U.S. Leading Economic Indicator: Either The Fed Was Wrong Before, Or The Fed Is Wrong Now

Summary

• The Leading Economic Index for the US has historically predicted economic growth fairly well.
• The Federal Reserve has recently changed the methods it uses to calculate one of the inputs to the index.
• This change has compromised both the timeliness and prediction accuracy of the Leading Economic Index.

Below is a chart of the Leading Index provided by the Federal Reserve:

It is important to note that the series is not actually historical data. The entire series is the output of a Vector Autoregressive (VAR) model for the current month. In other words, the statistical model recalculates history based on parameters determined by the entire data series through the current day, so each new release can have a small effect on the entire data-set back to 1982.

Each year, the Fed stops producing the USLEI for 2-3 months while it recalibrates its mathematical models. After it does this, there is a brief flurry of releases for the missed months, before things go back to normal.

6-month economic growth is measured by the Fed's Coincident Index. The Leading Index predicts the 6-month growth of the Coincident Index.

From the Fed's website:

"The leading index for each state predicts the six-month growth rate of the state's coincident index."

How accurately does the USLEI predict 6-month growth?

Until recently, very well. Very well indeed.

This year, the Fed changed the way it calculates the Coincident Index. A description of the changes made by the Fed can be found here.

Now, let's chart the 6-month forward growth rate of the Coincident Index, which is what the US Leading Index predicts. Bear in mind the Coincident Index, like the Leading Index, is the output of a mathematical model (Dynamic Single Factor) and is not historical data.

Below is a chart of the new and old versions for the Coincident Index.

As you can see, the new version is a little noisier than the old version, but the correlation between two versions is 98.8% and, more importantly, there is no appreciable time-lag between the new and old versions, particularly at the peaks and valleys.

So far, so good. It's hard to argue which version is a more accurate representation. On one hand, real-world data can be noisy. On the other hand, the US economy is huge and the growth rate is unlikely to change erratically, so it's anyone's guess which more accurately represents the real economy.

But the important takeaway from the chart above is that the change in methodology doesn't introduce any time-lag in Coincident Index growth.

Timeliness of the Leading Economic Index

Now let's compare the Leading Index using the new and old methods, shown in the chart below.

As you can see, the new methodology creates a time-lag of several months in the Leading Index for the United States. There appears to be a 2-3 month lag in general, and average lag around economic peaks is around 4 months.

So the US Leading Economic Index is less timely than it was, lagging the old index by about 3 months, and the Fed still claims it predicts the Coincident Index growth for the next 6 months. How can this be?

Was the Fed mistaken before, or are they mistaken now?

Accuracy of the Leading Economic Index

Now, let's have a look at how well the Leading Index predicts the 6-month growth of the Coincident Index. The chart below illustrates the correlation between the Leading Index and the growth it is intended to predict. This correlation represents how accurately each Leading Index predicted growth of the corresponding Coincident Index over the previous 3-year period.

It is readily apparent the new US Leading Economic Index makes less accurate predictions than the previous method did. More troubling, the new method is least accurate at predicting economic growth entering into a downturn in the economy, which is precisely when it is needed the most. In fairness, this is also true of the old method, but not to the same degree. For example, in 2013-2014, the 3-year correlation was negative for an entire year: For a 4-year period, the Leading Index would have performed (slightly) less well than random numbers. The old methodology correlation never dropped below 58%.

The Fed is aware of all of this. I've spoken to the Assistant Director of the Real-Time Data Research Center at the Philadelphia Federal Reserve, and they had noticed the time-lag in the Leading Economic Index. The reason for this time-lag (and potentially the correlation deficiency) is that they have unconstrained a filter in the model that previously ensured a constant time-shift, and it now allows a variable time-shift.

Yet the Fed still states the Leading Index predicts 6-month growth of the coincident index, even with a variable time-shift in the Leading Index relative to the Coincident Index.

Caveat Emptor.

Analyst’s Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.