Patterns of Intergenerational Mobility in Income and Earnings.” Review of
Economics and Statistics, Volume 74,pp. 456-466.
Pungello, E., F.A. Campbell, and S.M. Johnson. 2000. "Benefits of High Quality Childcare for
Low-Income Mothers: The Abecedarian Study." Poster presented at Head Start's Fifth
National Research Conference (June, 2000). Chapel Hill, North Carolina: Frank Porter
Graham Child Development Center.
Ramey, C.T. and F.A. Campbell. 1984. "Preventive Education for High Risk Children:
Cognitive Consequences of the Carolina Abecedarian Project." American Journal of
Mental Deficiency, Volume 88, Number 5, pp. 515-523.
Ramey, C.T., B. Dorvall, and L. Baker-Ward. 1983. "Group Day Care and Socially
Disadvantaged Families: Effects on the Child and the Family." Advances in Early
Education and Day Care, Volume 3, pp. 69-106.
Ramey, C.T., F.A. Campbell, M. Burchinal, M.L. Skinner, D.M. Gardner, and S.L. Ramey.
2000. "Persistent Effects of Early Intervention on High-Risk Children and their Mothers."
Applied Developmental Science,Volume 4, Number 1, pp. 2-14.
Ramey, C.T. and S.L. Ramey. 1998. "Prevention of Intellectual Disabilities: Early Interventions
to Improve Cognitive Development." Preventive Medicine, Volume 27, pp. 224-232.
Roosa, M.W. and L. Vaughn. 1983. "Teen Mothers Enrolled in an Alternative Parenting
Program: A Comparison with their Peers." Urban Education, Volume 18, pp. 348-360.
Ross, C. and D. Paulsell. 1998. Sustaining employment among low-income parents: The role of
quality in child care: A research review. Princeton, New Jersey: Mathematica Policy
Sander, W. 1998. "The Effects of Schooling and Cognitive Ability on Smoking and Marijuana
Use by Young Adults." Economics of Education Review, Volume 17, Number 3, pp.
Schweinhart, L.J., H.V. Barnes, D.P. Weikart, W.S. Barnett, and A.S. Epstein. 1993.
"Significant Benefits: The High/Scope Perry Preschool Study through Age 27."
Monographs of the High/Scope Educational Research Foundation, Number 10, Ypsilanti,
Michigan: High/Scope Press.
Singh, R. 1992. “Underinvestment, Low Economic Returns to Education, and the Schooling of
Rural Children: Some Evidence from Brazil.” Economic Development and Cultural
Change, Volume 40, Number 3 (1992), pp. 645-664.
Solon, G. 1992. "Intergenerational Income Mobility in the United States." American Economic
Review, Volume 82, pp. 393-407.
Thompson, M.S. 1980. Benefit-Cost Analysis for Program Evaluation. California: Sage
Tolley, G., D. Kenkel and R. Fabian. 1994. Valuing Health for Policy: An Economic Approach.
Chicago, Illinois: University of Chicago Press.
United States Bureau of the Census. 1998. “Money Income in the United States: 1997.”
Current Population Reports, Series P60-200. Washington, D.C.: Government Printing
United States Bureau of the Census. 2002. "Table PCT12 – Sex by age." Retrieved July 23,
2002, from http://factfinder.census.gov/servlet/DT…d=01000US&mt_name=
United States Department of Education (USDOE). 1998. Dropout Rates in the United States,
1996. Washington, D.C.: National Center for Education Statistics, Report 98-250.
United States Department of Health and Human Services (USDHHS). 1997. Fertility, Family
Planning, and Women’s Health: New Data from the 1995 National Survey of Family
Growth. Atlanta, Georgia: United States Department of Health and Human Services,
Centers for Disease Control and Prevention, Series Report 23.
United States Department of Health and Human Services (USDHHS). 1998. Tobacco Use
Among U.S. Racial/Ethnic Groups – African Americans, American Indians and Alaska
Natives, Asian Americans and Pacific Islanders and Hispanics: A Report of the Surgeon
General. Atlanta, Georgia: United States Department of Health and Human Services,
Centers for Disease Control and Prevention.
United States Department of Health and Human Services (USDHHS). 2000. Head Start Fact
Sheet, 2000. Washington, D.C.: Administration for Children and Families, Head Start
Vandell, D.L. and B.Wolfe. 2002. "Child Care Quality: Does it Matter and Does it Need to be
Improved?" in Early Childhood Care and Education: What are States Doing? edited by K.
Bogenschneider, B. Friese, K. Balling, and J. Mills. Madison, Wisconsin: University of
Wisconsin Center for Excellence in Family Studies.
Varden, S.A. 1982. “Public School Systems.” in The Costs and Financing of Adult Education
and Training. edited by R.E. Anderson and E.S. Kasl. Lexington, Massachusetts:
Lexington Books, pp. 33-56.
Wechsler, D. Wechsler Intelligence Scale for Children – Revised. 1974. New York: The
Wolfe, B.L. and J.R. Behrman. 1985. “Who is Schooled in Developed Countries? The Roles
of Income, Parental Schooling, Sex Residence, and Family Size.” Economics of Education
Review, Volume 3, Number 3, pp. 231-245.
Woodcock, R.W. and M.B. Johnson. 1977. Woodcock-Johnson Psycho-Educational
Battery. Hingham, M.A.: Teaching Resources Corporation.
Table 8.1: Preschool Program Effects Related to Economic Benefits.
Program Group No-Program Group
Outcome Variable Measure N Measure N
Education Effects (children)
IQ (Stanford Binet), age 3* 101 50 84 48
IQ ( McCarthey GCI), age 4.5* 101 49 91 46
IQ (WISC-R), age 15 95 48 90 44
Reading Achievement (WJ), age
15* 94 48 88 44
Math Achievement (WJ), age 15* 93 48 82 44
Ever retained in grade by age 15* 31% 48 55% 44
Special Services by grade 9* 25% 48 48% 44
High School Graduation by age 19 67% 54 51% 51
Ever enrolled in a 4-yr college by
age 21* 36% 53 13% 51
Employment Effects at participant's age of 54 months (teenage parents)
Teenage Mothers and post-
secondary training 46% 13 13% 15
Teenage Mothers and self-
supporting 70% 13 58% 15
Teenage mothers and additional
births 23% 13 40% 13
Notes: * Differences are significant at the .05 level of confidence. WISC-R is the abbreviation for the Wechsler
Intelligence Scale for Children Revised (Wechsler 1974). WJ is the abbreviation for the Woodcock-Johnson Psycho-
Educational Battery, Part 2: Tests of Academic Achievement (Woodcock and Johnson 1977). McCarthy GCI is
abbreviation for the McCarthy General Cognitive Index. Data on education effects are from Campbell and Ramey
(1995), Ramey and Campbell (1984), Clarke and Campbell (1998), and Campbell et al. (2002). Data on employment
effects for teenage mothers are from Campbell et al. (1986). A mother was considered to be self-supporting if welfare
funds were not used except in the cases where the mother was a student and had made 4 years of educational progress
in the 4.5 years since the birth of her child.
Table 8.2: Present Value of Per Child Benefits and Costs of the Abecedarian Early
Intervention. Data are in 2002 dollars.
3% 5% 7%
Program Cost FPG Settinga $35,864 $34,599 $33,421
Program Cost PS Settingb $41,916 $40,427 $39,041
Part. Earnings 37,531 16,460 6,376
Earnings of Future
Generations 5,722 1,586 449
Ages 26-41c 43,030 34,378 27,786
Ages 42-60d 30,578 17,561 10,299
Subtotal 73,608 51,939 38,085
K-12 Education 8,836 7,375 6,205
Smoking / Health
Higher Education Costs
Total Benefits $135,546 $76,035 $48,317
Net Present Value FPG
Setting $99,682 $41,436 $14,896
Net Present Value PS
Setting $93,630 $35,608 $9,276
aProgram cost is for the Frank Porter Graham Child Development Center.
bProgram cost is for replication in a public school setting.
cMaternal earnings through age 41 are estimated using actual data on maternal earnings at ages 32, 35, and 41.
dMaternal earnings from age 42 to age 60 are extrapolated based on estimates through age 41 and assumes no increase in
Table 8.3: Estimated yearly costs of the Abecedarian Program in three different
settings. Data are undiscounted and in 2002 dollars.
School Child Care
Year 1 10,799 11,710 6,847
Year 2 16,222 17,793 10,189
Year 3 16,222 17,793 10,189
Year 4 11,991 13,175 8,133
Year 5 11,991 13,175 8,133
Totals 67,225 73,646 43,491
Table 8.4. Present Value of Marginal Costs of the Abecedarian Program in three Cost
Settings. Data are in 2002 dollars.
Discount Rate Abec/FPG Public School Child Care
0% $37,826 $44,246 $14,092
3% 35,864 41,916 13,410
5% 34,599 40,427 12,923
7% 33,421 39,041 12,469
10% 31,799 37,135 11,843
Table 8.5. Estimated Program Effects on Participant Compensation, Earnings of
Future Generations, and Schooling Costs. Data are in 2002 dollars.
Compensation Future Generations Schooling
0% $144,998 $48,542 $11,605
3% 37,531 5,722 8,836
5% 16,460 1,586 7,375
7% 6,376 479 6,205
Table 8.6. Program Effects on the Value of Life due to Decreased Rates of Smoking
and Increased Longevity. Data are in 2002 dollars.
Discount Rate Program Effect
Table 8.7. Program Effects due to Increased Maternal Earnings. Data are in 2002
Discount Rate Program Effect
Ages 26-41 0% $61,690
Ages 42-60 0% 73,256
Total 0% $134,946