A lot of my  work is on test-scores. Here is a (very partial) list of what we know about test-scores, why they may be important (and why they may not). In future blogs we hope to take up each of these topics in greater detail. We cite the papers we know below.
Test-scores, Enrollment and Growth: Macro Picture
- Enrollment and growth are not related according to Pritchett’s classic paper on “Where has all the education gone”.  The missing ingredient may well be test-scores. There is a positive association between test-scores and growth as Hanushek and Woessmann argue.
- Further work by Hanushek and Woessmann suggests that these effects may be causal. Particularly interesting is the fact that home-country cognitive skills affects the earnings of immigrants to the United States.
Test-scores, Enrollment And Wages: Micro Picture
- Test scores are positively correlated with wages later in life.
- In country after country, higher years of schooling are linked to higher wages.The effect remains in studies that have examined causality. Non-labor market outcomes are also positively impacted through enrollment. Maternal education among currently adult women (who went to school 20-25 years ago) is causally linked with better child outcomes---in both education and health. This is true in the US, in India and in Pakistan. For low-income countries, the results suggest that even in poor learning environments, years of schooling matter.
Levels And Growth Of Test-Scores
- Test-scores in most low-income countries are abyssmal. Most low-income countries are one or more standard-deviations behind their OECD counterparts. At the same time, test-scores have not improved much in the United States and France since the 1960s. They must have got to those higher levels somehow, but we really don’t know how.
- Differences in cognitive outcomes build up in the early years, but then seem to remain stable through the schooling years.
- Learning fades quite rapidly. This comes from studies of summer learning loss, and recent literature on estimating test-score depreciation.
- Many randomized interventions try to improve test-scores. The majority show little increase, or a 0.2 standard deviation increase that disappears within a year of the end of the program. Most interventions that supplement school materials do not show improvements; most interventions that target teachers do. This is consistent with a large number of studies that show that teachers are the key ingredient in determining how much children learn.
- Keeping kids in schools tends to improve test-scores, although not necessarily for those who were likely to drop-out. In other words, kids who eventually drop-out are those who are not learning a thing in school.
- Longer term studies in the US also show that test-score increases disappear after a while. But the impacts of some programs on non-cognitive outcomes remain, with impressive differences in economic outcomes 20-25 years after the intervention.
Test-scores do seem to matter, both at the macro and the micro level, but improving them is proving to be a very difficult task. The improvements seen in randomized evaluations are small relative to the massive learning deficit in these countries and these improvements tend to fade away quite quickly. Having said that, more years of even the existing low-quality education leads to wage increases and improvements in other non labor-market outcomes. There is insufficient evidence on non-cognitive outcomes from low-income countries to say much.
Photo credit - the image ("Taking the test: these children in Jaura, India dligently apply themselves to the task in hand") comes from the Wikipedian named Yan  and is used according to the terms of its Creative Commons Attribution ShareAlike 3.0 License.