By Adeesh Setya, Head of Financial Research & Principal · Reviewed August 26, 2026
Bank Gap Index: The Academic Foundation
The Bank Gap Index has always rested on a simple formula: a representative balance times the spread between a top-available savings rate and the national average. That arithmetic doesn't change on this page. What this page adds is the academic grounding for why that spread persists, drawn from four banking-economics papers published in 2025 and 2026, plus an honest look at what the Index's balance assumption does and doesn't capture. This is a companion to, not a replacement for, the primary methodology page, which remains the canonical reference for the live formula and figure.
The unchanged formula
Bank Gap Index = Representative Balance × (Top-Available APY − National-Average APY)
= $25,000 × 3.75% = $938 per year (August 2026)
Why the gap persists: four findings
1. The gap is priced, not accidental
Xu Lu and Lingxuan Wu, in research awarded the 2026 Ieke van den Burg Prize for work on systemic risk, measured depositor inattention directly across millions of US bank accounts by comparing how quickly people move money that arrives on a predictable schedule versus money that shows up unexpectedly. Banks serving more inattentive depositors, by that measure, set lower rates, pass through less of any Federal Reserve rate move, and see less money leave even when their own rates fall further behind the market. The spread the Bank Gap Index measures is not an oversight either side is unaware of; it is a number banks actively manage.
2. The loss is concentrated, not evenly spread
Bronson Argyle, Benjamin Iverson, Jason Kotter, Taylor Nadauld, and Christopher Palmer, studying roughly 12 million accounts across 154 US credit unions, found that 10% of depositors hold about 70% of total deposits, and that these high-balance depositors are the least sensitive to rate differences of any group, using their accounts as a medium-run store of liquidity rather than a place to earn a competitive return. A single fixed balance necessarily understates the aggregate dollar amount at stake for this group. The sensitivity table below exists because of this finding, not as decoration.
3. The spread is a bank asset with a name
Itamar Drechsler, Alexi Savov, Philipp Schnabl, and Olivier Wang formalize the value a bank gets from paying below-market deposit rates as a "deposit franchise," an asset whose value rises with interest rates and vanishes if depositors leave. Their paper, published in the Journal of Finance in 2026, is written to explain bank-run risk, not consumer behavior, but it gives the Bank Gap Index's spread a name on the bank's side of the ledger: it is not just forgone interest to the saver, it is booked franchise value to the institution.
4. It is not a US-specific quirk
Fernando Cirelli and Arna Olafsson, using transaction-level data from a major Icelandic bank covering roughly a third of that country's population, found that most households barely respond to rate differences between otherwise-identical accounts, even when transfers between them are instant and free, with the exception of the wealthiest households, who were roughly ten times more responsive. A different country, a different banking system, and the same basic pattern.
The balance assumption, examined honestly
The Index has used a fixed $25,000 representative balance since launch, described on the primary methodology page as approximating "a common mid-range savings or emergency-fund balance." That framing is accurate. What is not accurate, and what this page corrects, is an internal citation that at one point described $25,000 as the actual median U.S. household liquid-savings balance per the Federal Reserve's 2022 Survey of Consumer Finances. It is not. The real 2022 SCF median conditional transaction-account balance is $8,000. $25,000 was, and remains, a deliberately round reference point rather than a literal median, and it is kept unchanged here. The correction is to the citation, not the figure.
Because Argyle et al. found deposits this concentrated, a single balance point understates the picture for the households actually holding the most money in the wrong account. The table below shows the same live formula at several reference balances, so the fixed-balance choice is transparent rather than a hidden assumption.
| Reference balance | Annual gap at current rates |
|---|---|
| $8,000 (actual SCF 2022 median) | $300 |
| $10,000 | $375 |
| $25,000 (Index default) | $938 |
| $50,000 | $1,875 |
| $100,000 | $3,750 |
Each row applies the same live 3.75-point gap (August 2026) to a different balance. This table is presented for transparency and does not change the published Index value, which remains fixed at the $25,000 balance.
A higher-balance reference point
A fully population-weighted version of the Index, weighting every dollar by where US deposits actually sit across the balance distribution, would require complete SCF liquid-savings percentile data that was not accessible while researching this page (the primary Federal Reserve bulletin's detailed tables did not render in the tools available for this research pass). What is available, from a single secondary source citing the same 2022 survey, is a median transaction-account balance of roughly $128,000 for households in the top decile of net worth. That figure has not been independently cross-verified against the primary Federal Reserve table and should be read as directional, not as precise as the $25,000 headline figure. At the current gap, that balance implies an annual cost of roughly $4,800 for a household holding that balance at a below-market rate. This figure will be tightened, or removed if it does not hold up, once a primary-source verification pass is possible.
Limitations
- This page establishes correlation and pricing behavior, not a controlled experiment. The four papers above document that inattention is priced and that balances are concentrated; none of them, individually or together, run a randomized test of what makes any single household switch. See the companion piece on disclosure and switching behavior for the field-experiment evidence on that separate question.
- The top-decile balance figure above is single-sourced. It is disclosed as such rather than presented with false precision.
- The literature is US- and Iceland-specific for the household-response evidence. Generalizing further requires caution; banking structure, deposit insurance regimes, and product availability all differ across markets.
- This page does not change the published Index value, formula, or public dataset schema. The academic grounding and sensitivity table are additive context.
The canonical figure and formula
This page is the academic companion. The primary methodology page remains the citable reference for the live formula, inputs, and figure.
View the primary methodology →