Can changing how people think about themselves and what they can do, change their employment outcomes? The answer is a subject of a neat new paper by Madeline McKelway.
McKelway is working in India, in the state of Uttar Pradesh. As in much of the world, India has lower female labor force participation in comparison with males. Indeed, in the area that McKelway is working this is 31.9 percent for women against 75.5 percent for their husbands. Are they not working because of family resistance? Or because they aren’t sure of their abilities? Or is there just a big information problem?
McKelway sets up an experiment to tease this out. She starts with two treatments. The first (“promotion”) consists of working with a large carpet manufacturer which is ramping up its efforts to train and employ women (interesting fact: this is a male dominated profession in this locale). In this treatment, both the target woman and her family members get a range of information about the job and also a six minute video which talks about the firm, but also has testimonials from women who have worked for it and a husband of one worker. In control households, folks only got job details.
The second treatment targets generalized self-efficacy(“GSE”). For those of you not familiar with this, the best way is to quote Bandura, who wrote the book on it: “perceived self-efficacy refers to beliefs in one’s capabilities to organize and execute the course of action required to produce given attainments.” This shouldn’t be confused with locus of control which is about how much you think you control things relative to fate or environmental factors determining everything. Both of these make up part of common (or at least useful) definitions of agency.
McKelway works with an NGO to build an intervention (which the research team then delivers) that helps to boost generalized (rather than domain specific) self-efficacy. This is delivered over the course of nine meetings and delivered through not only instruction but things like storytelling and group discussions. The control group for this intervention also gets regular meetings (to net out any purely social effects) but these meetings are devoted to taking a survey.
The design of the experiment is to look at the effects of the GSE treatment, the promotion treatment and the two of them together. Women whose families give them permission and who are of the age and ability to work are randomized into different groups for the training, with the randomization of promotion done at the household level. And McKelway collects a host of survey data, as well as some administrative data on these folks.
So what does she find?
First off, reassuringly, the GSE intervention increases self-efficacy. It goes up 0.178 points (on a scale of 1 to 4) immediately after the training and is still significantly higher (0.135 points) a year later. So the training did what it was supposed to do. In a neat measurement twist, McKelway shows us that this is channeled into believing one can do things and trying them: she finds that when presented with a sure payout, the GSE group is more likely to choose instead to take an incentivized puzzle task with a higher payoff for success but a lower payoff for failure. And she rules out risk taking by also offering them a straight lottery (which they don’t disproportionately take).
What does this increased self-efficacy mean in the job market? Women in the GSE group are more likely to sign up for the firms training and employment program by 5.8 percentage points (or 29.9 percent). Strikingly, the promotion intervention alone has an even bigger effect, boosting sign up by 11 percentage points. But, when she puts the two together, there is no additional effect. What’s more, the combined treatment has a lower effect than the promotion intervention alone, which might suggest some potential for backlash.
In terms of job outcomes, the GSE treatment leads to an increase of 8.7 percentage points (37.5 percent) at four months post-intervention. And promotion works as well, with employment up by 9.8 percentage points (42.2 percent). Again, offering both interventions isn’t better than offering either alone.
However, at one year post interventions, things get funky. Here the employment effects of both interventions slide back to zero. A closer examination shows a significant surge in the control group. With increased information and women getting empowered, spillovers are a distinct possibility. Alas, the design of the experiment doesn’t allow her to answer this.
McKelway crowns these results with a really nice finding – she shows that, in turn, employment seems to have increased self-efficacy. In a number of communities, jobs/training were over-subscribed, so the offers were randomized, and households were told they were assigned in a lottery. Three months after job offers women who got them had self-efficacy that was 0.166 points higher than those who signed up for the job but were waitlisted. So there’s a virtuous cycle here – higher self-efficacy increases the likelihood that women will enter the labor market and when they get jobs, self-efficacy increases again.
All in all, this is an intriguing and exciting set of results. First, her promotion intervention suggests that pushing information to the woman and the folks in her family who likely have some say over how she works can get significant and meaningful increases in female labor supply. And it’s important to note that this is not just the details of the job, but testimonials from workers and their spouses. But psychology also works – increasing women’s self-efficacy gets us a boost in employment, with this employment then boosting things further. So boosting elements of agency can lead to better employment outcomes.