DV491 SUMMARY
Outline
(1) HUMAN DEVELOPMENT
WEEK 1 - Economics of Health + SA 1
WEEK 2 - Economics of Education
WEEK 3 - Intra-HH Decision-Making + PS1
(2) STATE EFFECTIVENESS AND MARKETS
WEEK 4 - State Effectiveness
WEEK 5 - Credit Markets, Insurance, Risk
WEEK 7 - Private Sector Development + PS2
WEEK 8 - Employment and Migration + SA2
(3) SOCIAL NETWORKS AND BEHAVIOURAL ECON
WEEK 9 - Social Networks
WEEK 10 - Behavioural Economics + PS3, SA3
(1) HUMAN DEVELOPMENT
WEEK 1 - Economics of Health
LECTURE
Frontier questions in health economics:
(1) Supply side challenges
- Incentives for health care workers
- Organizational Inefficiencies in the Management of Health Systems
(2) Demand side challenges, due to..
- Income
- Information
- Behavioural Biases
(1) Supply Side
1
,Incentives for health care workers
Motivation: evidence that there is a correlation between low income per capita and absence rate
of health workers. The rate is quite high, between 30 to 60%. Within India (between indian
states) there is a high level of variation (Chaudhury et al., 2006). Given this evidence,
1) How can we select high performing health care workers?
Theory: ambiguous. Health care workers may place more value on community attachment and
altruistic capital or on civil service career benefits - should we select altruists or career-oriented
HCW?
Ashraf et al., (2019): test career vs altruism. They issue ads inviting people for a training
opportunity. Treatment: career-oriented vs Control: community attachment. Findings: career
incentives select less prosocial candidates who have more skill and ambition. Career-based
deliver more services, promote institutional childbirth, and reduce child malnutrition by 25% in
the communities they serve. However, the trade-off only exists at low levels of talent (so people
that are trained and are highly skilled perform equally in both pro-sociality and career-based). ??
ASK
2) How can we create incentives for HCW to actually deliver health services?
Theory: ambiguous. Pay for performance contracts can deliver improved performance or they
can crowd out intrinsic motivation for prosocial tasks with positive externalities for society. Not
sure how pay incentives that work well on the pvt sector would translate to the public/pro-social
sector. Should we focus on salary incentives or pro-social messages/motivation?
Ashraf et al (2014): SEE READINGS SECTION. Findings: social incentives performed better
than financial rewards that are small relative to overall earnings.
Open questions:
Organizational Inefficiencies
Albuquerque et al. (year): poor maternal and infant health and high rates of HIV-AIDS in
Mozambique. They found that with these two groups specifically, there was a very low rate of
retention (return visits). Challenge: coordination failure (btw doctors and patients) in terms of
patient inflow. First-come first-serve system, where most people tend to show up Monday and
Friday, while the rest of the week is empty, so high cost of attendance leads to no returns.
Intervention: RTC pilot (prenatal care), treatment gets low-tech scheduling system. Findings:
treatment smoothed time of arrival throughout the day (instead of early), reduced waiting time,
higher provider (nurses) satisfaction, increase in service utilization and no increases in waiting
time for other services (no negative spillovers). Potential empirical challenge: control individuals
spillover to treatment once they know that other clinics are more efficient.
Open questions:
2
,(2) Demand Side
Income
1) Should health products be provided free of cost or at positive price?
Theory: ambiguous. Free of cost (cost-sharing with government): because low-income HH have
liquidity constraints, so with positives they won’t be able to access the product. Positive price:
price increases confidence in the product + if one pays for it, he is more likely to actually adhere
to the product and use it for longer. → Seminar topic + READING
2) Can short-run subsidization compromise long-term willingness to pay due to
reference-dependence and anchoring? Since people form belief that they should be getting it for
free (fairness).
Dupas (2014): short–run subsidies do not compromise long-term willingness to pay for bednets.
From seminar slides.
Information
Misperception of benefits. Genderally, underestimation of preventative care (since benefits are
not obvious), while overestimation of curative care (immediate and clear benefits).
Open questions: importance of source of misinformation (how to debias?), rate of decay of
information, can we leverage peer effects to increase cost-effectiveness of informational
interventions?
Behavioural Biases
Time inconsistent preferences: Households make choices today (not to invest in health) that
their future selves would have preferred not to have made (resulting in poor health). Though
they may have a preference for better health in both periods, agents discount the future benefits
heavily when assessed against today’s costs.
Policy response: small subsidy to offset costs today or commitment device (Eg: compulsory
vaccination, health care provider outreach, package of lentils as a gift for getting vaccination).
EX: HIV testing nudging
Thornton (2008): Can small financial rewards persuade people to take an HIV test and learn
about their results? 2,182 people tested HIV-AIDS and they had to pick-up results. The
Financial Reward for pick-up was randomized between $0 and $3 Location of pick-up also
randomized (goal: vary cost of pick up). Result: when there was no financial incentive to pick up
results, less than 40% actually learned about their HIV results. Important: VERY small reward
doubled pick-up rate (which then increased only modestly for any further increases in rewards)
+ this did not covary with distance to pick up place -> problem is not cost of travelling per se but
the mental cost of overcoming procrastination (time inconsistency!).
3
, READINGS
Ashraf, Bandiera and Jack (2014) SUPPLY
RQ: what is the effect of extrinsic rewards on the performance of agents in a public health
organization? Compare the effects of monetary and non-monetary incentives + measure the
interaction between extrinsic rewards and pro-social motivation of agents.
Motivation: Understanding what motivates individuals to devote time and effort to work is
crucial to understand observed behavior and to design incentive mechanisms that align the
individuals' and organisations’ interest. However, previous papers have focused only on
financial rewards in a setting where employee effort only benefits the employer, but not on
non-financial incentives + in pro-social tasks (benefits are for other people that employee).
Theoretical prediction: the effect of extrinsic rewards on performance in private and pro-social
tasks might differ. Mission-driven organizations benefit from matching with workers whose
interests are aligned with the mission, and these individuals might respond less to incentives or
even deliver a weaker performance if incentives displace other sources of motivation. In
particular, to the extent that agents are motivated by the externalities generated through
pro-social tasks, this motivation may interact positively or negatively with extrinsic incentives.
Design + Results: Field experiment set-up: randomly assigns 205 distinct geographical clusters
containing 1222 agents to one of four groups that receive different rewards based on condom
sales. (1) Control group: no rewards, (2) Star group: non-financial rewards, performance
indicator, (3) Small financial reward: 10% margin on each condom sale, (4) Large financial
reward: 90% margin. Track performance for one year.
1st part: financial vs. non-financial rewards.
- agents in the star treatment sell over twice as many condoms as agents in any other
group, on average. Constant throughout the year.
- Financial incentives may be ineffective due to earnings from condom sales being a small
fraction of overall earnings, because both demand for the product and earnings from
each sale are low. Hypothesis: all rewards above that threshold would be more effective
at eliciting effort than non-financial incentive (supported by 3rd part).
4
, 2nd part: mechanisms. Treatment differs because of higher efforts of treatments affecting
demand directly? 3 checks:
(1) Show that gents in the star treatment behave differently on dimensions correlated with
sales effort
(2) survey a random sample of customers and find that they recall and correctly describe the
promotional posters given to agents in all treatments, but only a negligible minority
mentions the thermometer that is only given to agents in the star treatment.
(3) Placebo test: provide in some salons a placebo thermometer that does not measure
individual performance (so carries not reward), but average region sales + customers
can see it. Displaying the thermometer does not change condoms demand.
3rd part: study interaction extrinsic incentives x intrinsic motivation for the cause. Measure
motivation based on donations to an HIV/AIDS charity. Donation is a strong prediction of sales.
Thus agents who are motivated by the cause respond more strongly to financial and
non-financial rewards → direct contradiction to the theoretical hypothesis that extrinsic
incentives crowd out intrinsic motivation. Incentives are complementary to pro-social motivation.
4th part: responses to both financial and non-financial incentives is larger when utility
associated with them is higher.
- Financial: incentives increase sales for poorest participants, for whom relative value of
reward is higher
- Non-financial: effect of incentives is increasing in the number of neighbouring salons that
also received non-financial incentives (stars more valuable when visible to a larger
group).
Identification Strategy: Since agents choose whether to participate in the program after
learning about incentives, the coefficients δ0j capture the effect of incentives on sales
performance through both the margins of selection and effort. In this setting, however, the role of
selection is limited since almost all the agents who were exposed to treatment joined the
program.
Identifying assumption: ‘treat’ variable is orthogonal to error term. Thus, it would fail if the
decision to participate in the training program is (1) not orthogonal to treatment, and/or (2) if
there are spillovers between treatments - control group is not a proper counterfactual for how
agents in the treatment groups would have behaved in the absence of treatment.
(1) Since stylists were not informed about treatments (incentives) until the end of training,
selection ought to be orthogonal to treatment.
(2) Four design features were employed to minimize the risk of spillovers. Check: In over
7000 restocking visits, only one stylist asked about different incentive schemes.
Cohen and Dupas (2010) - Distribution or Cost-sharing? - DEMAND
5
Outline
(1) HUMAN DEVELOPMENT
WEEK 1 - Economics of Health + SA 1
WEEK 2 - Economics of Education
WEEK 3 - Intra-HH Decision-Making + PS1
(2) STATE EFFECTIVENESS AND MARKETS
WEEK 4 - State Effectiveness
WEEK 5 - Credit Markets, Insurance, Risk
WEEK 7 - Private Sector Development + PS2
WEEK 8 - Employment and Migration + SA2
(3) SOCIAL NETWORKS AND BEHAVIOURAL ECON
WEEK 9 - Social Networks
WEEK 10 - Behavioural Economics + PS3, SA3
(1) HUMAN DEVELOPMENT
WEEK 1 - Economics of Health
LECTURE
Frontier questions in health economics:
(1) Supply side challenges
- Incentives for health care workers
- Organizational Inefficiencies in the Management of Health Systems
(2) Demand side challenges, due to..
- Income
- Information
- Behavioural Biases
(1) Supply Side
1
,Incentives for health care workers
Motivation: evidence that there is a correlation between low income per capita and absence rate
of health workers. The rate is quite high, between 30 to 60%. Within India (between indian
states) there is a high level of variation (Chaudhury et al., 2006). Given this evidence,
1) How can we select high performing health care workers?
Theory: ambiguous. Health care workers may place more value on community attachment and
altruistic capital or on civil service career benefits - should we select altruists or career-oriented
HCW?
Ashraf et al., (2019): test career vs altruism. They issue ads inviting people for a training
opportunity. Treatment: career-oriented vs Control: community attachment. Findings: career
incentives select less prosocial candidates who have more skill and ambition. Career-based
deliver more services, promote institutional childbirth, and reduce child malnutrition by 25% in
the communities they serve. However, the trade-off only exists at low levels of talent (so people
that are trained and are highly skilled perform equally in both pro-sociality and career-based). ??
ASK
2) How can we create incentives for HCW to actually deliver health services?
Theory: ambiguous. Pay for performance contracts can deliver improved performance or they
can crowd out intrinsic motivation for prosocial tasks with positive externalities for society. Not
sure how pay incentives that work well on the pvt sector would translate to the public/pro-social
sector. Should we focus on salary incentives or pro-social messages/motivation?
Ashraf et al (2014): SEE READINGS SECTION. Findings: social incentives performed better
than financial rewards that are small relative to overall earnings.
Open questions:
Organizational Inefficiencies
Albuquerque et al. (year): poor maternal and infant health and high rates of HIV-AIDS in
Mozambique. They found that with these two groups specifically, there was a very low rate of
retention (return visits). Challenge: coordination failure (btw doctors and patients) in terms of
patient inflow. First-come first-serve system, where most people tend to show up Monday and
Friday, while the rest of the week is empty, so high cost of attendance leads to no returns.
Intervention: RTC pilot (prenatal care), treatment gets low-tech scheduling system. Findings:
treatment smoothed time of arrival throughout the day (instead of early), reduced waiting time,
higher provider (nurses) satisfaction, increase in service utilization and no increases in waiting
time for other services (no negative spillovers). Potential empirical challenge: control individuals
spillover to treatment once they know that other clinics are more efficient.
Open questions:
2
,(2) Demand Side
Income
1) Should health products be provided free of cost or at positive price?
Theory: ambiguous. Free of cost (cost-sharing with government): because low-income HH have
liquidity constraints, so with positives they won’t be able to access the product. Positive price:
price increases confidence in the product + if one pays for it, he is more likely to actually adhere
to the product and use it for longer. → Seminar topic + READING
2) Can short-run subsidization compromise long-term willingness to pay due to
reference-dependence and anchoring? Since people form belief that they should be getting it for
free (fairness).
Dupas (2014): short–run subsidies do not compromise long-term willingness to pay for bednets.
From seminar slides.
Information
Misperception of benefits. Genderally, underestimation of preventative care (since benefits are
not obvious), while overestimation of curative care (immediate and clear benefits).
Open questions: importance of source of misinformation (how to debias?), rate of decay of
information, can we leverage peer effects to increase cost-effectiveness of informational
interventions?
Behavioural Biases
Time inconsistent preferences: Households make choices today (not to invest in health) that
their future selves would have preferred not to have made (resulting in poor health). Though
they may have a preference for better health in both periods, agents discount the future benefits
heavily when assessed against today’s costs.
Policy response: small subsidy to offset costs today or commitment device (Eg: compulsory
vaccination, health care provider outreach, package of lentils as a gift for getting vaccination).
EX: HIV testing nudging
Thornton (2008): Can small financial rewards persuade people to take an HIV test and learn
about their results? 2,182 people tested HIV-AIDS and they had to pick-up results. The
Financial Reward for pick-up was randomized between $0 and $3 Location of pick-up also
randomized (goal: vary cost of pick up). Result: when there was no financial incentive to pick up
results, less than 40% actually learned about their HIV results. Important: VERY small reward
doubled pick-up rate (which then increased only modestly for any further increases in rewards)
+ this did not covary with distance to pick up place -> problem is not cost of travelling per se but
the mental cost of overcoming procrastination (time inconsistency!).
3
, READINGS
Ashraf, Bandiera and Jack (2014) SUPPLY
RQ: what is the effect of extrinsic rewards on the performance of agents in a public health
organization? Compare the effects of monetary and non-monetary incentives + measure the
interaction between extrinsic rewards and pro-social motivation of agents.
Motivation: Understanding what motivates individuals to devote time and effort to work is
crucial to understand observed behavior and to design incentive mechanisms that align the
individuals' and organisations’ interest. However, previous papers have focused only on
financial rewards in a setting where employee effort only benefits the employer, but not on
non-financial incentives + in pro-social tasks (benefits are for other people that employee).
Theoretical prediction: the effect of extrinsic rewards on performance in private and pro-social
tasks might differ. Mission-driven organizations benefit from matching with workers whose
interests are aligned with the mission, and these individuals might respond less to incentives or
even deliver a weaker performance if incentives displace other sources of motivation. In
particular, to the extent that agents are motivated by the externalities generated through
pro-social tasks, this motivation may interact positively or negatively with extrinsic incentives.
Design + Results: Field experiment set-up: randomly assigns 205 distinct geographical clusters
containing 1222 agents to one of four groups that receive different rewards based on condom
sales. (1) Control group: no rewards, (2) Star group: non-financial rewards, performance
indicator, (3) Small financial reward: 10% margin on each condom sale, (4) Large financial
reward: 90% margin. Track performance for one year.
1st part: financial vs. non-financial rewards.
- agents in the star treatment sell over twice as many condoms as agents in any other
group, on average. Constant throughout the year.
- Financial incentives may be ineffective due to earnings from condom sales being a small
fraction of overall earnings, because both demand for the product and earnings from
each sale are low. Hypothesis: all rewards above that threshold would be more effective
at eliciting effort than non-financial incentive (supported by 3rd part).
4
, 2nd part: mechanisms. Treatment differs because of higher efforts of treatments affecting
demand directly? 3 checks:
(1) Show that gents in the star treatment behave differently on dimensions correlated with
sales effort
(2) survey a random sample of customers and find that they recall and correctly describe the
promotional posters given to agents in all treatments, but only a negligible minority
mentions the thermometer that is only given to agents in the star treatment.
(3) Placebo test: provide in some salons a placebo thermometer that does not measure
individual performance (so carries not reward), but average region sales + customers
can see it. Displaying the thermometer does not change condoms demand.
3rd part: study interaction extrinsic incentives x intrinsic motivation for the cause. Measure
motivation based on donations to an HIV/AIDS charity. Donation is a strong prediction of sales.
Thus agents who are motivated by the cause respond more strongly to financial and
non-financial rewards → direct contradiction to the theoretical hypothesis that extrinsic
incentives crowd out intrinsic motivation. Incentives are complementary to pro-social motivation.
4th part: responses to both financial and non-financial incentives is larger when utility
associated with them is higher.
- Financial: incentives increase sales for poorest participants, for whom relative value of
reward is higher
- Non-financial: effect of incentives is increasing in the number of neighbouring salons that
also received non-financial incentives (stars more valuable when visible to a larger
group).
Identification Strategy: Since agents choose whether to participate in the program after
learning about incentives, the coefficients δ0j capture the effect of incentives on sales
performance through both the margins of selection and effort. In this setting, however, the role of
selection is limited since almost all the agents who were exposed to treatment joined the
program.
Identifying assumption: ‘treat’ variable is orthogonal to error term. Thus, it would fail if the
decision to participate in the training program is (1) not orthogonal to treatment, and/or (2) if
there are spillovers between treatments - control group is not a proper counterfactual for how
agents in the treatment groups would have behaved in the absence of treatment.
(1) Since stylists were not informed about treatments (incentives) until the end of training,
selection ought to be orthogonal to treatment.
(2) Four design features were employed to minimize the risk of spillovers. Check: In over
7000 restocking visits, only one stylist asked about different incentive schemes.
Cohen and Dupas (2010) - Distribution or Cost-sharing? - DEMAND
5
