Behavioral Science Dictionary

Flat-rate bias

Behavioral Economics

People prefer flat-rate plans even when pay-per-use would cost less.

What it means

The tendency to choose flat-rate over metered pricing even when observed usage makes the flat rate more expensive. Kenneth Train named the effect in 1991, after telephone billing records showed a pull toward the flat tariff that calling volume alone could not explain. Several motives have been proposed rather than established. Some subscribers may be buying insurance against a variable bill, some may prefer a single payment that decouples cost from use so that each use feels free, and some may overestimate their future usage. Measured rates vary widely with the market and with the criterion used to call a choice a mistake, and a mirror pay-per-use bias falls on those who underestimate their consumption rather than on light users. The usual counterfactual holds usage fixed, so published savings are probably overstated. It matters for pricing design, since a flat rate that subscribers keep after seeing their usage priced both ways is a different product from one kept because the comparison is hidden.

The original demonstration

The earliest systematic evidence came from telephone billing records. Kenneth Train, Daniel McFadden and Moshe Ben-Akiva are among the first to have reported the pattern, in a fully discrete model that treated service choice and calling volume as one joint problem rather than two — which is what makes the key quantity identifiable: a tariff-specific constant, a pull toward the flat rate that volume alone could not explain. Train later named that residual the flat-rate bias in a 1991 monograph on natural monopoly, and the label stuck. The most-cited early numbers come from Donald Kridel, Dale Lehman and Dennis Weisman, reported here as Lambrecht and Skiera summarise them, the original being paywalled. Close to 65 percent of flat-rate subscribers would have paid less on measured service, against only about 10 percent of measured-service subscribers who would have gained by switching. Kridel's paper is titled for the argument that has shadowed the literature since: a flat rate is an option on unlimited use, carrying value even in months when it goes unexercised. A dollar figure for the gap circulates, but the secondary literature reports it both as the average magnitude of the bias and as that option value — different quantities, so none is repeated.

Why it happens

Four mechanisms are usually proposed, not mutually exclusive. The insurance effect is a preference for a predictable bill: a subscriber who dislikes variance pays a premium for certainty. The taxi-meter effect concerns the timing of pain rather than its size. In Drazen Prelec and George Loewenstein's mental-accounting framework, a running meter attaches a fresh twinge of cost to every unit, while a flat fee paid once decouples payment from consumption, so later use feels prepaid, which is to say free. The convenience effect covers the effort saved by not monitoring a meter. The overestimation effect is a forecasting error: people expect to consume more than they will. The verdicts differ: insurance and convenience describe a rational purchase of something other than units, overestimation a genuine error, and the taxi-meter effect a real change in experienced utility produced by the accounting frame rather than by the service.

What the evidence shows

The most detailed field measurement is Anja Lambrecht and Bernd Skiera's study of a European internet provider. The choice there lay among three DSL tariffs — two with a fixed fee, an allowance and a per-megabyte overage charge, one a true flat rate — so what is measured is the pull toward the higher fixed fee and larger allowance, not flat versus meter. Under a loose criterion, any five-month bill above the cheapest plan, up to 46.6 percent of subscribers in a given tariff group overpaid, against at most 6 percent making the opposite error; requiring the mistake in every month, those figures fell to at most 29.3 percent and under 1 percent. Flat-rate bias did not significantly raise churn here; pay-per-use bias did. Their mechanism tests rest on weaker ground: four analyses over three data sets — transactional records for 10,882 of the provider's customers, a survey of 241 MBA students, and a second survey of 1,078 subscribers. The attitude scales for the taxi-meter, insurance and convenience effects went to the students, as did the vignettes behind the overestimation test. Insurance, taxi-meter and overestimation each pushed toward the flat rate; convenience did not. Field data establish the pattern; the mechanisms rest largely on self-report. Stefano DellaVigna and Ulrike Malmendier tracked 7,752 members of three United States health clubs. Those on a monthly contract above $70 attended about 4.3 times a month, above $17 a visit against a ten-visit pass at $10. Their headline $600 forgone over an average membership mixes two errors they treat separately: the signup contract, a tariff-choice bias, and failure to cancel an auto-renewing contract once attendance stops, which is inertia. Only the first belongs here. The record is not one-directional. Eugenio Miravete, using a 1986 Kentucky tariff experiment, found the imbalance reversed: about 12 percent of flat-rate subscribers were wrongly on that plan against roughly 67 percent of measured-service subscribers on theirs, both within-plan shares. His emphasis fell on ignorance and learning, not a standing preference for flat rates. Michael Grubb and Matthew Osborne, modelling a 2002 to 2004 cellular panel, found consumers underestimating the variance of future calling — close to the opposite of the variance aversion insurance requires. No meta-analysis appears to have been published, and the evidence base remains a few field data sets in telecommunications and fitness.

Limits and caveats

The core measurement rests on a counterfactual almost always computed the wrong way: analysts take observed usage and ask what it would have cost on the cheaper plan, holding usage fixed. Usage is not fixed. Eva Ascarza, Anja Lambrecht and Naufel Vilcassim showed that customers moving from two-part to three-part tariffs overused relative to what their changed budget constraint alone predicts. If the meter would have suppressed consumption, the advertised savings are overstated and some share of every published estimate is an artefact of arithmetic. The effect is also criterion-dependent, moving from 46.6 to 29.3 percent in one data set on a definition change alone, and market-dependent, reversing in Miravete's Kentucky data. Switching costs and lock-in produce persistent mismatch requiring no bias at all. The evidence is unevenly aged. The classic measurements come from telephone, dial-up internet and gym markets between the mid-1980s and mid-2000s, but modern markets have been studied, mostly with a different question in view. Aviv Nevo, John Turner and Jonathan Williams estimated residential broadband demand under three-part tariffs from high-frequency usage data. Liran Einav, Ben Klopack and Neale Mahoney, using a comprehensive payment-card panel, found that months in which a card is replaced, forcing active renewal, produce far higher cancellation rates, and estimate that these frictions alone roughly double subscription revenue. That is the sharpest caution in the literature: a subscriber who keeps paying because cancelling requires an act, not because a flat rate feels better, leaves the same billing record but calls for a different remedy.

Using it in practice

For pricing design, a flat rate is not simply a discount for heavy users. It is also a product bought by light users for reasons unrelated to units consumed, and it changes how consumption feels once bought. That looks commercially attractive, since overpaying flat-rate customers did not leave at elevated rates while overpaying metered customers did. But that comes from the one study that examined it directly, and the low churn of the overpaying group may itself be renewal friction rather than satisfaction. That is where the ethical line sits. The defensible test is whether a customer on a flat rate would still choose it after seeing their own realised usage priced both ways. Answering it takes three measurements: a signup usage forecast compared against realised usage; the share of subscribers who move plans once a bill makes the comparison visible; and the share of those who complete the change, since a menu easy to read and hard to leave earns what one that is neither earns. If forecasts run high and switching is rare only because the comparison is hidden or the exit slow, the revenue comes from error or friction. If subscribers keep the flat rate once shown the numbers, they are buying predictability, which is legitimate to sell.

Examples

Gym members on flat monthly plans often attend so rarely that pay-per-visit would have been cheaper.

A gym sells a monthly membership at a price that only pays off above about eight visits, alongside a ten-visit pass. Members who sign up in January projecting three visits a week average closer to one, and end up paying several times the per-visit rate the pass would have given them.

A mid-size firm's procurement team compares a metered plan for a cloud analytics tool, billed per query, against an unlimited-seat tier at roughly triple its current spend. It picks the unlimited tier so that no team has to justify individual queries, and internal query volume rises but never approaches the break-even point.

An all-inclusive holiday package folds meals and drinks into one up-front price. Guests order more than they would from a menu, because each order registers as costless once the package is paid, and the resulting consumption is often used, misleadingly, as evidence that the package was good value.

A household stays on a metered mobile plan because the monthly line looks small, consistently underestimates its own video streaming, and pays overage charges that exceed an unlimited plan's price in most months. This is the mirror error, and it falls on users who consume more than they think, not on light users.

First described in Train (1991); Lambrecht & Skiera (2006).

Key references

  1. Train, K. E., McFadden, D. L., & Ben-Akiva, M. (1987). The demand for local telephone service: A fully discrete model of residential calling patterns and service choices. RAND Journal of Economics, 18(1), 109-123. doi.org/10.2307/2555538
  2. Train, K. E. (1991). Optimal Regulation: The Economic Theory of Natural Monopoly. MIT Press. openlibrary.org/isbn/9780262200844
  3. Kridel, D. J., Lehman, D. E., & Weisman, D. L. (1993). Option value, telecommunications demand, and policy. Information Economics and Policy, 5(2), 125-144. doi.org/10.1016/0167-6245(93)90018-C
  4. Prelec, D., & Loewenstein, G. (1998). The red and the black: Mental accounting of savings and debt. Marketing Science, 17(1), 4-28. doi.org/10.1287/mksc.17.1.4
  5. Miravete, E. J. (2003). Choosing the wrong calling plan? Ignorance and learning. American Economic Review, 93(1), 297-310. doi.org/10.1257/000282803321455304
  6. Lambrecht, A., & Skiera, B. (2006). Paying too much and being happy about it: Existence, causes, and consequences of tariff-choice biases. Journal of Marketing Research, 43(2), 212-223. doi.org/10.1509/jmkr.43.2.212
  7. DellaVigna, S., & Malmendier, U. (2006). Paying not to go to the gym. American Economic Review, 96(3), 694-719. doi.org/10.1257/aer.96.3.694
  8. Ascarza, E., Lambrecht, A., & Vilcassim, N. (2012). When talk is "free": The effect of tariff structure on usage under two- and three-part tariffs. Journal of Marketing Research, 49(6), 882-899. doi.org/10.1509/jmr.10.0444
  9. Grubb, M. D., & Osborne, M. (2015). Cellular service demand: Biased beliefs, learning, and bill shock. American Economic Review, 105(1), 234-271. doi.org/10.1257/aer.20120283
  10. Nevo, A., Turner, J. L., & Williams, J. W. (2016). Usage-based pricing and demand for residential broadband. Econometrica, 84(2), 411-443. doi.org/10.3982/ECTA11927
  11. Einav, L., Klopack, B., & Mahoney, N. (2025). Selling subscriptions. American Economic Review, 115(5), 1650-1671. doi.org/10.1257/aer.20231612

Related terms

← All 1001 terms