Underinvestment in a profitable technology Bryan et al. (2014)
Aim Why people were not already migrating despite the high returns
Context Setting: Bangladesh seasonal-famine region: incomes drop during post-planting and pre-harvest season + higher
grain prices = seasonal famine (‘Monga’ )
RCT experiment: to incentivise HH temporarily out-migrate during monga season by randomly selecting and
assigning to:
Treatment: Incentivised to migrate (Cash – 600 takas conditional on migration,Credit – cash in 0 interest)
Control: not incentivised to migrate (information – on jobs available & control: no treatment arms)
Motivation Puzzle: Spatial misallocation of labor: Labor should flow to locations to where it is most productive, until marginal
products of labor (MPL) are equalized across locations
Inter-regional variation in income btwn Rangpur & Bangladesh larger than inter-seasonal variation within Rangpur
o Seasonal out-migration from monga-prone districts (Rangpur) is low despite migration in this setting is very
profitable, and in some sense underutilized
Data Survey data: Consumption modules, income, assets, credit, and savings and migration
Identificatio ITT : effect of treatment on consumption (expenditure on food, non-food, total calories): OLS is biased downwards
n bc rich households are not very likely to migrate, while those who migrate r poor
IV - LATE regression: effect of migration on households induced by the treatment to migrate (the compliers) using
random assignment to treatment group as an instrument for migration
- X= controls household characteristics
Results Result 1: Incentive led increase in seasonal migrants + more likely to remigrate in future years. (table II)
- Row 1: Incentivised HH (cash or credit treatments) = migration rate jumps to 59% and 56.8% = incentive
induces 22% of households to send a seasonal migrant + statistically significant
- Row 2: Treatment HHs continue to migrate after incentive removed- 2009: migration 9pp higher
Result 2: incentivised Migration by intervention increases consumption (food and non-food expenditures) of
migrants’ family members at the origin by 30–35% and improves their caloric intake by 550–700 calories per day
- Table III shows this^ in panel A. Panel B: consumption in 2009 as large + significant as 2008 consumption
Overall Result 1 and 2- migration is very profitable, and in some sense underutilized
Result 3: Qualitatively, model shows risk/incomes close to subsistence/learning ab migration explain low
migration
- Volatile background risk: high income + consumption variability which leads to Buffering behaviour: buffer
stock savings = Table IV: treated earn 7451 Taka on avg and save half = high savings/expenditure ratio
- Learning: treated learn about migration & have better migration experience- i.e., there is something to
learn, or makes connections) -- earn higher earnings & informs re-migration
- Table VI: strong persistence in migration: treated remigrate in 2009 but friends’ migration unaffected =
people learn from their own experience, but do not learn from the experiences of others (migration is
experience good)
- Migration risk : (fig 3): migration carries risk of a very adverse outcome of falling below subsistence for
those close to subsistence, as theyre less likely to have savings to buffer against this risk.
- Risk of falling below subsistence is an important deterrent to migration:
- Fig 5: - those closer to subsistence less likely to migrate in control group & treatment had largest effect on
subsistence households: they the ones induced to migrate by our incentive
Result 3: Quantitively, model shows we don’t we do not fully capture HH migration choices
Other results:
- Migration risk plays a role in migrating :Table V: providing migration Insurance: offer loan conditional on
migration explicitly to rainfall conditions at Bogra: hh 30 pp more likely to migrate under the rainfall
insurance
- rule out possibility that some failed to migrate due to a liquidity constraint: migration is more responsive
to incentives (e.g., credit conditional on migration) than to unconditional credit (table V col 1)
Mechanisms risk/incomes close to subsistence/learning ab migration explain low utilisation
Policy Migration support through cash and credit treatments allows people to cope with monga famine
, implications more cost-effective than subsidizing consumption & liquidity to induce migration
should provide migration insurance since migration risk plays a role in migration
skeptical to draw policy implications: quantitative work- can't provide a satisfying explanation for why
people in Rangpur had not saved up to migrate
External Pre-harvest famine common in South Asia/SSA
validity explain other migration patterns -lower out-migration rate among poorer Europeans
Limitations model cannot quantitatively account for findings: inducing large numbers of people in villages to migrate is unlikely
does not capture long-term psychological and social effects of migration, and general equilibrium effects in labor
markets if gov were to contemplate scaling up such a program
LABOR MARKETS AND POVERTY IN VILLAGE ECONOMIES BANDIERA et al 2017
Aim How women’s labour choices in village economies correlate with poverty
- whether enabling poorest women to take same work as wealthier counterparts (livestock rearing) can set
them on a sustainable path out of poverty
Context Setting: Rural Bangladesh villages 53% identified as ultra-poor, woman are sole earner in ultra-poor HH
- ultra-poor disadvantaged relative to wealthier counterparts in same village: largely assetless – own less
own livestock(cows- key livestock asset in village economies), less agri land
Women engage in 2 labor activities—agricultural/maid labor and livestock rearing
o livestock rearing : self-employed, cows, requiring capital input, higher returns (Wealthier women)
o agriculture/maids - women hired daily without any guarantee of future employment ultra-poor
women rely on unskilled casual labor, which requires no capital input + low returns, irregular
o Table II; women bunch their work into fewer days, but work more hours overall. wealthier women
smooth their labor supply over the year
RCT Experiment: BRAC’s Targeting the Ultra-Poor program, randomised at BRANCH LEVEL
- Randomly select eligible HH from BRAC villages and assign them to receive ultra-poor
program(treatment): one-off transfer of livestock assets which produce income-generating activity &
given skill inputs: weekly visit and support among non-economic dimensions
- Valid: attrition rates + characteristics of ultra-poor are balanced btwn treatment and control
Motivation Puzzle: poor dont allocate their labor to the activity with the highest return
o Billions in extreme poverty and labour is their main endowment understanding link between poverty
and labor markets to lift them out of poverty and whether policy interventions in fostering higher return
labor activities can set them on a sustainable trajectory out of poverty
Identificatio RCT: ITT effect of TUP on ultra-poor HH using diff-diff: compares changes in outcomes among ultra-poor residing
n in treated villages before and after intervention, to changes among ultra-poor in control
- exploiting variation caused by the random assignment of villages to treatment (receives asset transfer) or
control (doesn’t receive transfer)
- unit of randomisation: BRAC branch
- identifying assumption: random assignment of treatment and no spillovers between treatment and
control villages via markets
data Labour survey: hours and days worked, wages for labor activity
Results Result 1: Ultra poor women can engage more in livestock rearing increase labour supply and raise net
earnings and smooth their labour supply over the year (table II)
- Panel A: ultra-poor allocate 415 more hours to rearing, while casual labour hours decline = beneficiaries
continue to own livestock & maintain it once assistance is removed
- Panel B: earnings from livestock rearing increase to $115 postintervention = women becoming more
productive= smooth labour supply over year
Result 2: TUP = improves financial inclusion for ultra-poor: higher savings/more likely to get loan (table IV):
Aim Why people were not already migrating despite the high returns
Context Setting: Bangladesh seasonal-famine region: incomes drop during post-planting and pre-harvest season + higher
grain prices = seasonal famine (‘Monga’ )
RCT experiment: to incentivise HH temporarily out-migrate during monga season by randomly selecting and
assigning to:
Treatment: Incentivised to migrate (Cash – 600 takas conditional on migration,Credit – cash in 0 interest)
Control: not incentivised to migrate (information – on jobs available & control: no treatment arms)
Motivation Puzzle: Spatial misallocation of labor: Labor should flow to locations to where it is most productive, until marginal
products of labor (MPL) are equalized across locations
Inter-regional variation in income btwn Rangpur & Bangladesh larger than inter-seasonal variation within Rangpur
o Seasonal out-migration from monga-prone districts (Rangpur) is low despite migration in this setting is very
profitable, and in some sense underutilized
Data Survey data: Consumption modules, income, assets, credit, and savings and migration
Identificatio ITT : effect of treatment on consumption (expenditure on food, non-food, total calories): OLS is biased downwards
n bc rich households are not very likely to migrate, while those who migrate r poor
IV - LATE regression: effect of migration on households induced by the treatment to migrate (the compliers) using
random assignment to treatment group as an instrument for migration
- X= controls household characteristics
Results Result 1: Incentive led increase in seasonal migrants + more likely to remigrate in future years. (table II)
- Row 1: Incentivised HH (cash or credit treatments) = migration rate jumps to 59% and 56.8% = incentive
induces 22% of households to send a seasonal migrant + statistically significant
- Row 2: Treatment HHs continue to migrate after incentive removed- 2009: migration 9pp higher
Result 2: incentivised Migration by intervention increases consumption (food and non-food expenditures) of
migrants’ family members at the origin by 30–35% and improves their caloric intake by 550–700 calories per day
- Table III shows this^ in panel A. Panel B: consumption in 2009 as large + significant as 2008 consumption
Overall Result 1 and 2- migration is very profitable, and in some sense underutilized
Result 3: Qualitatively, model shows risk/incomes close to subsistence/learning ab migration explain low
migration
- Volatile background risk: high income + consumption variability which leads to Buffering behaviour: buffer
stock savings = Table IV: treated earn 7451 Taka on avg and save half = high savings/expenditure ratio
- Learning: treated learn about migration & have better migration experience- i.e., there is something to
learn, or makes connections) -- earn higher earnings & informs re-migration
- Table VI: strong persistence in migration: treated remigrate in 2009 but friends’ migration unaffected =
people learn from their own experience, but do not learn from the experiences of others (migration is
experience good)
- Migration risk : (fig 3): migration carries risk of a very adverse outcome of falling below subsistence for
those close to subsistence, as theyre less likely to have savings to buffer against this risk.
- Risk of falling below subsistence is an important deterrent to migration:
- Fig 5: - those closer to subsistence less likely to migrate in control group & treatment had largest effect on
subsistence households: they the ones induced to migrate by our incentive
Result 3: Quantitively, model shows we don’t we do not fully capture HH migration choices
Other results:
- Migration risk plays a role in migrating :Table V: providing migration Insurance: offer loan conditional on
migration explicitly to rainfall conditions at Bogra: hh 30 pp more likely to migrate under the rainfall
insurance
- rule out possibility that some failed to migrate due to a liquidity constraint: migration is more responsive
to incentives (e.g., credit conditional on migration) than to unconditional credit (table V col 1)
Mechanisms risk/incomes close to subsistence/learning ab migration explain low utilisation
Policy Migration support through cash and credit treatments allows people to cope with monga famine
, implications more cost-effective than subsidizing consumption & liquidity to induce migration
should provide migration insurance since migration risk plays a role in migration
skeptical to draw policy implications: quantitative work- can't provide a satisfying explanation for why
people in Rangpur had not saved up to migrate
External Pre-harvest famine common in South Asia/SSA
validity explain other migration patterns -lower out-migration rate among poorer Europeans
Limitations model cannot quantitatively account for findings: inducing large numbers of people in villages to migrate is unlikely
does not capture long-term psychological and social effects of migration, and general equilibrium effects in labor
markets if gov were to contemplate scaling up such a program
LABOR MARKETS AND POVERTY IN VILLAGE ECONOMIES BANDIERA et al 2017
Aim How women’s labour choices in village economies correlate with poverty
- whether enabling poorest women to take same work as wealthier counterparts (livestock rearing) can set
them on a sustainable path out of poverty
Context Setting: Rural Bangladesh villages 53% identified as ultra-poor, woman are sole earner in ultra-poor HH
- ultra-poor disadvantaged relative to wealthier counterparts in same village: largely assetless – own less
own livestock(cows- key livestock asset in village economies), less agri land
Women engage in 2 labor activities—agricultural/maid labor and livestock rearing
o livestock rearing : self-employed, cows, requiring capital input, higher returns (Wealthier women)
o agriculture/maids - women hired daily without any guarantee of future employment ultra-poor
women rely on unskilled casual labor, which requires no capital input + low returns, irregular
o Table II; women bunch their work into fewer days, but work more hours overall. wealthier women
smooth their labor supply over the year
RCT Experiment: BRAC’s Targeting the Ultra-Poor program, randomised at BRANCH LEVEL
- Randomly select eligible HH from BRAC villages and assign them to receive ultra-poor
program(treatment): one-off transfer of livestock assets which produce income-generating activity &
given skill inputs: weekly visit and support among non-economic dimensions
- Valid: attrition rates + characteristics of ultra-poor are balanced btwn treatment and control
Motivation Puzzle: poor dont allocate their labor to the activity with the highest return
o Billions in extreme poverty and labour is their main endowment understanding link between poverty
and labor markets to lift them out of poverty and whether policy interventions in fostering higher return
labor activities can set them on a sustainable trajectory out of poverty
Identificatio RCT: ITT effect of TUP on ultra-poor HH using diff-diff: compares changes in outcomes among ultra-poor residing
n in treated villages before and after intervention, to changes among ultra-poor in control
- exploiting variation caused by the random assignment of villages to treatment (receives asset transfer) or
control (doesn’t receive transfer)
- unit of randomisation: BRAC branch
- identifying assumption: random assignment of treatment and no spillovers between treatment and
control villages via markets
data Labour survey: hours and days worked, wages for labor activity
Results Result 1: Ultra poor women can engage more in livestock rearing increase labour supply and raise net
earnings and smooth their labour supply over the year (table II)
- Panel A: ultra-poor allocate 415 more hours to rearing, while casual labour hours decline = beneficiaries
continue to own livestock & maintain it once assistance is removed
- Panel B: earnings from livestock rearing increase to $115 postintervention = women becoming more
productive= smooth labour supply over year
Result 2: TUP = improves financial inclusion for ultra-poor: higher savings/more likely to get loan (table IV):
