In some of my recent presentations, I have been using a slide that shows how preschool’s effects, even when they fade in test score effects as students progress in K-12, can re-emerge even stronger in effects on adult outcomes.  (For example, see slide number 4, “Re-Emergence of Pre-K’s Effects Suggests Importance of Soft Skills”, in my presentation to the American Chamber of Commerce Executives convention in August 2011.) I want to explain the research behind this slide, relate it to other research, and explain why this research result is important.

This slide uses results from the Chicago Child-Parent program, specifically research led by Art Reynolds. It also relies on research by Professor Raj Chetty of Harvard and his colleagues.

I use research by Reynolds on the “effect size” of the CPC pre-K program on test scores in early and later grades. “Effect size” is researcher jargon for measuring effects on test scores relative to how much test scores typically vary across individual students (the “standard deviation”), which roughly gauges the size of the program’s effects on test scores relative to how difficult it is to change test scores due to other influences. In summarizing test score effects from Reynolds’s research, I find that the average test score “effect size” of the CPC pre-K program fades by about 40%, when comparing averages across various tests in kindergarten and first grade versus averages in grades 2 through 6.  This fading of test score effects over time is consistent with meta-analysis of numerous preschool studies by Steve Barnett.

I then translate these effects on test scores into predicted effects on adult earnings. These predicted effects rely on research led by Chetty on how test score predict adult earnings. It turns out that Chetty’s results imply that a given increase in test scores in “effect size” units has somewhat greater effects when it occurs at later grades than at earlier ages. As a result, the predicted effects of the CPC preschool program on adult earnings based on 2^nd through 6^th grade test scores, versus kindergarten through 1^st grade test scores, fades by somewhat less, by only one-third.

Specifically, based on the estimated effects of CPC’s pre-K program on kindergarten and 1^st grade test scores, we would predict adult earnings to increase by 8.1%. Based on the estimated effects of CPC’s pre-K program on 2^nd through 6^th grade test scores, we would predict adult earnings would increase by 5.3%, which represents a fading in predicted effects by about one-third.

However, when we look at actual effects of CPC’s pre-K program on adult earnings, based on research by Reynolds and his colleagues, we find the percentage effect on adult earnings, based on actual adult earnings data, is estimated to be 7.3%. The initial test score effects in kindergarten and first grade are better predictors of the program’s long-run effects than the more faded test score effects in later elementary school.

This fading and re-emergence of early intervention’s effects is also found in other research.  For example, Chetty and his colleagues find a similar fading and re-emergence for the effects of something they call “kindergarten class quality”. For example, in Figure VI of their paper, a specified improvement in kindergarten class quality would be predicted to increase annual adult earnings by $600 per person based on end of kindergarten test scores, but effects fade to less than $200 per person based on later grades’ test scores. However, in adulthood, adult earnings effects of kindergarten class quality are close to what was predicted based on end-of-kindergarten test scores.

Deming’s research finds a similar fading and re-emergence of Head Start’s effects.  (See his 2009 paper on Head Start, linked to on his home page.) His research finds significant effects of Head Start on test scores at ages 5 and 6. These test score effects decay by two-thirds, in effect size units, by ages 11 to 14, and are no longer statistically significant at these later ages.  However, effects on young adult outcomes, at ages 19+, are even greater in effect size units than was estimated at ages 5 and 6. (See panel A of his table 4 in his 2009 Head Start paper.)

What’s going on here? One plausible hypothesis is that this re-emergence of early intervention’s effects is related to these programs’ effects on “soft skills”. By “soft skills” I mean skills such as how well a student relates to fellow students and to the teacher, the student’s ability to plan and persevere and exert self-control, and the student’s self-confidence. These soft skills, as well as hard skills, may be developed in preschool and other early interventions. These soft skills may then lead to further soft skills and hard skills development later on in K-12, and lead to greater success in adulthood.

However, this soft skills development may not be fully reflected in results on most tests, which tend to emphasize “hard skills”. (By “hard skills”, I mean whatever is measured by reading, math, and other academic tests.) This does not mean that how students do on hard skills is not important. It simply means that looking at effects on hard skills alone understates the long-run effects of early intervention.

The significance of this research is twofold. First, this research points to the importance of soft skills development in early intervention programs. Second, this research suggests that a complete evaluation of early interventions must go beyond effects on academic test scores to effects on life outcomes such as high school graduation, college attendance, and success in the workplace.

This research should not be interpreted as meaning that hard skills development in preschool is not important. I suspect that an attempt to divorce hard skills from soft skills is probably a mistake. Soft skills development helps encourage hard skills development. And hard skills development is one important way in which students actually develop the self-confidence and other soft skills that they need. A balanced curriculum must develop both hard skills and soft skills.