Healthy Behavior: Essays in Health and Behavioral Economics
Name
shreekumar-adviks-phd-economics-2025-thesis.pdf
Description
Thesis PDF
Size
2.19 MB
Format
Adobe PDF
Checksum (MD5)
e4b51f3963fd2947601bd4510c6ac310
Author(s)
Shreekumar, Advik
Advisor(s)
Schilbach, Frank
Mullainathan, Sendhil
Rambachan, Ashesh
Date Issued
May 2025
Publisher
Massachusetts Institute of Technology
Abstract
These essays examine beliefs and decision-making in health settings, emphasizing the role of attention, information, and technology in shaping behavior. The first essay studies human error in chest x-ray interpretation, a common and consequential medical task. It casts radiologists as facing a classical decision-theory problem, derives a novel martingale test for optimal behavior, and implements this test through a prudent application of machine learning to anonymized health records from the Beth Israel Deaconess Medical Center. I find that 58 percent of radiologists make predictable mistakes when assessing cardiac health on chest x-rays. Roughly two thirds of errors are explainable as individual radiologists making inconsistent decisions, and one third reflect the possibility that algorithms detect novel or complex signals. The second essay studies app-based mindfulness meditation, which has grown popular due to claims about its effects on mental well-being, productivity, and decision making. We assess these claims an experiment with 2,384 US adults, randomizing access and usage incentives for a popular mindfulness app. App access improves an index of anxiety, depression, and stress at two weeks and four weeks, with persistent effects three months later. It also improves earnings on a focused proofreading task by 2 percent. The third essay studies a tradeoff governments face when making recommendations in an evolving crisis. We investigate the effect of taking an early position on how much people believe later recommendations, using an online experiment with 1,900 US respondents in early April 2020. We present participants with CDC projection about coronavirus death counts and randomize exposure to information that highlights how the President previously downplayed the threat. When the President’s inconsistency is salient, participants are less likely to revise their beliefs about death counts from the CDC projection. This aligns with a model of signal extraction from government communication, and has implications for changing guidelines in other settings. JEL Codes: D91, I12, C8
MIT Department
Massachusetts Institute of Technology. Department of Economics
Terms of Use
In Copyright - Educational Use Permitted
Copyright retained by author(s)
Persistent DSpace Link
