A regional perspective on the U.S. business cycle

Chicago Fed Letter, Nov 2002 by Kouparitsas, Michael

It is not difficult to come up with a definition of the business cycle that most of us would understand. A thornier issue is how to come up with an empirical measure of the business cycle. This article tackles the question and, in the process, offers a new measure of the U.S. business cycle, derived from regional data. A highlight of this approach is that it facilitates the identification of region-specific influences, such as technology booms or commodity price spikes.

On a recent flight to Australia, the traveler in the neighboring seat (upon learning that I was an economist) promptly asked me to explain what the business cycle was. I pointed out that I was an academic economist and that my answer might be a little vague. My neighbor quickly reminded me that we had 15 hours to work out the details. In fact, as I explain below, my definition of the business cycle is pretty easy to understand, so it took no time at all to answer her question. On the other hand, I did not complete my answer to her follow-up question on whether it was possible to come up with an empirical measure of the business cycle before touching down at Sydney's Kingsford Smith Airport, since she tended to doze off every time I uttered some statistical jargon. The mainstream academic view of the business cycle has its roots in the pioneering analysis of Burns and Mitchell conducted in the 1940s.' Burns and Mitchell's definition emphasizes that business cycles consist of expansions occurring at about the same time in many economic activities, followed by similarly general contractions.' In other words, the business cycle is not marked by a large upswing or downturn in a particular industry or economic region, but is the outcome of an upswing or downturn in many industries or regions.3 In this Chicago Fed Letter, I offer a new measure of the U.S. business cycle, derived from regional economic data, that is consistent with the narrower academic definition of the business cycle.4 I focus on the differences and common features of U.S. regional cyclical fluctuations, with common movements across regions measuring the U.S. business cycle and the remaining variation highlighting region-specific sources of disturbance.

Measuring cyclical fluctuations

The starting point for any business cycle analysis is the age-old problem of decomposing fluctuations in economic growth into trend and cyclical components. I use the unobserved components (UC) approach developed by Watson.5 Watson's approach explicitly assumes that current output (measured as the log of U.S. gross domestic product) depends on its most recent past observation plus some random component and a constant term. The constant term, typically called drift, measures the underlying trend growth rate. That is, in the absence of random fluctuations, trend output grows at a rate equal to the drift term. In contrast, positive random fluctuations lead to trend growth in excess of the drift, while negative random fluctuations cause the trend to grow by less than the drift. Using this method, Watson generated a cyclical component for aggregate U.S. output with peaks and troughs that closely matched those reported by the National Bureau of Economic Research's (NBER) Business Cycle Dating Committee. Following Watson's approach, I assume that log of per capita income for the eight U.S. Bureau of Economic Analysis (BEA) regions is composed of a trend, which is modeled as a random walk with drift and a stationary cyclical component. I build on his approach by assuming that the cyclical component of U.S. regional income is made up of two parts, a common cycle across regions, modeled as a second-order autoregression, and a region-specific cycle, modeled as a first-order autoregression. Following my own research, I also allow the drift to vary over time, with three discrete shifts in the trend growth rate occurring from the start of the sample, 1961:Q1 to 1972:Q2, the productivity slowdown era from 1972:Q3 to 1995:Q4, and the new economy era from 1996:Q1 to the end of the sample, 2001:Q4. The resulting common cyclical component is my measure of the U.S. business cycle, while the remaining region-specific components give us some insight into the difference sources of disturbances affecting U.S. economic regions.

U.S. business cycle

Figure I plots the common cyclical component of log per capita regional income of the eight BEA regions from 1961:Q1 to 2000:Q4, against the NBER's peak-to-trough dates. The common cycle generates turning points that match up closely with the NBER's business cycle dates. According to this measure of the business cycle, the U.S. economy has been operating below its trend for much of the 1990s. However, I estimate the average trend growth rate of per capita income across all regions to be about I percentage point higher in the second half of the decade than it was in the productivity slowdown period from the early 1970s to mid-1990s.

To test whether U.S. regions have different sensitivity to the U.S. business cycle, I allow the regional cyclical component to be a function of the common cyclical component scaled by a sensitivity coefficient. In order to identify these coefficients, I need to set the sensitivity coefficient for one region to be one.