Chicago Fed's Definition Change

by: Wildebeest

When the Chicago Fed released the latest National Activity Index (CFNAI) they added a new definition for the end of a recession. Previously in the guide to their index they had stated:

A CFNAI-MA3 value above +0.20 following a period of economic contraction indicates a significant likelihood that a recession has ended.

That line still exists, but an additional line was recently added:

A CFNAI-MA3 value above -0.70 following a period of economic contraction indicates an increasing likelihood that a recession has ended.

I've written a couple of articles in which I said that it didn't look mathematically possible for the index to get to +0.2 in the foreseeable future -- the index was being weighed down heavily by the housing and consumption component. So my first reaction was to conclude that since the index wasn't going to confirm an end to the recession, they decided to redefine how to interpret the index.

In late 2002 the Chicago Fed wrote a paper (.pdf) discussing, among other things, what value should be used to identify the end of a recession. Back then they concluded:

A more lax recovery threshold of +0.00, or return to trend growth, would have identified the end of the 1990-91 recession earlier. Had this threshold been in effect, the recovery would have been signaled in April 1992, or 19 months prior to the +0.20 threshold date. On the other hand, the weaker recovery threshold would also have generated false signals. In particular, a +0.00 threshold would have prematurely (by 11 months) signaled the end of the 1973-74 recession. Overall, then, the CFNAI-MA3 with a recovery threshold of +0.20 was able to identify all of the recoveries, signaling four out of the five recoveries within the first five months. Its identification of the erratic 1990-91 recovery, however, did not come until 32 months after the actual trough.

So they seemed happy with +0.2 back then. In order for my "outrage" over the Chicago Fed seemingly moving the recessionary goal posts to be valid, I had to understand why the original definition marking +0.2 as the end of a recession made sense.

The economic indicators used for the CFNAI are drawn from four broad categories of data: 1) production and income (23 data series), 2) employment and hours (24 data series), 3) personal consumption and housing (15 data series), and 4) sales, orders, and inventories (23 data series). The index is constructed to have an average value of zero and a standard deviation of one. The data also has skewness of -1.0 and kurtosis of 5.7. Skewness and kurtosis are best illustrated in the figure below created from the CFNAI data. The negative skewness means the data has a "fat tail" skewed to the negative direction and the kurtosis means the values in the center are taller and thinner than, for example, the normal bell shaped distribution (red line).

Put another way, if the CFNAI index was normally distributed we wouldn't expect two events in 35 years with standard deviations of 4 or more. But big deal, right? What counts is what it is telling us about the economy.

It turns out that one of the objectives of those who constructed the index was to create indicators that showed when recessions were started and ending. The National Bureau of Economic Research [NBER] is the body that determines when recessions start and end. However, the NBER dates recessions many months after they start and finish. The CFNAI would, in theory, enable the entry and exit from a recession to be evident from the chart. In other words, as well as providing an indicator of economic activity, it would provide a coincident indicator of the start and finish of a recession. So the -0.7 and +0.2 targets for entry into, and exit from, a recession were arbitrarily set by comparing the CFNAI with the NBER defined recessions.

Click to enlarge

The Chicago Fed also described the simulation work they did in 2002 to test the precision of the -0.7 mark for determining a recession. They concluded that the -0.7 mark would correctly call the start of a recession 72% of the time. They didn't do a study on the +0.2 mark.

The CFNAI is usually presented as a 3 month moving average, the CFNAI-MA3, because of the noise in the index. Of the 85 indicators in the CFNAI, 72 of them are converted into difference series for use in the index. Six other series are diffusion indices that are already effectively difference series. That leaves 7 series, housing starts, that aren't differenced by the Chicago Fed and don't appear to be received as differenced data.

The fact that this is basically a difference index accounts for the large amount of noise. Numerical differentiation, i.e. differencing, produces noisy data. On the other hand, integration and summing smoothes data. The other problem with difference charts and difference data is that they can be misinterpreted. The usefulness of a difference chart is that it enables you to detect inflection points in the underlying data. For example, when the rate of change of a decline in something decreases, the difference chart will begin to exhibit a V shape. However, as I'll show below, this doesn't mean that the underlying data has turned positive.

I doubt that the CFNAI would be misinterpreted by the Chicago Fed, but many people reading the press releases may interpret a recovery in a difference chart, i.e the V shape, as meaning a recovery in the underlying activity being measured, in this case the economy.

Summing the CFNAI produces a plot of the underlying data that produces the month-on-month CFNAI difference chart. Note that in this case this is inexact because, as I mentioned above, it appears that only 78 of the 85 data series are differenced. Nevertheless, the chart below shows a data series from which, if you take month-on-month differences, you get the CFNAI. The scale is irrelevant, what I want to point out is that when the CFNAI starts heading upwards after being deep in negative territory this doesn't equate to a recovery. The underlying data is still heading down, just at a slower rate. Eventually, of course, actual growth will occur.

Click to enlarge

Another example: one of the 4 CFNAI components is "Employment, Unemployment, and Hours". Of the 24 data sets in this component, 14 relate to employment numbers and contribute 63% of this component. Rather strangely, instead of taking the total number of people employed, i.e. one series, they take the total employed in various sectors, i.e. lots of series, and then add them up. It shouldn't be surprising then that the performance of this component tracks month-on-month employment numbers that the Bureau of Labor Statistics [BLS] releases. The chart below shows the MOM employment numbers from the BLS versus the Employment, unemployment & hours component of the CFNAI.

Click to enlarge

Again, this V shape doesn't equate to a recovery. What it says is that a trough has been reached or soon will be, i.e. things have stopped getting worse. You can satisfy yourself of this by comparing the MOM employment changes with the actual employment numbers, also released by the BLS.

Click to enlarge

No V shape there.

So was I correct in being "outraged" by the insertion of a new definition for the end of a recession? No. It is probably a reasonable call to declare that we have reached a trough which is what they are saying has occurred. Things have stabilized and stopped getting worse. The markers they use for the start and end of a recession are arbitrary anyway, designed purely to anticipate where the NBER will call the start and end. The problem, of course, is that it doesn't follow from the "V" in the CFNAI or other data that robust growth is around the corner and the consumption and housing component of the CFNAI shows no signs of life, a situation that presumably cannot continue indefinitely.

In summary, the V shape in the CFNAI chart signals a bottoming of the economy, an inflection point in the underlying data, and not a return to robust organic growth.

Disclosure: No positions