L11: Corruption

33%; The audit treatment reduced the percent of major items reported missing by about 9 percentage points, on a control group base of roughly 27 percent. In other words, corruption went down by about a third (33%.)

By approximately how much did the audit treatment in Indonesia reduce corruption by, measured by the percent of major items in roads (and ancillary projects) that were found to be missing? 33% 0.3% 3% It did not reduce corruption at all

80; This large discrepancy is one reason why Professor Olken cautions against taking the headline number of 13 percent too literally. There could be a host of reasons for the 13 percent figure in the first round of surveys, not least of which is accounting errors and reporting bias.

A couple of years after the finding that only 13 percent of the grants made it to the schools was published, the surveys were conducted again. This time, __________ percent of the grant was reported as having been delivered. (Enter the nearest whole number.)

False; As Professor Olken notes, there are a number of ways in which ethno-linguistic variation may affect economic growth and development. Ethno-linguistic variation thus fails the exclusion restriction; thus, it is more a strong reduced form fact and correlation that corruption is associated with lower levels of growth and national income.

Corruption is the principal channel through which ethno-linguistic variation can affect economic growth: a. True b. False

A, D; Options a and d are saying the same thing i.e. that richer countries tend to be associated with lower levels of corruption. Professor Olken illustrates this using a perceptions-based measure of bureaucratic efficiency and finds that per capita income strongly predicts the perceived level of corruption. Thus, option c is wrong, while the direction of the relationship in b is wrong as well.

If bureaucratic efficiency is used as a measure of corruption, then we see that: (Check all that apply) a. Lower corruption is correlated with higher per capita income b. Lower corruption is correlated with lower per capita income c. There is no relationship between corruption and per capital income d. Richer countries tend to have lower levels of corruption

B; Say there are two types of agents, poor agents (H) and rich agents (L). In this situation, H > L since a welfare-maximizing society would prefer for the richer agents to pay more in taxes. However, in this scenario, yh < yl since the rich can afford to pay more in bribes to avoid taxes than the poor can (and thus, h < y since supervisors get the most benefit from allotting the slot to rich agents.) Thus, we are likely to get misaligned incentives here.

If the "slot" that we are dealing with is not having to pay taxes, why might we be faced with a case where societal incentives and supervisor incentives in a society with corruption are misaligned? a. Agents with the highest ability to pay are also those for whom the societal benefit of getting the slot is the highest b. Agents with the lowest ability to pay are also those for whom the societal benefit is the highest c. All agents are likely to have the same ability to pay regardless of differences in the societal benefits d. None of the above

A; Professor Olken finds that more overweight trucks pay more in bribes than less overweight ones. This is seen in the plots of the non-parametric regressions of payment in bribes on the number of tons that the truck is overweight. We notice in these regressions that the intercept, which denotes the payment made when the truck is 0 tons overweight, is generally positive i.e. some amount is paid in bribes even for non-overweight trucks.

In his study of bribes paid by overweight trucks in Indonesia, Professor Olken finds that: a. The amount paid in bribes increases for each ton that a truck is overweight b. The amount paid in bribes decreases for each ton that a truck is overweight c. The amount paid in bribes is roughly constant for each ton that a truck is overweight d. Non-overweight trucks do not pay anything in bribes

B; Official line-item expenditure reports were obtained from village books and financial reports, which have to be filed in order to receive central government funds. These were used to measure reported expenditures. Working with a group of civil engineers, Professor Olken and his team then measured actual expenditures by taking core samples of road materials to measure the quantity of materials used, surveying suppliers in nearby villages to obtain prices of the materials, and interviewing villagers to determine the wages paid and tasks done by voluntary labor. These were then used to construct actual expenditures, which were then compared to reported expenditures.

In his study on monitoring corruption in Indonesia, Professor Olken measured corruption by: a. Asking members of the sample what they perceived corruption to be. b. Measuring the difference between reported road expenditures and actual road expenditures. c. Asking bureaucrats how much they had taken in bribes. d. Randomly giving villages funds for road repairs and then measuring the amount actually spent on road expenditures.

A, B; Recall that an agent is indifferent between taking a bribe or not when w - v = [(1 - p)/p](b) i.e. when w - v = [(c)/1 - c](b). Thus, the agent will take a bribe when w - v < [(c)/1 - c](b). When c = 0, nobody takes a bribe and when c = 1, everyone takes a bribe, both thus representing stable equilibria. When w - v = (c/1 - c)*b, small changes in c or in w - v will ether push all agents to take a bribe, or no one to take a bribe. Thus, the latter condition represents an unstable equilibrium and means that temporary wage or enforcement shocks can have permanent effects

In the figure below, b denotes the amount taken in bribes, c denotes the fraction of the population that is corrupt (with the probability of being caught taking a bribe given by p = 1 - c), w denotes the wage earned by the agent, and v denotes the outside market wage. Stable equilibria occur where (check all that apply): a. c = 0 b. c = 1 c. w - v = (c/1 - c)*b d. 0 < c < 1

A; If H > L, there is higher social benefit to the slot going to agents of type H instead of agents of type L. The slots will only end up going to agents of type H when h > l, and yh > yl, in which case societal and supervisor incentives will be aligned since those agents from whom the supervisor gets the highest private benefit and those agents with the highest willingness to pay are also the agents to whom society has the greatest interest in seeing the slots be assigned to.

In the formal model of corruption sketched out by Professor Olken, there are two types of agents, Type H and Type L. Let the private benefit to a supervisor from assigning a slot to agent X be given by x, the social benefit be given by Y, and the ability of the agent to pay be given by yx. Then, if H > L, h > l, and yh > yl: a. Societal incentives will be aligned with supervisor incentives b. Societal and supervisor incentives will be partially aligned c. Societal and supervisor incentives will be misaligned d. How aligned incentives are will depend on the specific value of h and l

B; We observe that the lines for the control group are to the right of the treatment group, suggesting that the gap between reported and measured expenditures is higher in the control group than in the treatment group. This is evidence of there being greater corruption in the control group. In addition, we also observe a number of observations for both groups below 0 i.e. where reported expenditures are less than actual expenditures.

In the graph below, we see the cumulative density function on the left and the probability density function on the right, plotting the difference between log(reported expenditures) and log(actual measured expenditures) on road construction in the Indonesia study. What do the two plots suggest? a. Both groups showed roughly the same amount of corruption b. There may have been more corruption in the control group c. There may have been more corruption in the treatment group d. Reported expenditures were always higher than actual measured expenditures

c; Each school is supposed to receive a block grant from the central government. In comparing the amount that schools report receiving to the amount that the central government sent out, surveyors find that only 13 percent of the grant is reported to have made it to the schools.

In their study of corruption in a central government transfer program in Uganda, Reinikka and Svensson find that schools report receiving __________ percent of what the central government sent out. a. 0.13 b. 1.3 c. 13 d. 31

C; If the agent doesn't take a bribe, he earns his wage w with a probability of 1. A bribe-taking agent has probability p of being detected and getting fired so that he only earns the outside wage v, and probability (1 - p) of going undetected and earning both the bribe b and the wage w. Thus, his expected wage is pv + (1 - p)*(b + w) if he takes a bribe, which would be equal to his wage w for him to be indifferent. The expression w = pv + (1 - p)*(b + w) is just another way of expressing the expression in option c above

Let the wage that an agent earns in a job be given by w, the amount he can earn in bribes by given by b, the probability of being caught and taking a bribe be given by p, and the wage he would earn from an outside job if caught and fired by given by v. At what level should the equilibrium wage w be set so that the agent is indifferent between accepting and not accepting a bribe? a. w = pv + (1 - p)(b) b. w = (1-p)v + (p)(b) c. w = v + [(1 - p)/p](b) d. w = v + [(1 - p)/p](b + w)

D; The process by which the engineers and the research team measure actual expenditures is not an exact science; as Professor Olken notes, there is substantial room for measurement error. This won't bias the estimates if the errors are distributed similarly across the treatment and control groups. However, measurement error in these techniques does make it difficult for auditors to get measure corruption, particularly when averaging across many roads.

One reason for why we might observe reported expenditures being lower than actual expenditures in the plots below is that those reporting expenditures may be making accounting errors. What is another reason suggested by Professor Olken? a. Missing data b. Small sample size c. Corruption amongst the engineers measuring actual expenditures d. Measurement error

A; Since we have posited that the system is corrupt, it is possible for people to get licenses more quickly by paying a bribe. The financial incentive would allow people to pay a bribe to obtain a license. However, if all corruption does is to grease the wheels and speed up slow-moving bureaucratic processes, we would only see this impact for good drivers and not bad drivers (since it would be an inefficient outcome if these bad drivers received a license.)

Say we have a scenario where there are good drivers and bad drivers trying to get drivers' licenses in a corrupt system. If all corruption does is to "grease the wheels" and not create any inefficiencies, what would offering drivers financial incentives to receive licenses achieve? a. It would allow good drivers to get licenses more quickly b. It would allow bad drivers to get licenses more quickly c. It would allow both good and bad drivers to get licenses more quickly d. It would have no impact at all

False; Consider a system with corruption in which bribes need to be paid in order to win a government contract. In theory, the most efficient contractor may be able to pay the highest amount in bribes since he has the largest amount surplus (by virtue of being the most efficient contractor) to spend. In this scenario, the most efficient contractor would win the government contract even in the presence of corruption.

True or False: One outcome of corruption is that the most efficient contractor will never win a government contract, resulting in inefficiencies and waste. a. True b. False

True: the researchers found that hirign of family members increased in response to audits.

The results of the study suggest that the intervention reduced one type of corruption (direct stealing from the project), but led to more corruption of another type (nepotism). True False

False; Those in the lesson group are more likely to obtain a license than those in the comparison group, while paying no more in fees or bribes. Thus, learning how to drive does seem to facilitate the process of obtaining a drivers' license by effectively lowering the cost of getting a license. Thus, even in this corrupt system, there is a differentiation between good and bad drivers. While the incentives to learn how to drive may be flatter than ideal, there is evidence that the system responds to people's ability to drive, thus creating some incentives.

True or False: Based on the results in the Bertrand et al. study, the drivers' license system seems to be creating no incentives for people to learn how to drive. a. True b. False

False; As Professor Olken notes, there are many potential reasons for multiple equilibria. The probability of detection is one of them, but selection into the bureaucracy - whereby more corrupt people select into the bureaucracy - and the issue of enforcement (where those responsible for punishing corruption can be corrupt themselves) are amongst the other reasons why these multiple equilibria might emerge. Indeed, we do not know for sure whether these multiple equilibria exist in corruption, but in so far as this is a feature of theoretical models, there are a number of reasons behind why multiple equilibria may emerge.

True or False: Empirical research indicates that changes in the probability of detection are the main reason behind multiple equilibria emerging in a model of corruption. a. True b. False

False; In the examples presented, where the willingness of an agent to pay a bribe is correlated with how much society wants the agent to get the slot for which a bribe is being paid, corruption will affect the allocation of rents within society, but not overall economic efficiency.

True or False: In the examples cited by Professor Olken, corruption affects overall economic efficiency. a. True b. False

False; Almost all of the change in the bonus group came about through the use of agents, or middlemen. As Professor Olken points out, direct corruption can be risky because those seeking to pay a bribe may not know how it will be received, while those seeking to receive a bribe cannot be sure that the bribe is being made as part of an undercover sting operation. Over time, agents - who provide a legitimate license processing service - can establish trust with government officials, while also establishing which ones are more receptive to bribery, thus cloaking the payment of bribes in the official (legal) payments made as part of the standard license processing structure.

True or False: Most of the change in how members of the bonus group obtained licenses came about through direct bribery of government officials. a. True b. False

A; Non-corrupt officials earn a wage that internalizes the amount that they would earn in bribes over and above the market wage, while corrupt officials earn their wage plus any amount that they earn in bribes. More formally, if the wage w is given by w = v + [(1 - p)/p](b) where v is the market wage, p is the probability of being caught taking a bribe, and b is the amount earned in bribes, w will be greater than v whenever b is greater than 0, which is the case in corrupt environments.

True or False: One implication of the Becker-Stigler model of corruption that Professor Olken sketches out is that in corrupt environments, officials may be paid more than their market wage. a. True b. False

True; The bonus in the bonus group seems to allow people to grease the wheels of the system regardless of driving ability since more people who obtain licenses in the bonus group report having never been taught how to drive, while also scoring worse on the license exam.

True or False: Results from the bonus group in the Bertrand et al. study tell us that corruption in the drivers' license system allows both good and bad drivers to obtain a license more quickly. a. True b. False

False; The research team did indeed run surveys to get initial estimates of labor, but then calibrated this data by building four small roads themselves and measuring input use on those four roads in order to get as unbiased an estimate as possible of labor input required.

True or False: Since the village financial reports did not include accurate measures of the amount of labor that goes into accomplishing a given task, the researchers used only survey data on wages paid and tasks done by voluntary labor for measures of the quantity of work involved in road construction. a. True b. False

True; The economists' estimates were higher than what the UN investigation into kickbacks was able to prove. As Professor Olken points out, this just underlines the fact that there is a difference between what estimates you can arrive at for corruption in economics research, and what you can actually prove in a court of law.

True or False: When the economists Hsieh and Moretti studied corruption under the Iraqi Oil-for-Food program, they estimated rents of roughly $3.5 billion through underpricing of Iraqi oil. This was higher than the figure estimated by the formal UN investigation into the program. a. True b. False

False; Through a series of examples, Professor Olken illustrates that while corruption may be observed in a number of different contexts, estimated magnitudes vary substantially, from 2% in the Oil-for-Food program in Iraq, to 80% in the Ugandan education system.

True or False: With corruption a pervasive problem in a number of different contexts, estimated magnitudes of corruption are fairly constant across multiple contexts. a. True b. False

C: There is indeed evidence that truckers pay bribes in Indonesia. Moreover, regressions of bribe amounts paid on tons overweight indicate that more overweight trucks do pay more in bribes. However, Professor Olken points out that these increases are quite small and thus, the increase in bribe amount paid is lower than the increase in social cost as trucks are more overweight. While it is true that even non-overweight trucks need to pay a bribe, this may not be incentive enough for trucks to drive overweight since we could imagine a scenario where the marginal bribe amount to drive overweight is still higher than the marginal benefit to driving overweight, thus providing no incentive for truckers to drive overweight.

What does Professor Olken point to as an explanation for why trucks in Indonesia may be driving overweight: a. There is no evidence of bribes being paid for driving overweight b. The amount paid in bribes does not have any relation to how overweight a truck is c. The marginal increase in the bribe amount is lower than the marginal social cost as trucks become more overweight d. Trucks need to pay a bribe even if they do not drive overweight

D; The lesson group shows us that the system is somewhat responsive to the quality of driving, thus it is not necessary to pay a bribe in order to obtain a license. However, paying a bribe can grease the wheels of the system and increase the speed at which licenses are obtained, for both good and bad drivers (as the bonus group shows us), thus providing some incentive for even good drivers to go through the agent channels to obtain a license. Ultimately, all of these results are consistent with a model where there exist two systems as Professor Olken posits, one official one through which a license can be obtained by passing the licensing exam, and a backdoor one whereby agents take a fee to facilitate the process by bribing corrupt officials.

What does Professor Olken propose as a model that would explain the results from the Bertrand et al. study? a. The system for obtaining a license involves always having to pay a bribe to obtain a license, regardless of driving ability. b. The process of obtaining a license can be speeded up by payment of a bribe/higher fees, but only for good drivers. c. Better drivers should not bother paying a bribe/higher fees since they are just as likely to obtain a license without doing so. d. There exist two systems, an official honor one through which good drivers can obtain licenses by taking the licensing exam, and a more corrupt one through which all drivers can obtain licenses by paying a bribe.

C; By randomly offering some people free driving lessons, the researchers are able to randomly vary the quality of drivers in the sample. This allows them to test whether better drivers are more likely to get a license than worse ones and thus, whether the system is at all responsive to the very thing it is supposed to be responsive to: the quality of driving.

What is the key question that the "lesson" group in the Bertrand et al. study in India seek to answer? a. Whether there is corruption in the process by which drivers' licenses are obtained b. Whether corruption in the drivers' license system leads to inefficiencies c. Whether the system is at all responsive to the quality of drivers d. Whether rich people are more likely to receive a license

A, D; When looking at absences in a number of Indian states, as well as Bangladesh, Ecuador, Indonesia, Peru, and Uganda, we see that average absence rates are 35 percent for health workers and 19 percent for teachers. In both cases, absence rates are negatively correlated with state/national per capita income.

When looking at the absence rates of teachers and health workers in Indian states and in a number of developing countries, we see that (check all that apply): a. Absence rates of teachers are negatively correlated with per capita income b. There is no correlation between absence rates for either teachers or health workers, and per capita income c. Absence rates of health workers are positively correlated with per capita income d. Absence rates of health workers are negatively correlated with per capita income

A, B, D; As Professor Olken mentions, directly asking people how corrupt they are rarely works from the perspective of measuring corruption. Groups like Transparency International survey people in various countries and ask them about what they perceive corruption to be, then using these perceptions of corruption as a measure of how much corruption there is. Comparing two measures of the same thing involves measuring an activity before and after corruption takes places, using any differences as a measure of corruption. Finally, corruption can sometimes be observed directly, or measured by say asking people how much they pay in bribes in countries where bribe payment is not as stigmatized.

Which of the following are used in empirical research to measure corruption? (Check all that apply) a. Perceptions of corruption b. Comparing two measures of the same thing c. Directly asking people how corrupt they are d. Direct measurement by say asking people how much they pay in bribes

A, C, D; Those in both the lesson group (12 percentage points more) and the bonus group (24 percentage points more) were more likely to obtain a license than those in the comparison group. Moreover, the bonus group had more people obtaining a license without anyone teaching them how to drive and more people obtaining a license with scores of under 50% on the licensing exam, indicating that the bonus group allowed more people who couldn't drive (if being taught how to drive and scoring well on the exam are used as proxies for good driving) to obtain a license. The license group on the other hand had fewer people who obtained a license without anyone teaching them how to drive in comparison to the comparison group.

Which of the following results did Bertrand et al. find in the study of corruption with drivers' licenses in India? (Check all that apply) a. Those in the bonus group were more likely to obtain a license than those in the comparison group b. Those in the lesson group were more likely to obtain a license than those in the comparison group c. Those in the lesson group were more likely than those in the comparison group to get a license without having anyone teach them how to drive d. The bonus group allowed more people who couldn't drive to obtain a license