Seeing the Cliff Before You Fall: The Cognitive Cost of Misperception

• In Event: New Research on Administrative Burdens

Abstract

Means-tested benefits phase out as earnings rise, so a raise can leave a lowincome
household with fewer resources, a situation known as a benefit cliff. In a
pre-registered experiment, 458 benefit recipients evaluate a hypothetical job offer
built from the program rules for their own household, and I randomize how
those same rules are delivered. Workers facing a cliff misperceive the monthly
change in resources by $768 on average, and the direction of the misperception,
rather than its size, is associated with recommending the offer. Delivering the
rules through a conversation with an AI assistant instructed not to state the net
change, rather than through the Atlanta Fed dashboard, reduces the misperception
by $303 a month, while the dashboard leaves beliefs largely unchanged.
The results suggest that the way already-public benefit information is delivered
affects whether workers can use it, and the pattern of corrections points to
guided reasoning as a plausible explanation.

Author

• Daniel Alejandro Tascon-Buchelly, University of Pittsburgh
  Presenting Author