A recent paper, by Susan Dynarski, Joshua Hyman, and Diane Whitmore Schanzenbach, has received some attention in the early education community, and elsewhere. The paper is entitled “Experimental Evidence on the Effect of Childhood Investments on Postsecondary Attainment and Degree Completion”. In addition to an article by Maureen Kelleher at Education Week, I received an email from a prominent early education supporter who was concerned that this paper might reflect negatively on early childhood programs.

The article examines the long-run effects on educational attainment of the Tennessee Student /Teacher Achievement Ratio (STAR) experiment. The STAR experiment lowered early elementary class size in kindergarten through third grade.

The article’s main conclusion is that STAR significantly increased college attendance and completion. The lower class sizes increased the percentage of those attending college by 2.7 percentage points, and the percentage of those completing college by 1.6 percentage points.

These significant effects occurred even though the test score effects of the STAR experiment tended to fade by middle school.  The early effects of STAR on test scores are better predictors of its long-run effects on college attendance than are later test score effects. This is similar to findings for many early childhood programs.  These early interventions appear to have some profound long-run effects on behavior that are not fully reflected in middle school and high school scores on standardized tests.

Why the concern from some early childhood supporters? The concern is with one small section of the paper. This section is entitled “Do Early Interventions Pay Off More Than Late Ones?” This section is explicitly addressed at the argument, sometimes made by Nobel-prize-winning economist James Heckman and his co-authors, that the rate of return to human capital investments is generally greater for earlier investments.

In the Dynarski/Hyman/Schanzenbach paper, they review the cost per additional student attending college for various policy interventions. These policy interventions include:  two preschool interventions, Head Start and the Abecedarian program; STAR; the Upward Bound program, which provides at-risk high school students with targeted instruction and counseling; college scholarship aid; help with completing financial aid applications.  The “cost per additional student” induced to attend college due to these policies are as follows:  Head Start, $133,000; Abecedarian, $410,000; STAR, $400,000; Upward Bound, $94,000; targeted scholarship aid, $21,000; help with completing financial aid applications, $1,000.  They conclude, based on this review, that “these results provide little support for Heckman’s assertion that early investments are the most cost-effective, at least if the desired effect is increased college attendance.”

To complement the authors’ analysis, I did some analysis using the most recent data on the Chicago Child Parent Center (CPC) program. CPC’s program design is similar to that of many state-financed pre-K programs. These recent data suggest a cost per additional student attending a four-year college of $224,000. (I divided the cost of the program per child, $8,512, by the program’s estimated effect of increasing the likelihood of students getting 0.5 credits or more from a four-year school by 3.8 percentage points.)

So, from these data, preschool appears to be potentially more cost-effective than early elementary class size reduction in raising college attendance. However, targeting special counseling, tutoring, or financial aid on at-risk high school students appears to be even more cost-effective in raising college attendance.

What should be said about these findings? First, as the authors note, these results focus on only one possible benefit of these programs: increased college attendance. The results might be quite different if we look more comprehensively at benefits.

For example, in benefit-cost analyses of early childhood programs, there are two kinds of benefits that loom large. The first is increased adult earnings. The second is reduced crime.

With respect to the first type of benefit, increased adult earnings, a consistent finding is that the effects of early childhood programs on adult earnings go well beyond what we would expect based on effects on educational attainment. For example, in analyses of the Perry Preschool Program, based on the program’s effects on educational attainment, we would only expect adult employment rates to increase by 2 or 3 percentage points. The actual observed increase in adult employment rates is 13 to 18 percentage points (see Table 11 in my 2006 paper).

Apparently, the increase in hard and soft skills due to early childhood programs increases the employability of former participants by more than we would expect based on number of years of school completed.  Among other things, this could be due to these programs’ effects in reducing crime among former participants. Obviously a criminal record is a significant impediment to employment and earnings, holding educational attainment constant.

Therefore, an important issue in analyzing the benefits and costs of these different interventions is how their benefit-cost ratios compare looking at all benefits and costs. For example, will targeted tutoring and scholarship aid for high school students have the same benefits in increasing earnings and reducing crime rates as is true for earlier interventions? We don’t know, but this seems doubtful. For example, by the time these high school programs intervene,  it will be more difficult to make a difference for students already involved in criminal activity. It seems likely that these high school interventions are more appropriate for more targeted groups of students who are high-risk, but have in other ways well-developed “soft skills”.

This brings up a second point. I don’t think it is the case that early interventions always trump late interventions. What is true is that early interventions are often capable of having more profound effects on a broader target group through a relatively straightforward intervention.

For example, as reviewed in chapter 7 of my book investing in Kids, I think some job training programs can have quite high rates of return. But these training programs have to be carefully designed to have close ties to employers with job openings, and have to be targeted at workers who already have many basic skills.

In contrast, preschool programs are fairly straightforward, and work for many students. Preschool in essence simply adds more learning time for students. If the preschool program has a reasonable curriculum and decent teachers, and a reasonable class size, it likely can affect many students from a variety of backgrounds.

Class size reduction also is a relatively straightforward intervention that benefits a wide variety of students.  Smaller class sizes mean less class disruptions, and hence more learning time, and more individual attention, hence more effective learning time. At early ages, many students can benefit from these changes.

As people age, they get more set in their habits and character, and harder to change. You can still invest and change their skills. But not everyone will be as open at later ages to such investments. Hence, the programs need to be more carefully designed and targeted.

For example, a well designed financial aid or counseling program may significantly help some students at modest costs. But there are likely to be whole groups of students for whom such an intervention will not work, because they are no longer readily open to such an intervention by the time they are in high school.

Because later interventions may need to be more tightly targeted on particular groups of students, this may limit the size of these programs at full-scale. Even with high benefit-cost ratios, a program with a smaller target group may not have as high net benefits when operated at full scale, compared to a program that can benefit broader groups. Tight targeting and more focus on specific barriers to success may also require more careful design, which is an additional challenge to maintaining program quality.

I don’t think early interventions and later interventions should be seen as being in competition. Rather, these programs should be seen as complements. I agree with Dynarski, Hyman, and Schanzenbach that “there are cost-effective programs at every point in the educational pipeline”. Earlier programs can have more profound effects on broader groups of students. But there will still be barriers to achievement by students at later ages that can be overcome by clever late interventions.  Both early and late interventions can pay off, but in different ways.