Motivation
As individuals in a growingly competitive world, we are in constant demand to improve and reflect on our capabilities. In job search and principal-agent settings, we often, as economic agents, compete for the opportunity to prove our worth to the economy — fulfill a job vacancy or obtain a promotion. Occasionally, we win these blood sports, but more often than not, we lose, leaving our hearts saddened and our minds in a rush to parse for a prospective truth. In a state of defeat, we have the convenience to request from the principal information about ourselves. That is, we can demand and avoid feedback about our performance.
Unfortunately, these signals do not come without a price. For the economic agent, we have to exert effort to sit down and write an email to the forgone principal requesting information that takes both cognitive and time-associated resources. As for principals, the production of signals does not benefit the firm since it was already determined the economic agent would not fulfill the job vacancy or the promotion. The lack of benefit on the firm’s behalf produces a situation where it is more likely that if signals are sent to economic agents that demand them, they could contain noise because they are cheaper to manufacture. Therefore, we as economic agents demand noisy rather than noiseless feedback about our performance in many situations. Conditional on the informational quality and the time spent obtaining feedback about ourselves, we can then update our beliefs that could improve or make worse decisions involving future effort provision or search behavior with micro or even macro implications.
This study looks at this domain of noisy information acquisition. However, it strips the design of any particular scope condition (application) to focus on the behavior of interest in isolation — acquisition rates and updating behavior.
Historically, a vast amount of literature focuses on the theory of informational preferences about self (Benabou & Tirole, 2002; Brunnermeier & Parker, 2005; Koszegi, 2006; Benabou & Tirole, 2011), but the experimental literature as lagged behind its theoretical counterparts. Seminal experimental works related to this study are (Eil & Rao, 2011) and (Robalo & Sayag, 2018). While (Eil & Rao, 2011) were interested in updating behavior and acquiring information, they studied the two under unrealistic parameters. In particular, the choice to obtain a noisy signal about performance was not a voluntary decision, nor was it costly. Second, the acquisition of performance feedback at a cost was featured as a noiseless signal, not noisy. Instead, I seek to make obtaining the noisy signals themselves costly and voluntary. To control self-selection effects, I adapt the treatment framework of (Robalo & Sayag, 2018) to fit an ego-relevant environment.
Broadly, this study asks two questions.
-
Who is willing to acquire noisy information at a price, by how much, and under what conditions?
-
Conditional on any informational investment, how do subjects update their beliefs about their performance? That is, do sunk cost effects have an impact on updating behavior about self (Thaler, 1980; Eyster, 2002)?
Experimental Design
In my design, participants engage in a four-part experiment similar in flavor to (Coffman et al., 2019b) that should, on average, take 22 minutes to complete. Parts 1, 2, and 4 are all windfall incentivized with similar expected earnings.
Part 1
In Part 1, subjects will complete a 20 question ASVAB quiz with similar characteristics to that used in (Exley & Kessler, 2019).
Part 2
In Part 2, subjects will be elicited their prior beliefs over the entire outcome distribution and a point-specific valence question relating to their subjective beliefs.
Part 3
In Part 3, participants face the central part of the experiment of acquiring a signal. Subjects face two rounds of acquisition decisions. Each decision task is preceded by instructions explaining the information structure that contains details about the likelihood of receiving their correct score versus some noisy score as a signal. After subjects read the instructions, they make a binary decision about whether they would be willing to pay a cost of c to observe the signal. After completing both rounds, one of the rounds is randomly selected to be the round-that-counts. In the round-that-counts, the subject’s decision will be executed based on that round’s contents. Participants will then pay the cost before obtaining the signal. The price is in the form of pressing a button to compliment the motivation of cognitive and time associated costs rather than explicit monetary costs.
Below you will find an outline of the treatment variation in Part 3.
Conditions
Within-Subject
Signal Precision (assigned in random order):
- p = 0.90: A 90% chance the signal participants acquire shows their actual score, and a 10% chance it shows a noisy score.
- p = 0.60: A 60% chance the signal participants acquire shows their actual score, and a 40% chance it shows a noisy score.
The noise is randomly drawn from a uniform distribution between -5 and 5 and additively attached to the subject’s actual score.
Between-Subject
Type of Cost (randomly assigned):
- Free: Participants can obtain a signal at no cost (c = 0).
- Costly: Participants have to pay a cost of c to observe the signal.
- Compulsory: Participants have to pay a cost of c irrespective of their decision to observe the signal.
Degree of cost (randomly assigned):
The degree of the cost will only apply to two treatments: Costly and Compulsory. The range of cost consists of:
c = {100k | k = 1, 2, …, 10}
Part 4
Once participants observe the signal, they update their beliefs in Part 4, analogous to the task in Part 2.
After subjects finish Part 1-4, I elicit risk preferences, demographics, external instrumentality, and stereotype questions.
Sample Size and Costs
Based on previous pilots, the sample would need roughly 1500 subjects to have enough variation in the exogenously assigned costs for both Costly and Compulsory treatments to have comparable sub-samples.
The overall cost of the experiment should equate to:
Average cost per subject = $2.80 ($7.63/hr) x 1.33(Prolific fee) = $3.56
1500 x $3.56 = $5340.00 or ÂŁ3827.45
References
BĂ©nabou, R., & Tirole, J. (2002). Self-confidence and personal motivation. The quarterly journal of economics, 117 (3), 871-915.
BĂ©nabou, R., & Tirole, J. (2011). Identity, morals, and taboos: Beliefs as assets. The Quarterly Journal of Economics, 126 (2), 805-855.
Brunnermeier, M. K., & Parker, J. A. (2005). Optimal expectations. American Economic Review, 95 (4), 1092-1118.
Coffman, K., Collis, M., & Kulkarni, L. (2019). Stereotypes and belief updating. Harvard Business School.
Eil, D., & Rao, J. M. (2011). The good news-bad news effect: asymmetric processing of objective information about yourself. American Economic Journal: Microeconomics, 3 (2), 114-38.
Exley, C. L., & Kessler, J. B. (2019). The gender gap in self-promotion (No. w26345). National Bureau of Economic Research.
Eyster, E. (2002). Rationalizing the past: A taste for consistency. Nuffield College Mimeograph.
Köszegi, B. (2006). Ego utility, overconfidence, and task choice. Journal of the European Economic Association, 4 (4), 673-707.
Robalo, P., & Sayag, R. (2018). Paying is believing: The effect of costly information on Bayesian updating. Journal of Economic Behavior & Organization, 156, 114-125.
Thaler, R. (1980). Toward a positive theory of consumer choice. Journal of economic behavior & organization, 1 (1), 39-60.
Pre-Registration
A pre-analysis will be available shortly at AsPredicted. I will comment on this post or edit the original with the relevant link when completed.
Open Science
All data and code files will be available after successful publication in an open-access peer-reviewed general-interest economics journal. Specifically, I will host all relevant data and code on the publishing platform and my website.
Ethics
This study is approved by the University of Exeter Business School (UEBS) IRB. The document showing proof will be attached: click here.