Are Non-Cognitive Gains in Education More Important than Test-Scores?

Most educational interventions are widely considered successful if they increase test-scores -- which indicate cognitive ability. Presumably, this is because higher test-scores in school imply gains such as higher wages later on. 

However, non-cognitive outcomes also matter---a lot.

In the graph above, from James Heckman's 2007 presentation, "Investing in Disadvantaged Young Children Is Good Economics and Good Public Policy," are IQ scores measured on the Stanford-Binet Intelligence scale -- for the treatment group and the control group for the children that took part in the Perry Pre-School project. [See above, from Figure 8a in Heckman's presentation by Terman & Merill, 1960]. This study examines the lives of 123 African-Americans born in poverty and at high risk of failing in school. From 1962–1967, at ages 3 and 4, the subjects were randomly divided into a treatment group that received a high-quality preschool program and a comparison group that received no preschool program. As the graph above clearly shows, by age 10, IQ scores in both treatment and control groups looked the same.

Here are the economic effects at age 27 from the Perry Pre-School intervention below:

Source: from Figure 8c in Heckman's presentation by Barnett, 2004.

In the study's most recent phase, 97% of the participants still living were interviewed at age 40. Additional data were gathered  from the subjects' school, social services, and arrest records.

The study found that adults at age 40 who participated in the Perry Pre-School program had higher earnings, were more likely to hold a job, had committed fewer crimes, and were more likely to have graduated from high school than adults in the control group who did not participate.

While measuring and estimating the impact of educational interventions on non-cognitive and behavioral outcomes is now mainstream in the US
(see: Cullen, Jacob and Levitt in Econometrica), it has received less attention in low-income countries.

Possibilities for why Non-Cognitive Outcomes have been Measured Less in Developing Countries:

One argument are that test-scores are very weak in many low-income countries and at those very low levels of learning, improving test-scores should remain the #1 aim of the schooling system - and non-cognitive outcomes can wait until students can read and write. This argument doesn't convince me.

A new paper by Pramila Krishnan and Sofya Krutikova, entitled, "Skill Formation in Bombay’s Slums," measures psychosocial skills (self-esteem and efficacy) among poor households in Mumbai found the following:

1. Psychosocial skills directly correlated with wages and additional controls for cognitive outcomes have very little impact on the wage psychosocial skills association (see the graph below).
2. These skills can be improved through appropriate interventions---the authors' evaluation of such a program in Mumbai suggests that the “effect of the intervention on these (psychosocial) measures is between 0.75 and 0.90 of a standard deviation.” That’s an enormous increase in wages (between 10 and 20 percent) according to graph below.

Further, while cognitive outcomes tend to fade-out very quickly (see a summary of fade-out in test-scores in Do Value-Added Estimates Add Value? Accounting for Learning Dynamics), non-cognitive improvements persist and indeed may become more salient over time.

What is the Takeaway?

It's hard to say what countries should focus on. But clearly, we should be thinking a lot more about a full package of cognitive and non-cognitive outcomes rather
than test-scores alone. Let me know what you think........