The Limits of Prediction

Forecasts about the future derive all of their information from the present.


Facebook Share Icon LinkedIn Share Icon Twitter Share Icon Share by EMail icon Print Icon

Predicting the future requires patterns and straight lines. To predict, we project today’s patterns into the future, and we extrapolate along today’s trend lines to imagine where those trends will take us. If the pattern breaks or the line changes course, our predictions do not come true.

That predicament highlights one of the most profoundly strange aspects of our lives in this world. Namely: We do not know the future. No one does. We each are stepping into the future during every instant of our lives, yet no one knows what we’ll find there.

That predicament also accounts for why every science fiction story is ultimately wrong. Futuristic stories are not really about the future. They are about the present, asking the reader to examine where present trends might lead. But no present trend indefinitely continues. Some lose momentum, some are halted and some trend lines ricochet off of other lines. Indeed, the entire study of history is arguably the study of events that happened differently than people expected.

Last year, Modern Machine Shop famously (infamously?) made the prediction of a strong increase in machine tool buying in 2015 as compared to 2014. The prediction was based on the results of our annual Capital Spending Survey. That is, it was a synthesis of our readers’ expectations about their own businesses. And it proved incorrect. This year’s machine tool buying is actually on course to come in lower than the previous year.

Why the discrepancy? Steven Kline details why in this month’s cover story, which offers a view on how to think about business conditions right now and going into next year. As he describes, the prediction did not pan out because the trends changed. Just after respondents filled out their survey forms last year, capacity utilization, oil prices and the strength of the dollar all changed. That represented enough shifting of trend lines that buying expectations when the forms were filled out did not prove indicative of actual buying.

Does this mean the forecast was wrong? In the most obvious sense, yes it does. Yet in another sense, that question represents a false view of what a forecast is and what it can do. At best, a forecast is an informed guess based on relevant trends at the moment of observation. The forecast is still based in the present, because we still lack any direct contact with the future. Thus, every forecast inherently comes with the caveat that the actual future might turn out differently.

We are all accustomed to this caveat in weather prediction. Weather forecasters do not predict rain; they instead state, say, a 60 percent chance of rain—meaning a 40 percent chance of no rain. Weather has been studied long enough and closely enough that meteorologists can gage how likely it is that today’s trend lines will continue.

We don’t have that kind of knowledge about the resilience of the forces affecting metalworking business, but this view is the right way to think about predictions. For the prediction from our most recent survey (see page 74), the question to reflect upon is this: What factors were the survey respondents reacting to, and what is the probability that those factors will persist?