Intertemporal Choice Experiments and Large-Stakes
Behavior
⇤
Diego Aycinena
Facultad de Econom ́ıa
Universidad del Rosario
diego.aycinena@urosario.edu.co
Szabolcs Blazsek
Escuela de Negocios
Universidad Francisco Marroqu ́ın
sblazsek@ufm.edu
Lucas Rentschler
Department of Economics and Finance
Utah State University
lucas.rentschler@usu.edu
Charles Sprenger
Department of Economics and Rady School of Management
University of California, San Diego
csprenger@ucsd.edu
December 8, 2019
Abstract
Intertemporal choice experiments are frequently implemented to make inference about
time preferences, yet little is known about the predictive power of resulting measures.
This project links standard experimental choices to a decision on the desire to smooth a
large-stakes payment — around 10% of annual income — through time. In a sample of
around 400 Guatemalan Conditional Cash Transfer recipients, we find that preferences
over large-stakes payment plans are closely predicted by experimental measures of patience
and diminishing marginal utility. These represent the first findings in the literature on
the predictive content of experimentally elicited intertemporal preferences for large-stakes
decisions.
JEL classification:
D1, D3, D90
Keywords
: Structural estimation, Out-of-sample prediction, Discounting, Convex Time Budget
⇤
We thank Michaela Pagel for helpful comments. Diego Aycinena grateful for the financial support provided by Fundaci ́on
Capital and the program “Inclusi ́on productiva y social: programas y pol ́ıticas para la promoci ́on de una econom ́ıa formal”, code
60185, which conforms the
Alianza EFI - Econom ́ıa Formal Inclusiva
,undertheContingentRecoveryContractNo.FP44842-220-
2018. We are grateful to Pablo Pastor, Alvaro Garcia and the field implementation team; Raul Zurita, Arturo Melville and Amy
Ben ́ıtez. Betzy Sandoval provided excellent research assistance, including field supervision, data handling and support in the project
logistics and design implementation.
1 Introduction
Intertemporal choice is prevalent in economic decision making. Explicit characterization of
structural discounting models has led to widely appreciated theoretical developments (
Samuel-
son
,
1937
;
Koopmans
,
1960
;
Laibson
,
1997
;
O’Donoghue and Rabin
,
2001
). Measurement of
the broad forces of these models and corresponding utility parameters has received deserved
empirical attention as well, with notable contributions in both laboratory and field settings.
1
A prominent discussion related to the measurement of intertemporal preferences has devel-
oped in the last decade.
Frederick, Loewenstein and O’Donoghue
(
2002
)noteacriticalissue: the
confounding e
↵
ects of diminishing marginal utility for making inference on patience. A decision
maker who is indi
↵
erent between $45 today and $50 in one month can be arbitrarily impatient
depending on changes in utility over this range.
2
Diminishing marginal utility confounds both
quantitative and qualitative predictions of measured discounting models. Erroneous inference
on the level of discounting will lead to poor quantitative out-of-sample prediction; while at-
tributing to discounting a behavior that is truly driven by diminishing marginal utility may
lead even qualitative predictions to be incorrect.
Recognizing this confounding e
↵
ect, recent work has developed experimental methodology
to identify diminishing marginal utility alongside discounting parameters (
Andersen, Harrison,
Lau and Rutstrom
,
2008
;
Andreoni and Sprenger
,
2012
).
3
Although this work shows that
estimates of discounting are indeed influenced by the shape of utility, it is unknown whether
the corresponding parameters meaningfully predict large-stakes behavior. Given that out-of-
1
Examples include
Hausman
(
1979
);
Lawrance
(
1991
);
Warner and Pleeter
(
2001
);
Harrison, Lau and
Williams
(
2002
);
Cagetti
(
2003
);
DellaVigna and Malmendier
(
2006
);
Laibson, Repetto and Tobacman
(
2007
);
Andersen, Harrison, Lau and Rutstrom
(
2008
);
Mahajan and Tarozzi
(
2011
);
Andreoni and Sprenger
(
2012
);
Fang and Wang
(
2015
).
2
An individual indi
↵
erent between $45 received now and $50 received in one month under exponential
discounting reveals
u
(45)
/u
(50) =