Let me follow-up on the “Goodbye, homo economicus“, by Anatole Kaletsky with a couple of quick thoughts.
First, academic economists have taken a lot of grief for not predicting the crisis, but realize that very few academic economists do forecasting. There are two uses of economic and econometric models, one is to use the models to understand how the world works, the other is to use the models to forecast. And while, of course, one of the goals of understanding the economy is to be able to predict it, it is simply not something most academic economists do (and the best models for forecasting are not necessarily the same as the best models for learning about how the economy works). Business economists do lots of prediction and forecasting, but academic economists? Not so much. We come along long after events have occurred – e.g. we’re still analyzing the Great Depression to some extent – and try to use those events (as well as data from normal times) to try to understand how the economic world works, how policy can improve performance, etc.
Second, the economists who do forecasting need better data. If we are we are going to forecast the immediate future accurately, we need data that are timely and informative. There can be big differences between forecasts made using the initial data releases, and those made once the data is revised, often months later. We simply do not have an accurate picture of aggregate activity over, say, the last month or two, even longer in some cases, due to data collection lags and other problems, and without accurate contemporaneous data, it’s not possible to produce accurate forecasts (e.g. GDP data are three months old when you get them, and they are revised three months later, then again even later than that though those adjustments tend to be smaller. So we don’t get a good picture of GDP until six months after it happens). It’s like being asked what the weather will be like tomorrow, but only having weather data that is a month or more old.
I’ve been wondering if I should call for enhanced investment in data collection as part of the response to this crisis. However, much of the problem is that the data come from reports that are filed quarterly, annually, etc., and the data collection agencies must wait for those reports before they can produce initial estimates or revisions. Those delays are fairly lengthy relative to our data needs, and it’s not clear to me that more money can overcome these lags in the process. Assembling these reports accurately, then interpreting them properly and then turning them into macroeconomic aggregates takes time.
But there must be some way we can get a better picture of what is going on contemporaneously than we have now, and I think we need to investigate how to make that happen. The data as they exist now are fine for academic researchers who are looking backward and do not necessarily need the very latest months worth of data, though even in this setting this is sometimes a problem, but for forecasting our data are inadequate. Maybe with some investment in the process there would be a way to go out and collect contemporaneous data from all the electronic sources of sales and other information that didn’t exist in the past, or to sample more traditional data sources in a more timely fashion. Obviously, I don’t have the exact answer to this problem, but I do think it’s something we should put some thought into, and if there’s a way to do substantially better at getting an accurate picture of the current economy than we have now, something that would be of great benefit to policymakers, forecasters, and people trying to make economic decisions in the private sector, it would be well worth pursuing.