link to the first part of a two-part primer), I just want to make some brief remarks about how it ties into the notion of forecastability (description). Should we be able to expect to forecast the business cycle?
(This article is fairly brief, and re-iterates points I have made in other articles. Travelling and major home renovations have cut into my writing time recently.)
In order to forecast profits, we need to forecast investment. And why do private firms invest? Because they expect future profits. Can we predict what private firms will do?
(One can presumably forecast various economic variables based on various relationships that hold in steady state. For example, we could probably pin down many of the components of the CPI index on a forecasting horizon of a few months. The problem is that longer term CPI forecasts will end up being conditional on avoiding a recession. I am not particularly interested in those short-term conditional forecasts, but the more intractable business cycle forecasting problem.)
A mathematical model that is based on measured economic variables builds in an assumption that firms will react mechanically to the variables in the model. One could spend days developing intricate mathematical models of recession probabilities based on the 2-/10-year slope, but we need to ask ourselves: why should any firm care about that slope? The reason why the 20-/10-year slope historically had predictive powers was because fixed income analysts priced those instruments based on their views about the cycle. There is no reason that those fixed income analysts will always be correct in their assessments. Furthermore, if you are a fixed income analyst (part of my target audience) you should not be using the 2-/10-year slope as an input into your model that determines the fair value of the 2-/10-year slope!
The 1990s tech cycle provides a good example of the limitations of such mathematical models. Firms were engaged in massive infrastructure investment to meet projected demand for products that did not even exist from a customer perspective. The 3G wireless and bandwidth capacity was eventually used -- but the take-up was too slow to save a lot of the firms behind the investment. Models calibrated against historical data could not capture this, as current capacity utilisation and profitability did not matter -- only the projections. Meanwhile, since these were new markets, there was nothing to calibrate those projections against.
One could try to model this by creating a variable that is not directly measured, but is instead inferred from historical data (like the output gap). This variable could be labelled "animal spirits." The model will then assume that firms mechanically react to "animal spirits."
This will work during an expansion -- investment does seem to follow trends. However, we have no way of predicting the direction of animal spirits. We could find out that our estimate for animal spirits falls in response to recession, but back-casting a recession is not exactly the most impressive analytical feat.
In summary, business cycle analysis mainly revolves around the determination of fixed investment (although we will need to keep an eye on other potential pitfalls, such as policy errors).
(c) Brian Romanchuk 2018