The art and science of economic forecasting

Here at the National Retail Federation, we are forecasting that holiday retail sales will grow 3.6 percent this year over last year.

Will that turn out to be the correct number? Maybe. Maybe not. Over the past several years, our forecast has occasionally been just about on the money, sometimes too low and, more often, too high.

Does that mean we don’t know what we’re doing? Or that NRF is just a cheerleader for the industry? We believe the answer to both questions is no.

The truth is that economic forecasting isn’t easy. While we try to be as careful, serious and scientific as possible, it is a mixture of art and science.

Even meteorologists have it better than economists. Weather, at least, obeys the laws of physics. There are solid, scientific measurements taken at thousands of locations across the country each day, numbers are updated hourly, weather radar gives real-time information and powerful computers make the weatherman at even the smallest local TV station look like an expert.

And when they get the weather forecast wrong — well, nobody ever expected them to be able to predict the weather anyhow.

Economic forecasting is a different story.

At NRF we use a sophisticated data-driven model as the starting point for our retail sales and employment projections. We examine data from all the official government agencies, and look at indicators such as income, savings, consumption and credit. We balance that against our own intuition and judgment based on decades of experience. And we compare the results we get against those of other professional economists — it is important to interact with others and understand what they are saying and why. In the end, we report numbers in which we have a confidence level of 90 to 95 percent. The fact is that the future course of the economy is uncertain, and no one can predict with a high degree of accuracy how things will transpire.

But the first challenge in all of that is that government data on the economy is no Doppler radar. It can’t tell you with any precision where the economy was yesterday, let alone this morning or right now. Instead, it is badly delayed, with a lag often as long as two or even three months and subject to updates and revisions. While this data is critical, it is information about where the economy was sometime in the past, not where it is at the moment. It’s better than working blind, but we have to grapple with imperfect information. That is part of the art.

Even if the data were perfect, economics is the study of human behavior. And humans are far from predictable.

Even if the data were perfect, economics is the study of human behavior. And humans are far from predictable. Unlike the physical sciences, economics deals with human actions, plans, motivations, preferences and so on, none of which can be easily quantified. Even if there were techniques to quantify these factors, behavior can change — tastes are often influenced by new products, events or other criteria, and data that seemed valuable one moment can become useless the next.

Part of our regular projections includes the identification of “wild cards” — trends or events that are risks to the estimates provided. A “fiscal cliff” coming on a certain date because of inaction by Congress or a trend of warm weather in the middle of winter is not necessarily factored in, even less so a traffic-snarling blizzard on the last shopping day before Christmas. These situations are difficult to quantify, but we do identify some of these unpredictable issues in our forecasts to the extent we can and attempt to describe the value of their uncertainties and their impact on the projection.

Could we attempt to do things differently? Sure. For one, instead of forecasting a specific figure like 3.6 percent, we could forecast a range, say 3.3 percent to 3.9 percent. But people like hard numbers, and would probably focus on the midpoint of the range.

The primary purpose of our forecasts is to provide a reasoned, intelligent projection that increases the odds of understanding what’s to come, and to establish a general sense of the future economic landscape. But we can only do that within the limits of economics — a field that is as much an art as a science.

At the end of the day, forecasting for the retail industry is like mapping a road trip.  You can identify the route you intend to travel and take into consideration the weather, known road closures and the condition of your car. But you can’t predict a flood, a bridge collapse or a breakdown, and you may need to accommodate the unexpected along the way. Still, it’s better to have a plan — a forecast with its mix of art and science — to be able to understand the path ahead and increase the odds of reaching your desired destination.