Momentum is real. You still can't eat it

If academic finance has a crown jewel of anomalies, it is momentum: the tendency of stocks that did well over the past year to keep doing well over the next few months. It has been documented across decades, countries and asset classes, and it embarrassed the efficient-market hypothesis badly enough to earn its own literature.

The standard recipe is mechanical. Rank stocks by their past twelve months' return, skipping the most recent month (short-term moves tend to reverse). Buy the top ranks. Re-rank and rebalance monthly. No forecasts, no stories — just a sort.

We ran the textbook recipe on our survivorship-free Japanese equity panel — the full point-in-time universe, including the companies that later delisted or went bankrupt — and got the textbook answer: before costs, momentum was clearly positive. The winners kept winning, on average, exactly as the literature says.

This matters because most things we tested did not survive even that first step. Momentum is not a data-mining artifact or a survivor-only illusion. As a description of how prices move, it is about as real as published anomalies get.

Momentum's defining feature is also its tax bill: chasing recent winners means the portfolio turns over constantly. Every month, some holdings fall out of the top ranks and must be sold — and in a taxable Japanese account, every sale with a gain hands roughly 20% of that gain to the tax office immediately.

Add the bid-ask spreads and market impact of all that trading, and the arithmetic turns brutal. The strategy must out-earn a buy-and-hold index not by a little, but by enough to pay a recurring toll the index never pays. The index, sitting still inside NISA, pays nothing at all.

In our after-cost, after-tax runs, the gross edge was ground down to zero or below. Not because the anomaly vanished — because the toll ate it.

Momentum is the cleanest example of the central finding of this whole project: a return pattern can be statistically real and still be worthless to a retail investor, because what you keep is the pattern minus your costs, your taxes and your mistakes.

Academic papers usually report gross returns; strategy marketing almost always does. The difference between the paper and your brokerage statement is exactly the part the backtest screenshot leaves out. When you evaluate any strategy claim, the first question is not 'is the effect real?' but 'is it real net of the tolls I personally would pay?'

Institutions with tax-exempt structures, internalised trading desks and basis-point costs can harvest some of what retail cannot — that is partly why the anomaly persists without being arbitraged to nothing. The crowding of professional money in the trade is also why what remains for everyone else is thin.

There is also a structural reason the pattern survives: momentum crashes. The strategy suffers violent reversals (2009 is the canonical example) that make it genuinely painful to hold. Persistent anomalies tend to persist because they hurt; the pain is the moat.

We published it as a negative verdict and moved on: momentum did not earn a place in any recommended default. The tax-free, low-turnover index holding it failed to beat remains the benchmark — and the takeaway travels well beyond momentum. When a strategy's pitch shows gross returns, mentally subtract the toll first. Most edges don't survive the subtraction.