Confidence in a forecast vs accuracy of a forecast

Recently, I saw a quote by Dean Williams who, in his keynote at the Financial Analysts Federation Seminar at Rockford College on 9 August 1981, said “Confidence in a forecast rises with the amount of information that goes into it. But the accuracy of the forecast stays the same.”

This quote summarizes much of his sentiment in that speech—which, by the way, was titled “Trying too Hard”—that revolves around the idea that we tend to forecast stuff that is not forecastable and in so doing we convince ourselves that we know something, except we only waste our time and energy (or something to that effect).

There is a fair bit of truth in this quote, especially considering the context in which it was stated. Williams gave the speech at a venue packed with investors as a senior vice president of an investing company, with decades of his own experience as an investor. In many ways, the speech was about the efficient markets, and our inability to forecast stock prices. To that end, I have absolutely no issue with this statement.

But the issue with quotes in general, and this quote in particular, is that they are rarely served with the context in which they were originally given. Rather, they are just thrown at you as a universal piece of wisdom. This quote is no exception.

There are, of course, many instances when the accuracy of forecasts increases with the amount of information that goes into it. Information is data. Not just any data. Information is data relevant to making an accurate forecast. At least that is how I view it. More information, to me, means more variables, more observations, or both.

There are no guarantees that either will necessarily improve the forecast accuracy. If a series follow a random walk process, such as stock prices, for example, then... well, nothing else matters. But many economic variables do not follow a random walk process. And they can be forecastable. Specifically, with more data, we can improve forecast accuracy and, therefore, our confidence in a forecast.

“Confidence” is an ambiguous term, anyway. In the context of forecasting. what does it mean exactly? Williams, I think, meant something along the lines of “cockiness.” With more information, we can be more comfortable calling the shots. Again, in some instances, that very well may be the case. Most recently, we have seen an abundance of it among cryptocurrency enthusiasts.

But one might also think of confidence as a measure of uncertainty surrounding a point forecast. Say, we may predict that the Federal Reserve will increase the interest rates by 0.5 percentage points next month. How confident can we be about this claim? This will depend on how well we understand the economy, and how well we understand the Federal Reserve’s “thought process.” Our forecast is likely to be more accurate now than a few months ago because we have observed a fair bit of action by the Federal Reserve in the past few months. And because of this, we can also be more confident in our forecast now than a few months ago.