Recent Posts

Wednesday, October 21, 2015

U.S. Economic Growth - Mediocrity Rules

Chart: Chicago Fed National Activity Indicator

Looking at U.S. data in isolation, I see very little to distinguish the current environment from the rest of the post-2010 period. (There was a strong bounce after the recession, which is unsurprising given the extent of the drop in activity. To use a phrasing popular amongst market commentators, it was a "dead cat bounce.") I believe that the unemployment rate is essentially broken as an indicator, and there is no reason to expect a pickup in inflation on any reasonable time frame. On the other hand, I still do not see enough weakness to justify the panic periodically seen in the risk markets over the past couple of months.

I believe the best way of looking at the current state of the economy is to use aggregated indicators, such as the Chicago Fed National Activity Indicator (technically, the 3-month average) depicted above. (I discussed this further here.) You will often see analysts showing individual indicators that "only did X when the economy entered into recession". Such analysis sounds impressive, until you realise the volume of economic and financial time series that are available. It is always possible to data mine a series that "predicts" the last few recessions; it would be statistically impossible that no such series exist.

The indicator dove at the beginning of 2015, but is now struggling near the zero mark. This is roughly what it has been doing throughout the whole cycle, which is unlike previous cycles, where it managed to stick at higher levels for a period of time during the expansion. Unless there is a strong negative impact from problems overseas, the U.S. economy appears that it is capable of muddling along.

I should note that I do not spend a great deal of time reading the tea leaves of the data in order to read the cycle; as always, I might be missing something.

For those with an interest in trying to forecast the cycle, I would recommend as a first step to look at central bank research. Although I am happy to snicker about the methodological failures of central bank theoretical research, the reality is that they have large research budgets and generally have a good handle on current conditions. You have to keep in mind that there are "political" constraints on central bank forecasts -- they invariably predict that the economy returns to trend growth at the end of their forecast horizon. By implication, they will never forecast recessions (and especially policy errors by the central bank, such as inadequate regulation of the banking sector). Therefore, as long as you take the medium-term forecast with a grain of salt, central bank forecasts are a good place to start.

Is The Output Gap Falsifiable?

That said, the Fed has not exactly distinguished itself in dealing with the current cycle. The tendency to assume that economic variables will automatically "normalise" and the economy will be dragged to "potential" has not served the Fed well over the past five years.

In my previous article, I complained that mainstream models were non-falsifiable because they just adjusted the non-directly measurable natural rate of interest to whatever level is needed to fit the model predictions. However, I was more generous with regards to another state variable, the "output gap." In principle, it should be possible to see whether the "output gap" has an effect on observed economic variables in line with theoretical predictions. (This makes the concept falsifiable, as we could run into data that do not conform with predictions.)

Before 2010, I was relatively comfortable with central banks' estimates of the output gaps in their respective economies. There were problems with real time estimation of the output gap, which is how many New Keynesian economists explain the 1970s inflation. As a result, I would not be overly shocked that the Fed got the output gap estimate for 2015 wrong at the time of writing. But the lack of clarity from the Fed on current conditions makes me less confident that they have a better handle of what the output gap was in 2013.

In summary, the dependence upon non-measured variables, of which we have little certainty as to their values, is a severe drawback of modern mainstream macro.

(c) Brian Romanchuk 2015


  1. Brian, I think Paul Ormerod is on the mark here: output gap is irrelevant for 21st century - note especially the steel mill example !

    1. Hello,

      I agree with his sentiments, but I argue that we can infer something that looks like the conventional output gap based on economic behaviour. I agree that the standard techniques of estimating the output gap are biased (Bill Mitchell wrote a lot about that), but there's still an underlying concept of the economy "overheating" which is not radically different in terms of observed behaviour from the way the mainstream thinks about the output gap.

      I will expand upon this later; I will probably work in the cited article - thanks!

    2. Are you going the route of using the inflation rate to infer the output gap ? Well, anyway I guess we'll see what you say.

      I am sure you know that the New Classicals have a whole different interpretation of the output/inflation nexus. Their model leans heavily on the role of unobserved expectations, just as much as the traditional Phillips curve, but in their case it is expectations about monetization of fiscal deficits I believe.

    3. Yes, you can define the hidden variable ("the output gap") by what inflation and other measured variables are doing. You need a good model for inflation, which I frankly lack. But the basic principle is there. ("Animal spirits" is another squishy variable similar to this.)

      The new classical interpretation of the mechanics of inflation is dubious. But in practice, I doubt that many economists actually believe those details, and so they treat the matter somewhat empirically. For those people, my interpretation of the "output gap" is not that radically different from theirs.

    4. Ashwin I think would agree with you that the New Classical model does not work

      "Whether the central bank monetises government debt or not is almost irrelevant (except from a signalling perspective) because the private sector can monetise government debt just as effectively. And when the government debt does not represent a ‘hot potato’, the private sector often does exactly that."


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.