The Truth About Dividend Investing — And the Metric That Decides Nothing

What we granted, and what we actually tested

Granted, not measured: that yield on cost is a correctly-computed number about a real past transaction. It is not a lie and it is not a fudge — a holder who bought years ago genuinely does collect that much cash per dollar he originally paid. Our claim is that it is redundant, which is a different word from false. We also grant the whole of the Canadian income-ETF educator's case: his figures are his own and we contradict none of them, his self-imposed sell rule is more discipline than this genre usually carries, and his tax argument — a realized gain at Canadian fifty-percent inclusion with no better opportunity visible to him — is genuinely correct.

Tested: two things, both on funds we measured ourselves from a survivorship-free store, month-end through 2026-06-30. One: whether yield on cost contains anything beyond today's yield and the price ratio — rendered month by month, 184 of 184 fund-months, 2 of 2 funds, in both directions: one fund whose price tripled and one whose price collapsed. Two: whether the two dividend-focused funds on our own reference table kept pace with the index over their whole lives.

§1 · The locked pre-registration

Committed e75442f on 2026-08-05, before the probe was run — the wager you cannot move after seeing the cards.

The falsifiable question

Do the two curves separate in the two directions we registered — yield on cost above market yield on the fund whose price rose, and below it on the fund whose price fell — on the engine we said we would use, without re-declaring a fund spec to get there?

• A yes looks like: both directions appear on the two named funds, computed natively, and the separation between the two lines equals the price ratio.
• A no looks like: either fund separates the wrong way, or the engine cannot address one of them without a hand-rebuilt spec — which would make the receipt a reconstruction rather than a measurement.
• Ambiguous residue: one fund native and one reconstructed. We pre-committed to report that as two different evidence classes rather than as one table.

Our stated prior: SCHD rising and ULTY falling — by ticker and by direction, 10 days before the first line of the lab

Both directions arrived exactly as registered. SCHD's yield on cost finished 8.30% against a market yield of 2.49% on a price ratio of ×3.3384; ULTY's finished 37.71% against 108.30% on ×0.3482. What we got wrong was the exhibit. We registered two curves as the payoff. The single strongest thing the lab produced is not a curve at all — it is a count, the 166 simultaneous yields on cost in section 2 — and we had not registered it, asked for it, or thought of it.

The primary configuration (locked)

Deliberately NOT tested — named in advance

1. The Canadian income-ETF educator's own fund. It is Canadian and our store is US equities. We hold no price, distribution or total-return record for it, and no number on this page is a measurement of it.
2. The dividend-payer cohort. Two funds are two funds. The size- and sector-controlled payers-versus-non-payers test on our own survivorship-free store has never been run.
3. Yield on cost as a forward predictor. The identity makes the question moot for a holder's decision; it does not answer the statistical question, and this report does not run that experiment.
4. Any fund's distribution ledger. Where a payout was funded from — income versus return of capital — is a different receipt. Return of capital appears here only as a modelled property of a wrapper class, never as a measured quantity of anyone's fund.
5. The price of the covered-call overlay. That is another episode's method-home, public on the channel; its one receipt is quoted in section 5 as a callback, never as this report's proof.

The goalposts — all required to proceed

5 of 5 — rows 1, 2 and 5 registered by name in the lock commit, row 3 by the engine it names, row 4 written with the lab — and that tally is worth less than it looks, which is why it is printed with the question it asked. These goalposts asked “did the lab measure what it registered, on the engine it registered, without touching the jurisdiction it cannot see?” They did not ask whether yield on cost is empty — that is an identity and it could not have come out any other way, which section 3 says at length. And three of this lab's seven checks were not registered in advance at all: the 166-purchase-month fan, the anchor re-derivation and the negative control were written in the same commit as the code they check, so version control cannot witness their ordering, and we are not going to imply that it can.

§2 · The headline arithmetic

§3 · Was it the path? The stress tests

The first stress test is the annualization convention, because that is where this corpus has been bitten before. The column our reference table prints as “CAGR” is an annualized arithmetic mean, not a geometric one, and the two are not the same number. Every rate this report could have shipped was recomputed both ways. SCHD's gap against the index moves from −2.08pp to −1.99pp (full precision; the receipt file's cell reads −1.98 because it subtracts a 2-dp intermediate — REINDEX_FIX_2026-09-05.md, rounding note); SCHD's price-only rate moves from 10.48% to 9.46%. The verdict survives both. But DIVO's gap moves by 0.51pp (−3.21 to −2.70 at full precision; the receipt cell's −2.69 would say 0.52, for the same reason), which is enough that its published rate is only true on one convention — so this report does not print that rate, and says about three points a year instead.

That is the whole reason this report's headline is a multiple and not a rate. $5.95 against $7.68 carries no annualization convention at all, so no convention can inflate it. Wherever a figure could be stated either way on this page, it is stated as a multiple.

The second stress test is the negative control, and it is the one that matters. An identity check that has only ever been seen green is not a check. So the lab corrupts a purchase basis on purpose and re-runs, and its results file records the outcome: 166 of 184 fund-months flip to fail and the 18 untouched months stay green — the guard bites exactly the fund it was pointed at, and nothing else. What that proves is less than it sounds: not the arithmetic, which cannot fail, but that the machine is reading the price it was told to read. This page's own generator carries the same kind of control — corrupt one ledger row and it refuses to write.

And the honest limit of all of it. Every figure here is downstream of one store. A store error would reproduce perfectly through every check on this page, and no amount of internal consistency would show it.

§4 · The controls

The index is the control, and it is the same instrument in every comparison. Over SCHD's own 14.7 years a dollar in SPY became $7.68; over DIVO's 9.5 years, $3.87; over QYLD's 12.5 years a dollar in QQQ became $9.26. Each fund is measured against the thing it is actually made of, over that fund's own window, never a shared one.

Is this lab standing on the same engine as the receipt of record? 8 of 8 published cells across 2 of 2 funds reproduce to ±0.06pp, the print precision of the page they came from. Separately, the whole reference table was re-derived for this episode: 96 of 96 master cells across 24 of 24 funds, 21 of 21 gap cells, 0 differences. That check asked whether the engine still produces the number the page prints. It did not ask whether the store's own records are right.

One scar from that re-derivation, recorded rather than buried. 6 of the 24 rows on our reference table had no live engine entry, because the file that declared them is no longer on disk anywhere. Their specs were rebuilt and they reproduce — but a reconstruction is a different evidence class from a measurement, and QYLD, cited on this page, is one of them. The two funds this report's identity work rests on are not: 2 of 2 natively specified, which is why that was goalpost 3 and not a footnote.

§5 · The best objection, pre-registered and answered

“You did not measure his fund, so what exactly are you telling him?” That is the right objection, and the answer is a boundary rather than a dodge. His fund is Canadian. Our store is US equities and we hold no record of Evolve BANK — no price, no distributions, no total return — so we never measured it, and nothing on this page is a measurement of it. Exactly three things transfer, and only these three:

1. The arithmetic. Yield on cost equals current yield times the price ratio in every currency and every tax code. It is an identity, so jurisdiction is irrelevant to it. That is the only reason this report works at all with a Canadian holder in it.
2. Covered-call class behaviour, offered as an analogue and labelled as one. Our US evidence is QYLD: a dollar became $1.76 while a dollar in the index it writes calls on became $9.26 over 12.5 years. That is a US fund and a US index. It is an analogue for a class, never a stand-in for his fund. Underneath that class sits a contract, and an earlier episode of ours priced one: a single share of that same index with one call written against it every week, 2020-07 to 2026-06, collected $327 of premium and handed back $429 of capped-away gains — you sold the ceiling, and that is where the gap between the two dollar figures goes. It is that episode’s receipt, quoted here as a callback, never as this report’s proof.
3. The only on-jurisdiction evidence in this episode — and it is not ours. Ben Felix measured 20 Canadian-listed covered-call ETFs against matched underlyings and found an average drag of −3.25pp a year (median −2.96pp), with only 2 of 20 outperforming; in his withdrawal study the equity investor ends about 26% richer at identical spending, an implied break-even fee of 1.5–2.7% a year. Those are his measurements of Canadian funds, not ours. They are cited here with his name on them, and they may not be blended into a sentence with anything we measured.

The second objection is bigger: two funds is not “dividend investing.” Correct. SCHD and DIVO are two funds with real, cycle-tested histories — on our own reference table only 7 of 24 funds have lived through a bear market at all, and these two are among them — but two is two. The controlled payers-versus-non-payers test on our own survivorship-free store: we own the data and have not run it, and that is reopen criterion 2 rather than a sentence we are hoping you skip.

§6 · Their own terms

The holder in this episode is a Canadian income-ETF educator. He is not named on this page, in the episode, or on the thumbnail. Two reasons, and they are independent: our own framing rule keeps both practitioners in this episode out of the title and the thumbnail; and he is a sincere retail practitioner selling nothing, so this page identifies him by role, jurisdiction and source rather than by name or video ID — an earlier draft of this page carried the ID, and it was removed before publication for exactly that reason. His video is credited, by channel, at the top of the episode’s description, which is where a credit belongs and where you can go and watch the original. His figures below are quoted as he stated them, and the arithmetic does not depend on who he is.

His numbers are his, and we contradict none of them. He reports a yield on cost of 22.6% against a market yield of 15.1% and calls it his secret weapon; he has a self-imposed rule to sell at +20% unrealized, and he breaks it. His numbers, our arithmetic: 22.6 divided by 15.1 is ×1.4967, which says the price is up 49.7% — and the unrealized gain he states separately, in his own video, is about 50%. The two agree to 0.3 of a point. The inputs are his and are marked as his; only the division is ours. His secret weapon returned the number his own profit-and-loss line had already shown him.

And he is right to hold. A realized gain of about $19k, as he reports it, at Canadian fifty-percent inclusion, with no better opportunity visible to him, has to clear its own tax haircut before a switch makes sense. That argument is correct, and it is the one that actually decides his case. We are replacing his reasoning, not his answer. He also says two things on camera we would not improve on: that Canadian banks “acted like tech IPO launches”, and that “what goes up quickly could come down just as quickly”.

The mirror runs the other way too, on a different holder in a different video, and the same arithmetic applies: ULTY (his cost basis) 35% on cost against 62% market = ×0.5645, a price move of −43.5% · WMAX (his cost basis) 28% on cost against 47% market = ×0.5957, a price move of −40.4%. Same vanity metric, opposite direction, and it decides nothing in either. That holder is not named here either.

§7 · The complete ledger — all 184 rows

§8 · What would change our mind

1. Make yield on cost decide something. Show us a rule that takes yield on cost as an input, is not reducible to current market yield and current price — which is the entire content of the metric — and beats the same rule with yield on cost removed, on a pre-registered out-of-sample test. We will run it here, first. We owe you a warning about this one: the identity makes it a hard bounty by construction, because a transform of two numbers you already have cannot carry information they do not. So the honest version is prove the arithmetic wrong, or show us the rule. We cannot see how, and we would like to be shown.
2. The two-fund scope limit — the likeliest of the three to move. SCHD and DIVO are two funds, not the dividend-payer cohort. A size- and sector-controlled payers-versus-non-payers total-return test on the survivorship-free store could show the two-fund gap is a selection artifact. We own the data and have not run it. Send a pre-registered cohort specification, or we will run it ourselves; until then the general claim stays scoped to two funds and says so on its face.
3. A Canadian covered-call ETF record we can measure. The whole jurisdiction section of this report is a boundary we cannot cross with the data we own. A survivorship-free Canadian price-and-distribution source would let us stop citing somebody else's study for the one market the holder in this episode actually lives in — and it would let us test Ben Felix's numbers instead of quoting them, which is a better outcome for everyone, him included.

§9 · What survives

The cash is real, and so is the compounding. Our own reference page said so before this episode existed, and in these words: “Only SCHD and DIVO are genuine equity portfolios … real, cycle-durable NAV growth.” A dollar of SCHD's price alone, with every dividend stripped out, became $3.76 in 14.7 years. That is a real equity portfolio doing a real thing, and it is not the covered-call machine — a dollar in QYLD became $1.76 over 12.5 years while a dollar in its own index became $9.26. Anyone who leaves this page thinking “dividend stocks are a trick” has been mistaught by us, and we would rather spend a paragraph than let that happen.

What does not survive is the metric's authority. Yield on cost is your current yield times your price ratio since purchase — one plus your unrealized gain. Both of those numbers are already on your brokerage screen; the third is what you get when you multiply them. It cannot tell you to hold, because it cannot know anything the other two did not already know — which is why one fund on one day can wear 166 of them at once and every one of them is correct.

And the question it lets you skip is the only one that pays. Not what does this yield against what I paid, but what did the money do. Over 14.7 years the answer for the strongest fund on our own table was $5.95 against the index's $7.68 — a real portfolio, really compounding, behind the thing it is made of. That is a decision you can actually make. Yield on cost never was.

§10 · Want to run a test like this yourself?

Everything we make is free — the reports, the ledgers, all of it. The only thing that costs us real money is the market data.

Get every report + the 7 Questions one-pager (free)