A recent paper by Nuria Rodriguez-Planas presents discouraging information on the long-term results of a program to help at-risk high school students. (The full paper requires subscription access to the journal, but working paper versions of many of the results can be found here and here.)

The paper looks at the “long-term” results of the Quantum Opportunity Program. This program provided extensive mentoring, educational services, and financial incentives to at-risk high school students, starting in 9^th grade and continuing for up to 5 years.  The long-term results are for when the participants were 24 years old. The estimated effects of QOP are based on a random assignment experiment, in which eligible students were randomly assigned to either receiving program services, or being in a control group. Because of the random assignment, we can be highly confident that the estimated effects are true causal impacts of the program on outcomes.

The upshot is that the program had short-run effects on educational outcomes, which faded over time. Females had some positive long-run effects on employment rates at age 24. But males showed increases in criminal activity and arrests at age 24.

Let me focus on one outcome, high school graduation. The point estimate is that the program increased the rate of high school completion for the entire sample by 2.1%. That is, the control group had a high school completion rate by age 24 of 61.6%, and the estimate is that the program increased high school completion from 61.6% to 63.7%. This 2.1% increase is not statistically significant, which means that we cannot reject the hypothesis that the true effect is zero, and the observed positive effect happened by chance.

This 2.1% effect on high school graduation occurred for a program that cost almost $25,000 per student.  Although high school graduation is beneficial, it seems unlikely that its social benefits are sufficient to justify that cost. For example, if we average these benefits across all program participants, the average present value of future earnings effects per program participant from these high school completion numbers would be a little under $4,000. (This multiplies the 2.1% effect times the present value of converting a high school dropout to a high school graduate of $175,234 estimated in chapter 12 of my book Investing in Kids.)

Let’s compare the cost-effectiveness of this “later intervention” with the cost-effectiveness of high-quality early childhood programs.  Let’s consider results for the Chicago Child-Parent Center program. CPC is a half day preschool program. About 55% of the preschool participants were in preschool at both ages 3 and 4, and 45% were only in the preschool program at age 4.

According to research by Art Reynolds and his colleagues, CPC increased high school graduation rates at age 28 by 6.4%. The comparison group had high school completion of 75.1%, and CPC is estimated to have increased high school completion to 81.5%.

This was done by a program whose average cost per participant was $8,512. This averages costs over one-year and two-year participants. The available evidence suggests the benefits per dollar spent are greater for one-year participants than two-year participants. That is, children participating in the program do better if they participate for two years rather than one year, but they don’t do twice as well.

Just considering high school graduation rate results by themselves, CPC clearly pays off. The 6.4% increase in high school graduation rates would be predicted to increase the present value of future earnings by over $11,000 per program participant. (This multiplies the 6.4% times the present value of converting a high school dropout to a graduate from my book Investing in Kids.) Other evidence suggests that pre-K programs often have effects on adult earnings that significantly exceed what one would expect based on their effects on educational attainment (see discussion on pp. 94-95 of Investing in Kids.) Reynolds et al estimate earnings benefits of over $22,000. And CPC also has large benefits in reducing crime rates.

But what I want to focus on here is the comparison of this early intervention of the CPC preschool program with the later intervention at 9^th grade of the Quantum program.  The early intervention, at one third the cost per participant, achieves a three times greater effect on high school graduation rates.  In other words, this preschool program’s cost-effectiveness, or “bang for a buck”, is 9 times as great as that of the Quantum program.

For the cost of the Quantum program for one student, we could serve 3 children in the CPC program. And the expected effect on high school graduation for each of these three children is three times as great as the results for the one student in the Quantum program.

Of course, this comparison is specific to these two programs. But I think it illustrates a larger point. It is more challenging to find interventions for highly disadvantaged children that work once these children get to 9^th grade.  By 9^th grade, many disadvantaged children have problems that are difficult to overcome.

There are later interventions that can work. For example, high school Career Academies, which provide more focused career education for students, have shown good evidence of long-run success.  But these later interventions are more difficult to design, and may be only suitable for a smaller proportion of all students.

If we want highly cost-effective interventions that have been proven to work for a wide variety of children, and when run at a large scale, we need to look to early childhood programs. These programs intervene early, when intervention can more easily make a difference to life prospects.