MARCUS ROEL
How Many Others Apply for Jobs I am Applying for?
The Effect of Perceived Labor Market Competition on Search
Job seekers’ expectations about labor market prospects are crucial in determining their search behavior and outcomes in search models, yet clear empirical evidence is lacking. We conduct a large-scale information-provision experiment on a Chinese job board to analyze the causal impact of beliefs about labor market competition on job search. We vary whether information about last month’s applications per vacancy in the job seeker’s preferred occupation is provided and whether a personal daily application goal is elicited. We document that job seekers have large misperceptions about competition. Being provided with information, job seekers update their beliefs about the upcoming month’s competition in a way that is consistent with Bayesian updating. While we find that the intended search effort (goal) increases in beliefs, neither job seekers’ real effort nor their tendency to switch to other, non-preferred occupations is affected by those beliefs. In contrast to the prediction from a sequential search model, we find that both the lowest wage offered (reservation wage) and the lowest required work experience among the jobs job seekers apply for in the month after the intervention increase in their beliefs. This surprising behavioral response is driven by currently employed job seekers, and can be rationalized by gambling for higher-quality jobs when landing a job becomes more difficult. For the unemployed, the reservation wage and lowest experience are rigid. We further show that neither this seemingly risky adjustment nor the rigidity affects their job-matching outcomes negatively.
-- link to working paper --
Strategy vs. Direct-Response Method:
Evidence from a Large Online Experiment
This paper studies the behavioral differences that arise from eliciting choices in sequential games by the direct-response method, where players observe the choices made by players who acted before them and make a single choice in response, or the strategy-method, where they make their choice(s) in response to all possible choices by those who acted prior regardless of whether such actions were actually taken. We conduct a 2x2 between-subject design, large-scale online experiment with over 8000 participants on Amazon MTurk, where, in addition to varying the elicitation method, we also manipulate the ex-ante beliefs of participants about player 1’s likely choices via an information-provision treatment. In two neutrally framed binary-choice sequential games, a sequential Prisoners’ Dilemma and a mini-Ultimatum Game, we document that the elicitation method does not alter player 2’s preferences (their tendency to reward cooperation or reject unfair offers) yet significantly reduces mistakes (rewarding defection or rejecting fair offers). Our results suggest that the more economical strategy method should be the go-to choice for eliciting behavior in sequential games, which may be particularly relevant for all experiments conducted online nowadays.
-- link to working paper --
The Benefits of Being Misinformed: Information Moderation under Misperception
We explore how two fundamental mistakes in information processing – incorrect beliefs and misperception of information – can be mitigated by a benevolent information moderator who has no superior access to information but is more skilled at interpreting it. We analyze a simple sender-receiver model in which a moderator (i.e., sender) can garble signals about payoff-relevant states. We characterize when such manipulation can be beneficial, both for a decision maker unaware of any interference (naïve), and one who takes it into account (sophisticated). We find that sophistication allows the moderator to beneficially intervene in more cases but can render moderation less effective. A particularly interesting case arises when the moderator and decision maker completely disagree about which action should follow which signal. If there are at least three states, this can be caused by only small differences in how information is interpreted. We provide necessary and sufficient conditions for the possibility of complete disagreement and examine the consequences for moderation and welfare. What might look to an outside observer like malicious misinformation can make the decision maker strictly better-off, yet completely misinformed.
-- link to working paper --
Information, Goal Setting and Performance: A Field Experiment
with Haoran He, Ronghai Su, and Qian Weng
We investigate how information impacts goal setting and performance by conducting a field experiment that varies whether students know or do not know their true relative ability when they set goals in a university physical education course based on relative performance rating. We document that most students set challenging goals, based on either their actual or estimated relative ability, proxied by their relative performance in a baseline test. However, only a small proportion of students estimate their performance accurately; under- and over-estimation are equally common. Consequently, consistent with our theoretical prediction, receiving information about one’s relative ability significantly improves goals – increasing (decreasing) goals set by students who underestimate (overestimate) their baseline relative test performance. We find that providing information significantly raises both relative and absolute final test performance of those who underestimate their relative ability. However, we do not find any effect on performance for students who overestimate themselves.
-- link to working paper --
The Silver Lining of Crises – A Loss Aversion Based Model of Reform
We explore how and when crises can help firms, organizations, and societies undertake beneficial reforms. In our model, a loss averse decision maker decides whether she should undertake a new project (a reform), characterized by a sequence of cash-flows, or stick with the status quo. In normal times, the decision maker may not pursue a beneficial project, a project with a positive net-present-value, for she places a greater emphasis on losses than on (equal sized) gains. We show that a sufficiently bad crisis guarantees that she undertakes the most beneficial project and characterize when a crisis begets change. When choosing between a single project and the status quo, a crisis can only shape preferences for the better. When choosing among multiple projects, it may distort choices. However, the crisis will always push the decision maker towards implementing a project that is better than the status quo. Implications for economic reforms and policy changes are discussed.
-- link to working paper --
Incomplete Adaptation of Reference Points
People get used to their environment, their consumption, the stimuli they face. However, evidence suggests that they do not adapt to all stimuli equally. We develop a theoretical model of adaptation in order to explore the behavioral implications of different amounts of adaption. In our model, a myopic, loss-averse decision maker repeatedly chooses from a set of outcomes. When making a decision, she compares those outcomes against her reference point. Her reference point either fully or incompletely adapts to her choice. Under full-adaptation, her current choice becomes the reference point for her next decision. Under incomplete adaptation, the reference point only adapts in some but not all dimensions to her choice. We show that both forms of adaptation result in complete, transitive, and consistent preferences and that preference can be cyclical. In settings that comprise outcomes with a benefit and a cost dimension, we show that a decision maker who does not adapt to costs chooses outcomes with weakly more benefits and costs than a decision maker who fully adapts. If the decision maker does not adapt to benefits, she will choose outcomes with weakly less benefits and costs than a decision maker who fully adapts. Applications include labor-choice problems, job switching, mental health, and whether people consume and work too much.
-- email me for an up-to-date version --
A Theory of Reciprocity with Trust
People are reciprocal if they reward kind actions and punish unkind ones. I propose a new theory of intention-based reciprocity that addresses the question of when a mutually beneficial action is kind. When both benefit from the action, a player's motive is unclear: he may be perceived as kind for improving the other player's payoff, or as self-interested, and not-kind, for improving his own. I use trust as an intuitive mechanism to solve this ambiguity. Whenever a player puts himself in a vulnerable position by taking such an action, he can be perceived as kind. In contrast, if this action makes him better off than his alternative actions do, even if it is met by the most selfish response, he cannot be kind. My model explains why papers in the literature fail to find (much) positive reciprocity when players can reward and punish. In particular, I show how negative reciprocity crowds out positive reciprocity. By allowing for interactions between rewards and punishments, my model provides a theoretical framework to analyze institutional design and incentive structures when people are motivated by reciprocity.
-- email me for an up-to-date version --