Why a roommate's slowly growing debt never gets confronted

Why does a debt that grows slowly get a pass?

Because people judge each step, not the distance travelled. In four laboratory studies, Gino and Bazerman put participants in the role of “watchdogs charged with catching instances of cheating.” Watchdogs who saw cheating grow in small increments approved it far more often than watchdogs who saw the same cheating appear all at once. The authors call this the slippery-slope effect, and they show that part of it happens without the observer noticing.

The task was simple. Participants looked at photographs of a jar of pennies worth about $10 and were asked to approve or reject another person’s estimate of the amount. In the abrupt condition, the estimates were honest for ten rounds and then jumped to about $14 in Round 11. In the slippery-slope condition, the estimates rose by 40 cents a round from Round 2 until they reached the same $14, then stayed there. From Round 11 on, both groups were approving identical estimates — inflated by about 40%. The only difference was the path.

Over Rounds 11 to 16 of Study 1, with 76 participants, the mean approval rate was 52% on the slippery slope and 24% after the abrupt jump. Study 3 removed the money participants earned for approving, and the gap held. Study 4 reversed the incentive and paid participants to catch fraud; even then, approvals ran 19% on the slope versus 3% after the jump. Participants on the slope also decided faster and produced fewer ethics-related words in a word-completion task, which the authors read as evidence that the slow version never fully registered as a violation.

The authors reach for the boiling-frog folk tale and then state the finding plainly: third parties “might grow accustomed to unethical behavior if the latter develops gradually, and, as a result, accept such behavior.” A roommate who is owed money is a third party to their own ledger in exactly this way. Each month they approve one more small estimate.

Source: Gino & Bazerman, “When misconduct goes unnoticed”, JESP 45(4), 2009 — Study 1 method and Table 2 (N = 76, mean approval over Rounds 11–16), Study 3 no-incentive condition, Study 4 fraud-detection task (19% vs. 3%), the word-completion and reaction-time results, and the general discussion. The “about 40%” is our arithmetic on the paper’s $10 and $14 figures.

What does the slippery slope look like on a roommate ledger?

It looks like six decisions that were each easy to make and one total that nobody ever decided on. The ledger below is illustrative: the amounts are invented to show the shape, not drawn from any survey or from splitty’s data. Picture two roommates who split utilities down the middle. One pays the provider each month and the other owes half back.

Now run the same money as an abrupt change: the roommate pays in full for five months and then, in Month 6, is $620 short all at once. If the lab pattern carries over, that version is far more likely to get a conversation the same week. It has a date, a single number, and no history of approvals to contradict. The gradual version has six approvals behind it. Each month the reasonable comparison was to last month, not to zero.

Note the entries in Months 3 and 4. The history of letting it go becomes its own reason to keep letting it go. Gino and Bazerman saw a version of this in their abrupt condition too: watchdogs who rejected the sudden jump in Round 11 drifted toward approving it, reaching 37% by the end of the phase. The authors offer two readings: participants “may adapt to unethical behavior and go along with it even if they rejected it initially,” or they may have second-guessed their own judgment as the inflated estimates kept coming. Either way, a change that was noticed can still be worn down by repetition.

This is the creditor’s side of a pattern we have covered from the other chair. The guide to why “it’s only five dollars” is the most expensive sentence at dinner explains why the person who pays a small overcharge dismisses it. This page is about the person who is owed, and why they fail to collect a sum they would never have agreed to lend.

The ledger is an illustration with invented amounts, and the abrupt-change comparison is our inference from the lab finding, not a measurement. The 37% figure and the quoted sentence are from Gino & Bazerman (2009), Study 1 discussion.

Why doesn’t the person who is owed just ask?

Because asking has a social price, and the person who is owed pays it up front while the money comes later, if at all. Sociologists who interviewed people about lending to family and friends found that the cost of a money conversation is awkwardness, and that people go to real lengths to avoid it. In their interviews, a lender’s silence is often a strategy, not an oversight.

Frederick Wherry, Kristin Seefeldt, and Anthony Alvarez gathered 57 in-depth interviews in 2015 with low- and moderate-income clients of a California nonprofit lending program. Their subject was the reverse situation, friends and relatives asking for loans, and people deciding whether to lend, not people trying to collect. The awkwardness they document is, we think, the same emotion that keeps a roommate quiet. A person “feels awkward about resisting a request from a close social tie, particularly when the request seems sincere.” To avoid that awkwardness, people “engage in obfuscation,” saying no without saying no. And when repayment is unlikely, lenders “may also disguise the fact that a loan request cannot reasonably (or appropriately) be repaid,” because sharing norms “make it inappropriate to pursue or to expect repayment.”

Turn that around and it plausibly describes the roommate who is owed $620. Every month they do not ask looks like the same obfuscation: performing a self that is not petty, not a nag, not “his mom.” The extension from refusing to lend to declining to collect is ours; the paper does not study roommates. The paper’s literature review quotes a couple from an earlier study, Donna and her husband Kevin, who lends to his extended family. Kevin: “Family members still owe me three, four thousand dollars,” and yet the couple’s good fortune “really overweighs what we have lost in being good-hearted.” That is a creditor converting an unpaid loan into a story about character, which is far less awkward than collecting it.

Mental accounting adds a second reason. Richard Thaler’s review of the field describes how people treat an account that is still open differently from one they have closed: “a realized loss is more painful than a paper loss,” and “because closing an account at a loss is painful,” people are reluctant to close it. Thaler wrote about investors who will not sell a losing stock. By analogy, an unpaid roommate balance is a paper loss. Asking for it forces a resolution: either the money arrives or the refusal becomes real. Not asking keeps the account open and the loss unrealized. We offer that as an analogy; Thaler’s paper does not discuss roommates.

Sources: Frederick F. Wherry, Kristin S. Seefeldt & Anthony S. Alvarez, “To Lend or Not to Lend to Friends and Kin: Awkwardness, Obfuscation, and Negative Reciprocity”, Social Forces 98(2), 2019 (57 interviews, summer 2015, clients of Mission Asset Fund; the Donna and Kevin passage is from the paper’s literature review, which cites Shapiro 2004 as quoted in O’Brien 2012). Richard H. Thaler, “Mental accounting matters”, Journal of Behavioral Decision Making 12(3), 1999, section on opening and closing accounts.

Does the roommate who owes drift too?

Probably, and in the same direction. Research on the person doing the sliding finds that small lapses make larger ones easier, because each one lowers the bar for the next. A roommate who was $95 short in Month 1 has, by Month 2, a precedent. Whether the creditor’s silence is what lowers the bar is not something the research tested; the studies below manipulated the task, not an observer’s reaction.

David Welsh and colleagues tested this across four multi-round studies published in the Journal of Applied Psychology. They found that people engage in “a slippery slope of increasingly unethical behavior,” and that the mechanism is moral disengagement: the self-justification that “can reduce ethicality over a series of gradually increasing indiscretions.” Their studies are multi-round studies of unethical behavior, not of late utility payments, so read the connection as a parallel rather than a proof. A roommate who keeps falling short may also simply be unable to pay, or may disagree about the split, and neither of those is moral disengagement. Where the shortfalls are drift, though, the slope has two people on it: the creditor approves a little more each month, and the debtor finds a little more acceptable to owe.

Bazerman and Gino, reviewing the field in 2012, put the observer side in one sentence: people “are more likely to ignore the unethical behavior of others when ethical degradation occurs slowly rather than in one abrupt shift.” Two separate literatures, one for the person sliding and one for the person watching, and neither describes a built-in moment at which the sliding stops. A roommate ledger has both roles in it.

Sources: David T. Welsh, Lisa D. Ordóñez, Deirdre G. Snyder & Michael S. Christian, “The slippery slope: How small ethical transgressions pave the way for larger future transgressions”, Journal of Applied Psychology 100(1), 2015 (abstract). Max H. Bazerman & Francesca Gino, “Behavioral Ethics: Toward a Deeper Understanding of Moral Judgment and Dishonesty”, Annual Review of Law and Social Science 8, 2012.

How do you bring up money a roommate has owed for months?

Make it abrupt. In the lab, one clear jump was rejected far more often than the same change spread over many small steps. Nobody has tested whether stating six months of arrears as one total gets them paid; the lab result says the abrupt shape is the one people react to, and that is the reasoning behind these steps. Give the debt the shape it should have had: a single number on a single day. Four steps, in order:

If the total is met with a refusal rather than a plan, the problem has changed from a slope to a dispute, and the options are different. The guide to what happens when someone refuses to pay their share covers that path, including when small claims is and is not worth it. And if the answer is “I’ll Venmo you,” the guide to why “I’ll Venmo you later” never happens explains why a request beats a promise.

Quoted advice: Becca Rogers, “Ask Splitwise”, The Splitwise Blog, September 29, 2026. The four-step structure is ours.

How do you keep it from starting again?

Stop carrying a balance. The slippery-slope effect, as the lab produced it, needs a quantity that grows in steps too small to object to. Between roommates that quantity is the running total. Remove the running total and there is nothing to erode. Two rules do it, and both are set once, before anyone is behind.

The second rule is the structural one. A request does not settle anything by itself; an unpaid request is still a debt, and it may need a reminder. What it does is put a dated, specific ask on the record early. Silence after that is not a refusal, but it is no longer ambiguous about what was asked. In Thaler’s terms, by analogy, the account is not left open for months: within days the money has arrived or the open request is visible. In Gino and Bazerman’s terms, every charge becomes an abrupt change rather than one more step on a slope: a bill arrives, a number is requested, it is paid or it is not, and the answer is visible that week. The guide to the four decisions to make when a new roommate moves in covers which costs to put on a standing rule. The guide to splitting utilities between roommates covers which bills to split equally.

For recurring bills with a fixed split, a tracker that holds the ledger is the honest tool. Use Splitwise for that. Some of the bills that start slopes arrive as receipts instead: the grocery run, the takeout order, the trip for cleaning supplies that one person paid for and meant to split later. Those have line items, and line items can be closed on the spot.

A bill that was requested on the day it was paid has a status by the weekend: paid, or an open request with a number on it. That is the whole method. The slope needs time and ambiguity, and a same-day request gives it neither.