Microeconomic Model Analyses
Microeconomic Model Analyses
Loading...
Files
(RESTRICTED ACCESS)
(RESTRICTED ACCESS)
Publication or External Link
Date
2023
Authors
Ellis, Keaton Hyuckmin Kweon
Advisor
Ozbay, Erkut
Citation
Abstract
The first chapter, joint with Dr. Shachar Kariv and Professor Erkut Ozbay, compares thepredictive performance of a standard economic model to a variety of machine learning models by
presenting nearly 1,000 subjects with a consumer decision problem – the selection of a bundle
of contingent commodities from a budget set. Our dataset allows us to compare predictions
at the individual level and relate them to the consistency of individual decisions with revealed
preference axioms. Using dual measures of completeness and restrictiveness from Fudenberg
et al. (2022a,b), we show that the economic model outperforms all machine learning models,
with a wider margin as choices align more with an underlying preference ordering. The second
chapter, joint with Professor Emel Filiz-Ozbay and Erkut Ozbay, empirically investigates the
consideration and choice functions behaviors of individuals under uncertainty. Our design elicits
these functions by repeating the decisions repeatedly questioning subjects in a rich lottery domain
and, hence, allows subjects to reveal their stochastic or deterministic consideration and choice.
Since most subjects act stochastically in both consideration and choice decisions, we focus on
testing well-known axioms defined for such behavior. Our analysis includes individual-level
testing of the logit model (Brady and Rehbeck (2016)), and the axioms of monotonic attention
(Cattaneo et al. (2020)), and attention overload (Cattaneo et al. (2021)) for the consideration
data. For the choice data, we test properties including the independence of irrelevant alternatives
(Luce (1959)), regularity (Block et al. (1959)) and consistency with the attribute rule (Gul et al.
(2014)). The third chapter, joint with Dr. Shachar Kariv and Professor Erkut Ozbay, extends
work frmo the first chapter. We make use of rich individual-level data sets from three budgetary
choice environments. The environments provide a strong test of both the intra-economic model
comparisons, as well as a comparison between economic models and machine learning models.
Overall, we find that the extension from two goods to three goods does not greatly reduce
completeness, but does greatly increase the restrictiveness. Both standard and behavioral economic
models see larger increases in restrictiveness compared to machine learning models, and a lower
drop in completeness when moving from two goods to three goods. Surprisingly, there is no
additional drop in completeness when moving from choice under risk to choice under ambiguity
in this environment; the completeness and restrictiveness scores of all models are nearly identical
across the two domains, and the minor differences that are present favor models under ambiguity.
We interpret these results as favorable for standard economic models in rich choice environments:
absent external factors, economic models with one parameter detailing risk preferences are sufficient
to capture individual-level behavior of choice under risk and choice under ambiguity. Additionally,
these models are more restrictive than machine learning models; along with the high completeness,
this result indicates that the assumptions of EUT and SEU capture the regularities in choice under
risk and ambiguity.