Getting more value from consensus forecasts

Yesterday I posted a piece about how business people can squeeze value from the consensus economic forecasts compiled by the NZIER, and along the way I mentioned that it's useful to look at the range of views around the consensus average: "it's very helpful to see the spread of views, as it gives you some feel about the degree of uncertainty ahead".

By coincidence I was fossicking around today in the AEA journals, and came across "Uncertainty and Economic Activity: Evidence from Business Survey Data" (American Economic Journal: Macroeconomics 2013, 5(2): 217–249). I can't find a free link to the paper (or to other versions - it also appeared as an NBER Working Paper) but if you're an AEA member or your organisation has a subscription to the AEA journals, you can access it via the AEA's journals page. In any event here's the gist of it.

It found that the spread of opinions (in this case across respondents to German and American business opinion surveys) did indeed convey useful information about the degree of uncertainty around the economic outlook.

The authors looked at two measures derived from the surveys: one was an ex post one (the variance in how closely businesses' expectations for production matched their actual outcomes), and one was ex ante (the range of views across the businesses about expected production). And they found, first, that the two measures were closely correlated (a wider range of views in advance did indeed signal more bumpy outcomes later, where expectations had not been met), and that there were clear links between higher uncertainty (on either measure) and economic activity, and in an inverse way. When uncertainty rose, subsequent activity fell: big rises in uncertainty occurred, for example, just before recessions.

There's a lot more in this paper than the usefulness of paying attention to the range of views across businesses or forecasters. The authors look, for example, at what the transmission mechanisms are between that rise in uncertainty and subsequent weakness in activity, and find that for Germany it seems to be that firms go into "wait and see" mode, deferring hiring and investment, while in the US it's not so clear what's going on.

Bottom line, though, for people trying to navigate better through the economic cycle, is this: there's value in knowing what the consensus expectation is. But there's also real value in knowing how wide is the range of views around that consensus - and in particular, in noting if the range of views has suddenly got wider. It's not a good sign when it does.