You’ve seen the scenario on “Law and Order” many times: the defense lawyer tosses out a wild accusation that the person on the witness stand (or someone else related to the case) is the real killer – with no evidence whatsoever behind it. Jurors have now heard about an alternative suspect for the crime. The judge proclaims that the jurors “must disregard the last statement.” But, can they?
Each year almost 4 million children die within the first four weeks of life, many from preventable or treatable causes. Much programmatic aid is now devoted to devising ways to ensure that simple effective health practices, such as ensuring a more sterile birth environment, are adopted on a wide scale. A number of recent evaluations from South Asia suggest that the active involvement of local women’s groups in problem solving can be among the most cost-effective interventions to prevent deaths.
coauthored with Alaka Holla
As we argued last week, we need more results that tell us what works and what does not for economically empowering women. And a first step would be for people who are running evaluations out there to run a regression that interacts gender with treatment. Now some of these will show no significant differences by sex. Does that mean that the program did not affect men and women differently? No. Alas, all zeroes are not created equal.
Today's post comes from guest bloggers Hanan Jacoby and Ghazala Mansuri...
New York City has suspended payments in a pay-for-performance program for teachers after an experiment found the program had not worked. From the New York Post:
Another attempt by the city to improve student performance through cash payments has failed, much to the surprise of Mayor Bloomberg.
The basic principles of ethical research as laid out in the Belmont Report include “respect for persons”, which stipulates that all individuals should be treated as autonomous agents. Typically this principle is translated into practice with a statement read to all study subjects concerning the voluntary nature of study participation and the freedom to withhold consent. These ethical guidelines largely derive from medical trials where individual targeting of an intervention such as an experimental drug is typical.
co-authored with Alaka Holla
Everyone always says that great things happen when you give money to women. Children start going to school, everyone gets better health care, and husbands stop drinking as much. And we know from impact evaluations of conditional cash transfers programs that a lot of these things are true (see for example this review of the evidence by colleagues at the World Bank). But, aside from just giving them cash with conditions, how do we get money in the hands of women? Do the programs we use to increase earnings work the same for men and women? And do the same dimensions of well-being respond to these programs for men and women?
The answer is we don’t know much. And we really should know more. If we don’t know what works to address gender inequalities in the economic realm, we can’t do the right intervention (at least on purpose). This makes it impossible to economically empower women in a sustainable, meaningful way. We also don’t know what this earned income means for household welfare. While the evidence from CCTs for example might suggest that women might spend transfers differently, we don’t know whether more farm or firm profits for a woman versus a man means more clothes for the kids and regular doctor visits. We also don’t know much about the spillover effects in non-economic realms generated by interventions in the productive sectors and whether these also differ across men and women. Quasi-experimental evidence from the US for example suggests that decreases in the gender wage-gap reduce violence against women (see this paper by Anna Aizer), but some experimental evidence by Fernald and coauthors from South Africa suggests that extending credit to poor borrowers decreases depressive symptoms for men but not for women.
The thought has occurred to me that there are more people than ever doing surveys of various sorts in developing countries, and many graduate students, young faculty, and other researchers who would love the opportunity to cheaply add questions to a survey. I therefore wonder whether there is a missed opportunity for the two sides to get together. Let me explain what I’m thinking of, and then let us know whether you think this is really an issue or not.
One of the most important things while designing an intervention is to try to ensure that your study will have enough statistical power to test the hypotheses you're interested in. Picking a large enough sample is one of a variety of things to increase power. Another is block stratified randomization, of which paired randomization is the extreme.