Optimistic Forecasting at the Fed

Unemployment varies quite a bit over time:


Not surprisingly, many people are interested in predicting unemployment in advance.

The Philadelphia Federal Reserve Bank is one interested party. They have conducted a survey every quarter for several decades. In this survey, they ask professional macroeconomists to make predictions about unemployment one quarter ahead, two quarters ahead, and so on up to five quarters ahead.

There are plenty of interesting things to see in the results of this survey. In this post, I will look at two interesting phenomena:

  • First, that forecasters are on average overly optimistic
  • Second, that their overoptimism increases with temporal distance (the further ahead the forecast, the more optimistic)


I replicated this result with a similar survey run by the European Central Bank. Here are the results:


I then replicated the result again with yet another survey run by the Wall Street Journal. Here are the results (note that this particular plot does not include all of the data - I cherry picked the data that looked the best - I can explain more if you contact me):


I think the results are striking. There are a few directions for future analyses:

  • First, what mistakes are driving this behavior. Are forecasters failing to predict recessions, are they predicting greater or quicker recoveries than actually happen, or are they predicting that static periods will suddenly improve?
  • Second, what is the psychological basis for this behavior?
  • Third, what are the downstream consequences of these predictions? Does overoptimism lead to poor investment outcomes, bad policy, or something else entirely?

Post by Bradford Tuckfield

I offer a full range of data analysis services. To set up a free consultation, email me at bradford@analytics-man.com.

(under construction)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>