Dividend investing after tax, and the question nobody asks

The argument in the wild is not really about tax. It is about control. One side says income arrives without a decision and does not require selling into a falling market. The other says a dividend is a forced partial liquidation you did not authorise and cannot defer, and that the tax code charges you for the privilege.

Both sides are arguing past a number that can be computed. This page computes it.

What the page does not claim

It does not assume dividend stocks earn more than the market, or less. Nobody can measure that reliably, and a study that quietly picks a side has answered its own question. Instead the pre-tax premium is the horizontal axis of the first chart, swept from −100 to +300 basis points a year. The deliverable is the crossing: the premium at which the two arms draw. Compare it against whatever premium you believe a dividend tilt actually earns. Supplying that belief is your job, not the page's.

The null, and why it is enforced rather than assumed

A dividend is a transfer from the share price into your pocket. The engine builds that in: the month's bootstrapped factor is total return, the distribution is paid monthly, and the price drops by exactly the distribution. With full reinvestment,

shares′ × price′ = shares × (price×G − D) + shares×D = shares × price × G

so any yield returns exactly the same total return. Raising the dividend yield moves money from the price into your hands and changes nothing else — until tax touches it.

Why the tax is not symmetric — the whole study in four lines

• The dividend investor receives the distribution and is taxed on 100% of it, every year, whether or not they wanted the cash.
• The market investor sells shares and is taxed only on the gain fraction of the sale. The basis comes back untaxed.
• The market investor's unrealised appreciation compounds untaxed until sold — and if the money is left as an estate, never taxed at all.
• The dividend investor cannot defer, harvest, or time any of it.

The two numbers most readers need first

A 65+ couple filing jointly with no other income owes exactly zero federal tax on qualified dividends up to $146,400 — a $47,500 standard deduction plus $98,900 of taxable income inside the 0% preferential bracket. For that household the entire dividend-versus-index tax argument is worth nothing.

Add $60,000 of Social Security and that free band collapses to $62,824, because benefits are dragged into taxable income behind the dividends. That second number is the one nobody expects, and it is why the Social Security control on the lab page is not decoration.

What was measured, and where it came from

Yields. Trailing-twelve-month distributions over month-end price, from the Yahoo chart API, computed in DIV-001/notes/calibrate-dividends.mjs:

Dividend cuts, and the folklore they contradict. In 2007–09, trailing distributions per share fell by:

The dividend funds cut harder than the market did, by 1.2 to 1.4 times, and the cut landed more than a year after the price trough — well into the recovery, by which time a retiree has already been selling.

The tax model

2026 federal, married filing jointly, both spouses 65+, standard deduction. The core is RET-CC-002's tax.js, unchanged: seven ordinary brackets, three preferential brackets stacked on ordinary taxable income, the OBBB senior deduction with its 6% phase-out, the 3.8% NIIT at its unindexed $250,000 threshold, and capital-loss netting by character with the $3,000 ordinary offset and carryforward. Constants sourced to IRS Rev. Proc. 2025-32.

Two things are added for this study, in tax-ext.js, additively — tax.js is not edited, and with Social Security and Medicare both off the extension returns bit-for-bit what the original returns:

1. Medicare IRMAA. A cliff, not a phase-in: one dollar of MAGI over a threshold costs the whole step, for each spouse, for a full year, on a two-year lookback. From the CMS fact sheet of 2025-11-14, for a couple both enrolled:

2. Taxation of Social Security via provisional income, under IRC §86. Dividends count toward provisional income, so forced distributions drag part of the benefit into ordinary income — a second tax on the same dollar. The $32,000 and $44,000 thresholds have not been indexed since 1983 and 1993 and are modelled that way on purpose.

Accounting decisions that matter

• The portfolio is charged only the tax it caused — the household's bill including this year's distributions and realised gains, minus what it would have owed on its other income alone. Other income positions the household on the bracket ladder; it is not a bill the investments settle.
• Tax is paid the following April, by selling shares. That is what actually happens, and it removes a circularity that would otherwise need a fixed point.
• Cost basis is tracked lot by lot, with FIFO the default because it is the least favourable to the market investor. In retirement at a 4% spend it usually changes nothing: a 1.2–3.0% yield never exceeds the withdrawal, no shares are ever bought, and there is only ever one lot.
• Distributions are paid monthly, not quarterly. Quarterly payment strands cash between the payment and the spending; the higher-yielding arm carries more of it and loses for a reason that has nothing to do with tax.
• Inflation is a stochastic AR(1) fitted 1948–2025, and indexed thresholds inflate along the same path, so bracket creep is not smuggled in as a result.

Where the dividend fund is given the benefit of the doubt

This study expected to conclude against dividend investing, so every unresolved choice was made in the dividend fund's favour. Its yield default sits at the low end of the measured range and the market fund's at the high end; its distributions are fully qualified; the market investor is charged the least favourable lot method; excess income is reinvested rather than assumed spent.

Verification

• Dividend irrelevance. Roth, equal fees, volatility 1.00: a 1.2% fund and a 3.0% fund end identical to 6.5×10⁻¹⁴ relative across every path.
• The closed form. At zero volatility and a flat 18.8% preferential rate, the algebra predicts a 33.8 bps gap and the engine produces 35.3 — erring high, because the tax is itself paid by selling shares and that sale realises gain too.
• The 0% bracket. A household inside the 0% preferential band shows a difference of exactly zero from a taxable account.
• Fees. 30 bps of fee difference compounds to exactly 30 bps a year over 30 years in accumulation.
• The extension is a no-op. With Social Security and Medicare off, householdBill() returns exactly ORLTax.bill() across 48 input combinations, so RET-CC-002's published figures cannot have moved.

Limitations, stated plainly

• No claim that dividend stocks earn more or less. That premium is the horizontal axis.
• Sector concentration is captured only through the volatility control, which is a thin proxy for a bet on a handful of industries.
• Return of capital is not modelled. Most high-yield option-income funds distribute it, and it is a basis reduction rather than income. That deserves its own study.
• Foreign withholding and ADR fees are ignored.
• State tax is a flat rate on all realised investment income. Real state treatment of preferential income varies.
• ACA premium subsidy cliffs are not modelled. For an early retiree under 65 they are larger than IRMAA.
• The step-up in basis is current law, not a law of nature. If the terminal treatment is what decides your answer — and it often is — then your answer depends on a rule Congress can change.

Everything on the lab page runs in your browser. Nothing is sent anywhere.