Are macro forecasters paid to make good forecasts?

I hope not. In the 1990’s, Macroeconomic Advisors GDP forecasts performed worse than random walk forecasts (basically guessing next year’s GDP will be the same as last year’s) in 6 out of 9 years (Fed WP, Anderson 1998). No sense paying someone to give you worse answers than a quick glance on a government website would give.

Brad Delong says we should believe these people’s forecasts because they get paid to make them. MA’s performance, on the other hand, suggest people pay them for services other than their GDP point estimates.

6 thoughts on “Are macro forecasters paid to make good forecasts?”

  1. I think it’s a random walk with drift, as GDP was growing strongly, no?

    And random walks beats almost everything in macro… (see exchange rates)

    The random walk performs decently versus everything it seems, if this is the paper:

    Are there any forecasts that can be used to see if the economic outlook has improved?

  2. gabe, yeah, univariate random walks are better forecasts than most models.

    ssendam, random walks are the best description of macro variables *because* people form (rational) expectations. Current variables (like stock prices) incorporate expectations about the future, so they’re good indications of what those expectations are. Because current variables incorporate expectations, only random unexpected outcomes move variables around. To repeat myself, unforecastability is an implication of expectation-based models.

Comments are closed.