Self-selection and peer-effects in experimental labor markets
In my dissertation, I use experimental methods to study the otherwise unobservable labor market interactions, social norms and peer-effects in non-cognitive skills acquisition.
In my first Chapter, I use experimental methods to better understand behavioral roots of corruption. Corruption is endemic to most developing countries. Most of the recent research in economics has focused on documenting the existence of corruption, measuring its economic impact and designing effective mechanisms to reduce incentives for bureaucrats to engage in corruption. Anecdotal evidence often suggests that improving the internal workings of bureaucracies is insufficient, unless corrupt officials are permanently replaced. It might be the case that people with a lower personal cost of engaging in unethical behavior choose to become bureaucrats because they are relatively more efficient at corruption. As a result, we might find a higher number of dishonest people opting for public sector jobs. If this were the case, corruption might be hard to fight, since corrupt institutions will attract more corrupt employees. In my dissertation I explore this self-selection of more corruptible individuals into public sector. Using a laboratory experiment, I answer the question of whether or not labor market sorting in the preference between public and private sector can be explained in part by propensity to engage in unethical behavior.
It is hard to get observational individual-level data on corruption because it is secretive in nature. Testing my self-selection hypothesis is particularly challenging because propensity to engage in unethical behavior is typically not measured in observational or survey data. Moreover, data on bureaucrats and politicians before they joined the civil service or became public officials is typically not available. Also, economic policies that are often designed to attract a better pool of bureaucrats, include increasing wages. It is difficult to separate selection based on propensity to engage in unethical behavior from the selection effects of increased monetary incentives. Therefore, I choose experimental methods to study my research question.
In Chapter 1, I provide experimental evidence of the sorting effect into the public and private sector. I conducted this experiment (from hereby called Experiment 1) at one of the largest public universities and a premier management institute in India. From these two institutes, I recruited a cross-section of students. I created a firm-like environment in the laboratory and divided subjects randomly into ``workers'' and ``supervisors.'' Subjects played a game that measured the propensity of supervisors to engage in unethical behavior. I then linked the propensity to engage in unethical behavior to future career choice. Related deception experiments in economics and psychology suggest that not all people engage in dishonest behavior. Therefore, I treat the willingness to engage in unethical behavior as an agent's preference characteristic.
The findings of this study suggest that both public sector and private sector aspirants are equally likely to cheat. In fact, 64 percent of the private sector aspirants and 70 percent of the public sector aspirants cheat. While there is no statistical difference in the proportion of cheaters in the two sectors, the amount of cheating in the two sectors is significantly different. Public sector aspirants cheat 51.5 percent more than private sector aspirants. Since the choice of sector comes from a survey question and is not measured through the experimental design and I also treat subjects' propensity to engage in unethical behavior as a preference characteristic, these experimental data provide evidence of sorting based on propensity to engage in unethical behavior.
While in Chapter 1 I address individual unethical behavior, in Chapter 2, I study group behavior, specifically social norms regarding bribe-giving and bribe-taking in India, using a novel experimental method introduced by Krupka and Weber (2008). Economists have emphasized the strong effect of social norms and legal enforcement on corruption. However, social norms are difficult to measure. Social scientists commonly use surveys to measure norms. There are pros and cons to using a survey. First, surveys are non-incentive compatible. Second, because corruption is a sensitive issue, the norms measured by a survey might be biased and not reflect participants' truthful response.
For Chapter 2, I used a unique experimental method to measure the social norms regarding bribe-giving and bribe-taking in India for a common corrupt situation (i.e., obtaining a driver's license in India). The subjects of this experiment were the ``workers'' who participated in Experiment 1. After Experiment 1, the workers were asked to elicit the social norms regarding bribe-giving and bribe-taking in India. In this norm experiment, we asked subjects to provide social acceptability ratings for five situations related to bribe-taking and bribe-giving encountered by a driver's license applicant in India. Subjects were provided with incentives to match their rating with the ratings of others. The subjects were required to think how others would respond, and not about their own preferences. This followed a coordination game related to the ``shared understanding'' of social norms.
Situation 1 depicts a driver's license applicant who fails the driving test and the public official who asks him to pay a bribe to obtain the license. The subjects rated the actions of the officer asking for a bribe and also the conditional actions of the citizen accepting or rejecting the officer's bribe request. There were four actions: low, medium, and high bribe amounts and honest action of the officer.
Situation 2 is framed from the perspective of the applicant. The driver's license applicant fails the driving test and offers a bribe to the public official to obtain the license. Situation 3 describes the exact situation the subjects encountered in Experiment 1. Here I obtained appropriateness ratings of the actions that the subjects encountered in Experiment 1. Situations 4 and 5 are similar to Situations 1 and 2 with one exception. To set reference points for the subjects, I provided the additional information that most officers at the drivers' license office ask for a side payment if the applicant fails the driving test.
This is the first experimental study that quantifies the social norms regarding bribe-giving and bribe-taking in India for a common corrupt situation. I found that social acceptability ratings are malleable to the amounts of bribes. Subjects are not able to coordinate on the socially acceptability ratings for lower amount of bribes. However, the higher bribe amounts are rated as socially inappropriate by the majority. In the socially acceptable treatment i.e. Situations 4 and 5, the social acceptability ratings of bribe giving and bribe taking were lower than in Situations 1 and 2. I found that bribe giving is more socially unacceptable than bribe taking in Situations 4 and 5. However, this difference is not significant in Situations 1 and 2.
I also addressed a methodological concern arising from the experimental method used to measure the social norms regarding corruption for a particular situation (i.e., obtaining a driver's license in India). The participants of the norm experiment were the ``workers'' in Experiment 1. They participated in the unethical behavior experiment before eliciting the social norms. One criticism of this subject pool is that their views might have been biased. To address this concern, we conducted the norm experiment on a separate pool whose only role was to elicit the social norms. There was no difference in the description of vignettes and experimental procedures between the two subject pools. I found no difference in the norms elicitation between the two subject pools.
For Chapters 1 and 2, I conducted my own experiments in India. For Chapter 3, I used the experimental data of Leider et al. (2009) to study the effects of social networks on non-cognitive skills, particularly self-confidence regarding one's ability. Recent research in psychology and economics has emphasized the importance of non-cognitive skills on labor market outcomes. We used economic experiments to measure participants' beliefs about their expected relative performance in a real-effort task using an incentive-compatible mechanism for belief elicitation. We then combined these self-confidence measures with a unique dataset on the social network of the participants to investigate whether there are peer effects in confidence.
We found that more confident participants tend to have more confident friends. Peer effects might be an important transmission mechanism for acquisition of self-confidence. We then divided our subjects into years at college and explored whether our finding was due to selection (confident participants choose confident friends) or social interaction (one's confidence is influenced by friends' confidence). We found that the results are significant for seniors and juniors. However, to separate the treatment effect from the selection effect, we need to carefully design a longitudinal study where we follow a cohort of students through years at college. Therefore, this cross-sectional experimental data provide preliminary evidence of the treatment effect in peer effects.
My dissertation research contributes to the literature in labor economics and experimental economics. To summarize, in the first chapter, I highlight the evidence of sorting based on propensity to engage in unethical behavior by public and private sector aspirants. The second chapter highlights the importance of social norms in corruption. In the third chapter, I study peer effects in non-cognitive skills. The findings of my first chapter provide evidence of self-selection into the public vs. private sector based on propensity to engage in unethical behavior. A caveat of this study is that the subjects were university students who were yet to join the public or private sector. In my future research, I want to track the subjects and collect data on their real-life career choices. Findings of my second chapter suggest that when corruption is prominent, government policy should be targeted toward changing the social norms around corruption. Stronger legal enforcement of rules can help change the perception of corruption. My experimental design is not sufficient to separate the exact channel which leads to a higher mean rating for the social acceptability of bribe-giving and bribe-taking under socially acceptable treatment. In my future research, I want to separate the effects of these channels. The findings of the third chapter provide evidence of peer effects in acquiring non-cognitive skills. I am now conducting a longitudinal experiment at the Indian Institute of Management, where I am following a cohort of management students for two years. This experiment will help us segregate the selection and treatment effect of peer effects not only in non-cognitive skills but also in other-regarding preferences and competitive preferences.