Household Liabilities Puzzles

The liabilities side of household balance sheets exhibits a parallel set of behavioral puzzles to the asset side. Three are central:

1. The credit-card debt puzzle. Many U.S. households simultaneously hold high-interest credit-card debt (APRs of $15$ – $25%$ ) and low-yield liquid assets (checking accounts, savings accounts paying $0$ – $2%$ ). Under any rational benchmark, these households should pay down the debt with the liquid assets and earn the spread. The puzzle generalizes: many households also hold credit-card debt alongside retirement assets that they could in principle borrow against or withdraw from at much lower effective cost. Gross & Souleles (2002) document the puzzle and find it persists even after controlling for liquidity needs, future borrowing constraints, and obvious institutional frictions.

2. Mortgage choice puzzles. Households frequently choose mortgages that appear suboptimal under standard analysis. Examples: choosing high-fee, no-points loans when low-fee, points-required loans would be cheaper given typical holding periods; failing to refinance when rates fall by enough to easily cover refinancing costs; choosing adjustable-rate mortgages (ARMs) just before interest-rate spikes and fixed-rate mortgages (FRMs) just before rates fall; and defaulting at rates inconsistent with any pure financial-incentive analysis (some households default with positive equity; others stay current with deeply negative equity).

3. Payday loans. Payday loans charge effective annual rates of $200$ – $700%$ , vastly above any other consumer credit. The standard rational defense is that payday loans serve a temporary liquidity need that no other product can meet. Behavioral critics argue that the loans are a debt trap — borrowers roll them over repeatedly, paying fees that quickly exceed the original principal — and that demand reflects present bias and over-optimism about repayment more than rational liquidity smoothing. Empirical work is mixed; some studies (Morse 2011) find payday loans help in genuine emergencies, others (Melzer 2011) find they worsen long-run financial outcomes.

Aggregate context. Total U.S. household debt rose from $50%$ of GDP in $1970$ to $80%$ of GDP in $2017$ , with credit-card balances and mortgage debt accounting for most of the growth. The behavioral puzzles are therefore not edge cases — they describe choices made by tens of millions of households about trillions of dollars of debt, and the welfare implications of getting them wrong are correspondingly large.

Intuition

A household’s balance sheet has two sides, and rational financial planning treats them jointly. If you have $ $5, 000$ in a savings account earning $1%$ and $ $5, 000$ in credit-card debt costing $20%$ , the rational move is to net them out and save $19%$ on $ $5, 000$ — a guaranteed $ $950$ per year. Households who do not do this are not making a complicated mistake; they are leaving money on a table that requires no skill, no risk-taking, and no information to claim. The credit-card debt puzzle is therefore a particularly stark challenge to rationality, because it does not require any sophisticated optimization. It just requires looking at both sides of the balance sheet at once.

The leading explanation is mental accounting — the tendency to evaluate each financial position separately rather than as part of a unified portfolio. The household perceives the savings account as “emergency reserve” and the credit-card balance as “this month’s spending overrun.” The two are kept in separate mental compartments and never compared, even though the unified-balance-sheet view shows that the savings account earns far less than it costs to maintain. Mortgage choice puzzles and payday loans reflect related forces: present bias (heavily discounting the future cost of high-interest debt), complexity aversion (refinancing requires effort), over-optimism (assumption that the next paycheck will repay the payday loan), and framing (evaluating each loan as a standalone transaction rather than as part of a multi-month sequence).

Properties

(Credit-Card Debt + Liquid Assets) The two coexist for tens of millions of U.S. households.

Gross & Souleles (2002) use the Survey of Consumer Finances to document that around one-third of U.S. households with credit-card debt also hold sufficient liquid assets to pay it off. The average “puzzle household” carries about $ $5, 000$ in card debt at a $14%$ APR while holding more than $ $5, 000$ in checking and savings at near-zero rates. The annual implicit cost is several hundred dollars per household, and the puzzle persists across the wealth distribution — not just among the financially constrained.

(Liquidity-Need Defense Fails) The pattern survives controls for emergency reserves.

A possible rational defense is that the household needs the liquid assets as a buffer against income shocks and cannot use them to repay debt because the debt would have to be re-incurred. Gross & Souleles test this directly by looking at households whose liquid assets exceed any reasonable buffer estimate (e.g., $6$ months of expenses). The puzzle persists even there, and even after controlling for the option value of having unused credit limits as future liquidity. Mental accounting, not rational liquidity management, is doing the work.

(Refinancing Failures) Households leave money on the table by not refinancing.

Agarwal, Driscoll & Laibson (2013) compute optimal refinancing thresholds for U.S. mortgage holders and find that a substantial share of households fail to refinance even when the interest savings vastly exceed any reasonable refinancing cost. Andersen, Campbell, Nielsen & Ramadorai (2020) replicate the result using Danish microdata where refinancing is essentially costless and find similar patterns. The estimated welfare cost runs into thousands of dollars per household over the life of a mortgage. The bias correlates with low financial literacy and with cognitive load measures like income shocks and medical events — exactly the markers a behavioral story would predict.

(ARM-vs-FRM Timing) Households choose the wrong mortgage type at the wrong time.

The choice between adjustable-rate and fixed-rate mortgages has a clear empirical pattern: ARM share rises just before interest rates spike (when locking in is most valuable) and falls just before rates decline. Households appear to extrapolate the recent rate environment forward, choosing whichever product worked best last year. This is the household-finance analogue of the extrapolative beliefs pattern documented in equity markets. The cost of the mistake compounds over the $20$ – $30$ year horizon of a mortgage and can reach tens of thousands of dollars in present value.

(Default Patterns) Some households default with positive equity, others stay current with negative equity.

A purely strategic-default model predicts that households default whenever the value of their home falls sufficiently below the mortgage balance. Empirically, many households default while still in positive equity (typically driven by income shocks combined with cash-flow pressure), while many others stay current despite deep negative equity (driven by social norms, attachment to the home, and the disutility of moving). The mismatch between observed defaults and strict cost-benefit calculations is consistent with reference-dependent preferences and norm-driven behavior, not pure financial optimization.

(Payday-Loan Debate) Real liquidity provision vs. debt trap.

Payday loans are the most extreme high-interest household credit product, with effective annual rates of $200$ – $700%$ . Two empirical camps disagree on whether they help or hurt borrowers. Morse (2011) uses California disaster data to show that access to payday loans reduces foreclosures and small-business failures during temporary shocks — consistent with rational liquidity smoothing. Melzer (2011) uses cross-state variation in payday-loan availability to show that access is associated with increased delinquency on other bills, increased food stamp usage, and worse self-reported well-being — consistent with a debt trap. The discrepancy may reflect heterogeneity: payday loans help genuinely-temporary cash crunches and harm households that roll over repeatedly. Behavioral mechanisms — present bias, over-optimism about repayment — explain why the same product can be rationally useful and behaviorally damaging at the same time.

Examples

A typical credit-card-debt-puzzle household

A household has $ $3, 000$ in checking earning $0.1%$ , $ $2, 000$ in a savings account earning $1%$ , and $ $4, 000$ on a credit card at $18%$ APR. Annual interest income: $ $23$ . Annual interest cost: $ $720$ . Net cost of carrying the debt instead of paying it down: $ $697$ per year. The rational move is to pay $ $4, 000$ from the savings/checking accounts toward the card balance, retaining $ $1, 000$ as a buffer. The household does not do this, partly because the savings account is mentally labeled “emergency reserve” and the credit-card balance is mentally labeled “to be paid down next month.” The mental compartments never communicate, and the household pays $ $697$ per year for the privilege.

Refinancing failure costs (Andersen et al. 2020)

Andersen, Campbell, Nielsen & Ramadorai (2020) study Danish mortgage refinancing, where the institutional environment makes refinancing nearly costless. They construct an optimal-refinancing benchmark and compare it to actual household behavior. They find that about $20%$ of households are below the optimal-refinancing threshold by enough to save substantial money but do not refinance — leaving an average of about $\$$30{,}000$ in present-value savings on the table over the life of the mortgage. The propensity to fail to refinance is concentrated among less-educated and lower-income households, suggesting that financial sophistication and attention costs (not pure economic optimization) drive the gap. The Danish setting is especially clean because it strips away the U.S. complications (refinancing fees, prepayment penalties, credit-quality concerns) that might rationalize the same patterns elsewhere.

tags	domain/finance, type/concept, level/intermediate, status/complete
aliases	Credit Card Debt Puzzle, Mortgage Choice, Payday Loans

Project Alexandria

Explorer

Household Liabilities Puzzles