Readings MT Flashcards
(Tackling youth unemployment in Uganda Alfonsi et al)
Special setting?
Misallocation of talent if the poor face barriers to education and training; youth unemployment key policy issue. Set in Uganda, where 60% of unemployed are under 25.
Types of interventions: vocational training (supply side), apprenticeships (demand side).
Firms in Uganda typically very small and lots of unemployed workers.
(Tackling youth unemployment in Uganda Alfonsi et al)
Theory?
Firms get ++ payoff from hiring a worker (output) but must pay a cost (wage). The probability that they will receive a payoff depends on the skill of the worker: lower skill workers raise the risk that the firm has to pay the cost but does not receive the payoff.
Apprenticeships insure the firm against this risk by paying the wage if the firm agrees to hire and train the worker. Apprenticeships benefit the firm more because they can teach non-transferrable skills and retain the worker for a lower wage
Vocational training makes it more worthwhile for the firm to hire the worker by providing him with better skills that increase the payoff. Vocational training benefits the worker more because they get transferrable skills and can bargain for higher wages each time they change the employer.
(Tackling youth unemployment in Uganda Alfonsi et al)
Empirical design?
Two-sided experiment involving both workers and firms which allows to compare supply and demand side interventions – vocational training (VT) and firm-provided training (FT) through apprenticeships respectively. As both interventions are fielded in the same setting we can directly compare their impacts on worker outcomes.
Supply-side subjects: disadvantaged youth entering the labour market.
Demand side subjects: small and medium size enterprises in both manufacturing and service sectors. Workers assigned to VT were randomly assigned into two treatments: 1) the first group completed their six months of training and then transitioned into the labour market; 2) the second group were matched to firms operating in the same sector as the worker had been trained in, and in the same region.
Workers not offered VT were randomly assigned as follows: (i) matched to firms; (ii) matched to firms and those firms offered a wage subsidy in order to hire the worker and train them on-the-job for six months; (iii) held as a control group.
Randomisation of matching: firms are allocated randomly to receiving workers from any of these treatments. As workers are observationally equivalent only at the point of application to vocational training, we mostly present ITT estimates for worker outcomes based on random assignment to treatment at the point of application. The job ladder model to pinpoint the exact mechanisms driving the results.
Assumption: if VT workers have more certifiable skills than FT workers, and their balance of skills is more tilted towards sector – rather than firm-specific skills, they will more quickly move up the job ladder and their wage profiles will diverge away from FT workers.
(Tackling youth unemployment in Uganda Alfonsi et al)
Data?
156 employed workers, 40 firms in each labour market. 13% attrition rate. Match an average of 8 workers per market.
(Tackling youth unemployment in Uganda Alfonsi et al)
Key findings?
- FT more likely to report being firm trained whereas VT no change: firms are less willing to train workers that have already been vocationally trained in sector specific skills.
- The ITT estimates show that VT workers and FT workers all report being significantly more likely to have relevant skills than those in the control group, but the figure is higher for VT workers.
- VT workers have measurably higher sector-specific skills and indeed report their skills to be transferable across firms; FT workers report having more firm specific skills; workers transitioning into the labour market through VT and FT training routes perform very different tasks in firms as measured even years post intervention.
Overall effect:
- FT and VT workers are more likely to be employed over the control group
- VT and FT workers have ++ hours worked
- both groups earnings rise but more for VT: this is driven by them having more stable employment and working more months over the year.
- Among those actually hired by the firm, the share of FT workers who were employed at the matched firm for at least 6 months is 57%. Hence for the majority of hired workers, retention with the firm lasts longer than the period of the wage subsidy itself, suggesting these firms were constrained to begin with.
- Workers are far more mobile: when unemployed they get back onto the job ladder more quickly than control
- FT workers tend to end up in firms with lower productivity than VT
(Tackling youth unemployment in Uganda Alfonsi et al)
Conclusions?
Benefit-cost ratio is above one for both interventions however if we were to start from scratch, the wage subsidy intervention would just break even & costs for VT would still need to more than double
FT skills give job right away but VT skills more portable. Over time, FT employment rate converges to control group but VT workers employment increases over time. FT workers do well initially.
Reason why firms/workers don’t invest in training is still unclear - maybe due to credit constraints. For policy implications, providing better training can be a starting point.
(Tackling youth unemployment in Uganda Alfonsi et al)
Things to consider in experiment?
- when we instrument for take-up by the assignment, the labour market outcome for VT and FT are actually more similar
- basically a lot more people took up VT
- the gap is driven by firm rather than worker - firm interest is key limiting factor - many firms don’t want to grow as they are usually ran by self-employed people who are just waiting to find a job so they don’t want to expand
(Beyond GDP, Jones et al)
Special setting/Motivation?
GDP is a flawed measure of economic welfare as many factors affect living standards within a country that are incorporated imperfectly if at all, in GDP.
Therefore we want to compare living standards of people in different countries taking into account welfare measures (e.g. consumption, leisure, inequality and mortality using the standard economics of expected utility.
(Beyond GDP, Jones et al)
Theory?
Comparing welfare across countries using a common spec for preferences
Utilitarian expected utility calculation gives equal weights to each person
The welfare metric could be an equivalent variation (what is the proportion to adjust to equal welfare)
or it could be compensating variation (by what factor to increase to raise welfare)
these two methods produce different results especially for poor countries and the paper in particular reports the equivalent variation
(Beyond GDP, Jones et al)
Empirical design?
Calculate consumption-equivalent measure of welfare to see what proportion of consumption in one country, given its values of these factors, would deliver the same EU as values in another country and make an individual indifferent between the two countries.
Utility from leisure: constant Frisch elasticity of labour supply (hold marginal utility of consumption fixed, the elasticity of labour supply with respect to wage is constant)
(Beyond GDP, Jones et al)
Data?
Micro data (Household surveys ) - 13 countries welfare provide publicly available multi-country datasets to construct some of these welfare measures omitted: morbidity (disease prevalence), the quality of the natural environment, crime, political freedom and intergenerational altruism
(Beyond GDP, Jones et al)
Key Findings?
GDP +ves:
- excellent indicator welfare across broad range of countries: correlation with welfare of 0.98
- difference still can be important - deviation is big - welfare seems dispersed than income
How do we reconcile the large deviations w the high correlations:
- scales are so different
- income varies by more than factor of 63,000 % in our sample whereas deviations are 25-50%
Average Western European living standards look v close to US when taken into account life expectancy, leisure, lower levels of inequality. Same w non-European being poorer bc of these same reasons.
Life expectancy v important to welfare growth
(Beyond GDP, Jones et al)
Conclusions?
Developing countries should focus on inequality issue
Developing countries not just to focus on boosting incomes but also health, life expectancy directly.
(Global Inequality of Opportunity. Milanovic)
Setting?
Less than 3% of world’s population live in countries where they weren’t born. More than 2/3rds global inequality between individuals is due to national income differences.
(Global Inequality of Opportunity. Milanovic)
Theory?
Income can be written as a fn of country-specific circumstances, own-specific circumstances whose effort also depend on country, person’s own effort and a random shock that can also be called luck
(Global Inequality of Opportunity. Milanovic)
Empirical design?
Every country divided into 100 groups of equal sizes to compare positions (e.g. 2nd percentile in China w/ 70th in Nigeria) and this allows us to define income classes in the same way across countries. Income within all percentiles except highest is homogenous (i.e. within all percentiles it doesn’t vary too much its all a similar amount)
Regression:
- annual ave. household per cap income in PPP is regressed on country’s GDP per cap PPP and inequality in income distribution
- both variables on RHS strictly exogenous to an individual effort
2 approaches about taking into account pop size:
- individual viewpoint (size is irrelevant) and world as it is (accounting for size)
To test robustness: replace GDP per capita with ave. number of years of education of population over the age of 15
To examine location premium:
run similar regression but with person’s own income ventile (each ventile contains 5% population ranked from poorest to richest) - for each ventile separately, regress ventile income on country’s GDP/capita and Gini coefficient
(Global Inequality of Opportunity. Milanovic)
Data?
2008 data from 118 countries’ household surveys representing 94% of world’s population and 96% of world dollar income
(Global Inequality of Opportunity. Milanovic)
Key findings?
Base case regression 1:
- elasticity of own income wrt to country’s GDP per cap is 0.866
- gini coefficient enters with a negative sign (i.e. more unequal country on average reduces one’s income)
- overall explains a lot of the variability of individual percentile incomes across the world
- the increase of country’s average educational level by 1 year + of schooling associated with huge increase in individual incomes
Regression 2 with ventiles:
- take all people in given ventile of country’s income distribution - some 90% of variability of incomes will be explained by GDP/cap and Gini coefficients of the countries where they live
- locational premium holds for everyone but premium is less for those in lower ventiles of income distribution
- more than 1/2 variability in income globally is explained by circumstances given at birth
(Global Inequality of Opportunity. Milanovic)
Conclusions?
Benefit of higher mean income is proportionately greater for the rich classes so policy implications should target within country inequality
(Returns to Capital Microenterprises. De Mel)
Special setting?
Sri Lanka small and informal firms
3 areas in Sri lanka
If returns to low level capital stock are low, individuals without access to capital will face big disadvantage (poverty trap) but if they’re big they could grow so experiment tests this
(Returns to Capital Microenterprises. De Mel)
Empirical design?
RCT to identify effect of cash /in kind investments on profitability of enterprises irrespective of if they apply for credit at market interest rates.
Examine heterogeneity of returns to test which theories can explain why firms have marginal returns well above i rate
Randomly allocate 4 types of grans (2 smaller of each and 2 bigger of each), randomization stratified within district and zone and there’s control group
(Returns to Capital Microenterprises. De Mel)
Things to be worried about in the experiment?
Random allocation of grants to ensure changes in capital stock not associated with entrepreneurial ability, demand shocks, and other factors affecting profitability
Treatment influences hours work not j capital stock so validity assumption violated - therefore adjust profits to reflect
LATE is weighted ave. of marginal returns to capital w/ marginal return to each firm weighted by how much that firm’s capital stock responds to treatment
(Returns to Capital Microenterprises. De Mel)
Key findings?
- all 4 treatments = higher capital stock, using logs (dampens effect of outliers), still +ve effect on capital stock, more able owners have larger impact
- shock associated with ROR 5.85% and these treatment impacts appear to be flat/decreasing - no evidence of increasing returns
- more treatment = more hours
- cash better than in-kind
- heterogeneity of returns supports the view that high marginal returns from treatment reflect credit constraints rather than missing insurance markets
(Returns to Capital Microenterprises. De Mel)
Conclusions?
Outcome of interest is profits - likely to be under/overreported in response to treatment have to think abt this.
Policy: need to understand how these entrepreneurs make investment decisions - why don’t they take advantage of high ROR (i rate maybe)?
(Profits of Power, Goldstein)
Special setting?
Connection between explicitly negotiable property rights over land to agricultural investment and agricultural productivity.
Set in Ghana where lots of land is farmed under shifting cultivation so fallowing (permit soil fertility to regenerate), remains most important investment in land productivity
During fallow period, one’s rights over plot can be lost
(Profits of Power, Goldstein)
Theories behind fallowing and property rights?
Cultivators’ fear of expropriation or loss of control over land on which investments have been made might deter such investment. (ADAM SMITH)
Access to credit may be hindered if property rights not sufficiently well defined for land to serve as collateral for loans.
Inability to capture potential gains from trade in improved land might reduce investment incentives.
(Profits of Power, Goldstein)
Empirical Design?
Examine: effect of individual’s position in local/political/social hierarchies on his/her fallowing choices on plot, conditional on plot characteristics and household fixed effects.
Instrument: security of tenure over that plot - however this can’t be observed so they collect set of variables that represent cultivator’s position in hierarchies that might influence their tenure security and these are the IV’s
Estimate: determinants of duration of last fallow period treating current wealth as endogenous, using occupational background of parents as instruments for wealth.
(Profits of Power, Goldstein)
Findings?
Women vs men:
- women much lower profits than husbands: poss explanation is that women farm plots of lower exogenous quality however they are even lower on plots within 250m from each other so not likely
Officeholders?
- fallow longer than others even within same household
- estimate is stronger for inherited offices than non-inherited but not stat sig
- not j because of position but because through their position they are able to secure rights over plots that they obtain through the explicitly political process of land allocation
(Profits of Power, Goldstein)
Conclusions?
- concern for unobserved variation in exogenous plot characteristics (are they correlated with social/political status then IV is inconsistent)
- —> i.e. could be that officeholders get land that is better than average is higher on these plots
- removing fallowing inefficiencies in Ghana could increase national GDP level and reduce poverty therefore policy would be to improve security of land tenure so ++ investment in land fertility
(Labour markets and Poverty in Village Economies. Bandiera)
Special setting?
Rural Bangladesh - v v poor
Wants to see if asset transfers can mean more productive occupational choices and better growth trajectory
Poor are engaged in unproductive activities w things stopping them from doing productive activities: talent, preferences, constraints
For women: choice is v limited - maid services, agricultural labour, live-stock rearing
Inequality in asset holdings is much more marked than inequality in consumption.
(Labour markets and Poverty in Village Economies. Bandiera)
Empirical design?
RCT average treatment effects
Random assignment of some villages to treatment and control.
Randomization at branch rather than village level to mitigate spill-overs between treatment and control villages either through markets or program officers.
Treatment: women presented with menu of assets and given 1, also program provided health support and training in legal, social and political rights.
Partial population experiment: all ultra-poor and near-poor households and 10% middle/upper class so that we can see indirect treatment effects on ineligible households at diff points of wealth distribution - how they close in on the next class. Program induces ultra-poor women to take on same work activities as richer women (more competition now for the richer women) - allows us to estimate treatment effect on gap between wealth classes - so distributional consequences of intervention
Difference in differences:
Compares changes in outcomes among ultra-poor residing in treated villages before and after intervention, to changes among counterfactual ultra poor in control villages in same sub-district.
To create counterfactuals used QTE (quantile treatment effects) :
- test for heterogeneous effects (innate abilities etc)
(Labour markets and Poverty in Village Economies. Bandiera)
Key findings?
Labour activity choices;
- transforms for ultra-poor women - more hours to livestock rearing and fewer to agricultural labour and maid services, shift is permanent too
Hours worked:
- positive effect - before were idle and now more productive
Asset and skills transfer basically relax the barriers to entry for the productive labour activities.
Earnings:
- higher than counterparts in control villages and consumption expenditure higher
Assets:
- increase in value of assets owned
(Labour markets and Poverty in Village Economies. Bandiera)
Conclusions?
Reallocations of labour supply and diversification of asset base espec through accessing land sets poor on more productive trajectory
External validity - similar experiments done in other countries and similar results
Limitations:
- assets at baseline may be correlated w specific traits - those closer to threshold may be inherently different to those further away
- can’t distinguish between asset transfer and asset specific training
Policy: programme only effective if it brings poor people above the threshold
(Nominal Wage Rigidity in Village Labour Markets. Kaur)
Setting?
India.
Labour is mostly casual labour in agriculture.
Rainfall generates shock to production. Wages are negotiable between worker and employer so reflects market conditions.
Survey results show nominal wage cuts are seen as unfair.
(Nominal Wage Rigidity in Village Labour Markets. Kaur)
Theory behind wage rigidity?
- rigidities prevent wages fully adjusting to shocks
- if wages don’t fall with negative shocks this may increase layoffs deepening impact of recessions
- labour rationing generated by rigidities could give rise to disguised unemployment (employed but are redundant - aren’t utilized) or forced entrepreneurship
(Nominal Wage Rigidity in Village Labour Markets. Kaur)
Empirical Design?
Small economy with decentralized wage setting and exogenous product prices.
Rigidities arise because workers view nominal wage cuts as unfair and retaliate to such cuts by decreasing efforts.
Identification strategy: response of wages to transitory shocks generated by rainfall in current and following period.
Assumptions: rainfall shocks are transitory - monsoon rainfall affects TFP in current year but doesn’t directly affect TFP in Future years - shocks uncorrelated with other determinants of wages; shocks are serially uncorrelated.
Ratcheting: when a positive shock one year is followed by non-positive shock next year, wages do not go down.
(Nominal Wage Rigidity in Village Labour Markets. Kaur)
Key findings?
- wages adjust upwards but not downwards - relative to no shock, wages are higher if there is a positive shock and no lower if there is a negative shock
- ratcheting effect present - wages stay at higher level after positive shock and employment just falls
- inflation mitigates rigidity and ratcheting
- when inflation is higher, nominal and real wages are lower under negative shocks than if there is no shock - allows for downward adjustment
After positive shock:
- landless experience largest decrease in wages
- ratcheting distorts firm size and misallocates labour
Rigid wages due to fairness norms: workers perceive wage cuts as violation of fairness norm, employers fear workers reduce effort in response to wage cut and inflation-induced wage cuts not seen as unfair.
(Nominal Wage Rigidity in Village Labour Markets. Kaur)
Conclusions?
2 concerns:
- violation of assumption that shocks are transitory (Rainfall improves soil and leads to higher output in current and following period), ,and this would explain ratcheting. But then employment should also be higher next period and inflation should not reduce persistent effect
- rainfall may affect labour supply or demand through other channels (OVB) bc of migration or capital accumulation
External validity - in other settings wages less reflective of market conditions so fairness norms could be context specific
- in other places even more like this eg OECD subjects and their unions, collective bargaining
Policy: poorest and most vulnerable bear most of labour market effect. Have to address unemployment that comes from rigidity: fiscal policy (employment programs), monetary policy (allowing inflation)
(Who becomes an Inventor. Bell, Chetty)
Setting?
Patent grants from 1996-2014 and applications from 2001-2012 as well as federal income tax returns covering US population from 1996-2012
Patent data linked to tax data by inventor name, city and state at the time of patent application.
Based in USA. Followed potential inventors from their conditions at birth, growing up and finally labour market.
(Who becomes an Inventor. Bell, Chetty)
Empirical design?
Dependent variable is whether the child grows up to be an inventor.
Then you have measures of ‘exposure’ to innovation: whether parent was inventor, how innovative industry where parents work, neighbourhood characteristics. For all 3, they examine TYPE of innovation by exploiting the information on patent technology class.
Measure ability using data on test scores for all children in NYC public schools and how parental income affects these. Then looks at patent rates vs 3rd grade test scores - more patents with higher ability.
Then they look at factors at birth - race and ethnicity, gender. How does income and citations vary with characteristics.
(Who becomes an Inventor. Bell, Chetty)
Key findings?
Low-income parents effect:
- less likely to be inventors than high income counterparts: as are minorities and women
- decomposing this using 3rd grade test scores indicate this income difference meant diff human capital acquisition
3rd grade math scores:
- positive relationship with becoming an inventor for both rich and poor children however for children in top decile, rich kids have a higher invention rate than poor kids
- no difference between genders however there doesn’t seem to be gender gap with ability - just income
- amongst the most gifted, those of minorities have lower inventor rates
Older scores:
- as kids get older, test scores account for more of the gap
Wealthy family and innovative college:
- parents being inventors greatly increases inventor chance
- more likely to invent in same tech class as their parents if parents are inventors
- if workers in your parent’s industries are inventors, you’re more likely to be inventor
- growing up in inventive neighbourhood, more likely to be inventor
(Who becomes an Inventor. Bell, Chetty)
Conclusions?
Lack of opportunity for those who’ve had comparative advantage in innovation but for their financial position leading to loss of innovation and output - misallocation of talent - space for policy makers to remove such barriers.
Supply side educational policies:
- increase rate of entry into innovation for children from low-income families
Redistribution
- changing marginal tax rates on high incomes would not substantially affect occupational choice to become an inventor
(Tax Evasion and Inequality. Zucman)
Theories?
Tax avoidance vs evasion
Avoidance we care about because it means current laws are ineffective and need to be changed. High tax evasion says we are having a difficulty enforcing the law.
Poor people more likely to evade taxes as they are more informal/self-employed and can fly under radar.
Wealthier individuals also evade more taxes because they face much higher tax burdens as they have specialised services.
(Tax Evasion and Inequality. Zucman)
Measuring tax evasion how to do so?
Random audits perhaps but they aren’t satisfactory because of small sample size and they ignore sophisticated forms of evasion involving intermediaries.
(Tax Evasion and Inequality. Zucman)
How does paper measure evasion?
3 pieces of data:
- admin records on income and wealth in Norway and Sweden and Denmark
- leaked data on assets in tax havens (HSBC switzerlandd, panama papers)
- tax amnesty records of individuals who voluntarily declared previously hidden assets
(Tax Evasion and Inequality. Zucman)
Empirical design/how do they use the data?
- HSBC switzerland was leak so random event
- 520 individuals who declared themselves taxable in Scandinavia where matched to admin data
Panama paper was also a leak - list of owners of shell companies. Here, they were only able to match data to Norway ad Sweden and sample size further reduced by difficulties linking shell companies to ultimate owner. Difficult to disentangle legal vs illegal shell company usage.
Tax amnesty: challenge is who selects into amnesty and declares their hidden wealth.
(Tax Evasion and Inequality. Zucman)
Key findings?
Hiding assets:
- prob of hiding assets offshore rises sharply and significantly with wealth
- if assets have been hidden, fraction of one’s true wealth hidden abroad is very high and does not vary at wealth level
- extreme concentration of who holds wealth in these tax havens and more concentrated in Panama than HSBC
- 14% declared assets in tax amnesty and around 1/3 of their assets were previously hidden
(Tax Evasion and Inequality. Zucman)
Conclusions?
At the top, majority of people’s income comes from wealth so when people are hiding 40% of wealth its close to 40% of income whereas less wealthy evaders will account for less of their total bill.
This means personal tax rate paid by wealthiest is substantially lower than implied - regressive at the top.
External validity?
Scandinavian countries have some of the strongest respect for rule of law and some of highest ‘tax morale’ so this is the lower bound.
In countries with more self-employed including developing countries, size and distribution of tax evasion is likely to be very different. Where rule of
(Underinvestment in profitable technology. Bryan, Chowdhury)
Setting?
North western Bangladesh were 5 million people live below the poverty line and cope with regular pre-harvest seasonal famine.
Lack of job opportunities and low-wages during pre-harvest season and the coincident increase in grain prices.
However, there is inter-regional variation in income between Rangpur and rest of Bangladesh.
(Underinvestment in profitable technology. Bryan, Chowdhury)
Theory?
Monga season leads to lower employment opps and higher prices. Urban areas offers earning opps even during monga season. Therefore temporary migration of fam member allows household to increase consumption.
Why people don’t migrate and cause Poverty trap:
- high risk, everything is uncertain about getting earnings elsewhere
- migration failure v costly
- no info about gains from migration
Therefore inducing the inaugural migration by insuring against bad outcome would lead to long-run benefits
(Underinvestment in profitable technology. Bryan, Chowdhury)
Empirical design?
Random selection of 100 villages, conduct village census.
Then select 19 households from each village.
Treatments: cash (conditional on migration), credit, information and control.
(Underinvestment in profitable technology. Bryan, Chowdhury)
Key findings?
Control group
= about 1/3 providing info has no effect but cash and credit increased rate
Migration benefits:
- increased food and non-food expenditure
Long run effects:
- those treated continue to migrate at higher rate in next seasons
Low utilisation because of risk, incomes close to subsistence, learning about info.
(Underinvestment in profitable technology. Bryan, Chowdhury)
Identification strategy?
We want to show it’s not failure to migrate due to liquidity constraint by showing migration is more responsive to incentives (credit conditional on migration) vs unconditional credit
(Underinvestment in profitable technology. Bryan, Chowdhury)
Policies to help migration?
- hard to draw it from this paper
- policies encouraging temporal migration better than those subsidising consumption through Monga season
- connecting rural and urban areas could be important
- incentive and insurance policies
(Cultural proximity and loan outcomes. Fisman)
Setting and theory?
India
Asymmetric info model:
- cultural proximity means lenders have better info about borrowers reducing cost of communication
- less defaults and lower cost of capital as bad borrowers can be screened out and repayment enforced
- more variation in loan amounts (targeted to the borrower)
Favouritism model:
- means lenders derive utility from favouring their group
- misallocation - divert at expense of profit
- worse loan outcomes in long term
- lower default rates in short term (ever greening - rolling over of bad credits), but no persistent effects in long run when in-group officer removed
(Cultural proximity and loan outcomes. Fisman)
Empirical design?
Difference in difference estimator:
- dependent variable: loan outcome of group in branch-quarter relative to overall outcome in same branch
- errors clustered at branch level (serial correlation in lending)
Identification approach:
- rotation policy induces variation in the matching between officer and borrower group identity that is plausibly uncorrelated with demand for credit
- changes in cultural proximity affect loan outcomes
- changes that don’t involve cultural proximity do not affect loan outcomes
Fixed effects:
-group branch to control for time invariant characteristics
Reverse causality - one group thrives at one location and an in-group loan officer is sent there.
(Cultural proximity and loan outcomes. Fisman)
Findings?
Officer assigned to branch of same group:
- new loans to borrowers increases
- increases new loan recipients and more dispersion of loan size
- better repayment performance ex-post
- improvements are persistent (default stays low) - inconsistent with favouritism theory and ever-greening
Heterogeneity between groups - cultural proximity stronger for religion than caste, stronger for minority religions.
(Losing prosciality in quest for talent? Bandiera)
Setting?
India, creation of new community health workers post.
Community health workers are located in rural areas where measurement of performance v difficult and they work where they are recruited (treatment effect in district)
(Losing prosciality in quest for talent? Bandiera)
Empirical design?
1st stage - open selection channel:
Randomly allocated villages to treatment and control. Treatment (job posting advertises career prospects, stresses civil service identity). Control (job post usual stressing social identity)
Spill-over is minimised at randomising at district levels where information unlikely to flow
2nd stage:
- closing incentive channel, providing same actual incentives to all
Performance:
- effect of career incentives on each step of causal chain: inputs, facility utilisation, health outcomes
(Losing prosciality in quest for talent? Bandiera)
Identification assumption?
Treatment and control differ only in information at application stage and no different incentive-structure on the job.
Violated if control group works harder/is demoralised once they find out about career incentives.
(Losing prosciality in quest for talent? Bandiera)
Findings?
Average health worker is more talented and less pro-social. Treatment attracts higher cognitive skills and career motivations. Effect on pro-sociality is 0 for top ranked candidates but negative for lower ranked : the ‘infra-marginal candidate’ will be less pro-social.
Most talented health worker:
also most pro-social
Selection mechanisms make a difference - panels choose most talented applicants in both treatment and control and so choosing most talented will also pick most pro-social.
Treatment workers outperform control group
(Losing prosciality in quest for talent? Bandiera)
Policy Implications?
Welfare:
- promote health worker can lead to welfare loss if they can’t be replaced
- output in other sectors
Career incentives improve performance through self-selection of applications.
Public health jobs shouldn’t keep rewards low to screen out individuals without altruism
External validity:
- if economy is at full employment, attracting health workers always removes tem from another sector
- in economy more based on agriculture, taking away talented workers may be welfare decreasing
(Incentives work: getting teachers to come to school. Duflo)
Setting?
India, NGO runs experiment to reduce absenteeism.
Schools taught by para-teachers - flexible contracts, limited tenures, may be lower quality. Ideal as these can be financially incentivised and face risk of being fired.
(Incentives work: getting teachers to come to school. Duflo)
Empirical design?
RCT
Treatment: schools receive cameras with instructions and financial incentives.
Control: random checks and flat salary.
Absenteeism: regress no. of days school is open on treatment.
Performance: regress test scores on treatment, control for previous performance.
Measurement error - compare effect of attendance on scores in different specs - OLS spec w random checks data at baseline, at end line, with camera data.
(Incentives work: getting teachers to come to school. Duflo)
Results?
- Treatment reduces teacher absenteeism. Long lasting. Greater affect for below-median teachers.
- Treatment increases instructions per student.
School is open more so more kids receive instructions - no lack in student demand. - Improves test performance.
- Weak correlation between absenteeism and performance can be due to measurement error.
- Financial incentives main driver of result. Discontinuous jump when teacher is out-money and no jump when in-money
(Incentives work: getting teachers to come to school. Duflo)
Conclusions/policy?
Bias:
If treatment induced schools with higher attendance to open more often, student attendance rate would be biased upwards.
Identification assumption requires teachers outside option and monitoring effect not to change.
Social norms changing about teacher absenteeism.
External validity/policy:
- financial incentives for teachers
- programme could fail with government teachers
- hiring para-teachers and incentivising them
- scale up depends on applicability in other schools and cost-effectiveness
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Setting?
Nigerian civil service (health, education, environment, water resources)
OPEN initiative - activities of public bureaucracies subject to detailed and independent scrutiny by independent teams of engineers and members of civil society.
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Theory?
- think of management like technology (input in production function)
- capabilities of state shape the process of economic development (poverty, inequality, economic growth)
- how does management practices for bureaucrats in civil service organisations correlate with public service delivery in developing country
Incentives and autonomy:
- incentives might crowd out intrinsic motivation of those self selected, bureaucrats have multiple effort types which aren’t all measurable so hard to give right incentives
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Empirical design/data?
Double-blind interviews to public sector managers across 9 diff topics and focused on autonomy provided to bureaucrats and provision of incentives/monitoring of them.
regression:
y = CSautonomy + CSincentivesmonitoring + CSother + PC + OC + lamda + error
PC is project characteristics. OC is organisational controls. lamda is project type fixed effects.
Unit of observation i, type j in organisation n.
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Limitations?
Correlation not causation. Management practices could be endogenously chosen or there is reverse causality (those organisations with higher completion rates might choose to grant their bureaucrats more autonomy and those with lower completion rates might respond by increasing monitoring and incentive provision to bureaucrats).
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Results?
Management practices correlate with bureaucratic outcomes.
Two dimensions of management practice related to autonomy and incentives/monitoring have opposing correlations with public services delivered. (despite practices being positively correlated w each other)
- increase in autonomy for bureaucrats means increase project completion
- increase in incentives/monitoring means less project completion
Management practices correlate with quality-adjusted project completion rates in similar ways.
(Management of bureaucrats and public service delivery in Nigeria. Rasul.)
Why do we get this negative correlation with project completion and incentives and monitoring but in private sector settings results this is not true?
- multi-tasking environment
- e.g. more negative for more complex projects and better for organisations with good IT facilities (target incentives to productive efforts) - incentives/monitoring management practices picking up elements of subjective performance evaluation (SPE)
- more negative in organisation staffed by more experienced bureaucrats bc here civil servants may be trying to engage in influencing activities with supervisors so don’t focus on most productive projects - crowd out of bureaucrats’ intrinsic motivation
- this is offset by the share of intrinsically motivated bureaucrats in the organisation
Challenges for optimising management practices?
- Weberian view: organisations do already optimise practices according to their true objective which is pleasing politicians
- best management practices may be heterogeneous across organisations so there’s no specific management practice for all organisations
- lack of mobility in labour market for bureaucrats slows down rate at which practices spread through civil service
- large fixed costs of adopting better practices
- lack of competitive pressure enables poorly managed public sector organisations to survive
