Periodic Reporting for period 2 - BEHAVFRICTIONS (Behavioral Implications of Information-Processing Frictions)
Reporting period: 2019-12-01 to 2021-05-31
A good example of this approach is my paper titled “Attention Please!”, written jointly with Olivier Gossner and Colin Stewart as a part of this grant project, which is forthcoming in Econometrica. In this paper, we propose a novel mechanism that explains why directing the attention of a decision maker to a particular option increases the chance that she will choose this target option, even if the target is of low quality and her attention is thus likely to lead to negative impressions. Our model is motivated by attention-grabbing techniques used in marketing and advertising. This new procedural model explains how such attention manipulations can affect demand. If broadly accepted by the economic profession, the model could become a framework for discussing regulations on advertising and marketing.
As a response to the Covid crisis, I have devoted a part of my research time in 2020 to Covid testing strategies. Similarly to the information-processing frictions in individual decision-making, information-acquisition resources – Covid tests—are scarce at the societal level. Medical tests are mathematically conceptualized in the same manner as information-acquisition strategies in economic theory – both are modelled as Blackwell experiments. I have thus been able to apply the modelling methods of my grant proposal to Covid policy problems. I have produced three academic papers on the allocation of various types of Covid tests to individuals and on rotation schemes that minimize the risk of Covid spread when monitoring of Covid outbreaks is limited.
1. “On the Cost of Misperception: General Results and Behavioral Applications”, joint with Olivier Gossner, published in the Journal of Economic Theory.
We explain perception biases as second-best adaptations when perception mistakes are unavoidable but their distribution can be optimized. The framework offers unified explanations of the so-called illusion of control, over-precision, and overweighting of small probabilities. This paper contributes to Pillar 1 of my proposal.
2. “Selective Sampling with Information-Storage Constraints”, joint with Philippe Jehiel, published in the Economic Journal.
We model sequential learning by a decision-maker with bounded memory. The optimal procedure of such a decision-maker features a bias in favour of news confirming her prior belief, whereas information that contradicts prior belief is likely to be forgotten. Thus, the model lays micro-foundations for confirmation bias. This paper contributes to Pillar 4 of my proposal.
3. “Habits as Adaptations: An Experimental Study”, joint with Ludmila Matyskova, Brian Rogers, and Keh-Kuan Sun, published in Games and Economic Behaviour.
We test the dynamic rational-inattention theory of Steiner, Stewart and Matejka (Econometrica 2017). Our paper experimentally confirms that habit-formation – the tendency to repeat one’s own actions – can be explained by optimization of information acquisition. Since one’s own action has likely been optimal in a past period, it is likely to be correlated with the current optimal action if the incentives are serially correlated. Repeating one’s own action thus can provide a decision-maker with a good current payoff without costly information gathering about current incentives. This paper contributes to Pillar 2 of my proposal.
4. “Attention Please!”, joint with Olivier Gossner and Colin Stewart, forthcoming in Econometrica.
We provide a novel mechanism explaining how directing attention to a particular item increases the chances that this target item will be chosen, even if the item is of low value and thus its inspection is likely to deliver disappointing information. The effect arises when the choice process is based on approval learning – that is, when the decision-maker chooses the item once she has formed a belief that it is good enough. In such cases, and when the decision-maker has a unit demand, attention-grabbing improves the chance that the target of the manipulation will be approved before other items in the choice set. This paper fits Pillars 3 and 4 of the grant proposal.
The following three papers were written in reaction to the Covid crisis. They combine the authors’ experience with Covid policy advisory activities and information-economics theory. The testing aspect of Covid mitigation fits the topic of my grant well, since the problem of rational inattention and the problem of optimal test allocation both consist of allocating scarce information-acquisition resources in a manner that induces the best possible decisions. Given the relevance of the BehavFrictions research methods to Covid testing, I have spent part of 2020 doing research on Covid testing. Papers (5.) through (7.) are outcomes of this effort.
5. “Optimal Test Allocation”, with Jeff Ely, Andrea Galeotti, and Ole Jann; revision requested by the Journal of Economic Theory.
We study optimal Covid testing when the authority holds a portfolio of tests which are heterogeneous in terms of precision, and the individuals to be tested differ in their chances of being infected. The paper introduces the economic concept of opportunity cost to this pressing policy discussion. When tests are in shortage, then assignment of a test to a particular person should be based on the comparison of the benefit of the test to this person with the tests’ opportunity cost – the benefit from the test under its second-best assignment.
6. “Rotation as Contagion Mitigation”, joint with Jeff Ely and Andrea Galeotti, forthcoming in Management Science.
This paper focuses on organizations such as schools that wish to organize the attendance of its members using rotation schemes. The paper studies how the frequency of such rotations impacts the epidemiological risk of contagion when testing is imperfect and Covid outbreaks are noticed with a delay. The paper derives optimal rotation frequency.
7. “Merit of Test: Perspective of Information Economics”, joint with Andrea Galeotti and Paolo Surico, published in Health Policy and Technology.
This short note aimed at medical professionals explains the value of testing from the perspective of economic theory.
(a) A paper titled “The Endogenous Bernoulli Utility Function”, joint with Nick Netzer and Arthur Robson, is in the first-draft stage. The project explains risk attitudes described in prospect theory – the s-shaped Bernoulli utility function and over-weighting of small probabilities – in a unified model that features encoding and decoding of stimuli generated by risky prospects in the decision-maker’s choice set.
(b) A joint project conducted with Blair Shevlin and Ian Krajbich will experimentally test my paper (4.). The experiment will use eye-tracking techniques and controlled attention manipulation, and will aim to verify whether attracting attention to an element of the decision-maker’s choice set increases the probability that she will choose the target. One of the aims of the project is to communicate ideas from paper (4.) to the Brain-Science community, into which the two co-authors are well integrated. The project is in the pilot-preparation stage.
Depending on progress with (a) and (b), I may be able to pursue additional projects:
(i) If possible, I will return to Pillar 2 and will seek empirical applications of the dynamic rational-inattention model.
(ii) Colin Stewart and I are developing a novel approach to construction of the demand and supply curves from transaction data. The standard construction of demand and supply curves assumes that the market participants are heterogeneous in their reservation prices but they all perfectly observe homogenous market prices. Instead, one could assume that some of the heterogeneity in trading decisions is explained by incomplete and heterogeneous information about the market prices. We attempt to derive what an econometrician can conclude about demand and supply curves from transaction data when the model allows for incomplete information about market price.
(iii) It is typically assumed that the parameters of the available Blackwell experiments are perfectly observed by economic agents. However, as with medical tests, the conditional signal distributions of various Blackwell experiments are often not known to the agents, and must be learnt using other Blackwell experiments. What can the agent learn about available Blackwell experiments under various assumptions about the available data?