Doing it right: job training and education

There has of late been much hand wringing over what to do about the emerging new labor market, in which the real wages of high-skilled, well-educated workers have increased, while the real wages of low-skilled and poorly educated workers have decreased. Unless something is done soon about this growing inequality, argue politicians and pundits, serious political and social consequences will follow. The solution favored by most is actually an old one - policies and programs that will supposedly help America's unskilled workers gain socially productive skills. It is further thought that increasing the supply of skilled workers will help to dampen the rising wage gap between the skilled and the unskilled. Indeed, most recent welfare reforms simply assume that adult-job-training programs can give participants the necessary skills to take advantage of the growing number of high-skilled jobs.

In this article, I reexamine the foundations of our current policies regarding skill formation and closely scrutinize the conventional wisdom on these matters. I draw on a body of recent scholarship that challenges many of the premises that govern public policy discussions. This scholarship suggests a broader view of the way skills are produced in a modern economy. Once this view enters the mainstream, the public's thinking on appropriate policies to promote skill formation will be substantially altered.

Three blind spots

Current policies regarding education and job training are based on fundamental misconceptions about the way socially useful skills are produced and precisely what those skills are. The conventional wisdom, held by most politicians, educated laypersons, and even many academics - especially professors in schools of education - is that formal educational institutions contribute most to skill formation. But this neglects the crucial role of families and firms in fostering skills.

The current emphasis in policy discussions on formal schooling, to the exclusion of informal noninstitutional sources of learning, is a consequence of three blind spots. The first is factual in nature and arises from a failure to recognize that learning is a lifetime affair. Formal or institutional education is only one aspect of the learning process, and recent research indicates that it is not necessarily the most important one. Learning starts in infancy long before formal education begins and continues throughout life. Recent research in psychology and cognition demonstrates the vital importance for skill formation of the early infant and preschool years when human ability and motivation are shaped by families and noninstitutional environments. Success or failure at this stage translates into success or failure in school, which, in turn, leads to success or failure in post-school learning. In other words, early learning begets later learning and early success breeds later success. In addition, since the publication of the Coleman Report, we have known that families and environments - not schools - play the crucial role in bringing about educational success as measured by test scores. Students from failed families do not succeed in school. Thus policies directed toward families more effectively improve school performance than expenditures on teacher salaries or new computers.

At the other end of the educational process is the postschool experience. Postschool learning, which includes learning by doing and workplace education, is an important source of skill formation that accounts for as much as one-third to one-half of all skill formation in a modern economy. However, because much of this learning takes place in informal settings outside of educational institutions, it is neglected by educational technocrats and the politicians who commonly equate skill formation with classroom learning.

A second blind spot of most educational planners and policy makers results from their belief that achievement tests and various measures of cognitive skill indicate the success of an educational intervention. It is certainly true that cognitive ability is important in life, and the return to cognitive ability has increased over time. However, this narrow focus on cognition ignores the full array of socially and economically valuable noncognitive skills produced by schools, families, and other social institutions. Enriched early intervention programs do not substantially alter IQ. However, they substantially raise noncognitive skills and the social attachments of participants.

A third blind spot of most educational planners and policy makers is their failure to recognize that parents can choose wisely if offered choices about their children's education. Such policy-making elites close their eyes to the fact that competition and incentives can improve the performance of schools. More generally, there is substantial mistrust of unregulated informal learning, whether in the workplace or the home. Although the failure of social planning in Eastern Europe and elsewhere is widely acknowledged, the social-planning mentality is alive and well in the local monopolies of American public education at the primary and secondary levels in most states. Everything we know suggests that most parents care dearly about the education of their children and can tell the difference between a good school and a bad one. When offered choices and the opportunity to experiment, most parents generally choose wisely, at least after they have gained some experience. They can distinguish the good teachers from the bad ones, the good schools from the bad schools.

Policy myths

In addition to these blind spots, policy makers persist in promoting programs that do not work. For example, much educational policy regarding college attendance assumes that a large number of bright but poor students are being denied a college education by the inadequate financial resources of their families. This powerful myth is the cornerstone of conventional wisdom about educational policies. Though it drives current thinking about educational policy, it is only superficially supported by the evidence.

The entire field of educational research is littered with myths like this - myths that guide public policy. We are told that reducing class size or increasing per pupil expenditure in primary and secondary schools will substantially improve educational outcomes. Yet most of the evidence points to the contrary. We are told that education produces substantial externalities - benefits harvested by others beyond those obtained by the people who are educated. These externalities are said to justify further subsidies to education. Yet careful analysis yields little evidence of such externalities, at least in the modern American economy.

Myths abound about job training as well. Presently, the economy has a large group of unskilled workers who have lost their jobs because of the rapid shift in demand toward more skilled workers, and because of the floor on wages in the markets for unskilled labor. One dangerous myth that motivates welfare reform and training policy is that it is relatively easy to adapt adult unskilled workers to the modern economy. According to this view, most low-skilled labor can be easily retrained to become skilled labor. In fact, the costs of doing so are high. Resources are diverted away from the young and the more trainable for whom a human-capital-investment strategy is likely to be more effective. In an era of tight budgets, it is far from obvious that investments in such low-skilled workers are justified on any but political grounds.

Also missing from current policy discussions of education and training policy is any consideration of priorities or recognition of the need to prioritize. In an era of tight government budgets, it is impractical to consider active investment programs for all persons. The real question is how to use the available funds wisely. The evidence supports the following policy prescription: Invest in the very young to improve their basic learning and socialization skills; at a given level of expenditure, reduce investment in the old and severely disadvantaged, shifting to wage subsidies for them.

In analyzing the evidence for these claims and their relevance for educational policy, it is important to distinguish statements about a world in which there is no educational policy from the world in which we live. The current subsidy of direct costs to students at major public universities is around 80 percent. So the appropriate question is whether that already high subsidy rate should be increased, not whether there should be any subsidy at all.

More college students needed?

A dramatic shift against unskilled labor has occurred in the American labor market. Since the early 1980s, the real wages of unskilled males have declined, while the real wages of skilled workers have increased. Accompanying the decline in the wages of the unskilled is a decline in labor-market activity, especially among the unskilled. A variety of labor-force measures show increasing joblessness and longer unemployment spells for persons with low skills.

Can we reverse this trend by increasing the supply of college-educated labor? Certainly, it would be enormously expensive. From 1979 to 1989, the wages of college graduates relative to those of high-school graduates changed roughly 14 percentage points. To reverse this 10-year trend by increasing the supply of skilled workers would require a once and for all increase of approximately 20 percent in the number of high-skilled persons in the workforce. College equivalents are 40 percent of the workforce and high-school equivalents are 60 percent. For a 1990 workforce of 120 million, about 5.4 million people would have to become college equivalents to reverse the decade-long erosion of real wages. To maintain this gain in relative wages against the secular trend operating against unskilled labor, about one million additional skilled persons would need to be added to the workforce each year on top of the once and for all change of 5.4 million.

As a benchmark, the annual supply of high-skilled equivalents to the U.S. economy in the early 1990s is approximately 1.8 million persons. Maintenance of existing skill gaps alone would require that the percentage of persons acquiring post-secondary skills rise by 55 percent. To phase in the additional 5.4 million people required to restore wage parities to their 1979 levels over a 10-year period would require a total annual expansion of 70 to 80 percent in the supply of high-skilled persons. To engineer such an increase through a tuition policy, it would be necessary to reduce tuition substantially. In fact, college tuition would have to be zero, and people would have to be paid to attend school.

But is it really the cost of college tuition that reduces the quantity of college-educated labor? To see if there is any reason for further increases in subsidies to promote skills, let's consider evidence on the sources of skill formation and the claims made about market and family failures that impede such skill formation.

Conventional wisdom

The strong positive relationship between family income and participation in college has attracted an enormous amount of attention among academics. Most commonly, this phenomenon is thought to mean that short-term family financial constraints prevent children from low-income families from attending school. The Pell Grant program, the recently enacted HOPE fellowship program, and many governmental educational programs are premised on this interpretation of the data. It is also the most popular explanation for why whites have responded more quickly than minorities to the new market for skills by enrolling in college. Since minority families are concentrated in the lower end of the family-income distribution, their failure to respond to the increase in the economic return of schooling is widely viewed as a manifestation of the more general phenomenon of a market failure. The failure is attributed to borrowing constraints that prevent poor people from going to college and taking advantage of the increased returns to higher education.

This common interpretation also notes that real tuition costs have increased over the past 16 years. At the same time, family incomes have declined among the bottom quartiles of family-income distribution. The real wages and employment of unskilled males have declined since the late 1970s. Minorities have been affected more than whites because more minority families occupy the lower portion of the family-income distribution. In addition, more families at the bottom of the family-income distribution are headed by females with dependent children. Such families have lower earnings and income levels than families headed by males.

In sum, it is held that the combination of rising tuition costs and declining family resources has prevented many children of low-income families from attending college. What's needed, advocates claim, are more subsidies to educate children from low-income families.

Cost no barrier

Certainly, real tuition costs have risen over the past 16 years. Between 1980 and 1997, average tuition levels rose by 100 percent at public four-year colleges and universities and 77 percent at public two-year colleges. At the same time, it is important to recognize that government subsidies to higher education are huge. In fact, substantial college loan and aid programs already exist for students from poorer families. At public two- and four-year institutions in the United States, in 1996, average costs of tuition and fees were about $2,300. To offset such costs, both the Pell grant and subsidized student loan programs are available to children of low-income families.

A first-year dependent college student was eligible in 1997 for a maximum of about $8,000 in Federal grants and subsidized loans. No minimum contribution toward estimated college costs was required from the student or the parents. In round figures, the $8,000 maximum benefit consisted of $2,500 in Pell grants, $2,500 in Stafford loans, and $3,000 in Perkins loans. The maximum rose to $11,000 for third- and fourth-year college students as subsidized loan limits rose. A host of other Federal programs direct money, in some form or another, to students. In 1996, total institutional subsidies alone were approximately the size of the entire Pell grant program.

At the same time, discussions of college tuition policy pay too much attention to the elite public and private institutions, ignoring the broad spectrum of community colleges and public four-year colleges. About half of all current college students are enrolled in community colleges. These colleges offer low tuition (typically about $1,300 per year in 1997 dollars) and flexible schedules that allow students to work while attending college. There are many more community colleges (1,036) than four-year colleges (604). Furthermore, most prospective students live near to one and can avoid room and board costs by living with their parents. And it is important to note that, in studies by Thomas Kane of Harvard and Cecilia Rouse of Princeton, the economic return to a year at a community college was found to be the same as the economic return at a four year college. Their evidence suggests that attending a community college for the first two years of the college experience has no adverse consequences in terms of future earnings.

All told, individuals attending public institutions of higher education pay, on average, less than 20 percent of the total direct cost of attending college. (This direct cost does not include foregone earnings.) Moreover, a substantial fraction of the 20 percent of the total costs borne by individuals is actually paid by private foundations and charities that extend aid directly to students.

In the current environment, with the institution of the community college in place, and with generous loan and grant programs available, the argument that tuition costs and commuting are major barriers to college attendance by the poor is implausible. It is useful to note that the take-up rate on Pell Grants and Perkins Loans is low. Many more funds are available than are spent. Borrowing constraints are not a plausible explanation for the lack of utilization of these resources. It is more likely that many eligible persons decide that even with a substantial tuition subsidy, the returns to college education for them would be too low to pay for the foregone earnings required to attend school.

Family financial constraints reconsidered

But advocates for more subsidies put forward additional evidence to make their case. The argument for the importance of family financial constraints in explaining the disparity between minority and majority college participation rates starts by noting correctly that human capital is different from physical capital. People cannot sell the rights to their future labor earnings to lenders in order to secure financing for their human-capital investments. Even if they could, there would be substantial problems enforcing the performance of contracts on future earnings given that individuals control their own labor supply and the quality of their work effort. The lack of collateral and the inability to monitor effort are widely cited reasons for current large-scale government interventions to finance education.

If people had to rely on their own resources to finance all of their schooling costs, the level of educational attainment in America would eventually decline. To the extent that subsidies do not cover the full costs of tuition, people must raise tuition through private loans or part-time work, or they must forego consumption. Children from families with higher incomes have access to resources that children from families with lower incomes do not have, although children from higher-income families still depend on the good will of their parents to gain access to funds. Limited access to credit markets means that the costs of funds are higher for the children of the poor, restricting their enrollment in college.

Current tax law exacerbates these problems. Until recently, borrowing costs for education could not be written off against taxes. Even if they could be itemized, few poor families would find it profitable to itemize deductions. Until recently, the only exception to the nondeductibility of interest in the current tax law was mortgage interest. Under the new law, enacted in 1.997, deductions for student loans are available. These deductions benefit persons who itemize and are unlikely to have much effect on college attendance for poor children. Of course, all families can itemize mortgage-interest payments in declaring their taxes. However, poorer families are much less likely to own homes, and hence, fewer of them are eligible to use mortgages on homes to finance the schooling of their children.

The same principles that govern the purchase of other goods regulate the purchase of education. There is, undoubtedly, a consumption component to education. Families with higher incomes buy more education for their children - and buy higher-quality education as well. This factor partly explains the strong positive relationship between family income and college attendance.

An alternative, and not necessarily mutually exclusive, interpretation of the same evidence is that long-run family and environmental factors play a decisive role in shaping the ability and expectations of children. Families with higher levels of resources produce children who are better able to perform in school and better able to take advantage of the new market for skills. Such children have access to higher quality primary and secondary schools, and better educated parents who are able to assist and direct their studies. These family influences are present from birth through adolescence, accumulating over the years, producing scholastic ability and college readiness.

This alternative interpretation stresses the role of family and the environment and does not rule out the importance of short-term borrowing constraints. However, if the finances of poor but motivated families hinder them from securing decent elementary and secondary schooling for their children, resulting in a low level of college readiness, a government policy aimed at reducing short-term borrowing constraints for the college expenses of those children is unlikely to be effective. Policy that improves the environments that shape ability may be a more effective way to increase college enrollment in the long run. The issue can be settled empirically, but surprisingly little data have been brought to bear on this matter until recently.

The distinction between long-run family factors that promote college readiness and short-term borrowing constraints can be sharpened by imagining the following experiment. A random sample of families is drawn, and some of them win a million-dollar lottery at different points in the life cycle of their children. Those who win as their children are finishing high school have little opportunity to make cumulative long-run family investments that would contribute to college preparedness. Thus the newly acquired wealth would have little effect on the college attendance of their children (assuming college readiness and ability are the decisive elements in producing enrollment in college). For lottery winners with young children, in contrast, one might expect an increase in the college attendance of their children, provided that the parents invest in better schools and greater academic opportunities for their children over the longer horizon. If short-term credit-market constraints are the significant factor governing college attendance, then one would expect a large increase in college enrollments by children of previously poor families, irrespective of the age of their children, at the time the lottery was won.

If our aim is to encourage college attendance, improving the environments of children and improving preparation for college will be more effective than grant or loan programs to economically or cognitively disadvantaged children in their late teenage years. The reason is this: Cognitive ability is formed relatively early in life and becomes less malleable as children age. By age 14, basic cognitive abilities seem to be fairly well set. Since ability promotes academic progress, successful interventions early in the life cycle of learning lead to higher overall achievement. By the time individuals finish high school, scholastic ability is determined, and tuition policy will have little effect on college attendance.

Only to the extent that the family income of able high-school graduates falls below levels required to pay for college will short-term credit-market constraints hinder college entry. But given the current college-financial-support arrangements that are available to low-income and minority children, few bright students in this country are being denied access to college because of lack of credit.

The importance of the long run

In my joint research with Stephen Cameron, a professor of economics at Columbia, I dealt with a number of these general issues and posed, in particular, two questions: 1) Which factors have the most influence on school attainment? (2) Is the estimated influence of family income on college attendance a consequence of long-run family effects or short-term borrowing constraints?

To answer these questions, we compared the estimated effects of family background on college attendance when scholastic ability is included (and when it is not) as an explanatory variable in a statistical analysis of college enrollment. Measured scholastic ability is the outcome of long-term family and environmental factors - and these are, in part, the product of long-term permanent income of families. To the extent that the influence of family income on college attendance is diminished by including scholastic ability in an empirical analysis of college attendance, one concludes that long-run family factors, which are reflected in scores, are the driving force behind school attainment, and not short-term credit constraints.

We examined what portion of the gap in college attendance between minority youth and whites is due to family income, to tuition costs, and to family background. Not controlling for ability measured at an early age, about half of the 11-point gap between black and white college-attendance rates of high-school graduates is due to family income; more than half of the seven-point difference between Hispanics and whites is due to family income. When scholastic ability is accounted for, only one-half of one point of the 11-point black-white gap is explained by family income. For Hispanics, the majority-minority gap actually widens when family income is controlled for and when ability is introduced into the analysis. More importantly, equalizing ability more than accounts for minority-majority college-attendance gaps. Similar differences show up in high-school graduation rates and overall college attendance rates that do not control for the effect of high-school graduation.

Ability produced by families and environments accounts for pronounced minority-majority differences in school attainment, not financial resources available at the age prospective students are considering college enrollment. The disincentive effects of college tuition on college attendance are dramatically weakened when ability is entered into an analysis of college attendance. When family background, as measured by parental education, is introduced in place of measures of cognitive ability, similar findings result.

An important study by Jeanne Brooks-Gunn of Columbia and Greg Duncan of Northwestern demonstrates that the income available to the family when the child is young determines a child's schooling attainment, not family income at later ages. Their analysis suggests that long-run factors, not short-term borrowing constraints, explain the disparity in levels and trends between majority and minority college attendance. Programs like HOPE scholarships that operate late in the life cycle are likely to be ineffective in promoting college attendance and wasteful of public funds.

The main feature of the HOPE program is a tax credit of up to $1,500 for two years of college. The estimated enrollment response to the HOPE program is a 4.2 percentage point increase in two-year enrollments and a 0.9 percentage point decrease in four-year enrollments. Most of the response to the program will come in enrollment in two-year schools. At least 91 percent of the total expenditure of this program will be spent on people who would enter college in the absence of the program.

But what about other policy proposals to promote skill formation? Do they have better empirical foundations than those based on short-term credit constraints?

Do class size and money matter?

State and local governments heavily subsidize primary and secondary education. Virtually all direct operating costs are completely subsidized through high school; only the opportunity cost of the time that students spend in school, rather than at work, remains unsubsidized. Many have questioned whether the amounts spent are adequate. Should teachers be paid more? Should class sizes be reduced?

Many policy makers and politicians claim that class size reductions promise big gains in test scores. (Test scores are commonly used as a measure of the success or failure of educational reforms.) Yet the link between test scores and later outcomes is at best weak, especially in the early years of schooling. Fortunately, in recent years, a series of studies have appeared that link measures of schooling quality to measures of lifetime earnings and occupational achievement. From this literature, there is a growing consensus that within current ranges of expenditure, measured inputs such as class size and spending per pupil have little, if any, effect on the future earnings of students.

Even if we take the most favorable estimates from this research and combine them with the best-case scenario regarding the costs of raising school quality at the secondary level, increasing spending per pupil is not a wise investment. In fact, the available evidence indicates that we may be spending too much on students. For example, let's take the liberal estimate of a 2 percent increase in future earnings for a 10 percent rise in per pupil spending: This yields a negative net return for such expenditure increases for all schooling levels when benefits are weighed against costs. For high-school graduates, the net loss is about $3,800; for college graduates, the net loss is over $4,400. Unless the same increase in spending raises future earnings by 5 percent or more per year - a number far higher than shown by any study - the financial costs of raising school quality through increased spending far outweigh the returns. In order to justify additional spending on primary and secondary schools, we would need to appeal to social benefits not captured by earnings.

What about the effect of reducing class size on the test scores of primary-school children? When Eric Hanuschek of the University of Rochester examined the widely celebrated STAR program, he found little evidence that it benefited children, except for those in kindergarten. His analysis of the STAR program supports the case for certain kinds of early interventions (which I will get to momentarily), but not a wide-scale reduction of class sizes for all ages.

Such evidence does not prove that school quality does not matter - surely, it does. We know that if we increase quality from very low levels, it matters greatly. But the evidence does not indicate that marginal improvements on current levels of school quality will be effective. Pouring more funds into schools to lower class sizes by one or two pupils, or to raise spending per pupil by a few hundred dollars, will not solve the problems of our primary and secondary schools. While the effects of quality vary across environments, and additional funding for some schools may be justified, more fundamental changes are required if we hope to see a significant improvement in our educational system.

The benefits of competition and incentives

It is widely perceived that American primary and secondary schools are failing. And by many measures, they are. American high-school students do very poorly on standardized achievement tests compared to students from other nations. Then there is the slow growth in achievement test scores. And, in the recent past, there have been periods of declining scores in the face of rising real per pupil expenditures. As a nation, we are spending more and more on public schools and apparently getting less and less.

In contrast, consider the American college system. Students from around the world flock to the United States to study in our colleges and graduate schools - yet our secondary schools are second rate. How can we reconcile the apparent contradiction of mediocre high schools and great colleges?

The answer lies in the way public schools are currently organized. Public school systems are local monopolies with few competitors. The American high-school system is a creation of the twentieth century and is a world unto itself. Within it, an artificial adolescent culture that discourages academic achievement flourishes, as the sociologist James Coleman showed 40 years ago. Principals and teachers have few incentives to provide students with useful knowledge, or any knowledge at all. Accountable to no one, because it is not easy to monitor them, teachers are often unresponsive to the changing demand for skills or to market realities that will confront their students when they leave school. One valuable source of information - parental and student perception of the qualities of teachers and schools - is rarely used. Educational technocrats dismiss such evaluations as "subjective" and unreliable. And the parents and students in these schools have limited ability to act on valuable private information about bad teachers and schools. Poor families especially have only limited ability to choose alternative teachers and schools.

An emerging body of evidence indicates that competition and choice does improve the quality of schools, as measured by test scores and by parental and student satisfaction. Contrary to the view that competition siphons resources away from the public sector, Caroline Hoxby, a professor of economics at Harvard, has demonstrated that when public schools are subject to greater competition, both from parochial and other private schools, the performance of all schools increases. And thus higher levels of achievement are produced at lower cost. Meanwhile, the economist Derek Neal of the University of Wisconsin has demonstrated that the higher achievement scores of Catholic school students compared to public school students is largely a consequence of gains registered by inner-city students who choose Catholic schools over inferior inner-city public schools. (In the suburbs, where districts are smaller and competition among school districts is more intense, the Catholic schools have little advantage over the public schools, and the performance of both school systems is higher than in the centralized inner-city schools.)

Everywhere it has been investigated, the effects of competition in education and training (and in virtually every other sector of life) are found to be beneficial for the students and trainees. For example, the celebrated German Apprenticeship System has been recommended as a model for the United States. That system gives high-school-age students the choice of integrating work and learning as an alternative to strictly formal academic education. When stripped to its essentials, the German system differs from the U.S. public school system in several crucial ways: First, it breaks down the artificial separation between the world of work and the world of learning; second, it gives students and the firms that apprentice them choice among a variety of learning situations; and third, it motivates students to perform well in order to secure the most desirable apprenticeships, while motivating firms to provide valuable training opportunities. The high level of competition among firms offering apprenticeships and among students is a major source of the German system's success.

Once it is recognized that the public schools, especially inner-city public schools, are a virtual monopoly, while the U.S. university system is highly competitive, the mystery of the poor performance of the former, and the great success of the latter, vanishes. It is remarkable that, in a society as committed to consumer sovereignty and choice as ours, there should be so much resistance to choice in education. The conventional argument of the educational planners is that parents and students are not able to make wise choices. But all the evidence points in precisely the opposite direction.

Consider for a moment recent research suggesting that the early years of childhood, prior to schooling, are crucial for later success in schooling and life. It is parents who shape the environment of the preschool years, and, in most circumstances, we rightly trust them to do so. Yet, the logical extension of the paternalistic argument that parents cannot make wise choices in schooling would be that the state should play a more active role in the preschool life of the child - a position few would be willing to accept.

While I reject the paternalistic argument for most children, such interventions may be appropriate for some disadvantaged families. If we are to violate the principle of consumer sovereignty anywhere in the life cycle of learning, the case for doing so is strongest at the infant and preschool stage, at least for some. In fact, the evidence backs such interventions.

Invest early

Recent studies of early childhood investments indicate that the early childhood years are important for learning and can be beneficially manipulated. Moreover, early-childhood interventions of high quality have lasting effects. Disadvantaged, subnormal IQ children randomly assigned to the Perry Preschool program, one of the early childhood programs that has been most studied, were administered intensive treatment at ages four and five. Treatment was then discontinued, and the participants were followed over their life cycle. (They are now about 35 years old.) The evidence indicates that those enrolled in the program have higher earnings and lower levels of criminal behavior in their late twenties than do comparable children randomized out of the program. Reported benefit-cost ratios for the program are substantial. Measured through age 27, the program returns $5.70 for every dollar spent. When returns are projected for the remainder of the lifetimes of program participants, the return on the dollar rises to $8.70. A substantial fraction (65 percent) of the return to the program has been attributed to reductions in crime.

Another program, the Syracuse Preschool program, provided family development support for disadvantaged children from the prenatal stage through age five. Reductions in probation violations and criminal offenses 10 years later were as large as 70 percent among children randomly assigned to the program. Girls who participated in the program also showed greater school achievement. Studies of early-intervention programs have found short-term increases in test scores, less grade retention, and higher high-school graduation rates among enrolled children. Of those studies that examine predelinquent or criminal behavior, most have found substantially lower rates of deviant behavior among program participants.

These programs achieve their success more by affecting the motivation and noncognitive skills of participants than boosting IQs. Yet these enhanced noncognitive skills help children to adapt to society's norms, promoting a variety of beneficial behavior down the road, such as lower crime, greater educational attainment, and reduced participation in welfare. Evidence of success in the larger Head Start program is less clear. But the program is quite heterogeneous, not to mention less well funded than the Perry Preschool program. Janet Currie and Duncan Thomas, professors of economics at UCLA, found short-term gains in test scores for all participating children; however, most of those gains decayed quickly for African-American children. They concluded that either differences in local program administration or in subsequent schooling quality are at the root of the contrary outcomes for black and white children. In fact, the schools attended by the Perry students after the program was over were of higher quality than those attended by the typical Head Start student. Significantly, however, similar declines in test scores were found for programs like Perry Preschool, though their long-term evaluations are quite favorable. The fade-out effects in test scores do not imply that long-term beneficial effects of Head Start are not present; Head Start may improve the lifetime prospects of its participants, despite yielding only short-term gains in test scores.

However, Head Start may have smaller long-term impacts than more intensive programs. Studies of Head Start's impact on special-education placement and grade retention have produced fewer dramatic results than have been achieved from smaller demonstration projects. This is not surprising given the much lower spending per child and lower quality of service provided by the Head Start program. Unfortunately, there are no reliable long-term evaluations of the Head Start program that link interventions to conventional socioeconomic outcome measures like occupations and earnings.

The weaknesses of Head Start can be attributed to its shorter period of intervention, lower intensity, and less qualified staff than is typical of more ideal programs. With Head Start, as with most other things in life, you get what you pay for. For example, children enrolled in the Perry Preschool program received high quality full-time preschool services for one to two years (most received two years), and their parents benefited from weekly home visits by their children's teachers. In contrast, Head Start offers a much lower quality (and lower paid) staff, only part-time classes for children, and limited opportunities for parental involvement. Improvements in Head Start, proponents argue, would produce effects closer to those observed in more successful small-scale programs. Given the potential for success, more studies of the long-term impacts of various types of small-scale and broad-based early intervention programs are certainly warranted.

The more successful preschool programs intervene in the lives of the child and the parents, cultivating improved parenting skills and enhanced labor-force activity by parents. In this way, they create a permanent change in the home environment, reinforcing the early intervention received by the child. Head Start does not attempt to alter family behavior, and so its effects may be more transient. Provocative calculations recently published by John Donohue of Stanford Law School and Peter Siegelman of Yale Law School indicate that if Perry Preschool-type programs were developed for high-risk, disadvantaged minority male youth, the substantial costs of these enriched programs would be more than repaid by the expected savings in incarceration costs alone.

Private training vs. public training

I will now turn to the other end of the lifecycle of skill formation: job-training programs for adults. Because of a lack of data, and a bias in favor of the funding of studies of government training, the returns to private-sector training are less well understood than the returns to public-sector training. Many studies find that the effects of private-sector training are considerable. In comparison with studies of public-sector training, most of these studies do not attempt to control for the effect that more able persons are more likely to take training; and thus, estimated rates of return to this activity would overstate the true returns to training by combining them with the returns to ability. In other words, part of the measured returns may be due to more motivated and able persons taking the training. But, even adjusting for this, returns to private-sector training are substantial. Estimated initial returns range from 10 percent to 20 percent - a good investment by any measure.

An important feature of private-sector training is that the more skilled do more investing even after they attain high skill levels. To the extent that effective training can be produced on the job, it is produced in the private sector, not in the public sector. The best hope of getting reasonable returns from job training is to encourage private-sector investment. Firms are also more sensitive to changing market demands for skills than are government bureaucracies.

It is important to note, however, that private-sector training typically excludes low-skilled persons. Firms can be exclusive in a way that government training programs for disadvantaged workers cannot be. The lack of interest of private firms in training disadvantaged workers indicates the difficulty of the task and the likely low rates of return to this activity. Public-sector training programs are an inefficient transfer mechanism and an inefficient investment policy for adult workers with few skills.

Recent studies of "learnfare" and "workfare" programs provide the relevant evidence. An evaluation of two Wisconsin programs is of interest, especially given the claims made for their success by Governor Tommy Thompson. One program, the Community Work Experience Program (CWEP), required mandatory participation in unpaid community-service jobs for nonexempt AFDC participants. A second program, Work Experience and Job Training, provided AFDC clients with assessment, job-search activities, subsidized employment, job training, and community-work experience. Participants who failed to find employment after completing their education and training were also required to participate in CWEP jobs.

Using randomized trials for one county and nonexperimental methods for the rest, John Pawasarat and Lois Quinn at the University of Wisconsin, Milwaukee, found that these programs had no effect compared to existing program alternatives. The reduction in AFDC participation that is widely cited as a consequence of these programs is essentially due to the improvement in the Wisconsin economy during the time the programs were in place. These results are disappointing but consistent with previous studies of the efficacy of such programs by the Manpower Demonstration Research Corporation. Mandatory work-experience Programs produce little long-term gains in employment or earnings. No cheap training solution has yet been found that can end the welfare problem. Lifting a welfare woman out of poverty by increasing her earnings by $5,000 per year ($100 per week) will cost at least $50,000, assuming a wildly optimistic 10 percent rate of return to the investment. There is no "quick fix" or low-cost solution to the welfare problem, and politicians who promise otherwise are deluding themselves and the public.

The fact is that policies that seek to alleviate poverty by investing in low-skill workers conflict with policies that raise the wealth of society at large. Taking the available evidence at face value, the most economically justified strategy for improving the incomes of the poor, especially low-ability low-skilled adults, is to invest more in the highly skilled, and then to tax them and redistribute the tax revenues to the poor.

However, because many people view the work ethic as a fundamental value, they argue that cultivating a large class of transfer recipients would breed a culture of poverty. If value is placed on work as a necessary aspect of individual dignity, then society may be prepared to subsidize inefficient jobs. But it must be emphasized that job subsidies are not the same as investment subsidies. The evidence strongly suggests that investing in low-skilled, disadvantaged adult workers makes no economic sense. Wage subsidies offer a more efficient alternative for raising their incomes. The economist Edmund Phelps has argued this point in a recent book. Both Phelps and economist Lance Lochner argue that by encouraging work, rather than unemployment and crime, wage subsidies may also provide social benefits that extend beyond their impact on individual earnings.

Sensible policy

Most policy makers today assume that formal educational institutions are the major producers of skill. In truth, families and private firms are equally important, if not more important. Noncognitive skills fostered by functioning families are essential for success in life. The conventional wisdom also ignores the life-cycle nature of learning and the current high level of subsidy to education. The relevant question today is whether these high subsidies should be increased further, not whether there should be none at all. Certainly, the evidence does not indicate that further subsidies to college education are warranted. The disparity in college attendance between the poor and the middle class is not the result of cash constraints operating on poor children in their teenage years. Since the problem is to be found early in the life cycle of learning, throwing additional money at public schools will not improve performance. The problem in public education is primarily due to muted incentives, not to inadequate resources.

Politicians and the public must come to recognize the importance of informal sources of learning, the value of noncognitive skills, the cumulative character of learning, and the value of incentives in formal education. Only then will we be on our way toward crafting policies that promote the formation of skills suitable to a modern economy.

The author would like to acknowledge the David Kinsley Lecture at the University of Illinois at Urbana-Champaign where an earlier version of this article was first given, and he would like to dedicate it to the memory of the late T. W. Schultz.

JAMES J. HECKMAN is Henry Schultz Distinguished Service Professor of Economics and in the Harris School of Public Policy.

COPYRIGHT 1999 The National Affairs, Inc.
COPYRIGHT 2004 Gale Group