Forecast Error

If you’re in the business of making predictions, you’re everywhere and always bedeviled by an old adage.

Making predictions is hard. Especially when those predictions pertain to the future.

Those whose job it is to forecast financial assets and economic outcomes get around the futility inherent in fortune-telling by way of a rolling mark-to-market exercise, whereby analysts incrementally move erroneous forecasts closer to reality over the course of the forecast horizon.

That’s standard operating procedure. If you ever need to justify it (and don’t worry, you won’t), you can just avail yourself of that old saw often misattributed to Keynes: “When the facts change, I change my mind.”

As appealing as that is, it’s actually quite silly when taken to extremes.

If your 12-month price target for an equity benchmark is, say, 3,500, and you raise it a half-dozen times over the course of a given year in a transparent attempt to keep pace with an inexorable rise to 4,500, there should be some threshold beyond which the erroneous forecast is treated as separate and distinct from subsequent “revisions.”

But there is no such threshold. Analysts use words like “new” to describe updated projections, but rarely are the boundaries explicitly delineated. It’s not often that someone says, “Our old forecast turned out to be totally wrong, which by extension means our assumptions were mostly wrong too, so here’s a brand new forecast with new assumptions. You can disregard the old forecast in its entirety because, again, it was wrong and may have cost you money.”

Instead, one gets a constantly evolving set of projections, tweaked at regular intervals to account for what’s already happened. Like getting to the third quarter of a sports contest and changing the over–under.

In a rising equity market, that opens analysts up to derision, disaster and, after that, more derision.

First, you’re lampooned for raising your target after the fact. It doesn’t take a CFA to know that when a major benchmark is, say, 15% above your year-end number and it’s already mid-September, refusing to raise your target could mean looking up in December and being 25% off the mark. So, you raise your target. And people chuckle, because you were wrong, and it’s obvious you’re only admitting as much due to the proximity of year-end.

Of course, if stocks are rising inexorably, the risk of a drawdown is usually higher, which means raising your target on a delay chances a scenario where the correction you spent the last six months implicitly predicting (by refusing to raise your target) finally plays out. Then the chuckles aren’t chuckles anymore, they’re laughs. Clients are meanwhile exasperated.

At that point, there’s nothing you can do. You were right about November in June, but wrong about July and August. You marked to market in September to account for the previous two months’ gains, but by that time, the storm clouds were gathering, presaging the drawdown you “knew” was coming months ago.

So, you try to thread the needle, where that means your mark-to-market exercise finds you adjusting your year-end target to some middle ground, equidistant from the highs hit during the rally you failed to predict and the lows reached during a hypothetical drawdown you’re still convinced is right around the corner.

If you’re (extremely) lucky, your middle ground will roughly approximate levels reached after a theoretical rebound from your hypothetical drawdown.

Meanwhile, you have to start writing. Because November is when everyone publishes a 150-page tome laying out next year’s forecasts, almost all of which will be proven disastrously wrong, usually by mid-February. At which point you’ll need to start marking to market again.