There’s a pernicious rumor making the rounds that economic and market forecasts are accurate. There’s also a misguided notion embraced in some corners that all predictions are worthless at all times under all conditions. Both of these extreme views are unproductive, bordering on dangerous. Yes, peering into the future is hazardous work and it’s a field that’s burdened with an excess of landmines. But avoiding what can be a nasty business is impossible when it comes to the money game. Thinking clearly about forecasting, as a result, is crucial. Unfortunately, a sober-minded perspective on this hot-button topic seems to be the exception to the rule in a world that’s awash with projections.
There’s a fine line between prudent and reckless behavior with generating and consuming predictions. Defining one vs. the other can be tricky at times. It’s even tougher to steer a middle course between those two extremes, in part because we’re all bombarded with a mix of useful and not-so-useful predictions on a daily basis. With that in mind, what follows are some thoughts on how to manage expectations on matters of prognosticating, guesstimating, and otherwise developing a healthy relationship with a treacherous subject. The list is hardly definitive, but it’s a start, if only as a reminder that predictions typically don’t come with instructions.
1. Recognize the inevitable. Imagining yourself as immune to forecasting is the equivalent of thinking that you can live without breathing. Everyone is swayed by predictions of one form or another, and by necessity. We all need guidance about the morrow, and for good reason: that’s where we’ll spend the rest of our lives. Some folks like to assume otherwise. Even a dyed-in-the-wool passive investor with a strict buy-and-hold investment regimen relies on an assumption — a forecast. An investment, regardless of time horizon, is at its core a prediction. Whether you’re buying shares as a short-term speculation or on the basis of a 30-year outlook, you’re making a bet that a transaction today will yield a profit tomorrow. How do you know that you’ll end up in the black? You don’t, but your forecast gives you confidence that you’ll triumph. You may be wrong, of course — in fact, you can count on no less, at least some of the time. But there’s no alternative. Investing sans forecasting is a conflict with no solution. A similar rule applies to macro, starting with the big-picture view that economic growth will, through time, endure, albeit punctuated with bouts of volatility. How do we know that such a future awaits? We don’t, but we have a forecast.
2. Not all forecasts are created equal, of course, and many (most?) can and should be classified under the technical heading of trash. Entertaining trash, perhaps, but trash abounds. Separating trash from its useful counterparts is an art as much as it is a science, but basic principles apply. One is to be suspicious of analysts with a taste for drama. These folks are easy to spot, as they favor headline-grabbing words like “crash” and “bubble”. They also have a habit of offering few if any details on how they arrived at their visions of the future. For some wizards of finance, there’s always another catastrophe lurking around the next bend. But when forecasts are routinely in the extreme, that’s usually a good sign that credibility is lacking.
3. Another questionable practice is when some self-proclaimed deep thinker declares that some far-off point in time will yield big, big changes, either positive or negative. Rolling out a prediction for 2020 may bring invitations for interviews from the usual suspects, but efforts at divining events years if not decades down the road is little more than rank speculation. Forecasting tends to be more reliable as the time horizon shortens. It’s still hard to estimate what will happen next month, but trying to figure out what’s likely to unfold in ten years is beyond the pale.
4. All else equal, forecasting frequently trumps the occasional or irregular prediction schedule. All forecasts are wrong, but some are useful, as George E.P. Box famously observed. Deciding which ones are more useful than others requires data. Imagine two sets of equally flawed forecasts on next year’s GDP for the US. One economist publishes a new forecast every Monday; his competitor updates his outlook once a year, on January 1. For what should be obvious reasons, the weekly updates are vastly superior, if only because we have a better chance of analyzing the degree accuracy with any number of techniques (AIC, mean squared error, etc.). Delphic comments from on high that arrive every now and again, by contrast, should be treated with caution.
5. Transparency is a big issue with evaluating the value of forecasts. Actually, it’s everything. Judging the validity of a prediction is wholly dependent on understanding the model that produced the outlook. Black boxes and vague commentary that only hint at the underlying mechanics behind the forecast is a warning sign that tells us to look away.
6. Every point forecast has a prediction interval — a range of expected values based on the model. You may not see it, or the forecaster may try to tell you differently, but every prediction falls into a range of possible outcomes. I like to think of prediction intervals as measures of plausibility. Although convention favors the point forecast — a single estimate — a robust prediction has a spectrum of estimated outcomes, recognized or not. This reality complicates the interview du jour, but modeling the future can’t be reduced to one number… except in a 30-second sound bite. It’s impractical to publish the prediction intervals every time, but it’s important to remember that there’s always a range of estimates behind a forecast.
7. Setting reasonable objectives for the use of forecasting is arguably more important than the forecasting process. Whether you’re crunching the data yourself or relying on someone else to perform the heavy lifting, it’s essential to develop practical expectations. That includes recognizing that different forecasts serve different agendas. It’s obviously foolish to assume that any prediction is accurate per se, but some forecasts work well as a benchmark for deeper analysis. One quick example: the monthly risk premia forecasts published each month at CapitalSpectator.com. Yes, they should be taken with a grain of salt, but these long-run equilibrium estimates are useful as baseline predictions for comparison with “superior” tactical estimates for the short run. In the end, it’s folly to think that one forecast methodology reliably towers over everything else. Accordingly, you should have a clear understanding of what a given forecast is designed to deliver and make a habit of looking to a variety of estimates based on different models.
Overall, a healthy relationship with forecasting requires ongoing care and maintenance. Professor Francis Diebold summarizes this point quite well in Elements of Forecasting (3rd Ed.):
Forecasts are not made in a vacuum. The key to generating good and useful forecasts… is recognizing that forecasts are made to guide decisions. The link between forecasts and decisions sounds obvious – and it is – but it is worth thinking about in some depth. Forecasts are made in a wide variety of situations, but in every case, forecasts are made and are of value because they aid in decision making. Quite simply, good forecasts help to produce good decisions. Recognition and awareness of the decision-making environment is the key to effective design, use, and evaluation of forecasting models.