Let’s start with how the program works. Within the targeted communities, folks participate in a community wealth ranking exercise. BRAC officials then work through this group of eligible folks, excluding those who are getting government assistance or microfinance, and households that have no adult women. They focus on including households with small amounts of land, no male adult income-earner, adult women working outside the house, kids who are working and/or households with no productive assets. This should net them the poor, and it does, the households ultimately selected for the program in Bandiera and co.’s sample get by on an average of 93 cents a day.
These women (yes the program targets women) are then offered a menu of training/capital program combinations, plus a small stipend to sustain them while they get their new activity up and running. In Bandiera and co.’s sample, all of the women choose the livestock option (mostly cows, with some poultry and goats as well), which amounts to an asset transfer of around $140 (ten times the baseline livestock assets for these households). They then get fairly intensive training – first some classroom training, followed by home visits by a livestock specialist every 1-2 months, and a program officer every week, for a year. BRAC encourages beneficiaries to hang on to the new livestock, but doesn’t enforce this in any way.
Bandiera and co. set up the evaluation nicely. Working with BRAC, they randomize at the branch level. Then within both the treatment and control communities, they work through the beneficiary selection process I talked about above. This, coupled with the fact that treatment status is not revealed to beneficiaries until after the baseline gives them a nice comparison group and knocks out the chance of anticipatory behavior. They have a huge sample -- 7,953 eligible women at baseline (with 87% take up) and they also go out and survey all the ineligible poor households as well as a random 10% sample of the rest of the community for another 19,012 observations. Another neat thing is that they go back to these households twice – once at two years and a second time four years after baseline – to capture more medium term effects.
So what happens? Four years later, there is a marked shift in employment patterns. Women who participate in the program are 17 percentage points (65% of the baseline mean) less likely to be in wage employment. This is important for two reasons: 1) wage work is strongly correlated with poverty in these communities, and 2) wage work tends to be seasonal, with attendant issues for income smoothing. Rather than wage work, these women are specializing more in self-employment (which includes rearing livestock) – this jumps by 15 percentage points (and the fraction that engage in both goes up by 8 percentage points).
And they are working more hours. As might be expected, hours in wage employment goes down by 26% relative to baseline but self-employment hours go up by 92% relative to baseline (interestingly, the wage labor reduction at four years is double that of two years – suggesting that women may be holding on to their wage jobs while waiting for their self-employment activities to get off the ground). Overall, the aggregate increases – women are now working 19% more hours relative to baseline levels.
Total earnings show a concomitant increase, going up by 34% relative to baseline by year two and 38% by year four. Earnings per hour are intriguing – after two years these are not significantly different from the baseline, but at four years they are: 15% higher than baseline. This suggests productivity gains that manifest over time – but not right away.
So what does this mean for poverty? Expenditure is increasing – with larger percentage gains in non-food expenditure relative to food. Total per capita expenditure is up 7% relative to baseline by year 2 and 8% by year 4. This translates into a poverty reduction of 11% relative to baseline (a 9 percentage point drop). This expenditure boost not helps the poor close the gap with those classified as near poor in the baseline, but actually pass them.
What is particularly heartening about these changes in expenditure is that they are accompanied by increases in other metrics of longer term welfare. First, access to land, which is a “security asset” as well as a measure of social status is up relative to baseline -- by 11 percentage points for land rentals and 3 percentage points for ownership by year four. Second, savings increase 10-fold relative to the baseline. And finally, livestock holdings appreciate, moving above the level of the transfer by year four.
Taking all of the components together, this program costs about $300 per beneficiary. So not a small investment. But it shows that a well thought out and targeted big chunk of investment in individuals can have a fairly high payoff in terms of poverty reduction and putting women on a sustainable upwards wealth trajectory.