Thursday, March 21, 2019

The Classical Liberal diagnosis trumps the postmodernist ideology in the data arena

I am deeply skeptical of social programs which seek to benefit one population over another unless they are very targeted on very specific and well documented root causes - which they almost never are. Instead, we end up with well-intended but very loose interventions to improve things for girls, or for racial minorities or for immigrants, or some other population.

The history of such interventions is littered with failures with just sufficient successes to fan the flame of social justice. And, as often as not, there are all sorts of unanticipated down-sides to such interventions. Not infrequently, the intervention intended to ameliorate perceived differences ends up exacerbating them. My general inclination is that all improvements ought to be available to all people unless we are seeking to solve a very specific, very well documented known root cause with a very specific, tailored solution.

This new report provides some support for that skepticism. From We Can Learn a Lot about Improving Girls’ Education from Interventions That Don’t Target Girls by David Evans and Fei Yuan.
So what’s the best way to help girls to stay in school and to make sure they get the most out of it? A number of previous efforts to answer that question have searched for evidence on programs that seek to keep girls in school and to help them learn. Reviews search for studies that have terms like “girls” and “education” in their titles and bring together the most promising evidence from among that literature. Seems sensible, right?

But consider an alternative: What if the programs that help the girls the most are not the programs that target girls? Imagine two hypothetical programs. One is targeted towards girls, and it finds a big impact on girls’ learning. It even finds some impacts for boys, although those are much smaller. The other program is a general intervention (in other words, it doesn’t target girls specifically). Let’s imagine that it finds even larger impacts on girls, and that those impacts are roughly the same as the impacts for boys. If the objective is to increase girls’ learning as much as possible, the general intervention is clearly the most effective. But whoever writes the evaluation of that general intervention will most likely not mention girls in the title, the abstract, or the introduction. They might not even show the learning results separately by gender, since they may not find the lack of a difference between boys and girls very interesting.

[snip]

Strikingly, the example isn’t just hypothetical. A merit scholarship program targeted to girls in Kenya led to significant gains for girls (and even had some smaller benefits for boys), while a program that facilitated parent-teacher conferences in Bangladesh had comparable impacts for boys and for girls, with larger impacts for each than the scholarships in Kenya.

Click to enlarge.
Based on this question (relative value of general versus gender targeted), the researchers go in search of data.
In the end, we looked at 179 studies of 274 total interventions to improve either access to school or learning in school. We then compared the impacts of girl-targeted interventions versus general interventions, and we looked at which interventions have proven most—and least—effective for girls.
One of their primary findings is that across the population of studies:
Girl-targeted interventions and general interventions deliver similar gains

The impact of the middle-performing (median) girl-targeted intervention improved learning by 0.13 standard deviations; the middle-performing general intervention improved learning by 0.12 standard deviations. That’s between half a year and a year of what kids normally learn in school. Even if we look at the best or worst interventions, general and girl-targeted interventions perform similarly. The same is true for increasing access. So girl-targeted interventions don’t fare any worse than general interventions, but they also don’t fare better.

Because the gains are comparable and more general interventions have been evaluated, policymakers have a longer list of tested options to draw from among general interventions. Furthermore, scaling up programs that benefit boys as well as girls may be more politically palatable, even if the primary objective is to benefit girls.
So why go to the bother of targeted solutions when a general solution is just as effective?

Another finding is globally true and globally ignored, especially in the West.
The best interventions for improving girls’ access to school reduce the cost
Again, it is not gender specific. In the US we have funneled huge subsidies from the Federal and State governments into higher education, with no clear improvement in education outcomes but with a clear increase cost. Education is among the top five categories of inflation over the past several decades. And high cost is one of the highest barriers to wider dispersion of education attainment. Subsidies for education is viewed as one of the most progressive social policies and yet its actual impact is among the most regressive.

Another key finding -
The best interventions for improving girls’ learning improve pedagogy for all children
Fearing a social justice jacobin backlash, the researchers are careful to emphasize that there are cases where sex specific strategies are appropriate, and the importance of educating girls, and the need for further research. And all of which is true based on specific and narrow circumstance.

However, the conclusion appeals to common sense, comports with the data, and is broadly true for most people under most circumstances.
But in the meantime, one of the best ways to help girls overcome the learning crisis may be to improve the quality of school for all children.
The Classical Liberal diagnosis trumps the postmodernist ideology in the data arena. As it always does.

No comments:

Post a Comment