Can Fund Managers Time the Market? Treynor and Mazuy Built a Test

There are two completely different ways a fund manager can claim to add value.

The first is stock picking: buying the good companies and avoiding the bad ones, regardless of what the market does. Jensen's alpha was built to measure that.

The second is market timing: getting aggressive before rallies and defensive before crashes. Nothing about picking individual winners, just calling the direction of the whole market. This is what managers mean when they say things like "we reduced equity exposure heading into the quarter."

Jensen's alpha cannot detect timing. It assumes the fund's market exposure is a fixed number. Treynor and Mazuy asked the obvious follow-up question: is there a way to test for timing specifically? And they found a clean, geometric answer.

The problem: how would a timer's returns look different?

Suppose you are a manager with no timing skill. Your fund holds a constant exposure to the market. If the market is up 10%, you are up roughly 10% times your exposure. If it is down 10%, you are down by the same proportion. Plot your fund's return against the market's return and you get a straight line. Constant slope. Same sensitivity in good times and bad.

Now suppose you can genuinely see rallies coming. What would you do? You would increase your exposure before the good times and decrease it before the bad times. So in strongly positive markets your slope is high, and in negative markets your slope is low or even negative.

Plot that, and you do not get a straight line. You get a line that bends upward, a curve. It is steep on the right where the market is rising, and flat or gently sloping on the left where the market is falling.

That upward bend is the fingerprint of market timing, and it cannot be faked by taking more risk uniformly, because uniform risk-taking just makes the straight line steeper. It stays straight.

The key idea, via analogy

Think of a sailor. A sailor with no skill puts up the same amount of sail regardless of the wind. Their speed is proportional to the wind: a straight-line relationship.

A skilled sailor reads the wind ahead of time, hoisting more canvas as a strong breeze arrives and reefing before a squall. Their speed-versus-wind chart is not a line. It curves: they gain disproportionately in favorable conditions and lose less in bad ones.

The curve is the skill. And you can measure it without ever knowing what the sailor was thinking.

Treynor and Mazuy's test is exactly this. Rather than fitting a straight line of fund return against market return, they fit a curved line, adding a squared term to the regression. The coefficient on that squared term is the timing measure:

• A positive curvature coefficient means the fund gets more market-sensitive as the market rises. That is successful timing.
• A zero coefficient means the fund's exposure is constant. No timing, good or bad.
• A negative coefficient means the fund is more exposed in falling markets than in rising ones, which is timing skill in reverse, and the manager would have been better off flipping a coin.

Beautifully, the model measures both things at once. The curvature term captures timing, and the intercept still captures selection skill, the Jensen alpha. One regression, two verdicts.

What they found

Treynor and Mazuy took their test to the mutual fund data of the era, and the answer was largely disappointing for the industry. They found little evidence that funds were successfully anticipating market moves. The upward bend simply was not there for most funds. The relationship between fund returns and market returns was, for the overwhelming majority, a straight line, which is exactly what you get from a manager whose exposure is essentially constant and who is not calling the market at all.

The claim that professional managers add value by getting defensive at the right moments did not survive contact with the data.

Why it mattered

• It separated two skills that people constantly conflate. "Beating the market" is not one talent. Selection and timing are different abilities, they show up differently in the data, and a manager can have one without the other. This decomposition is now standard in performance attribution.
• It gave a testable geometric signature. The idea that a shape rather than a level reveals skill was novel and has been enormously generative. Later refinements, most notably Henriksson and Merton's version, use a similar logic with a different functional form.
• It reinforced the case for indexing. Combined with Jensen's finding that stock selection did not pay for itself, the absence of timing skill left very little for active management to stand on.
• It anticipated the option-like view of skill. A successful market timer's payoff curve looks like holding a call option on the market. This paper is an early, informal statement of a connection that later work made explicit and rigorous.

The honest limitations

• Curvature has innocent explanations. A fund that holds actual options, or convertible bonds, or does covered-call writing, will show curvature in its returns for purely mechanical reasons that have nothing to do with a manager forecasting anything. The test cannot tell forecasting apart from a nonlinear payoff structure.
• Curvature can also be manufactured. Any strategy with an option-like return profile registers as timing. This means the measure can be gamed, deliberately or accidentally, by a manager who simply buys convexity.
• It has weak statistical power. Detecting a modest curve in noisy monthly return data requires a great deal of history. A manager with real but modest timing skill can easily go undetected, so "we found no timing" is not quite the same as "there is no timing."
• Timing at the wrong frequency is invisible. If a manager times over days but you measure monthly, the effect washes out. The test only sees timing at the frequency of your data.
• The benchmark is still the CAPM. Everything here is measured against a single market factor, so all the usual objections about the market proxy, and Roll's critique in particular, still apply.

The one-line takeaway

Treynor and Mazuy realized that a genuine market timer's returns would bend upward against the market rather than track it in a straight line, built a simple regression to look for that bend, and found that mutual funds mostly did not have it.