I recently received a request for some information from some early childhood advocates. They were talking to a state legislator in their state. (I leave the state unnamed because I think it doesn’t matter – there are similar debates going on in many states.) This legislator believes that the Nurse Family Partnership program is an effective program. (NFP provides first-time, disadvantaged mothers and their child with nurse home visitation services from the pre-natal period to age 2.)

However, the legislator did not believe that preschool programs were sufficiently proven. One concern was that the effects of state-funded pre-k programs on test scores would fade by 3^rd grade. The legislator in particular was concerned that state-funded pre-k, by itself, would not affect high school graduation rates.

Another concern was that the Perry Preschool program, for which the legislator thought there was evidence of long-term success, was too expensive. (The Perry Preschool program, an experimental program begun in the early 1960s, provided half-day school-year preschool at ages 3 and 4 to children from low-income families in Ypsilanti Michigan. Participants have been followed through age 40. There is indeed strong evidence of large effects of Perry Preschool in increasing educational attainment, increasing earnings, and reducing crime. But the program was expensive, at an estimated cost in 2006 dollars of $17,759 per child.)

In sum, the legislator believes that preschool is either: (1) ineffective in the long-run, or (2) too expensive to be politically feasible. (Even if preschool modeled after Perry is expensive, it is still clearly economically feasible and sensible if the benefits exceed the costs. But it might not be politically feasible.)

I think it is interesting that the legislator is concerned about the research underlying the case for preschool, but is not concerned about the research underlying the case for the Nurse Family Partnership. I think the research on both is equally strong.

I wonder if part of the issue is that the NFP is much cheaper than universal pre-k. For example, in chapter 4 of my book, Investing in Kids, I estimate that at full-scale, universal pre-k would cost $14.3 billion, versus $3.7 billion for the NFP. Or, to focus on current spending, the National Institute for Early Education Research estimates state funding for pre-k at slightly over $5 billion. As for NFP, with an estimated 21,494 families in the program, and an annual cost of $4500 per family, total current spending on the program is a little less than $100 million.

In other words, it is much cheaper to make major expansions in NFP than to expand state pre-k by the same percentage. A legislator who wants to prove his or her commitment to children by increasing the funding for some early childhood program by 50%, or by saving funding for some program, would find it easier to meet that commitment with NFP than with preschool.

The early childhood advocates asked me what response might be made to the legislator’s concerns. Here is my response, lightly edited to preserve anonymity.

There is very good research on the Chicago Child-Parent Center Program’s long-term effectiveness. This program is run by Chicago Public Schools.  It has a price tag of about $5500 per school year. Some of the kids participated for 2 years and some for one year. As my book reviews (p. 146), the evidence suggests that both groups benefitted, but that there were “diminishing returns”: the benefits of 2 years are not twice the benefits of one year, although they are somewhat greater.

The latest results for CPC were published in early 2011. (Arthur Reynolds, Judy Temple, Barry White, Suh-Ruu Ou, and Dylan Robertson, “Age 26 Cost–Benefit Analysis of the Child-Parent Center Early Education Program”, Child Development, Jan/Feb 2011, pp. 379-404.)  This latest research follows CPC participants through age 26 . The program reduced the drop-out rate from 27% in the comparison group to 20% in the treatment group, which is a large effect given that we’re talking about effects 14 years later. The program reduced special education assignments from 25% in the comparison group to 14% in the treatment group. It reduced grade retention from 38% in the comparison group to 23% in the treatment group. Juvenile crime was reduced from 25% of the comparison group to 17% of the treatment group, and adult felony crime from 18% of the comparison group to 13% of the treatment group.  You can find all these stats in the paper’s Table 3.

Reynolds and his co-authors used these results to calculate benefits and costs (Table 4). The preschool program has overall benefits of about 11 times its costs. The program’s ratio of benefits and costs from the viewpoint of taxpayers, ignoring effects on participants, is about 7 to 1.

This report release in February 2011 generated a lot of PR. Here’s the NIH press release:  http://www.nih.gov/news/health/feb2011/nichd-04.htm

CPC is quite similar to many state-funded pre-k programs.  The only real difference is that most state-funded pre-k programs only provide services at age 4, whereas in Chicago, half the kids went for one year to the program, and half the kids went for 2 years.  However, as mentioned above, the evidence suggests that the benefit to cost ratio for a one year program probably exceeds the benefit to cost ratio for a two year program.

I modeled most of the simulations in my book after the effects estimated for CPC.  I was doing projections based on the Age 21 results, and the Age 26 results are generally consistent with the Age 21 results.

What about the reported fade-out for long-term effects on test scores of some state pre-k programs? With respect to this fade-out, there are two issues: (1) does this study showing fade-out have a good comparison group?; (2) do effects fade only for “hard skills” or also for “soft skills”? (By hard skills I mean the academic skills that might be measured by reading and math test skills. By soft skills I mean various social and personality skills such as the ability to deal with peers and authority figures, and self-confidence and the ability to plan and defer gratification.)

On the first issue, there is very rigorous evidence with excellent comparison groups of short-term effectiveness of many state pre-k programs in affecting kindergarten readiness. These studies use an evaluation technique called “regression discontinuity”, which is generally regarded as being as unbiased as random assignment experimentation, although with a bit more statistical noise in it. The intuitive idea here is that we give the same test to entering preschoolers and entering kindergartners, and then compare two kids who are similar age, but one is just old enough to have participated in the preschool program last year, that is they just “made the age cut-off”, whereas the other kid is starting preschool this year because they “just missed” the age cutoff last year. These studies show large effects of some state pre-k programs on kindergarten readiness in terms of test scores, in many cases much larger effects than are found on average for Head Start.

Longer-term studies of most state pre-k programs are more difficult. Many state pre-k programs use some criteria where they target students at risk.  Because of this selection, it is difficult to find a good comparison group. We would expect children selected because they have risk factors to do worse over time.  Therefore, it is difficult to determine whether any observed “fade-out” of effects on test scores is due to actual decline in preschool effects, or is due to other risk factors that have a cumulative effect over time.

On the second point, there are many studies showing that some early educational investments have fading hard skill effects over time, but their soft skill effects go up over time or at least are reflected in much better adult outcomes, for example much higher earnings. Soft skills are at least as important as hard skills for success in life.

For example, take the recent study by Harvard Professor Raj Chetty and his colleagues on kindergarten and life success. (Raj Chetty, Nathaniel Hilger, Emmanuel Saez, Diane Whitmore Schanzenbach, and Danny Yagan, “How Does Your Kindergarten Classroom Affect Your Earnings? Evidence from Project Star”) This is the study that was widely publicized as “showing” that a good kindergarten teacher was worth $320,000.

One of the interesting findings of this study is that the effects of ”kindergarten class quality” on test scores dramatically fade over time. Yet kindergarten still has effects on adult earnings later on (mid to late 20s) that are large. This is clearly shown in Figure 6 in the paper. The upper panel shows that the effects of kindergarten class quality on test scores fades almost to zero by grade 4. But the lower panel shows large effects on adult earnings. For comparison, the bottom panel shows the effects of kindergarten class quality on earnings that would be predicted based on its test score effects in each grade. What this shows is that if you just looked at test score effects in grade 4 of kindergarten quality, you would predict almost no effects of kindergarten quality on adult earnings. Yet when we look at actual adult earnings, the effects are quite large.

As another example, there’s a well-done paper by Professor David Deming at Carnegie Mellon who looks at Head Start. (“Early Childhood Intervention and Life-Cycle Skill Development:  Evidence from Head Start”). His comparison group is good: he compares siblings in the same family, with some kids in the same family who attend Head Start and some who don’t.  This comparison group is good because it controls for many family risk factors that would influence long-run outcomes for children.

Deming shows that test score effects of Head Start drop by more than half by ages 11-14. Despite this, Head Start has large effects on a variety of outcomes by early adulthood. For example, Head Start increases high school graduation rates by a little over 8 percentage points.

This fade out and re-emergence of the effects of early interventions may be because some of the aspects of “soft skills” such as self-confidence, ability to deal with peers and authority figures, and ability to plan, are not well measured by cognitive test scores. However, such “soft skills” heavily affect how productive a worker is in the labor market.

This is the argument pushed by Nobel prize-winning economist James Heckman.  His views can be found in his slides from a recent December 10 lecture, and in a New York Times article by James Warren based on that lecture.

In sum, here is how I would respond to a legislator with the stated concerns:

(1)   A high-quality preschool intervention can be achieved at a cost of around $5,000 per child, less than one-third the cost of the Perry program;

(2)    There is considerable evidence of the effectiveness of many state funded pre-k programs;

(3)    Long-term effectiveness of preschool is consistent with some fading of effects on school test scores, as success in life depends at least as much on soft skills.

I should mention that I would be interested in hearing other “frequently asked questions” about the effectiveness of early childhood programs. Maybe I can address some of these in future posts at this blog. Please post any such questions in comments on this blog post. Or email me at bartik@upjohn.org .