[Paper Review] Debt Aversion: Theory and Measurement

Thomas Meißner, David Albrecht|arXiv (Cornell University)|Jul 15, 2022

Decision-Making and Behavioral Economics4 citations

TL;DR

This paper proposes a formal model of debt aversion and uses a lab experiment with real-money debt and saving contracts to jointly estimate debt aversion alongside time, risk, and loss preferences. Structural estimation reveals that 89% of participants are debt averse, requiring a median 'borrowing premium' of 16% of the principal to accept debt, establishing debt aversion as a distinct, quantitatively significant preference dimension.

ABSTRACT

Debt aversion can have severe adverse effects on financial decision-making. We propose a model of debt aversion, and design an experiment involving real debt and saving contracts, to elicit and jointly estimate debt aversion with preferences over time, risk and losses. Structural estimations reveal that the vast majority of participants (89%) are debt averse, and that this has a strong impact on choice. We estimate the "borrowing premium" - the compensation a debt averse person would require to accept getting into debt - to be around 16% of the principal for our average participant.

Motivation & Objective

• To determine whether debt aversion is a standalone preference or an emergent property of other preferences like risk or time discounting.
• To develop a formal model of debt aversion that isolates it from confounding factors such as loss aversion, risk preferences, and time preferences.
• To design and conduct a lab experiment with real financial incentives to elicit and jointly estimate debt aversion and related preference parameters.
• To quantify the economic impact of debt aversion by estimating the 'borrowing premium' individuals require to accept debt.

Proposed method

• Develop a structural model of intertemporal choice that incorporates debt aversion as a separate preference parameter.
• Conduct a three-session lab experiment involving real-money debt and saving contracts with 196 participants.
• Use maximum likelihood estimation to jointly estimate debt aversion alongside time, risk, and loss preferences from observed choices.
• Control for confounding factors by including choices involving future gains, losses, and intertemporal trade-offs in the same experimental design.
• Employ simulated maximum likelihood estimation to model the joint distribution of preference parameters and test for correlations across domains.
• Use a staircase mechanism in hypothetical debt choices to elicit thresholds of acceptable debt.

Experimental results

Research questions

• RQ1 Is debt aversion a distinct preference in its own right, or merely an emergent property of other behavioral biases like loss aversion or time discounting?
• RQ2 What is the magnitude of the 'borrowing premium'—the compensation required—for individuals to accept debt, controlling for other preferences?
• RQ3 How is debt aversion related to individual characteristics such as cognitive ability, financial literacy, gender, age, and personality?
• RQ4 Is there a correlation between debt aversion and other preference domains, such as loss aversion, risk aversion, or time preferences?
• RQ5 Does the duration of indebtedness increase the level of debt aversion?

Key findings

• 89% of participants are estimated to be debt averse, indicating that debt aversion is a prevalent and significant preference dimension.
• The average borrowing premium required to accept debt is 16% of the principal, meaning debt-averse individuals demand substantial compensation to incur debt.
• Debt aversion is positively correlated with loss aversion but not significantly related to risk aversion or time preferences.
• There is a weak negative association between cognitive ability and debt aversion: higher cognitive ability is linked to lower levels of debt aversion.
• Debt aversion increases with the length of time individuals remain indebted, suggesting a cumulative psychological cost of debt.
• The results are robust across multiple alternative model specifications, confirming the reliability of the estimated debt aversion parameter.