Borrowed Discipline: Leverage, Holding Periods, and the Limits of the Commitment-Device Case

The study does not exist, and it could not have existed. No edition of DALBAR’s Quantitative Analysis of Investor Behavior (QAIB, an annual report on mutual-fund investor behaviour published since 1994), no edition of Morningstar’s Mind the Gap, and no academic source traceable through David Chilton’s published work has ever reported investor holding periods segmented by whether the investor borrowed to invest dalbar-2016 qaib-glossary mind-the-gap-2025. The reason is architectural: QAIB’s sole input is aggregate fund flows, which carry no field recording where a purchased dollar came from. Searching harder cannot produce the missing finding, because the instrument cannot take that reading.

The unconditional version of the recalled claim is also false where it has been tested: leveraged investors trade more, not less, and liquidate hardest when their constraint binds — which, in a falling market, is the bottom. A narrower thesis survives. What the causally identified evidence supports is not that borrowing produces patience, but that transferred exit control with a calendar-indexed trigger does. Forced mortgage amortization raises household net worth with an elasticity near 1 and, decisively, without crowding out other saving mortgage-amortization-and-wealth-accumulation — the closest thing in the record to proof that leverage’s documented wealth benefit runs through a savings-quantity channel operating whatever the leveraged asset’s return does evaluating-homeownership-as-the-solution-to-wealth-inequality.

Two conditions separate the cases that work from the ones that don’t, and both must hold: the program must transfer real control over exit away from the investor, and the exit trigger must be indexed to the calendar rather than to the collateral price. A mortgage satisfies both; discretionary margin fails both; a fixed-rate term investment loan without a margin-call provision satisfies the second and the first only weakly — and has never been measured. That untested cell, not the missing Chilton citation, is the real gap in the record.

1. The Empirical Landscape on Investor Holding Periods

Almost every claim in circulation about how long ordinary investors hold anything traces to one of two commercial research products. DALBAR’s QAIB restates industry-wide mutual-fund flows as though the entire investing public were a single synthetic investor, and reports for that investor an annual retention rate in years qaib-glossary. Morningstar’s Mind the Gap computes a dollar-weighted internal rate of return (IRR — the return actually earned by the money, accounting for when each dollar arrived and left) and compares it against the fund’s time-weighted total return, what a lump-sum buy-and-hold investor would have earned. The difference is the investor return gap, reported at −1.1 percentage points a year for the decade to 2023 and −1.2 for the decade to 2024 mind-the-gap-2024 mind-the-gap-2025.

A. DALBAR’s retention rate is an inverse-turnover ratio, not a tracked holding period

DALBAR’s own glossary is explicit: holding period (retention rate) “reflects the length of time the average investor holds a fund if the current redemption rate persists. It is the time required to fully redeem the account” dalbar-2016. That is assets divided by an annualized redemption rate. No investor is ever followed from purchase to sale, and nothing in the published method adjusts for share class, load structure, or distribution channel qaib-glossary.

The arithmetic consequence appears in DALBAR’s own series. Asset Allocation retention runs 4.86 years for 2014, 4.54 for 2015, 4.09 for 2016 and 4.65 for 2017 — a 0.77-year swing inside three years dalbar-2016 for-the-period-ending-december-31-2017. A genuine per-investor mean drawn from a stable population of tens of millions cannot move fifteen percent that fast. A ratio whose numerator is market-inflated assets can, because a bull market raises the numerator without any investor deciding anything. DALBAR documents the mechanism while misreading it, tying retention milestones to preceding index moves: retention “surpassed the 4 year mark in 2004, after the S&P rose over 28% in the previous year,” and resumed the streak “in 2013 with an S&P surge of over 32%” dalbar-2016. Its thirtieth annual edition reports equity-fund retention rising again in 2023 dalbar-inc-2024, which neither disturbs the ordinal ranking nor rescues the ratio’s validity.

The companion behaviour-gap headline is in worse shape still. DALBAR divides summed dollar gains by summed cost basis rather than computing a proper IRR over an ongoing contribution stream, which mechanically understates returns whenever money arrives mid-period investors-do-not-underperform-their-investments-articles-advisor-perspectives a-warning-to-the-advisory-profession-dalbars-math-is-wrong-articles-advisor-pers the-fallacy-behind-investor-versus-fund-returns-and-why-dalbar-is-dead-wrong-art. A saver depositing $1,000 monthly for twenty years at a steady 10% would be scored by DALBAR’s own method at 3.57% — a 6.43-point shortfall attributable to no behavioural mistake whatsoever; an independent reconstruction of smooth flows into a market returning exactly 8.19% a year yields a 2.29-point gap with zero behaviour in the model dalbars-measure-of-investor-underperformance-is-wrong.

DALBAR’s category ordering is worth something as ordinal evidence; its magnitudes are not, and this analysis uses the first and discards the second. The population the calculation penalizes hardest is the steady payroll-deduction contributor — precisely the disciplined, mechanized investor the forced-savings thesis holds up as its ideal can-average-investors-really-be-as-bad-as-studies-say. The instrument most often cited to prove investors need commitment devices is the one that most reliably mis-scores investors who already have one.

B. Morningstar corroborates the ordering and supplies a different mechanism

Morningstar’s dollar-weighted method uses a different dataset and a different formula, and it reproduces DALBAR’s ordinal result. For the decade to 2023, allocation funds — diversified, multi-asset, the closest industry analogue to a 60/40 portfolio — “had the narrowest gap (a negative 0.4% gap per year),” against −2.6 points for sector equity mind-the-gap-2024. The 2025 edition tightens it:

• Allocation −0.1
• Alternative −0.1
• US Equity −0.6
• Taxable Bond −1.0
• International Equity −1.1
• Municipal Bond −1.2
• Sector Equity −1.5

Allocation investors captured the largest share of their funds’ aggregate total returns — nearly 97% — while bond-fund investors captured about half mind-the-gap-2025.

Two independently constructed methodologies agree on the ordering — and Morningstar’s explanation is not DALBAR’s. DALBAR credits diversification with curbing bad behaviour: asset-allocation structure “tends to curb negative behavior and lead investors to stay more committed to that strategy” dalbar-2016. Morningstar credits channel and automation: “Context counts: Allocation funds are often used in defined-contribution plans. This mechanizes investing, avoiding the potentially large timing costs investors can incur when making large, ad hoc transactions” why-investors-missed-out-on-15-of-total-fund-returns-morningstar. Its companion commentary adds that the gap also reflects fund mergers, liquidations and non-conscious cash flows, not purely market timing investors-still-need-to-mind-the-gap-in-their-funds-returns-morningstar.

Morningstar’s account is the one supported by its own data. The category with the longest measured retention and the narrowest measured gap is the category whose money arrives automatically and sits because nobody is actively managing it. A vehicle-level comparison isolates structure from asset class: open-end mutual funds show narrower gaps than exchange-traded funds (ETFs) “even when we controlled for asset class,” −1.0 points against −1.1, and index funds narrower gaps than active funds, −0.8 against −1.2 why-investors-missed-out-on-15-of-total-fund-returns-morningstar. ETFs trade intraday; open-end funds price once daily. Same assets, different exit friction, different measured behaviour, no leverage involved.

C. The calibration correction that reframes the entire debate

The headline gap magnitudes should be read as an upper bound rather than a measurement. Morningstar revised its total-return calculation across three consecutive editions, and each revision shrank the reported gap for the identical 2015–2024 window: −2.14%, then −1.64%, then −0.93%. In the Alternative category the same data yields −5.54% under one weighting and −0.06% under another full-article-bad-timing-does-not-cost-investors-15-of-their-funds-returns-an-exa. A peer-reviewed decomposition in the Financial Analysts Journal (Fulkerson, Jordan, Riley and Yan, 2026) reproduces Morningstar’s results to within 0.2 points, then splits the gap into a timing component — the real cost of when money moved — and a hindsight component, an artifact of scoring investors against a return path knowable only after the fact, which costs no wealth. Investors’ actual timing decisions cost 0.10% per year — about 1.2% of total return rather than 15% — with roughly half the reported category-level gap attributable to hindsight construction full-article-bad-timing-does-not-cost-investors-15-of-their-funds-returns-an-exa.

That figure is the low end of a contested range. A US Department of Labor–commissioned replication finds load-fund investors losing 1.12 percentage points a year in US equity and sector funds to timing errors, against their no-load counterparts the-ability-of-investors-to-time — a difference between two arms of one dataset, so the hindsight artifact largely cancels out. The headline magnitudes are contested across at least a fivefold range; the timing cost is real but far below the 15%-of-returns framing.

This demolishes the narrative that investors destroy 15% of their returns through panic, while leaving the category ordering intact. The behaviour gap is the premise the entire leverage-as-behavioural-fix argument rests on: investors are supposedly bleeding several points a year to psychology, and a leveraged program is the structural repair. If the real timing cost is a tenth of a point, the problem is a small fraction of the advertised size while the risk added by the repair is unchanged. A leverage program sold as a cure for panic selling is repairing a much smaller leak than its sales material claims; one sold as a way to save more does not depend on the behaviour gap being large at all.

D. Why the leverage figure is missing, and why it will stay missing

Three separate exhaustive searches — full-text across DALBAR’s 2016 and 2018 editions, full-text across Morningstar’s 2024 and 2025 editions including methodology appendices, and the canvass of Canadian industry sources described in Section 2A — return no holding-period or return-gap data segmented by leverage or borrowing use dalbar-2016 for-the-period-ending-december-31-2017 mind-the-gap-2024 mind-the-gap-2025. The words leverage and borrow do not appear in the body of the 2016 QAIB report. The single appearance of leverage across the QAIB editions reviewed is the 2018 glossary’s definition of an “Average Alternative Strategies (Alt-) Fund Investor” — funds using “long/short, market neutral, leveraged, inverse, or commodity strategies” — which is fund-level strategy leverage rather than investor borrowing, and for which no retention figure is published at all for-the-period-ending-december-31-2017. The absence is structural, not editorial. No study built on aggregate fund-flow data, in any year past or future, could report a holding period segmented by whether the investor borrowed qaib-glossary. The limit is specific to that variable and that class of instrument: holding periods for a leveraged product have been estimated outside the fund-flow vendors, and Section 4B reports them.

2. Holding Periods by Strategy: Fixed Income, Equities, 60/40, and Leveraged

A four-way comparison across fixed income, equities, 60/40 balanced portfolios and leveraged strategies can populate three of its four cells from primary sources. The fourth cannot be populated, and filling it with a proxy would be a category error.

A. The comparison

Strategy	DALBAR retention rate, years (2014 / 2015 / 2016 / 2017)	Morningstar investor return gap, 10 yrs to 2024	Share of fund return captured	Independent measurement?
Fixed income	2.94 / 2.93 / 3.09 / 3.45	−1.0 pt taxable bond; −1.2 pt municipal bond	~50%	Yes — shortest of the three in every year dalbar-2016 for-the-period-ending-december-31-2017 mind-the-gap-2025
Equities	4.19 / 4.10 / 3.80 / 4.03	−0.6 pt US; −1.1 pt international; −1.5 pt sector	—	Yes — middle in every year dalbar-2016 for-the-period-ending-december-31-2017 mind-the-gap-2025
60/40 balanced	4.86 / 4.54 / 4.09 / 4.65 (DALBAR’s Asset Allocation category)	−0.1 pt (2025 ed.); −0.4 pt (2024 ed.)	~97%	No. The circulating “60/40 investor gap” figures trace back to this same data, relabelled dalbar-2016 mind-the-gap-2024 mind-the-gap-2025 can-average-investors-really-be-as-bad-as-studies-say
Leveraged / borrow-to-invest	No borrow-to-invest figure exists in any edition from any vendor	None	None	Not from the retention vendors: they cannot see borrowing, and the parties that can do not publish retention qaib-glossary mind-the-gap-2025. Outside them, one leveraged vehicle has been estimated — leveraged and inverse ETFs, 5.3 to 22.7 trading days, fund-level strategy leverage rather than investor borrowing (Section 4B) securities-litigation-and-consulting-group-inc-2010

Two scope limits belong beside the table. Every row is United States mutual-fund industry aggregate data. No Canadian equivalent to QAIB or Mind the Gap was located, and searches of Canadian industry sources returned no strategy-segmented holding-period data — an absence recorded against the published sources reachable here, not a canvass of every Canadian vendor. Every strategy-level holding-period number available to a Canadian investor is therefore imported from a US dataset, which becomes actionable later: Canada does collect the one variable the US datasets structurally lack.

B. What the populated rows actually show

One pattern dominates both methodologies. Fixed income, the lowest-volatility and most conservative category, has the shortest retention and among the widest gaps: its investors have not remained invested in their funds for longer than four years at any time in the past twenty years, and exceeded three years only once in the past eight dalbar-2016, capturing about half their funds’ aggregate return mind-the-gap-2025. Asset allocation, a riskier portfolio, shows the longest retention and captures 97%.

Holding period does not track portfolio risk. It tracks the friction between the investor and the exit. That ordering is inexplicable on any theory that investors hold what feels safe, and immediately explicable on Morningstar’s channel account: allocation funds are the default vehicle in defined-contribution (DC) retirement plans, where contributions arrive by payroll deduction and nobody places trades, while bond funds are disproportionately bought à la carte by investors making an active decision — who can therefore make an active decision to leave why-investors-missed-out-on-15-of-total-fund-returns-morningstar.

C. The 60/40 row is the asset-allocation row under another name

The figures in popular circulation about 60/40 investors look impressively specific — trailing a basic balanced portfolio by 2.4 percentage points a year over ten years, 3.8 over twenty, and 5.1 over 2002–2011, reproduced from J.P. Morgan’s Guide to the Markets. Their own footnotes resolve them: the ten-year figure is “based on an analysis from Morningstar,” the twenty-year figure “based on an analysis by Dalbar Inc.” can-average-investors-really-be-as-bad-as-studies-say. These are the same allocation-category data already in the table, renamed downstream — its implausible jump between decades (5.1 points for 2002–2011 collapsing to 2.4 for 2012–2021, despite half the twenty-year window overlapping the ten-year one) is the signature of the underlying calculation error, not a finding about investor behaviour can-average-investors-really-be-as-bad-as-studies-say.

D. The one genuine named-cohort retention figure

One adjacent measurement is the closest anything comes to a real per-investor retention statistic. The Employee Benefit Research Institute and the Investment Company Institute followed 700,000 401(k) participants fully invested in target-date funds at the end of 2016 through the end of 2022: 85% remained fully invested across the entire six-year window, with mid-tenure workers highest at 88%, against 82% for those with two years’ tenure or less and 80% for those above thirty years. Its authors attribute the persistence partly to the target-date fund’s status as the default investment under the Pension Protection Act of 2006 new-research-finds-strong-target-date-fund-retention-among-retirement-investors. That is a named-cohort, multi-year retention number, roughly double the persistence DALBAR’s whole-market ratio implies for any category.

Default status, again. The longest holds in the record belong to the vehicles investors never had to choose and rarely have to think about — an argument for automatic payroll contribution, not for borrowing.

3. Tracing the Chilton-Referenced Study

The recalled claim has a specific shape: a pre-2015 David Chilton reference to a Dalbar, Morningstar or academic study showing that borrowing to invest correlated with the longest investor holding periods. Five independent paths lead away from it — the two named institutions, the academic literature, Chilton’s own corpus, and the Canadian personal-finance journalism of the period. All five close.

A. Five paths, all closed

DALBAR is ruled out structurally: QAIB has never had a leverage category and its fund-flow input cannot produce one dalbar-2016 qaib-glossary for-the-period-ending-december-31-2017.

Morningstar segments Mind the Gap by category group, vehicle, management style, fee quintile and volatility quintile. Full-text searches of the 2024 and 2025 editions, methodology appendices included, return nothing on leverage, borrowing or margin loans mind-the-gap-2024 mind-the-gap-2025.

The academic literature was searched for investor-level holding-period or retention data segmented by borrowing status: the Barber–Odean discount-brokerage panel family, the retail margin-trading literature, and behavioural-finance work on fund retention. No pre-2015 paper reporting holding period by borrowing status was located, and the investor-level leverage-versus-trading-behaviour evidence that does exist runs the opposite way leveraging-overconfidence the-drivers-and-implications-of-retail-margin-trading. One pre-2015 paper does pair leverage with holding periods in its title — Guedj, Li and McCann’s 2010 SLCG study of leveraged and inverse ETFs — and it is not the recalled study on any reading: it estimates holds for a fund’s internal leverage rather than an investor’s borrowing, and finds the shortest holds in this corpus rather than the longest securities-litigation-and-consulting-group-inc-2010.

Chilton’s recoverable corpus contains no traceable citation of any study correlating borrowing with longer holds. What was searched: the 2011 Globe and Mail interview at the launch of The Wealthy Barber Returns the-wealthy-barber-takes-a-snip-at-debt-the-globe-and-mail, the November 2025 interview accompanying the updated edition the-wealthy-barbers-david-chilton-has-a-message-for-younger-generations-you-can, and three published reviews book-review-the-wealthy-barber reviewed-the-wealthy-barber-returns-canadian-business rockstar-book-review-the-wealthy-barber. What was not reachable: the full text of either book, and his 2012–2014 broadcast, column and speaking-circuit output — the most likely surviving home for the recollection, since a claim made aloud leaves no indexable record. His documented positions run the other way: the 2025 edition endorses all-in-one asset-allocation ETFs precisely because they require no monitoring, and warns that consumer credit is “so widely available now. So aggressively and persuasively marketed. It’s downright scary” book-review-the-wealthy-barber. His enduring framing — set aside 10% of each paycheque the moment it lands the-wealthy-barbers-david-chilton-has-a-message-for-younger-generations-you-can — is itself a forced-savings heuristic, with the borrowing removed.

The Canadian journalism cluster produced one promising lead and it dissolved. A 2015 Canadian MoneySaver article names a Tim Cestnick Globe and Mail column of 26 January 2012, “Does Leveraged Investing Really Work,” comparing $5,000 invested through a Registered Retirement Savings Plan (RRSP), a Tax-Free Savings Account (TFSA), or a leveraged strategy borrowing-from-peter-to-invest-in-paul-part-2-leveraged-investing-options-and-st. Direct search, Wayback Machine queries and checks of Cestnick’s own site found no such article — only two adjacent wrong candidates: his confirmed 18 January column, a pure RRSP-versus-TFSA piece with no leverage content, and Rob Carrick’s 26 January leverage column, which is not a three-way comparison. The MoneySaver citation is itself a relay, a title and date supplied verbally by an advisor three years after the fact; the parsimonious reading is that it conflated two real but distinct January 2012 pieces.

B. The nearest primary text, and what it actually says

The closest articulation of the remembered claim anywhere in the recovered Canadian record is a two-sentence sequence in Carrick’s January 2012 column: “Rather than investing bit by bit, leveraging allows you to put a big chunk of money to work immediately. Studies show this can achieve better results than gradual investing. Also, repaying the investment loan imposes financial discipline. You may balk at investing in uncertain times, but you have to make your regular investment loan payments” leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2.

It is two claims welded together. The studies show clause is a return claim about lump-sum versus gradual deployment, not a holding-period claim, and it names no study, author, year or institution. The discipline sentence that follows is Carrick’s own editorial voice with no citation attached. Compressing that paragraph into a study showing that borrowing to invest produces the longest holding periods requires exactly one operation: carrying the word studies from the first sentence across to the claim in the second — an ordinary and non-culpable feature of how people remember reading. The column’s own thesis, meanwhile, is hostile to the practice it describes: it defines leveraging as “an investment industry term that is best defined as borrowing trouble” and closes with “Leveraging? Why put yourself through it” leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2.

The only behavioural argument in the column comes from a named investing-industry consultant, Talbot Stevens, described as having “for years worked with advisers to develop guidelines on how to make leveraging work for investors”: “Behaviour trumps math 9.9 times out of 10,” he said. “We get scared, we hit the panic button and sell” leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2. A 2020 profile of the same practitioner’s Smart Debt Coach framework restates the case in tax-efficiency and market-timing terms — an RRSP gross-up borrowing technique and historical TSX return data — citing no Dalbar, Morningstar or academic source borrowing-to-invest-folly-or-fortune. This is practitioner assertion and carries no evidentiary weight. It matters for provenance: the behavioural-over-mathematical framing was circulating in Canadian personal-finance media in exactly the pre-2015 window, from a named source, attached to no dataset.

C. The most defensible reconstruction

No source states the following; three sourced findings jointly imply it, and it is offered in that shape. DALBAR’s real, repeatedly published finding is that its Asset Allocation category shows the longest retention, which DALBAR glosses as structure curbing negative behaviour dalbar-2016. Morningstar corroborates the ordering while attributing it to defined-contribution-plan mechanization — removal of easy exit — rather than diversification why-investors-missed-out-on-15-of-total-fund-returns-morningstar. And the only place a leverage framing appears in the contemporaneous Canadian record is uncited columnist commentary pairing loan repayment with discipline leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2.

The most defensible reconstruction is that a real finding about structured, non-discretionary investing producing the longest holding periods was retold across roughly a decade with leverage substituted for structure, aided by newspaper prose that put an uncited “studies show” one sentence away from a discipline claim. The substitution is wrong as attribution, but not entirely wrong as mechanism: both structures share the one property that does the behavioural work. Under exactly what conditions, then, does a leveraged program share that property?

4. Does Leverage Lengthen Holding Periods — and Why?

Leverage in the retail context covers instruments with almost nothing behaviourally in common. A margin account is a revolving credit facility attached to a brokerage account: the balance can be raised, lowered or unwound at will, there is no repayment schedule, and the broker can force a sale if the collateral falls below a maintenance threshold. An amortizing mortgage is the opposite: a fixed schedule of payments the borrower cannot skip, secured against an asset that takes months and substantial transaction costs to sell. A term investment loan — a fixed-sum, fixed-term borrowing used to buy securities, which is the structure a borrow-to-invest program actually sells — sits between them, carrying the mortgage’s payment schedule but the margin account’s liquid collateral. Asking whether leverage lengthens holding periods is therefore like asking whether contracts make people reliable; the useful question is which structural features do the work. The margin evidence below, while the best-measured material available, concerns the instrument furthest from the one at issue.

A. The counter-evidence

Where investor-level leverage has actually been measured against trading behaviour, leveraged investors traded more, not less. In the standard US discount-brokerage panel — non-retirement accounts of more than 43,000 investors, with every pairwise turnover difference statistically significant and robust to controls for age, gender, marital status, income and wealth — monthly turnover runs 6.9% for cash investors, 7.8% for margin-account holders with no observed margin use, and 15.2% for investors with margin experience leveraging-overconfidence. The doubling belongs specifically to the margin-experience subgroup, not to margin-account holders generally — a distinction that turns out to be diagnostic. Speculative trades, a gain-sale followed by repurchase within three weeks, follow the same ordering at 15.2%, 17.9% and 29.7% leveraging-overconfidence. The parent study, across 66,465 households, found the highest-turnover households earning far less than infrequent traders despite nearly identical gross performance — trading costs, not stock selection trading-is-hazardous-to-your-wealth.

Despite a 2023 publication date, that turnover analysis reuses the 1991–1996 discount-brokerage dataset: a re-analysis with an added overconfidence lens, not a modern replication, and no post-2010 North American replication exists leveraging-overconfidence. A modern international analogue partly closes the gap: Chinese account-level data from 2014–2015 show shadow-financed margin investors at 21.4% average daily portfolio turnover, brokerage-financed margin investors at 11.7%, and non-margin investors at 10.6% the-drivers-and-implications-of-retail-margin-trading. Two independent account-level panels, a continent and twenty years apart, agree that leverage users trade more. A third market supplies the only causal identification of the set. Kahraman and Tookes exploit the margin-eligibility thresholds of India’s National Stock Exchange in a regression-discontinuity design over 2004–2012, finding a stock’s liquidity improves when it becomes margin-eligible, driven by margin traders’ own activity the-journal-of-finance-vol-lxxii-no-4-august-2017 — stock-level margin intensity rather than account-level holding periods, so corroboration without a third direct measurement.

B. Leverage manufactures forced exits, and the mechanism has been measured

The Chinese data supply something the older US panel cannot: a direct measurement of involuntary unwinding. Defining each account’s distance to its forced-liquidation threshold, accounts very close to a margin call sell an additional 23.95% of their stock holdings relative to unlevered margin accounts, with selling pressure rising monotonically as the distance falls and the coefficient for the closest bin almost six times that of the safest. The nonlinearity is the tell: investors are not simply reacting to losses, they are reacting to the constraint. Shadow-financed accounts ran average leverage of 6.4 times equity against 1.4 for brokerage accounts, and through the 2015 crash window margin investors did not outperform — brokerage margin −17.2 basis points a day, shadow margin −64.0, against −21.0 for non-margin investors the-drivers-and-implications-of-retail-margin-trading.

The Indian panel shows a reversal of its own at market level. In ordinary conditions margin capital there is contrarian: margin traders buy declines and sell rises, improving liquidity. During the 2008 crisis that reverses into liquidity-draining deleveraging the-journal-of-finance-vol-lxxii-no-4-august-2017. Higher turnover under leverage is not uniformly worse; what inverts is crisis behaviour, precisely where the constraint binds.

No source measures holding-period length by financing structure — by whether the investor borrowed, and on what terms. One measures it for a leveraged product. Guedj, Li and McCann infer holding-period distributions from daily trading volume for five leveraged and inverse ETFs and find average holds of 5.3 to 22.7 trading days, with a persistent long tail — 6% to 24% of purchases held beyond a month — in instruments whose own prospectuses describe them as one-day vehicles securities-litigation-and-consulting-group-inc-2010. Those are the shortest holds anywhere in this evidence base: days, against the years DALBAR reports for every unleveraged category. They run against the recalled claim, not toward it.

Two scope limits bound them: the holding periods are inferred from volume through a proportional-trading model rather than observed in accounts, and daily-reset fund leverage is a property of the product rather than of the investor’s balance sheet — the same distinction the QAIB Alt- category forced in Section 1D. The paper does establish the inverse turnover-to-holding-period relation directly. Two findings then imply the direction for financed leverage: leveraged accounts sustain persistently higher turnover than unleveraged ones leveraging-overconfidence the-drivers-and-implications-of-retail-margin-trading, and liquidate sharply and involuntarily as they approach their constraint. Discretionary leverage therefore shortens effective holds twice over — elective overtrading in rising markets, constraint-driven selling in falling ones — and the second fires hardest precisely in the drawdowns where the commitment-device thesis claims leverage does its work.

C. Selection, not treatment

A separate line of evidence undercuts the causal reading of any positive correlation that might someday be found. In FINRA’s (the Financial Industry Regulatory Authority, the US brokerage self-regulator) National Financial Capability Study for 2018 and 2021, investors holding a margin loan were 17 percentage points more likely to own cryptocurrency, and those who had experienced a margin call 23 points more likely, with controls leaving the direction unchanged margin-trading-and-cryptocurrency-investment-among-us-investors-evidence-from-th. Chinese margin use shows the same pattern, driven by lottery preferences the-drivers-and-implications-of-retail-margin-trading. The overconfidence channel is measured directly: a one-standard-deviation increase in measured overconfidence raises the probability of holding a margin account by 7.9 percentage points and of using margin by 9.4 leveraging-overconfidence.

The turnover gradient’s internal structure is, with that in mind, quietly favourable to the structural thesis. Merely holding a margin account raises monthly turnover from 6.9% to 7.8%; the doubling appears only among investors who actually use margin leveraging-overconfidence. That is not what credit access corrupting investors predicts; it is what a certain kind of investor seeking out margin predicts. The same logic cuts the other way, and proponents rarely confront it: if a future study found that investment-loan borrowers hold longer, the selection reading — disciplined people take investment loans — would be at least as well supported as the causal one, and the observational design cannot distinguish them.

D. The two conditions that sort every case

Every documented forced-savings effect in the record involves an instrument where control over the decision to keep saving has been transferred from the investor to a contract — on the asset side, on the payment side, or both. Bernstein and Koudijs, exploiting a 2013 Dutch reform that forced amortization on new mortgages, find an amortization-wealth elasticity near 1 with little-to-no crowding-out of other saving (Section 5C) mortgage-amortization-and-wealth-accumulation.

Beshears and colleagues, in a randomized-contract experiment holding returns equal, found that “when the two accounts pay the same interest rate, the most illiquid commitment account attracts more money than any of the other commitment accounts” self-control-and-commitment-can-decreasing-the-liquidity-of-a-savings-account-in. Brune, Giné, Goldberg and Yang, randomizing Malawian smallholder farmers across ordinary savings accounts, ordinary-plus-commitment accounts and a tracked control, found that “only the commitment treatment had statistically significant effects on subsequent outcomes,” with large effects on deposits, input use and crop revenue commitments-to-save-a-field-experiment-in-rural-malawi.

None of these is about borrowing. All are about an instrument that takes the saving decision out of the saver’s hands each period. Asset-side illiquidity is the most legible way to do that, but the evidence does not make it necessary: Save More Tomorrow produced the strongest persistence result in this literature with no illiquidity whatsoever (Section 5D), through an actively chosen, employer-administered recurring obligation. Exit cost on the asset side strengthens the transfer; third-party administration of the payment can substitute for it.

Transferring exit control is not sufficient, because contracts differ in what makes them fire. A mortgage amortization schedule fires on the calendar: a payment is due on the first of the month regardless of what the housing market did. A margin call fires on the collateral price. The Canadian advisor continuing-education text describes the sequence plainly: when the security’s market value falls, the investor must post cash or marginable securities to cover the shortfall, and if they cannot, the dealer sells, needing permission from the investor neither to sell nor to give notice continuing-education.

That is the boundary. The causally identified forced-savings literature studies calendar-triggered instruments exclusively mortgage-amortization-and-wealth-accumulation commitments-to-save-a-field-experiment-in-rural-malawi, while the one leverage instrument whose behavioural record has been measured is price-triggered and shows the opposite effect leveraging-overconfidence. A calendar trigger fails for reasons largely uncorrelated with the portfolio — unemployment, not rates, dominates mortgage arrears across a 35-year window (Section 6E) staff-analytical-paper-document-analytique-du-personnel2026-12. A price trigger fails when and only when the portfolio falls, which is simultaneously the moment holding on would have mattered most and the moment every other leveraged investor is being called too. One is an idiosyncratic risk; the other is a correlated one that concentrates its failures at the bottom.

One further disanalogy bounds how far the mortgage result travels, and the Cleveland Fed’s own footnote makes it: housing leverage is “much less risky in comparison” to leverage on other assets because “leverage decreases to zero over the life of the mortgage” evaluating-homeownership-as-the-solution-to-wealth-inequality. An amortizing mortgage’s leverage ratio declines monotonically, so its risk profile improves mechanically as the holding period lengthens; an interest-only or bullet term investment loan holds leverage constant for its full term. The mortgage evidence therefore flatters any structure that does not amortize the borrowing itself — an argument for amortizing the investment loan, not merely for holding the portfolio a long time.

E. Mapping real programs against both conditions

Program type	Exit control transferred?	Trigger indexed to	Commitment-to-hold effect
Amortizing mortgage	Yes — home equity is genuinely illiquid	Calendar	Strong, causally identified: 0.993 elasticity, no crowding-out mortgage-amortization-and-wealth-accumulation
Smith Manoeuvre (readvanceable home equity line of credit, or HELOC)	No — the re-advance converts paid-down equity into a liquid, marketable position	Calendar on the mortgage leg; the investment leg has no schedule	Predicted weak-to-negative breaking-the-commitment-device welcome-to-the-smith-manoeuvre
RRSP gross-up loan	Partially — by the tax wrapper, not the loan: withdrawal is taxed and the room is lost	Calendar, typically one year or less	Real but short-dated; commits contribution, not holding borrowing-to-invest-folly-or-fortune
Term investment loan, no margin call	Weak — the portfolio stays sellable; only the obligation persists	Calendar	Plausible and untested — the structure the thesis needs continuing-education
Term investment loan, collateral-tested	Weak on the upside, absolute on the downside	Price	Inverts — forced unwind at the worst moment continuing-education
Discretionary margin	No — raised, cut or unwound at will	Price	Measured negative: 15.2% vs 6.9% monthly turnover leveraging-overconfidence

The Smith Manoeuvre is the diagnostic case, because it inverts the mechanism it markets. The strategy converts non-deductible mortgage interest into deductible investment-loan interest through a readvanceable credit line tied to the mortgage paydown schedule: each principal payment’s equity is re-advanced and invested outside registered accounts — a loop inferred here from the standard readvanceable-mortgage structure rather than narrated on the program’s own site, which markets it as a “proven debt conversion strategy” and publishes testimonials but no data welcome-to-the-smith-manoeuvre. Structurally, it preserves the amortization schedule but re-liquefies the equity that schedule builds — keeping the feature that does not drive commitment and removing the one that does.

The closest test already exists. Kovacs and Moran validate a life-cycle model of illiquid housing as commitment device against Texas’s 1998 legalization of cash-out home-equity withdrawal: following legalization, nondurable consumption rose 3% among homeowners (no effect on renters) and the savings rate fell 2.7 points — with the amortization schedule unchanged throughout, so only the liquidity of accumulated equity changed. Welfare declines by 2.6% in consumption-equivalent terms due to access to home equity, as the cost of weakened commitment outweighs the benefit of greater flexibility breaking-the-commitment-device — consumption-equivalent terms expressing a welfare change as the share of lifetime consumption a household would give up to avoid it.

Two qualifications temper the parallel: Texas households withdrew equity to consume rather than invest, and the amortization obligation keeps binding even as the asset-side lock is removed, making the Smith Manoeuvre a partial rather than failed commitment device. But the removed half is the half doing the measurable work, so whatever case survives for the strategy is a tax-arbitrage case — and that its official site publishes no adherence or dropout data after four decades is itself informative welcome-to-the-smith-manoeuvre.

The no-margin-call loan is the structure the thesis requires, and it is purchasable. A 2008 Canadian advisor training text names it: “Investors can also avoid margin calls by taking a no-margin-call loan,” at a higher interest rate continuing-education — the industry’s own recognition, seventeen years ago, that the forced-liquidation trigger is a separable, priced design variable independent of loan size. Nobody publishes take-up rates, and nobody has measured whether buyers hold longer.

5. The Behavioral-Benefit Thesis: Forced Savings vs. Return Optimization

Does a leveraged program’s real benefit lie in behaviour — a forced-savings commitment device raising retirement-savings adherence — rather than in return optimization? Two things the literature runs together must be separated: whether the return case for retail leverage is strong enough to be the main event, and whether the behavioural case is strong enough to replace it. A commitment device, in the technical sense used here, is an arrangement a person enters voluntarily that restricts their own future choices, adopted because they expect their future self to choose badly. The question is whether a loan qualifies.

A. The return case is genuinely narrow

Start with arithmetic the practitioners themselves publish. On $10,000 of personal capital plus $10,000 borrowed at 4%, at a 50% marginal tax rate, a 7% investment return produces a 12% effective return on personal capital — while a symmetric 7% loss produces a 16% effective loss, because interest costs layer on top of the price decline leveraged-investing-can-boost-returns-but-do-it-prudently-the-globe-and-mail. Rising borrowing rates raise the break-even return needed while simultaneously pressuring equity valuations leveraged-investing-three-risks-of-borrowing-to-build-wealth-the-globe-and-mail. Losses magnify more than gains, structurally. The strategy “works best if you’re in the highest marginal tax bracket” because the interest deduction is worth more there leveraged-investing-can-boost-returns-but-do-it-prudently-the-globe-and-mail, and the Canadian advisor training text is blunter still: “the tax benefit of interest deductibility is not a sufficiently compelling reason for an investor to use leverage” continuing-education.

The return case for retail leverage is therefore conditional on a high marginal bracket, low and stable borrowing costs, a long horizon and conservative underlying holdings — four things simultaneously. Anyone selling leverage on the upside arithmetic is selling the weaker half of the case. That is not itself an argument for the behavioural thesis, but it means the behavioural thesis is not competing against a strong alternative.

B. Two different behavioural jobs, routinely conflated

Commitment-to-contribute is the problem of getting money into the investment at all, against a present-biased self who would rather consume now. Commitment-to-hold is the problem of keeping money in through a drawdown, against a frightened self who wants out at the bottom. These are different failures with different mechanisms, and virtually every source treats forced savings as one undifferentiated thing.

On contribution, leverage has a formal theoretical case stronger than most advocates realize. Karna Basu proves that “simultaneous borrowing and saving cannot be optimal if agents have time-consistent preferences” — preferences that weigh the future consistently as it approaches. Sophisticated present-biased agents, who over-weight the present against their own stated plans and know it, rationally hold a loan alongside savings because it “generates incentives for future selves to invest optimally by punishing over-consumption” munich-personal-repec-archive. Borrowing while holding investable assets looks irrational under standard assumptions; Basu’s theorem says that appearance is the signature of a commitment device, not evidence against one.

The savings literature already has a category for it: Barrafrem, Tinghög and Västfjäll distinguish discretionary savings from contractual savings — regular contributions made in accordance with an agreement, such as a mortgage repayment, chosen as a method of self-discipline frontiers-behavioral-and-contextual-determinants-of-different-stages-of-saving-b. A leveraged program with a repayment schedule is structurally contractual savings with the contribution front-loaded — precisely what an RRSP gross-up loan does: gross a $3,000 after-tax contribution to $4,687.50 at a 36% marginal rate, borrow the difference, apply the $1,687.50 refund against the loan borrowing-to-invest-folly-or-fortune. The commitment sits in the tax wrapper, not the loan. Bryan, Karlan and Nelson’s canonical review names the mechanism — the use of microcredit has also been posited as a commitment device to save, since by borrowing, an investor raises the price of failing to save commitment-devices — as a posited explanation, not one extended to mortgage or investment debt in its own taxonomy.

That is as far as the literature goes: a targeted search across the commitment-device corpus, including Laibson’s illiquid-asset golden-eggs framework, located no source placing borrow-to-invest leverage within the established hard/soft commitment taxonomy. The classification used in Sections 4 through 6 is therefore this report’s own construction rather than an existing academic placement.

On holding, leverage has no supporting evidence and one measured contrary finding: the commitment-to-hold literature is uniformly about instruments that block withdrawal, none of which is a loan mortgage-amortization-and-wealth-accumulation self-control-and-commitment-can-decreasing-the-liquidity-of-a-savings-account-in commitments-to-save-a-field-experiment-in-rural-malawi, and the measured leverage finding runs the other way leveraging-overconfidence. A Swedish savings-app study of 2,619 goals sharpens the split: self-control predicted how much accumulated during ongoing saving but not whether saving was initiated, while financial knowledge predicted initiation but not accumulation frontiers-behavioral-and-contextual-determinants-of-different-stages-of-saving-b. A mechanism that solves one stage does not automatically solve the other.

The recalled claim is a commitment-to-hold claim. The defensible version of the thesis is a commitment-to-contribute claim. That is the reconciliation.

C. The housing evidence: wealth without a superior asset

Two sourced findings jointly imply what no single source states.

On asset-level returns, owner-occupied housing is unremarkable. The Cleveland Fed compiles the estimates: Demers and Eisfeldt find a 9.8% average nominal return for 1986–2014; Goodman and Mayer estimate 12.3% annualized nominal for a 2002 purchase sold in 2013, falling to roughly 7% once generous tax assumptions are removed; and Jordà and colleagues find housing and equities have roughly the same return over the full sample, though equities have outperformed housing on average since 1950 evaluating-homeownership-as-the-solution-to-wealth-inequality. In 20 of the 24 largest US metropolitan housing markets, median stock-portfolio returns exceeded median home-price appreciation over 2010–2016 homeownership-doesnt-build-wealth-study-finds.

Those estimates do not all measure the same thing. Demers and Eisfeldt’s and Jordà’s are asset-level returns; Goodman and Mayer’s 12.3% is explicitly the return on taking out a 30-year fixed-rate mortgage and buying a home — a return on the owner’s own equity, levered roughly four-to-one at entry. The same source credits that amplification in its own headline finding: “the opportunity for leveraged returns can lead to wealth gains among lower-income households” evaluating-homeownership-as-the-solution-to-wealth-inequality. Leverage does add a return-amplification channel. What it does not add is a superior underlying asset.

On wealth accumulation, forced amortization delivers a near one-for-one increase in net worth. Bernstein and Koudijs estimate the amortization-wealth elasticity at 0.993, “statistically different from 0, but not from 1 (95% confidence intervals of 0.88 to 1.10)” mortgage-amortization-and-wealth-accumulation. The importance lies in what does not happen. If households treated mortgage repayment and other savings as fungible, forced amortization would crowd out voluntary saving and net wealth would be unchanged. Instead there is “little-to-no change in the accumulation of non-mortgage savings, even five years later when amortization-induced home equity is larger than all other accessible savings,” with the additional wealth “financed via increased labor supply and reduced expenditures” mortgage-amortization-and-wealth-accumulation.

The documented wealth advantage of leveraged homeownership is therefore not explained by return optimization alone. What isolates the savings-quantity channel is the elasticity near 1 combined with the absence of crowding-out: a contractual channel calling new saving into existence, operating alongside whatever return amplification the leverage supplies.

The result is robust in three ways: instrumenting purchase timing with life events such as childbirth yields an elasticity of 0.864 (95% CI 0.54 to 1.19); a within-household comparison exploiting an annuity amortization schedule’s upward slope again yields about 1, ruling out one-time wealth effects from a concurrent tax change; and the effect holds across buyers with substantial liquid savings and across the age spectrum mortgage-amortization-and-wealth-accumulation — disposing of the objection that forced savings only helps people who would not have saved anyway.

The Cleveland Fed’s own selection warning, that homebuyers may simply be more inclined to save evaluating-homeownership-as-the-solution-to-wealth-inequality, is about selection into homeownership; Bernstein and Koudijs identify something narrower and better posed — selection into an amortization schedule, conditional on having bought. Even a source hostile to housing’s return case concedes the mechanism: Redfin’s Nela Richardson grants that “the forced savings of a monthly mortgage is a key reason why housing has served as an engine of growth for the middle class” homeownership-doesnt-build-wealth-study-finds.

The housing literature also supplies the one place where holding period under leverage has been measured against wealth outcomes at scale. Differences in housing equity accumulation across racial and income groups “were tied to the lower valued homes and the shorter duration of ownership for lower-income and minority households,” not to differences in appreciation rates joint-center-for-housing-studies. Of low-income minority households who transition out of owned status, only 37% return to owning against roughly 58% for high-income white households, and median non-housing wealth accumulation over nine years was approximately zero for the former against $2,650 for the latter us-department-of-housing-and-urban-development. The Cleveland Fed adds the compounding reason a short hold hurts: returns are lower or negative in early years because mortgage financing front-loads fees and interest evaluating-homeownership-as-the-solution-to-wealth-inequality. That asymmetry is a mathematical property of amortizing leverage, and the mechanism-consistent reason prudent practitioners converge on long minimum horizons.

D. What the retirement-savings literature says about adherence versus optimization

The adherence-versus-optimization split is the retirement-savings literature’s own central finding. Madrian and Shea examined a Fortune 500 firm that switched, effective 1 April 1998, from opt-in to opt-out 401(k) enrollment at a 3% default rate, with no change to match rate, vesting or menu: “Although none of the economic features of the plan changed, this switch to automatic and immediate enrollment dramatically changed the savings behavior of employees” nber-working-paper-series-2. Choi, Laibson, Madrian and Metrick extended this across three firms: participation exceeded 85% at all three under automatic enrollment, against 26–43% after six months of tenure under opt-in. But 65–87% of new participants anchored at the low default rate and default fund, declining only to 40–54% after two years. The net result: “automatic enrollment has little impact on average long-run wealth accumulation” — the participation gain and the low-default drag “are roughly offsetting on average” — while it “does increase saving in the lower tail of the savings distribution by dramatically reducing the fraction of employees who do not participate” nber-working-paper-series.

The most successful structural savings intervention in modern finance is, on its own authors’ evidence, a powerful adherence lever and a weak optimization lever. It gets people in and keeps them in; it does not make their portfolios better. That is precisely the shape of the claim advanced here about leverage.

Automatic enrollment’s weakness is that it is passive, and the persistence evidence is genuinely contested. Using the 2008 and 2010 waves of the US Health and Retirement Study, automatic enrollment is associated with a large participation increase and is particularly effective at reaching lower-income, less-educated and minority individuals — “however, automatic enrollment is not positively associated with longer-run contribution status in our sample—those who opt-in are more likely to continue making contributions over time” automatic-enrollment-in-retirement-savings-vehicles-evidence-from-the-health-and.

Thaler and Benartzi’s Save More Tomorrow (SMarT) fixes the passivity, and its persistence numbers are the best available: employees commit in advance to allocate part of future salary increases to retirement savings, exploiting loss aversion, present-biased discounting and inertia together. In the first implementation, 78% of those offered joined, 80% of enrollees remained through the fourth annual raise, and average savings rates rose “from 3.5 percent to 13.6 percent over the course of 40 months” save-more-tomorrowtm-using-behavioral-economics-to-increase-employee-saving precommitment-behavioraleconomicscom-the-be-hub — with no illiquidity at all, purely through an actively chosen, recurring, third-party-administered obligation. Active choice, not lock-in, is the variable that tracks persistence here, and a structured investment loan has that shape: actively chosen unlike a default, administered by a third party on a recurring schedule unlike a personal resolution, and abandonable only through a positive and sometimes costly act.

Third-party administration carries a measured advantage of its own: making a commitment observable to an identified third party rather than keeping it private raised the share of participants demanding any commitment from 41% to 65%, and the average staked from $5.17 to $8.87, across thirteen volunteer-workshop cohorts this-article-was-downloaded-by-12913349130-on-04-august-2016-at-0729. But intermediation is not automatically adherence-positive: the one compensated channel measured in this corpus cuts the other way — load-fund investors time their trades worse than self-directed ones the-ability-of-investors-to-time, the same incentive problem Section 6B documents in leverage-specific form.

The intention-action gap these devices close is the number the honest case rests on. Two-thirds of surveyed 401(k) participants report their own savings rate is “too low”; 35% express an intention to increase it within a few months; 86% of those well-intentioned savers have made no change four months later save-more-tomorrowtm-using-behavioral-economics-to-increase-employee-saving. The argument is not that borrowing produces superior returns. It is that the counterfactual to a contractual obligation is, for most households most of the time, not disciplined voluntary investing but nothing at all. Set that against a measured timing cost of 0.10% per year full-article-bad-timing-does-not-cost-investors-15-of-their-funds-returns-an-exa and the ranking is not close.

E. The dominance objection, and the narrow ground that survives it

The strongest objection to everything above is not that leverage shortens holds. It is dominance. If the load-bearing mechanism is illiquidity plus automaticity, then payroll deduction, automatic escalation and locked registered accounts deliver that mechanism with a worst case of having saved less than was optimal, while leverage delivers it with a worst case of still owing the money on an asset you no longer own. Every commitment device in this evidence base has a bounded downside commitments-to-save-a-field-experiment-in-rural-malawi save-more-tomorrowtm-using-behavioral-economics-to-increase-employee-saving nber-working-paper-series. Leverage does not: the obligation does not shrink when the asset falls, it can force involuntary termination, and the debt survives even total loss of the collateral.

CIRO states this to investors in its own words — the client “has a continuing obligation to repay principal and interest even if the value of the investment goes down,” and in the worst case “you can be forced to declare personal bankruptcy” guidance-on-borrowing-for-investment-purposes-canadian-investment-regulatory-org risk-of-borrowing-to-invest-canadian-investment-regulatory-organization. No commitment-device study located here models an instrument whose enforcement mechanism can compound the saver’s loss and terminate participation against their will. Transferring conclusions from that literature to leverage without flagging the unbounded-downside disanalogy is the central methodological error available in this domain.

The objection is correct, and the honest reply narrows the thesis rather than rebutting it. For a household with workplace plan access and unused contribution room, payroll deduction dominates outright: same mechanism, bounded failure, no interest cost, no suitability question. The leverage case only begins where that channel is exhausted or unavailable — not a rare corner. Roughly 78 million US workers, about half the workforce, have no access to an employer retirement plan, while the share at risk of inadequate retirement funds rose from 31% to 53% between 1983 and 2010 behavioral-economics-and-the-retirement-savings-crisis-chicago-booth-review. For the self-employed professional, the maxed-out registered-account saver and the household with no plan at work, advice to use payroll deduction is not available. That population is much narrower than the one to whom leverage is typically marketed.

The commitment-device analogies need calibrating before they are leaned on, because the two most-cited hard-commitment experiments are individually fragile. The Philippines SEED study reports an 81-percentage-point relative increase in savings balances after twelve months, but on only 28.4% uptake, and against a marketing-only arm the estimate is statistically indistinguishable from zero at both six and twelve months tying-odysseus-to-the-mast-evidence-from-a-commitment-savings-product-in-the-phi. The RAND hard-versus-soft trial’s headline 82% retention in its hard arm is definitional, not behavioural — withdrawals were expressly prohibited — while its genuine behavioural contrast, a soft pledge at 64% against a traditional account at 53%, fell short of significance working-paper.

Brune and colleagues’ Malawi trial carries the mechanism instead, independently replicated with “only very minor discrepancies” savings-revisited-a-replication-study-of-a-savings-intervention-in-malawi-3ie — though even there the operative mechanism is protection of funds from sharing-norm pressure within the kin network rather than self-control, which has no leverage analogue commitments-to-save-a-field-experiment-in-rural-malawi.

6. The Responsible-Sizing Condition

The behavioural benefit is explicitly conditional: it holds only when leverage is sized responsibly within the investor’s financial and emotional risk capacity. In Canada, that condition is regulated through a suitability framework administered by CIRO (the Canadian Investment Regulatory Organization, formed in 2023 from the merger of IIROC, the Investment Industry Regulatory Organization of Canada, and the Mutual Fund Dealers Association). What the framework does not contain is a number.

A. Regulators name the dimensions and specify no thresholds

IIROC’s 2021 Guidance Note GN-3200-21-001 and its CIRO successor GN-3200-22-001 require that any borrow-to-invest recommendation be individually suitable, and enumerate the dimensions:

• annual income
• liquidity needs
• financial assets
• net worth
• loan amount and term
• interest rate
• collateral
• the loan’s pro-forma impact on the client’s finances
• total monthly debt-service costs against monthly income
• the resulting debt-to-net-worth leverage ratio
• whether other assets are encumbered
• whether the client is using leverage to finance securities purchases at all

A leverage risk disclosure with positive client acknowledgement is mandatory borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen guidance-on-borrowing-for-investment-purposes-canadian-investment-regulatory-org. The investor-facing material adds qualitative screens: do not borrow to invest with low risk tolerance, a short horizon, or an intention to rely on investment income to repay risk-of-borrowing-to-invest-canadian-investment-regulatory-organization.

Every one of those is a dimension. None is a threshold. Across the Canadian regulatory record reviewed here — two guidance notes, the investor-education material, two FAIR Canada comment letters and the New Brunswick regulator’s annual report — not one numeric limit on retail leverage suitability appears borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen mfda-leverage-suitability-fair-canada annual-report. One cell went unchecked: MFDA Bulletin #0487-P, the single document that could plausibly carry a numeric minimum, could not be retrieved as a primary source and is known here only through FAIR Canada’s critique. The only hard number in the regime, IIROC Rule 5100’s margin caps, governs how much a dealer may lend as a credit-risk control, not how much leverage suits a given client. Off-book third-party borrowing, which the guidance concedes is “more difficult to detect and/or supervise,” carries no numeric cap at all borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen. Responsible sizing therefore collapses into individual dealer policy, unauditable from outside and inconsistent across firms.

FAIR Canada, the investor-advocacy body, has argued since 2011 that the bar is too low. It was responding to an MFDA proposal that did contain numeric minimum standards, objecting that those minimums were set too low — yet it answered with no counter-number, which is the more interesting fact: the advocacy body judged any threshold approach inferior to a presumption that leverage is unsuitable for retail investors, shifting the burden of proof onto whoever recommends it. It also flagged compensation-incentive misalignment as a driver of bad recommendations, arguing bluntly by 2012 that no reasonable basis exists for concluding a highly leveraged product sale is suitable for any but the most sophisticated, risk-tolerant investor mfda-leverage-suitability-fair-canada regulators-need-to-act-on-leveraged-investing-fair-canada fsco-unsuitable-recommendations-to-borrow-to-invest-fair-canada.

A conditional thesis whose condition has no operational definition is very hard to falsify. Any leveraged investor who panics and unwinds can be reclassified after the fact as improperly sized; any who holds can be counted as proof the condition works. That asymmetry is why the thesis has survived fifteen years without an empirical test.

B. The harm record, and why it is a base rate rather than a tail

The regulatory record speaks to suitability and loss rather than duration, and the two dependent variables should not be conflated. But on suitability and loss the Canadian record is consistent and long-running. Complaints to Canada’s Ombudsman for Banking Services and Investments where leverage was the main issue rose from 12 in fiscal 2009 to 16 in fiscal 2010 to 30 in the twelve months to 31 October 2011 leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2. MFDA enforcement data for July 2010 to July 2011 show 33 of 453 cases (7.28%) involving leverage suitability problems, and IIROC enforcement cases document oversized investment loans sold to clients in their 50s, 60s and 70s to buy deferred-sales-charge equity funds regulators-need-to-act-on-leveraged-investing-fair-canada. Two Ottawa investor-rights lawyers reported leverage matters as roughly half their complaint caseload, hearing clients say over and over that they did not know leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2.

The New Brunswick Securities Commission’s June 2010 leverage sweep found a high correlation between leveraged investing, unsuitable investments and losses to consumers, and that “68 percent of cases where the use of leverage was aggressive were in a loss position” regulators-need-to-act-on-leveraged-investing-fair-canada. The primary sweep report is unrecoverable, but the figure is restated across FAIR Canada’s October 2011 letters to the Canadian Securities Administrators and the MFDA mfda-leverage-suitability-fair-canada, and the New Brunswick regulator’s own annual report confirms the sweep and its suitability findings annual-report — what is missing is the primary text, not the claim’s provenance.

The standard reply — complainants are by definition the badly-sized cases — fails on the record’s persistence. IIROC reported in 2021 that it continues to uncover inappropriate borrowing-to-invest strategies borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen; the 2022 successor guidance restates the same framework guidance-on-borrowing-for-investment-purposes-canadian-investment-regulatory-org; and IIROC’s 2012 guidance had already reported the same increasing pattern fsco-unsuitable-recommendations-to-borrow-to-invest-fair-canada.

The same problem recurring across 2011, 2012, 2021 and 2022 is a base rate, not a tail — and the incentive structure explains why. FAIR Canada documents financing companies offering preferential loan rates tied to fund-family purchases, and registrants able to recommend leverage, process the loan and place the client into the fund company’s products within a single day regulators-need-to-act-on-leveraged-investing-fair-canada. When the party sizing the leverage is compensated on its volume, responsible sizing is not a neutral technical parameter.

The DOL-commissioned timing study reaches the same channel from another direction: commission-bearing load-channel investors time trades measurably worse than self-directed no-load investors — 1.12 points a year in US equity and sector funds — despite the advice the channel exists to supply the-ability-of-investors-to-time, an adverse prior for any program whose behavioural claim depends on that same distribution channel improving investor conduct.

C. Every practitioner guardrail is an exit-trigger rule wearing amount clothing

In the regulatory vacuum, the operative numeric norms come from practitioners. The recurring figures:

• Borrow no more than about 20% of net worth.
• Take a horizon of ten years or more, and “if you’re hoping to strike it rich within a mere year or two by using leverage — don’t bother” leveraged-investing-can-boost-returns-but-do-it-prudently-the-globe-and-mail.
• Maintain an emergency fund covering loan payments for a year.
• Borrow roughly half of what monthly cash flow can support.
• Never rely on the leveraged investments themselves to fund the payments.
• Prefer a fixed-rate term loan to a floating one leveraged-investing-three-risks-of-borrowing-to-build-wealth-the-globe-and-mail.

The Canadian advisor training text independently prescribes a horizon of not less than ten years, borrowing less than debt-load formulas permit so that “any margin calls or losses incurred as a result of leverage can be easily accommodated,” and a no-margin-call loan continuing-education.

Read that list through the trigger framework. The one-year payment reserve exists so the calendar trigger can be met without touching the portfolio. Never funding loan payments from the leveraged investments severs the loop between portfolio price and payment capacity. The fixed-rate term loan stops the calendar obligation from repricing. The no-margin-call feature deletes the price trigger outright. Even the 20%-of-net-worth cap works behaviourally by keeping the payment stream small relative to independent income — that is, by making the calendar trigger survivable.

Not one of these guardrails is really about the amount borrowed; each is about making the exit trigger non-binding. If that is right, sized responsibly within the investor’s financial and emotional risk capacity is already, in substance, a forced-liquidation-avoidance condition. The framing was closer to correct than its vocabulary suggested.

The emotional half resolves the same way. A Scotia Capital advisor who borrowed $10,000 to invest in equities in 2008 described watching it fall for six months as an emotional roller coaster that cost him sleep — despite having far more unleveraged money at stake in other holdings — and now dissuades clients who cannot honestly say they’d sleep through a downturn the-pros-and-cons-of-borrowing-to-invest-cbc-news. That within-person comparison, leverage isolated from position size, is the cleanest evidence here that the distress is a property of the borrowing itself. Practitioners echo the pattern: short-term turbulence overwhelms many investors into selling at a loss leveraging-where-math-and-emotion-are-a-bad-mix-the-globe-and-mail-2, and staggered entry is recommended precisely because a partial position is easier to unwind without damage borrowing-from-peter-to-invest-in-paul-part-2-leveraged-investing-options-and-st. Emotional risk capacity is the investor-side exit trigger, running in parallel with the contractual one. A program that eliminates the contractual trigger and ignores the emotional one has closed one of two doors.

D. Sizing is insufficient, because the risk is path-dependent

The Ontario Securities Commission has publicly told investors that “the risk of leverage declines as the time horizon grows, so you should be investing for the long term.” FAIR Canada rejects that as inaccurate, arguing investors should instead be asked whether they understand the risks, have high risk tolerance, have the net income to service interest, principal and any possible margin calls, are approaching retirement, and can withstand the loss osc-alerts-investors-about-risk-of-leveraged-investing-fair-canada.

FAIR Canada is right, and it does not need the contested academic time-diversification debate to be right. That debate is genuinely open — Bodie argued in 1995 that the cost of insuring against stocks underperforming the risk-free rate rises rather than falls with horizon, “even if stock returns are mean reverting in the long run” on-the-risk-of-stocks-in-the-long-run-financial-analysts-journal-vol-51-no-3; Dempsey, Hudson, Littler and Keasey rebutted him the following year on the grounds that identifying market risk with put prices is circular inside arbitrage-free option-pricing models on-the-risk-of-stocks-in-the-long-run-a-resolution-to-the-debate; and it never resolved, the camps partly measuring different things — probability of shortfall versus magnitude of shortfall time-diversification-controversy-bodie-1995-revisited-eco3min.

The deepest sceptical statement predates both: Samuelson proved in 1963 that an expected-utility maximizer who would reject a single favourable-but-risky bet at every wealth level in a range cannot rationally accept a long sequence of them — “no sequence is acceptable if each of its single plays is not acceptable. This is a basic theorem” risk-and-uncertainty. The distinction is between subdividing a fixed risk among many parties, which genuinely reduces exposure, and repeating a risk over time, which does not, because the maximum possible loss grows in full proportion to the sequence.

That matters because the most prominent academic case for retail leverage rests on exactly the borrowed move. Ayres and Nalebuff argued that buying stock on margin when young reduces retirement risk by spreading market exposure more evenly across a lifetime life-cycle-investing-and-leverage-buying-stock-on-margin-can-reduce-retirement-r. Their mechanism is diversification across time, precisely the object of Samuelson’s theorem risk-and-uncertainty, and their instrument is margin — the one instrument on the wrong side of the boundary drawn here, price-triggered with a measured record of elevated turnover leveraging-overconfidence continuing-education. A terminal-wealth model that assumes the position is held to the horizon assumes away the exact failure the empirical record documents.

Whichever way the terminal-wealth question resolves, the leverage-specific point is mechanical and independent of it. A leveraged position carries interim obligations — interest servicing, principal repayment, margin maintenance, loan maturity — that can terminate it before the long horizon works. A longer holding period does not reduce the chance of striking a forced-unwind event during the loan’s life; it supplies more market cycles in which to strike one. Leverage can therefore lengthen an investor’s intended holding period while shortening the realized one in exactly the states of the world where holding on would have mattered most. The response is to absorb that into the design specification: responsible sizing must mean structured to avoid forced-liquidation triggers, not merely modest in amount.

E. How often does each trigger actually fire?

For the calendar trigger the answer is: rarely, and mostly for reasons unconnected to markets. Canadian mortgage arrears — payments 90 or more days overdue, against a debt stock representing over 70% of all Canadian household financial liabilities — stood at 0.2% before the pandemic and reached 0.5% at the height of the 2008–09 crisis. Unemployment is the dominant driver across a 35-year window; a 100-basis-point rate rise adds roughly 3 to 5 basis points to arrears after two years staff-analytical-paper-document-analytique-du-personnel2026-12. US foreclosure base rates are higher and far more cycle-dependent, ranging twenty-fold from 2021’s low to 2010’s crisis peak us-foreclosure-rate-history-2005-2024-housing-almanac.

The binding constraint on a calendar-triggered leveraged program is therefore the borrower’s employment income, not the interest rate and not the portfolio’s drawdown. A sizing rule derived from this evidence stress-tests job loss. That is not the rule most leverage guidance gives.

For the price trigger, no source located here reports a comparable margin-call frequency statistic for Canadian retail leveraged accounts. The single most decision-relevant number for anyone weighing a leveraged program — how often the event that inverts the mechanism actually happens to suitably recommended clients — is unavailable, while the arrears analogue for the safer trigger type is published quarterly by the central bank.

F. The design specification

A leveraged program qualifies for the behavioural-benefit claim only if it satisfies all of the following. Each is sourced rather than stipulated.

1. Repayment serviceable from income independent of portfolio performance, stress-tested against employment interruption rather than market decline, since unemployment rather than the interest rate drives arrears staff-analytical-paper-document-analytique-du-personnel2026-12. Operationally: a segregated reserve covering at least twelve months of principal and interest; a loan sized to roughly half of demonstrable surplus monthly cash flow; servicing never funded from the leveraged portfolio; a fixed rate, so the calendar obligation cannot reprice; and a maturity no later than the employment income that services it leveraged-investing-three-risks-of-borrowing-to-build-wealth-the-globe-and-mail guidance-on-borrowing-for-investment-purposes-canadian-investment-regulatory-org.
2. No margin-call or forced-liquidation trigger. A term loan without a collateral-value covenant qualifies; a margin account does not continuing-education leveraging-overconfidence.
3. A genuine exit cost on the asset side — strengthening rather than strictly necessary. Commitment demand rises with illiquidity self-control-and-commitment-can-decreasing-the-liquidity-of-a-savings-account-in and re-liquefying a locked asset measurably erodes the mechanism breaking-the-commitment-device; but Save More Tomorrow achieved near-total persistence with none of it, riding on employer payroll machinery a privately arranged investment loan does not have save-more-tomorrowtm-using-behavioral-economics-to-increase-employee-saving. Where that institutional friction is absent, asset-side exit cost substitutes for it.
4. Conservative magnitude — the practitioner norm is no more than 20% of net worth, in the absence of any regulatory figure leveraged-investing-can-boost-returns-but-do-it-prudently-the-globe-and-mail.
5. A ten-year-plus horizon, because amortizing leverage front-loads costs and back-loads equity evaluating-homeownership-as-the-solution-to-wealth-inequality.
6. Conservative underlying assets, since losses magnify more than gains leveraged-investing-can-boost-returns-but-do-it-prudently-the-globe-and-mail.
7. Emotional risk capacity assessed independently of financial capacity, since a small position can produce disproportionate distress the-pros-and-cons-of-borrowing-to-invest-cbc-news. No validated quantitative instrument exists in this evidence base, so the test is procedural: a pre-commitment drawdown rehearsal stated in dollars rather than percentages — what the borrower would owe, and still owe, after a 40–50% decline — and staggered entry, so a partial position can be unwound “with minimal damage” if the rehearsal proves optimistic borrowing-from-peter-to-invest-in-paul-part-2-leveraged-investing-options-and-st.

Miss condition 2 and the program is not a commitment device regardless of how it scores on the others — a price trigger inverts the mechanism outright, where a missing asset-side exit cost merely weakens it. Miss condition 1 and it is not a commitment device at all — it is a bet with a deadline.

Opinionated Synthesis

A. The ranked verdict

The specific study does not exist, and could not have. QAIB derives behaviour from aggregate fund flows with no field for the source of an invested dollar qaib-glossary dalbar-2016; no citation of such a study appears in any Chilton source recoverable here the-wealthy-barber-takes-a-snip-at-debt-the-globe-and-mail book-review-the-wealthy-barber.

The closest verifiable equivalent is DALBAR’s Asset Allocation retention finding, corroborated by Morningstar’s independently constructed allocation-fund gap and paralleled by 85% six-year retention among target-date fund participants dalbar-2016 mind-the-gap-2025 new-research-finds-strong-target-date-fund-retention-among-retirement-investors. All three are delivered by payroll deduction and plan defaults: the finding is about mechanized delivery, not asset mix and not borrowing why-investors-missed-out-on-15-of-total-fund-returns-morningstar. The recollection is mislabelled, not misconceived.

The unconditional claim is declined by name. That leveraged investors hold longer than unleveraged ones is unsupported and contradicted where tested leveraging-overconfidence the-drivers-and-implications-of-retail-margin-trading. The Canadian harm record is not noise around a good idea; it is what the mis-structured version of this idea reliably produces regulators-need-to-act-on-leveraged-investing-fair-canada.

The conditional, mechanism-specific claim is endorsed with real confidence. Where leverage is a third-party-enforced recurring obligation with a calendar-indexed trigger, attached to an asset carrying real exit cost, the evidence supports treating it as a genuine forced-savings commitment device mortgage-amortization-and-wealth-accumulation self-control-and-commitment-can-decreasing-the-liquidity-of-a-savings-account-in commitments-to-save-a-field-experiment-in-rural-malawi savings-revisited-a-replication-study-of-a-savings-intervention-in-malawi-3ie commitment-devices munich-personal-repec-archive.

On behavioural benefit versus return optimization, the evidence favours the behavioural framing. Forced amortization raises net worth roughly dollar-for-dollar without crowding out other savings, and it is that absence — not any housing-versus-equities return comparison — that isolates a savings-quantity channel, since a levered position’s return on owner equity is never the asset’s return mortgage-amortization-and-wealth-accumulation evaluating-homeownership-as-the-solution-to-wealth-inequality. Retirement-savings research lands the same way nber-working-paper-series save-more-tomorrowtm-using-behavioral-economics-to-increase-employee-saving: structure beats optimization, and not narrowly.

But the defensible version is a contribution claim, not a holding claim — and dominance narrows it further. Where a household has workplace plan access and unused contribution room, payroll deduction delivers the same mechanism with a bounded failure mode and no interest cost. The leverage case survives only where that channel is exhausted, roughly half the US workforce behavioral-economics-and-the-retirement-savings-crisis-chicago-booth-review — not the population to whom leverage is typically sold.

On sizing, go further than the conditional does. A modest, well-sized position with a margin-call trigger is still not a commitment device: a forced unwind at the bottom is the precise inverse of the behaviour the device exists to produce. Canadian regulators name every relevant dimension and set no numbers borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen mfda-leverage-suitability-fair-canada annual-report, which makes the seven-point design specification the investor’s own job.

B. What remains unknown, and what would change the verdict

No study tests whether investment-loan takers are simply already-committed investors, which would drain the causal claim for securities leverage while leaving the mortgage evidence intact evaluating-homeownership-as-the-solution-to-wealth-inequality. No holding-period data exists for any structured investment-loan cohort in Canada or the United States; the Smith Manoeuvre’s own promotional materials publish no adherence or dropout statistics welcome-to-the-smith-manoeuvre; and the margin-turnover finding has no modern North American replication leveraging-overconfidence.

Two findings would overturn the position argued here: a cohort study showing structured investment-loan clients unwind at rates comparable to margin traders, or evidence that forced liquidations are common even within the practitioner-recommended envelope. Two would confirm it: a quasi-experimental study of investment-loan take-up finding longer holding periods after controlling for prior conviction, or a long-run extension of hard-commitment savings designs showing durable post-lock persistence rather than merely in-lock balances working-paper. Most decisively, recovery of a pre-2015 primary source reporting retention for a margin or investment-loan cohort would convert the central claim here from a misremembering into something verified but mislabelled.

C. What comes next

Three developments will decide whether any of this matters, and the first is actionable now.

The data to settle the original question already exists, uncollected, in Canadian dealer systems. CIRO’s know-your-client rule requires dealers to record whether a client is using leverage to finance securities purchases guidance-on-borrowing-for-investment-purposes-canadian-investment-regulatory-org — the missing variable. Every Canadian dealer holds a leverage flag attached to an account whose transaction history reveals exactly how long positions are held. Two limits bound what a linkage could see: it would capture on-book and dealer-arranged borrowing reliably but under-count the off-book third-party loans the guidance concedes are harder to supervise, and because the flag’s recording is a matter of individual dealer policy, inter-firm inconsistency would add noise borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen. A regulator-run linkage of that flag to holding period and liquidation behaviour would settle in one study what three decades of vendor reports structurally could not — for the term-loan structure the thesis actually concerns rather than the margin proxy, and with the first real test of the selection confound. DALBAR’s gap is architectural and permanent; Canada’s is administrative and closable.

Product design is converging on the mechanism from the wrong end. The trend in retail leverage runs toward more liquidity, not less — readvanceable credit lines, instant margin, frictionless app-based borrowing. Every one of those innovations subtracts behavioural value while adding return volatility. The behavioural case for a leveraged retirement program lives or dies on features currently sold as risk add-ons: fixed-rate term structure, no-margin-call provisions, a segregated payment reserve, servicing decoupled from the leveraged portfolio continuing-education leveraged-investing-three-risks-of-borrowing-to-build-wealth-the-globe-and-mail. Whether clients who buy the no-margin-call feature hold longer than those who don’t is a cheap, direct test of the claim.

The macro backdrop is not static, and regulation may move. The forced-savings base rates cited here come from an era of historically benign household distress, and calendar triggers are cheap to survive only while unemployment stays low — the Bank of Canada’s own scenario analysis projects trade-war-escalation stress could surpass 2008–09 levels staff-analytical-paper-document-analytique-du-personnel2026-12. Regulators have found unsuitable leverage recommendations in every examination cycle since 2010 borrowing-for-investment-purposes-suitability-and-supervision-canadian-investmen; if regulation moves, FAIR Canada’s preference for burden-shifting over numeric caps mfda-leverage-suitability-fair-canada would make the design specification above a compliance asset — a dealer able to show its product cannot force-liquidate a client has an easier suitability case than one selling margin.

D. The bottom line

The mechanism the recollection intuited is real, better evidenced than the missing citation would have made it, and narrower than the missing citation would have claimed. Borrowing does not make investors patient. A binding obligation makes exit expensive, and expensive exits keep money invested — which is what an amortizing mortgage, a payroll deduction and a locked commitment account each do, and the opposite of what a margin account does. Calling all four of those leverage, or none of them, is the error at the root of the whole question. The right question was never whether leverage lengthens holding periods. It is whether a given financing structure makes quitting costly enough to be unlikely and cheap enough to be survivable.