Thursday, August 15, 2013

Close encounters with the third phase (of the business cycle)

Macroeconomists analyze business cycles in two different ways:

1. Using a model where crises and booms are driven by shocks (unexpected events).

2. In terms of a a succession of phases (expansions and recession) where the key is to define turning points that represent the transition between phases.

Using a model with shocks allows for a much richer description of business cycles. We can allow for shocks of different size, talk about the dynamic response to shocks, etc.

But it happens to be that some economies (and the US in particular) behave in a way that can be approximated by periods of calm and stable growth rates (expansions) that get interrupted by sudden drops in growth (recessions) that are short in nature. And this is why we summarize the volatility of economic variables around the notion of infrequent and negative events that we call recessions.

Such a simple (simplistic?) description of cycles is used by the NBER business cycle dating committee and has become a very common way to analyze the business cycle in other countries. Not all recessions are alike, some are worse than others. One way to distinguish the severity of recessions is by looking at their length; this is the NBER approach to this issue, longer recessions are seen as "more severe" recessions. How good is this approximation? It might be that the number of quarters is not a good-enough indicator, as you might have two recessions that last for four quarters that look very different. But it happens (in the case of the US) that recessions are mostly short and their shape is similar so the length of a recession has been considered a good-enough approximation to its severity.

In my recent work with Ilian Mihov (summarized in this post at VOX), we question the use of the length of the recession as a good indicator of the severity of crisis by putting forward an additional argument that requires redefining the phases of the business cycle.

When a recession is over, the economy has reached a turning point, the trough. It represents a move away from a period of "declining economic activity", the NBER definition of a recession. But what happens afterwards? How long does it take to go back to normal? Where by normal we mean a level of output, employment which is consistent with the notion of trend or full employment. The NBER methodology has never dealt with this question. Once the recession is over, an expansion starts and there is no further communication from the business cycle dating committee until another recession starts. No explicit description of a recovery phase.

Is this approach good enough? It could be if recoveries were symmetric to recessions. If the length of the recovery was directly linked to the length of the recession then we could use the length of the recession phase as an indicator of the full consequences of the crisis. But our research shows that this is not the case and, in addition, the length of the recovery phase is becoming longer over time (at least according to the last two or three recessions).

To understand our point, let's compare three similar US recessions. According to the NBER, the three worst (longest) US recessions after WWII were: the 1973 recession (4 quarters), the 1981 recession (4 quarters) and the 2007 recession (4.5 quarters). The three are almost identical in length. In our research we have dated a recovery phase, defined as the time it takes for the US economy to go back to trend (or "full employment") and we can see that the three recoveries look very different. In particular, the 2007 recovery is already 16 quarters long (and we are not done) while the 1981 recovery was only 5 quarters long and the 1973 recovery was 6 quarters long. In other words, the cost of the 2007 shock is much bigger than that of the 1973 or 1981 ones even if the three recessions looked almost identical.

In conclusion, the use of the recession dates and their length can only provide a very partial and possibly misleading view of the shape, severity and length of crises. If we are using a framework (such as the NBER one) that summarizes business cycles around infrequent and negative shocks that we call recessions, we need to be explicit about what happens in the quarters of years that follow the trough, the turning point as we exit the recession. We are aware that dating the recovery phase can be difficult and subject to uncertainty (and endless debates among academics), but this difficulty should not stop further research from trying to establish consensus around some basic stylized facts about the shape of this third phase of the business cycle.

Antonio Fatás