Full employment and cycle’s end

One of the single most compelling indications that the risk of recession has risen beyond a one- or two-year horizon is that the US economy is now within hailing distance of full employment.

When there was ample spare capacity in the labor market, the Fed’s intentions for the economy were simple: to deliver as much aggregate demand growth as possible without making promises it knew it would end up regretting. *

It would have been odd for the economy to dip into recession in such an environment. And it was odd that the Failure Caucus thought they were being bearish (on markets) when they repeatedly insisted that there was more slack in the labor market than people realized. Once achieved, slack is bullish, guys.

The policy environment is now slightly more complicated.  The consensus seems right to believe that the Fed has no intention of delivering a nasty squeeze to force demand growth below-trend and reverse inflation pressures that have not even yet developed.  However, for the Fed it is no longer a case of pedal to the metal. They are now debating how quickly to raise the funds rate, rather than looking around for new ways to deliver stimulus.

This means the medium-term recession risk has probably risen a bit, which means in turn that the “easy” part of the equity rally is now behind us. This note is not about equity strategy. But there is an obvious link, which I leave aside here.

What I want to do with this note is put some empirical bones on the intuition by looking at the relationship between the unemployment rate and subsequent developments in the economy.  Having run some numbers, there is no major revelation, but my fuller look makes me inclined to put a bit more emphasis on the idea that the recession risk has risen beyond the one- or two-year horizon.  There is certainly no near-term signal blaring here, at least from the labor market.

Before I move on to the data I cobbled together, let me get clear on two details.  First, full employment is a “good” thing.  It is just that from a forecasting perspective the achievement of success can mean that the next move is down.  The fully employed economy is the happier economy.

Second, there is no secret sauce in the unemployment rate or anything else I know of. I am just looking at the numbers behind an idea that seems to me to be relevant.

I look for three possible relationships here. Based on the US experience since the late 1950s:

Is there a reliable pattern in the link between contemporaneous movements of the unemployment rate and the approach of recession?

Taking a slightly longer view, can I confirm Goldman’s rule that it is ominous when the 3-month moving average of the unrounded unemployment rate has risen more than 30 basis points?

Is there empirical support for my intuition that the approach of full employment is a yellow flag?

In approaching these questions, I use revised data on the unemployment rate, rather than contemporaneous vintages.  This is not a major issue as the only source of difference is changes in seasonal factors.  The estimate of the natural rate of unemployment, relevant for assessing the third possible relationship is a bigger issue, as I discuss below.

Screen Shot 2016-09-02 at 11.54.40 AM

The first relationship is trivial and may seem not to be even worth mentioning. When the unemployment rate is falling, the economy is not in recession. Since 1960, no recession has begun without a rise of the unemployment rate. I guess this is no surprise, as the unemployment rate should rise merely with a deceleration to below-potential growth.

Ideally, there would be a reliable lead, but the evidence there is mixed. In the 1973 recession, the first uptick of unemployment was contemporaneous with the start of recession.  Typically, there is a longer lead, but it has varied quite a bit.

In my view, the more recent experience of the past three recessions is more likely the more telling, because these were slow moving developments (at least initially) that resulted more from balance sheet pressures than from inflation shocks, but that would be a case of imposing some of my own priors. Based on the data, what we can say is that if the unemployment rate is falling we are not in recession. You knew that.

One obvious problem with this first pattern is that I discern it with lookback bias.  I define an unemployment trough as that which comes before big rise of unemployment, which happens to be associated with recession. When living through this in real time, you can’t really tell if a couple upticks of unemployment represent a trough or just noise within an expansion.

Screen Shot 2016-09-02 at 11.54.50 AM

The economists at Goldman control for this lookback bias by proposing a test. How well would we fare at forecasting recession if we called a recession only once the 3-month moving average of the unrounded unemployment rate had risen 30 bps or more above its low for the current cycle?

I like the guys at Goldman but we are all human and I cannot exclude the possibility that they choose their trigger retrospectively based on what works. Not judging.  But at least this approach does not just assume the result.

The rule works pretty well.  Since 1960, every time the 3-month moving average of the unemployment rate has risen 30 bps, the economy has subsequently moved into recession with a very steep further rise of the unemployment rate.  Moreover, here has never been a recession without this rule triggering.

While the signal generated by this approach is by design “late” in the employment cycle itself, it still allows the forecaster to “miss” (a good thing) almost 90% of the cyclical rise of unemployment and more than 90% of recessions, measured by time.

Regarding the latter, the signal has been given ahead of the recession on 5 of 8 occasions, allowing the forecaster to “miss” the entire recession. On the other three occasions, the signal was given after the recession had begun, but not by long.

One problem here is that there is no signal from unemployment to forecast recovery – or to get long stocks again, if that is your purpose. That is a serious issue and one I leave for another time.  Hint: unemployment is not very helpful at catching the bottom of the cycle or the market.

Screen Shot 2016-09-02 at 11.55.02 AM

I turn now to the last question, the one for which I have an intuition that I wanted to flesh out “empirically.” That is, how troubling for the survival of the expansion is the achievement of full employment?

To “test” this, I look at how long expansions have survived after the unemployment rate has dipped below the CBO’s estimate of the long-term “natural” rate of unemployment.  There is, alas, an issue here too. The CBO’s estimate of the natural rate is subject to revision and, according to Staiger, Stock, and Watson (1997) the real time standard error of the estimate is about 60 bps. Sorry.

But I am not going to use real time data, because a) I don’t get paid for the effort and b) the meaning of the natural rate has evolved through time. One advantage of the revised data is that its meaning is at least stable across the historical periods. This partly compensates for the lookback bias in the historical relationship between it and recession.

In any event, I arbitrarily give myself 50 bps of leeway and ask, how troubling to the expansion’s prospect has it been when the (unsmoothed) unemployment rate has dipped 50 bps (or more) below what we now consider to have been the natural rate of unemployment for the period involved?

Screen Shot 2016-09-02 at 12.03.59 PM

Based on this interpretation of the history, with the caveats I have put in place, it tells you that the expansion is probably getting a bit long in the tooth.  The last three times this has happened, which I think is the most relevant, the economy was on average about two years from a business cycle peak.

But two years is a fairly long time and this condition of over-full employment has not yet even been achieved.  CBO’s current estimate of the natural rate is 4.8%.  Fifty bps below that is 4.3% and so the economy can probably grow about a percentage point above potential for about another year before we get there. And only thereafter would the clock start ticking.

I am stacking assumption on top of assumption here, which weakens my result. But that is arguably the point. My intuition says that the near achievement of full employment is an issue, and I am going to stick with that – at least to the extent of believing that the “easy” part of the expansion and (more to the point) equity rally is behind us. But this is a yellow, rather than red, flag – if you will pardon the metaphor and somewhat weak conclusion.

* I don’t mean anything particularly deep by this. I mean only the mundane point that they would not apply optimal control. By the time they got focused on it, the economy was too close to full employment and the short-run advantages of optimal control were too small to be worth the effort.  It was just a deflection from their very sensible decision not to follow ANY policy rule.