Equity backtests often mishandle dividends in one of three ways: they treat an adjusted price series as executable, credit cash on the wrong date, or ignore the dividend entirely. Each choice can produce a plausible chart with the wrong account value.
The useful distinction is between the stock’s traded price and the cash or shares created by a corporate action. Track both. A dividend changes what the shareholder is owed; it doesn’t make the ex-date price drop disappear from the market record.
Mistake one: trading an adjusted close as if it were a quote
Historical vendors commonly adjust prices for dividends so that a chart can represent reinvested total return. That series is useful for some kinds of analysis. It is not the price at which an order could have filled.
Consider a stock that closes at $100 before going ex-dividend for $1 per share. If it opens at $99 and nothing else moves, an unadjusted price chart shows a 1% drop. A dividend-adjusted series may remove that discontinuity by rescaling historical prices. A backtest that uses the adjusted series for signals and fills has mixed a return convention with a tradable quote.
The wreckage depends on the strategy. A threshold based on a recent high might fire on a price that never traded. A stop can trigger at an adjusted level. Returns calculated from the same adjusted series may look smooth while fills are priced in a different universe.
| Price field | Useful for | Unsafe assumption |
|---|---|---|
| Raw trade or quote price | Signals and simulated execution, with suitable timing | It already includes shareholder cash flows |
| Dividend-adjusted close | Some total-return comparisons | It was available as an executable price |
| Adjusted OHLC bars | Analysis when adjustment methods are understood | Intraday highs and lows match historical quotes |
Mistake two: crediting the dividend on the ex-date
The ex-dividend date determines who is entitled to the distribution under the applicable market rules. The cash usually arrives later, on the payment date. Those dates serve different purposes in an account simulation.
Suppose you hold 100 shares through an ex-date for a declared $0.50 dividend, payable two weeks later. The strategy may have earned a receivable of $50 on the ex-date. It does not yet have $50 of spendable cash. If the simulator immediately adds cash to buying power, it can fund a purchase the real account couldn’t make with that distribution.
For a long-only buy-and-hold return calculation, recording the entitlement on the ex-date may be enough. For a strategy that reinvests distributions, tracks buying power, or trades around settlement, keep the receivable and payment separate. Use the payment schedule as it was known at the time; a later data correction shouldn’t rewrite what the system could have anticipated.
For a basic long position, an ex-date price drop near the dividend amount is not automatically a trading loss. The account has fewer dollars of stock and a dividend receivable. Whether those values offset exactly depends on market movement, taxes, withholding, and the simulation’s conventions.
Mistake three: ignoring dividends because the price already dropped
Leaving the price unadjusted and omitting the cash flow gives the opposite distortion. In the $100-to-$99 example, a holder of 100 shares now has $9,900 of stock instead of $10,000, plus a $100 dividend entitlement. Counting only the shares reports a $100 loss where the simplified total value is unchanged.
That error can be especially large in a long sample or a portfolio of high-distribution stocks. It can also affect signals: a strategy that interprets every ex-date gap as ordinary downside may sell at exactly the wrong time. The dividend may explain the gap, but it doesn’t dictate the next market move.
How should an equity backtest model dividends?
Start with raw, point-in-time prices for decisions and fills, then represent corporate actions as explicit account events. The required detail depends on the question the backtest is answering.
- For total-return measurement: include entitled distributions in portfolio value, with a documented reinvestment assumption.
- For strategy simulation: keep shares, receivables, cash payments, and buying power distinct where timing matters.
- For signals and execution: use price fields that correspond to actual historical market levels and their availability times.
Also check splits, special dividends, and withholding separately. A $1 regular dividend and a $10 special distribution can create different price adjustments and very different strategy behavior. A vendor’s single “adjusted” flag rarely tells you enough.
When a backtest crosses an ex-date, ask what the account owned before the date, what it became entitled to, when the cash arrived, and which price series drove the decision. If those answers are all implicit in one adjusted close column, the equity curve is carrying assumptions you can’t see.
← All posts
