The Credit Card Debt Puzzle: Why People Hold Savings and Debt at Once
The Credit Card Debt Puzzle: Why People Hold Savings and Debt at Once
A household with $4,000 saved at 4% and $4,000 owed at 24% pays $800 a year for the arrangement, with net worth unchanged either way. The standard model says this should not happen, and it happens constantly. Three explanations compete. Priced, the liquidity explanation is worth about $74 a year. Self-control explains far more, and it changes whether co-holding is a mistake at all.
What you'll learn
The puzzle
A household holds $4,000 in a savings account earning 4% and carries a $4,000 credit card balance at 24%. Paying off the card with the savings would leave its net worth unchanged and save $800 a year in interest. It does not do it.
Economists call this co-holding, and it is not rare. Gross and Souleles, working with account-level credit card data, documented that a substantial share of households revolving card debt at high rates simultaneously held liquid assets that could have retired it. Under the standard model of household finance, in which money is fungible and a household minimizes interest cost, the behaviour should not exist. It is common enough that it has its own literature, usually called the credit card debt puzzle.
Two explanations compete, and they lead to different conclusions about whether the household is making a mistake. One is that money is not fungible in the household's mind. The other is that money is not fungible in the household's circumstances. This article sets out both, prices each one, and argues that the second explains less than it appears and the first explains more than it is usually given credit for.
Mental accounting
Thaler's account of mental accounting holds that households assign money to separate categories, and treat money in one category as unavailable for the purposes of another even when it is physically identical. Savings is labelled for emergencies, a down payment, or simply as savings. The card balance belongs to a different account, the one for paying off over time. Moving money between them feels like a transaction with a cost, even when the arithmetic says it has none.
Shefrin and Thaler extended this into the behavioural life-cycle hypothesis: households treat current income, current assets and future income as three mental accounts with different propensities to spend, and the asset account is the least available. Draining savings to repay a card moves money out of the account the household is most reluctant to touch.
- The label carries protective value. A balance labelled as savings is harder to spend on impulse than a balance that has been used to free up a credit line.
- The categories have reference points. Per our present bias analysis, a household that has watched its savings rise has a reference level it is reluctant to fall below, whatever the interest arithmetic says.
- The cost is not salient. $800 a year arrives as twelve interest charges of about $67, none large enough to trigger a decision.
On this account, co-holding is a framing error with a known price. The household would, on reflection, prefer the $800, and the prescription is to collapse the accounts.
The liquidity explanation
The alternative explanation is that the household is not confused at all. Telyukova argued that co-holding reflects a genuine need for liquidity in a form credit cards cannot supply: rent, many utilities, some bills and a range of informal payments cannot be made by card, or can be made only at a fee. A household that pays off its card with its savings has converted cash, which pays for everything, into unused credit, which pays only for some things.
Price that. Suppose that in any given year there is a 15% chance of a shock that requires $4,000 of cash-only payments. A household that has emptied its savings into the card would cover that need with a card cash advance: a 5% fee and about 29% interest for three months, roughly $490.
| Item | Amount |
|---|---|
| Annual cost of co-holding | $800 |
| Cost of the cash-only shock without savings | $490 |
| Expected annual value of holding the cash, at 15% | $74 |
The insurance is priced at more than ten times its expected value. Make the assumption harsher: suppose the shock also cuts off the card entirely, so the household has no access to credit at all. For co-holding to break even, the consequence of being without liquidity in that event would have to exceed $5,333, which is $800 divided by the 15% probability. That is possible, for a household whose car, job and housing depend on paying one bill on time, and per our correlated failure analysis, that kind of cascade is real. It is not the typical case.
There is a second liquidity risk the household cannot control. Per our credit limit analysis, issuers cut lines in downturns, precisely when households need them, so a strategy of paying down the card and relying on the freed limit depends on that limit still being there. This is a real risk and it raises the value of cash. On the figures above, it still has to be very large to justify the premium.
The explanation that survives: self-control
Bertaut, Haliassos and Reiter proposed a third reading that combines the first two. Picture the household as two agents: an accountant, who manages money and would like to repay the card, and a shopper, who spends whatever credit is available. If the accountant pays off the card with savings, the shopper sees $4,000 of fresh credit and spends some of it. The household ends with a card balance again and no savings.
On this view, keeping the card balance outstanding is a deliberate way of keeping the credit line full, so the shopper cannot use it. The $800 a year is the price of a commitment device, and the comparison is not against the interest saved but against what the freed line would have cost.
- If the freed limit is never re-spent, paying down the card is simply correct, and the household loses $800 a year by not doing so.
- If the freed limit is fully re-spent, paying down the card leaves the same $4,000 card balance and no savings at all, which is strictly worse than co-holding.
- In between, the decision turns on how well the household knows itself.
This is why co-holding cannot be read as a mistake without knowing the household. Per our withholding analysis, the same logic justifies over-withholding taxes as a commitment device, and per our default settings analysis, people who know their own tendencies frequently choose arrangements that look irrational from outside and are rational given the self they will be tomorrow.
The self-control explanation is not a licence
That co-holding can be rational does not make it costless, and the cost compounds. The $800 a year continues for as long as the balance is held, and a balance kept revolving as a commitment device is also a balance accruing at 24%, which per our revolver analysis tends to grow rather than hold steady. If the household's real problem is that freed credit gets spent, the cheaper fix is to remove the credit, by lowering the limit or closing the card once it is repaid, rather than to rent the commitment at 24%.
Adjudicating between the explanations
The three explanations predict different things, which is what makes them testable.
| Prediction | Mental accounting | Liquidity need | Self-control |
|---|---|---|---|
| Co-holding among households with mostly cash-only expenses | No difference | Higher | No difference |
| Co-holding after a credit line cut | No difference | Higher | Lower |
| What happens when a household is nudged to pay down | Stays paid down | Re-borrows when a cash need arrives | Re-borrows steadily |
| Effect of automatic savings plans | More co-holding | More co-holding | Much more co-holding |
The third row is the experiment that decides the policy question. If households nudged to pay down their cards stay paid down, the framing explanation dominates and a nudge is a pure gain. If they re-borrow steadily, the self-control explanation dominates, and per our financial education analysis, telling them to pay down is advice that makes them worse off.
What households should do
- Keep a small cash buffer for cash-only obligations, roughly a month of rent and utilities. That is the liquidity the explanation genuinely justifies.
- Pay the card down with anything above the buffer. Per our prepayment analysis, the return is the card rate, which no savings account matches.
- If freed credit tends to get spent, remove the credit, not the payment. A lower limit costs nothing; co-holding costs $800 a year.
- Use a structure where you need one. Per our balance transfer analysis, a fixed schedule enforces the discipline that co-holding only rents.
- Watch for a line cut. If an issuer reduces your limit, the case for holding cash strengthens, and it is worth re-running the arithmetic.
Item one reconciles the explanations. A household with a modest cash buffer captures almost all of the liquidity value, and per our emergency savings analysis, a buffer of that size is what most households lack, while paying down the rest captures almost all of the interest saving. The puzzle is mostly a puzzle about the amount beyond that buffer, and on the figures above the amount beyond it should go to the card.
Frequently asked questions
Mental accounting, a genuine need for cash that cards cannot replace, and self-control, since paying down a card frees credit that may be spent again.
About $800 a year on $4,000 each side. The liquidity insurance it buys is worth roughly $74 a year on typical assumptions.
Yes, for a small buffer for cash-only bills, and where freed credit would be spent again. Lowering the limit is a cheaper commitment than paying 24%.
Keep about a month of cash-only expenses and pay the card with the rest. If freed credit gets spent, lower the limit rather than keep the balance.
Key takeaways
- Holding $4,000 at 4% while revolving $4,000 at 24% costs about $800 a year with no change in net worth.
- Mental accounting treats savings and debt as separate categories, which makes the cost invisible.
- The liquidity explanation is real but small: the cash buffer is worth about $74 a year on typical assumptions.
- Even if a shock also cut off the card, being without liquidity would need to cost over $5,333 to justify co-holding.
- Self-control explains more: paying down a card frees credit that may be spent again, leaving no savings and the same debt.
- That makes co-holding a commitment device, rational for some households, and an expensive one at 24%.
- Lowering the limit or closing the card is a cheaper commitment than keeping a revolving balance.
- Keep about a month of cash-only expenses as a buffer and put the rest toward the card.
References
- Thaler, R. H. (1985). Mental Accounting and Consumer Choice. Marketing Science, 4(3), 199–214.
- Shefrin, H. M., & Thaler, R. H. (1988). The Behavioral Life-Cycle Hypothesis. Economic Inquiry, 26(4), 609–643.
- Thaler, R. H. (1999). Mental Accounting Matters. Journal of Behavioral Decision Making, 12(3), 183–206.
- Gross, D. B., & Souleles, N. S. (2002). Do Liquidity Constraints and Interest Rates Matter for Consumer Behavior? Evidence from Credit Card Data. Quarterly Journal of Economics, 117(1), 149–185.
- Bertaut, C. C., Haliassos, M., & Reiter, M. (2009). Credit Card Debt Puzzles and Debt Revolvers for Self-Control. Review of Finance, 13(4), 657–692.
- Telyukova, I. A. (2013). Household Need for Liquidity and the Credit Card Debt Puzzle. Review of Economic Studies, 80(3), 1148–1177.
This analysis is educational and does not constitute financial advice. All figures are stylized illustrations using assumed balances, interest rates, shock probabilities, and cash advance terms, and do not describe any particular household or product. The cited studies are summarized for context; readers should consult the original works for their methods and findings. Consult a qualified adviser about your own circumstances.