Syndicate content

What happens when business training and capital programs get caught in the web of intrahousehold dynamics?

Markus Goldstein's picture
Two weeks ago, I blogged about a new paper by Arielle Bernhardt and coauthors which looked at the idea that when women receive a cash infusion from a program, they may give it to their husbands to invest in their business.
This week, we take a trip over to Uganda, where an interesting paper by Nathan Fiala looks at what happens with a small business promotion program in households where husbands and wives hide money from each other.  
Let’s start with the program setting.    This is taking place in northern Uganda and participants are randomized into a host of different treatment arms: a loan only, a loan plus training, a grant only and a grant plus training.  The loans are around $200, with a somewhat subsidized interest rate and reduced collateral.    Grants are $200 each.   The loan and grant size are about equal to average monthly revenues for these firms.  The business training was one that has been evaluated (alone) before in other contexts with not much, if any, impact on profits. 
When Fiala looks at the straight impacts in a two year follow up, there isn’t much for almost all of these treatment combinations.   There is a significant (at 10 percent) increase in an aggregate index of male respondent income in response to the grant plus training arm, and this seems to be driven by an increase in household assets.    For women, there is no significant results for their personal or total household outcomes.   However, in the vein of the Bernhardt and co. paper, the provision of a loan plus training to women does result in a significant (at 10 percent) increase in their spouse’s income (as reported by the husband only) relative to the impact when the program is given to men.  
Now for a tweak. In this follow up survey work, based on some patterns uncovered in qualitative work, Fiala and his team played a behavioral game.   The game gave men and women (who knew their spouse would be interviewed the next day) a choice:  invest 2000 Ugandan shillings into a sure, private return of 4000 or a return of 6000 that would be given to their spouse.   Sixty-seven percent of respondents chose the 4000 – to take less and hide the money from their spouse.   And when he does a regression controlling for a bunch of things to explain this choice, women are a lot less likely (26 percentage points) to decide to take the private, concealable return.   Empowerment and the number of kids also matter (with the latter differently for men and women).
In the survey, Fiala also asked the husband and wife to report on each other’s income.   Similar to other work on this, he finds that they have no clue.    The correlation between spousal reports is zero and only 47 percent of respondents get to within $20 of their spouse’s report.   So this is a world where spouses have little information on each other’s income, and offered the choice, they will pay to keep it this way.  
Fiala then uses the game result – those that hide versus those that don’t – to look at whether this makes a different in the treatment outcomes.   Now the first issue is that the game is taking place two years after treatment, so it could be actually capturing outcomes, not some fixed characteristic by which it would make sense to look at heterogeneity.   When Fiala regresses the different treatments on the game outcome, nothing is significant.  So he uses this to argue that this is a reasonable way to look at the heterogeneity of treatment.   
And now the results get more interesting.    Looking at aggregate household and main respondent economic indices, there are no significant impacts from all of the program variants for either women or men for those who hide income.   However, when he turns to those who did not hide income, there is a striking positive impact (for both indices) for men who got the loan only.  The signs on the female interaction for the loan only treatment, as well as the grant plus training are significant and negative.   So for women, a different story.  
These are the aggregate results from the indices.   But the details are interesting because they suggest the programs are having different effects in different realms (assets versus income versus expenditures).   For folks who hide income from their spouse: women who get the grant plus training show higher household assets (at 10%) and household income, as well as individual income (as reported by the entrepreneur – true to form, the spouse coefficient is negative but not significant). 
For folks who do not hide income, there is a lot more going on.   First, women who get the grant plus training now show a range of negative effects relative to men (who show positive and significant effects in household assets).   The loan alone is positive and significant for men in terms of household assets.    On the other hand the females have a negative and more than offsetting effect on assets, plus a negative coefficient on household expenditures as well. 
Fiala sums up the policy implications nicely: “These results present mixed news for policy makers. Women and men can benefit from programs that deliver capital, but only under certain conditions within the household.”   He’s given us some ideas what those conditions might be and I am excited to see us learn more!


Submitted by Anonymous on

Excellent study highlighting the impact of dynamics of the private sphere on socio-economic outcomes. Thank you for sharing

Add new comment