Reference-Dependent Effort Provision under Heterogeneous Loss Aversion: Pre-Analysis Plan
How to cite
Kellogg, Alexandre; Campos-Mercade, Pol; Goette, Lorenz; Sprenger, Charles. 2021. "Reference-Dependent Effort Provision under Heterogeneous Loss Aversion: Pre-Analysis Plan." AEA RCT Registry. AEARCTR-0007277. https://doi.org/10.1257/rct.7277-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In this study, we design a two-stage experiment to test the comparative static predictions of a leading model of reference-dependent preferences. In particular, we derive predictions from Koszegi and Rabin (2006, 2007) (henceforth KR) for both loss averse and gain loving individuals, and demonstrate how inconclusive average treatment effects from prior studies could be confounded by heterogeneity in gain-loss preferences. We then prepend a first stage to the classic experimental tests of KR in the real effort domain (Abeler et al, 2011; Gneezy et al, 2017) intended to measure individual gain-loss preferences, and examine the difference in treatment effects by type in our second stage.
Intervention
Our intervention shifts the expectations of lab participants regarding their earnings. In particular, they will be presented compensation schemes (in random order) such that they earn 20 cents per task with probability 0.5, $0 with probability p, $10 with probability q. Our intervention is shifting q from 0.05 to 0.45, which should effect the effort participants are willing to put forth as predicted by KR.
Experimental design
Our experiment asks participants how much they are willing to work at various wages. To measure preferences, we first offer a menu of deterministic wages (eg 10 cents per task) as well as stochastic (eg 0.5 chance of 0 cents per task, 0.5 chance of 20 cents per task). Then, we present our intervention described above. In particular, they will be presented compensation schemes (in random order) such that they earn 20 cents per task with probability 0.5, $0 with probability p, $10 with probability q. Our intervention is shifting q from 0.05 to 0.45, which should effect the effort participants are willing to put forth as predicted by KR. Finally, we elicit preferences using standard monetary Multiple Price Lists, and pick a decision that counts at random.
Design details
Cited by
No citation data available for this registration yet.