Myopic Forecasting?

A professor in grad school told me to forget about building models to forecast economic variables. He said,

Find out what the variable is today. Predict that value for the next month or quarter and you won't go far wrong.

His point was that

1. Big-time models are expensive and high opportunity costs (if you think model-builders have valuable alternative uses)
2. Big-time models, despite the major investments in them, were not outpredicting simple myopic predictions much, if at all.

It was rumoured that when Alan Greenspan was Chair of the Council of Economic Advisors, he had a five-equation model that he used and that was just as good as the FRB-MIT 473-equation model (or however many equations it had).



And so it should come as no surprise that (from a review by Michael Kinsley in NYTimes) [h/t to Greg Mankiw],

You gotta love a guy whose idea of an important life lesson is: “I have always argued that an up-to-date set of the most detailed estimates for the latest available quarter is far more useful for forecasting accuracy than a more sophisticated model structure.” Words to live by.

I.e., reducing errors in measurement would have more and bigger payoffs than building better models.



Nowadays most economists use myopic first-differencing to make their-seat-of-the-pants forecasts: predict the growth rate of most flow variables (e.g. GDP, etc.) will continue. It's not a bad way to make rough predictions. But if it works, what does that say about rational expectations??



Another way that many (most?) business economists make their forecasts, though, is at least as interesting. A business economist shared this technique with me several years ago:

Subscribe to at least 20 forecasting services and take an average.

Update: Phil Miller says that when he was hired to do forecasting, he also always adjusted his forecasts according to what other forecasters had predicted:

Back in the day when I used to do forecasting, I used a big simultaneous equation SAS model to generate my forecasts for GDP, personal income, etc. for the US and for personal income, employment, etc. for Missouri, Kansas City, and St. Louis. The growth rates for the National Income Product Accounts often would turn up to be some ridiculous number. So I compared what I had to what the Blue Chip Economic Indicator folks (a survey of forecasters) said and I made some adjustments to my national forecasts based on the average "Chipper" growth rates.



Now I prefer the Magic 8 ball and a dart board.