Summary
The screening issue focuses on a decision-maker who vets a broad set of elements using noisy unbiased assessments to maximize chosen items’ value. With this in mind, the individual creates a threshold level (also called an acceptance criterion) that the screened rudiments must meet to be embraced. When the screening is stricter, the quantity of options – and their average value – reduces. In other words, setting the bar high does not guarantee the quality, while a lower one offers a win-win situation regarding the (alleged) quality and quantity trade-off. Unbiased noise’s influence over different values remains the main driving force behind this state of affairs. A collection of average elements produces relatively more noisy scores that pass a higher bar when subjected to unbiased noise. Over a tinier collection of more superior rudiments, a similar effect remains relatively mild. In this regard, the screening action seems to bias the effect of balanced noise.
Here, the critical assumption is that the decision-maker has implemented a threshold strategy. Unfortunately, scientists have not yet established such strategies’ optimality despite their continued utilization in theory and practice. It leads to the paper’s second primary outcome; screening’s optimal threshold strategies’ characterization. The authors relate (conditionally on different-level noisy assessment) optimal level strategies to the original valuation’s first- and second-order stochastic dominance. They then apply their characterization to previous findings on the Peter Principles and derive policy implications about affirmative action. The authors’ description suggests that eradicating affirmative action could result in a suboptimal outcome. The rest of the paper provides a screening model, beginning with the robustness of the involved biases. Their first theorem is that every bounded impact variable contains tremendous screening biases. They also claim that every impact variable has a never-ending noise variable that produces screening biases.
Two assumptions underpin the authors’ works in creating and proving the theorem, claim, and lemma. The first one is that the decision maker’s goal is the maximization of the impact variable’s value. The second one is that the decision-maker utilizes a threshold strategy in the selection of elements. Even so, at their discretion, the decision-makers can utilize more sophisticated and complex policies to maximize an impact variable-dependent utility function. Thus, the authors also extend the basic model they offer in the beginning of the paper by considering a supposed utility maximizer decision maker facing a screening problem. Here, the authors’ goal is to understand the conditions under which threshold strategies remain optimal. They also examine practical implications in cases where these threshold strategies are not optimal. They establish the correlation between the optimality of threshold strategies used by decision-makers and the initial analysis of screening biases.
Evaluation
Novelty and Contribution
The primary screening model presented by the authors has no close resemblance to any recently published work. Indeed, the only publication closely related to what the authors provide is the seminal work of Stiglitz and Weiss (1981), which considers the function of interest rates regarding screening. Like the authors, Stiglitz and Weiss (1981) found no necessarily monotone results concerning interest rates. For instance, once rates rise, some safe borrowers may stop taking loans, which could negatively impact the bank’s bottom line. The difference between the two studies is that while Stiglitz and Weiss (1981) hold in equilibrium each time a customer reacts strategically to the interest-rate mechanism, the authors show that non-monotonicity may arise when clients lack better information than lenders. The second part of the authors’ work also correlates strongly with Lazear’s seminal publication (2004). Like the authors’ work, Lazear (2004) offers a theoretical basis for the Peter Principle. Thus, both prove that consecutive noisy screening causes an upward bias.
Validity of Assumptions
The two assumptions underpinning the authors’ work are valid based on a simple mathematical fact. Without a doubt, while threshold criteria are intuitive, they perform somewhat poorly as screening strategies. For example, although a higher bar strategy may provide a threshold that job applicants must meet to progress to the next phase of recruitment (or get accepted at a company), it would reduce the company’s (decision maker’s) acceptance ratio while simultaneously lowering the expected average value of applicants moving to the next stage. As such, the higher bar does not contain quality assurance or guarantee. Indeed, a lower one may prove better as it produces a win-win situation when it comes to dealing with the issue of quality and quantity trade-off.
Comment
Although the article contains a reliable model and realistic assumptions, it focuses extensively on creating and proving the model’s theorems and claims. In the future, the authors should consider creating models and explaining their application in real life. Using real-life examples to test models may result in better application of techniques to reduce screening bias. Notably, situations in the real world do not always follow a mathematical formula or model as they are diverse and varied. Therefore, practical solutions to the screening bias may only come from the careful consideration of real-life situations in the future and offering recommendations based on these observations. Another issue the authors should solve is explaining plainly how individuals can avoid the screening bias problem.
Works Cited
Lagziel, David, and Ehud Lehrer. “A Bias of Screening.”American Economic Review: Insights, vol. 1, no. 3, 2019, pp. 343-356.
Lazear, Edward P. “The Peter Principle: A Theory of Decline.”Journal of Political Economy, vol. 112, no. 1, 2004, pp. 141-163.
Stiglitz, Joseph E., and Andrew Weiss. “Credit Rationing in Markets with Imperfect Information.”The American Economic Review, vol. 71, no. 3, 1981, pp. 393-410.