Skip to main content

Eliciting Preferences over Saving and Borrowing

Periodic Reporting for period 1 - DEBT (Eliciting Preferences over Saving and Borrowing)

Reporting period: 2018-09-01 to 2020-08-31

Borrowing and saving decisions are among the most important and economically significant choices people face in their lifetime. An unwillingness to save may have severe economic implications such as insufficient retirement savings. In the same way, borrowing too much or too little can have negative economic implications. Debt aversion, defined as an unwillingness to take on debt even if economically beneficial, has received increased attention by researchers lately, for its adverse effects on financial decision-making, such as preventing optimal portfolio choice or a failure to invest in profitable investment projects. While recent experimental studies do find evidence supporting the general existence of debt aversion on an aggregate level, this phenomenon remains poorly understood on the individual level. No study so far explicitly measures individual differences in the degree of debt aversion. Moreover, debt aversion is likely to interact with other domains of preference, such as preferences over time, risk and losses. These links have not yet been analyzed, neither theoretically nor empirically.
The overall aim of this project was thus to significantly enhance the understanding of borrowing and saving decisions. Specifically, this project aimed to establish whether preferences over saving and borrowing (such as debt aversion) are preferences in their own right, that cannot be explained with other preferences such as preferences over time, risk and losses. Further, this project aimed to explore how preferences over saving and borrowing interact with other domains of preferences, such as risk and time preferences.
To achieve the objectives of this project, work package 1 consisted of the design of experimental protocol that allows to cleanly elicit and estimate preferences over borrowing and saving. Moreover, in this work package, a battery of survey items on debt aversion as well as borrowing and saving behavior was created (partly new creations, partly compiled from the literature). The answers to these questions were used to identify a quick and short module of survey questions that best predict behavior in the incentivized laboratory experiment. This module can be used in the future to test debt aversion in situations where complex and time-consuming, incentivized laboratory experiments are not feasible.
Work package 2 consisted of formulating the theoretical backbone of the project: the aim was to write a model that extends prospect theory to include differential preferences over saving and borrowing. Agents in this model may dislike debt. This model is crucial in identifying debt aversion using the data collected in work package 1, as it is the basis of the structural maximum likelihood estimation. This model was then used to estimate preference parameters for risk aversion, present bias, time discounting, loss aversion and debt aversion jointly.
The preliminary data analysis suggests that participants on average dislike being in debt. Importantly, the employed methods allow to conclude that this phenomenon cannot be explained with time discounting, present bias, risk aversion or loss aversion or a combination of all: debt aversion appears to be a preference in its own right.
The experiments conducted for this action allow for the first time to cleanly identify debt aversion, and to establish debt aversion as a preference in its own right. Formalizing preferences over saving and borrowing theoretically and estimating these preferences has also not been tried before. The theory of debt aversion developed for this project may therefore serve as a benchmark model for future work on this topic.
The project has also introduced methodological innovations. Previous studies (including my own work) on debt aversion usually feature highly stylized experimental implementations of life-cycle models of consumption. These experiments typically have no “real” time dimension, as they are conducted in one session only, and consumption smoothing motives (and thus motives to save/borrow) are induced artificially. While this setup is well suited to make a qualitative inference whether debt aversion exists or not, it is not well suited to estimate subjects’ corresponding inherent preference parameters. In contrast, in the experimental developed in this project, subjects choose over real saving and debt contracts defined over real money and time. This is a novel approach and should substantially increase external validity compared to the existing literature.
Regarding the broader socio-economic impact, this project may be a starting point to understand when and why people may refrain from borrowing even when economically beneficial. Understanding this is important for policy makers who use loan programs to achieve some desired investment goals. Examples include subsidized loans for investment in energy efficient technologies or subsidized student loans. The success of these programs depends crucially on the uptake of these loans. A better understanding of the mechanisms behind debt aversion may therefore help to design better policies.