JOLTS data and the "2019" recession

Another month, another JOLTS data release. However it looks like this time I can with a certain level of confidence say that the "2019" recession [1] is underway. It's still not as visible in the hires or quits data (only as biased model error), but job openings (vacancies) are definitely showing a deviation. Job openings appears to have lead the early 2000s recession (but that conclusion is uncertain as JOLTS data is only available from December 2000). Since this is a bold prediction, let me show the updates for all the JOLTS data series I've been watching. Click to enlarge the images.

Job Openings

Separations

Hires

Quits

And here are a couple of animations of counterfactual recession centers from June 2018 to June 2019:

To be specific, my prediction is that the current JOLTS job openings data is going to continue to deviate forming a shock (a logistic step function after subtracting the log-linear component) that will become visible (i.e. detectable with e.g. this algorithm) in the unemployment rate as originally described here but also in my paper. The exact timing of the NBER recession is uncertain (since it seems to depend more on the unemployment rate, which lagged the JOLTS indicators in the previous recession), but the time scale appears to be 2-4 quarters (6 months to a year). The unemployment shock center matches up with the NBER recession centers within a month or two on average.

One issue is possible data revisions; they appear to come with the March update (per ALFRED) with February data with the big Fed March meeting, so we won't see any until March 2019. However, the revisions all appear to be on the order of the model error so the only worry would be biased errors that shift all the data one way (this happened last March for the quits and hires rates). But overall, I'd say I'm at least 80% confident in this prediction inasmuch as I can put a qualitative Bayesian prior on the model.

...

Update:

Here's the Beveridge curve also discussed in my paper:

...

Footnotes:

[1] I put quotes around the 2019 because the recession is technically already visible in the Job Openings data, but NBER will likely say it began (i.e. the business cycle peaked) in some quarter of 2019 as the unemployment rate shock is probably at least 6 months in the future.

Consumption over investment

Steve Roth looked at Yaneer Bar-Yam (of NECSI) et al's paper [pdf], writing some notes on it on his blog. I'm not sure about the rest of the paper, but the ratio of consumption to investment as a leading indicator of recession piqued my interest.

In the dynamic information equilibrium framework, if we posit that consumption and investment are in information equilibrium ($C \rightleftarrows I$), then the ratio $C/I$ should follow (per my paper):

$$
\frac{d}{dt} \log \frac{C}{I} \sim \alpha + \sum_{i} \sigma_{i}(t)
$$

which in basic terms means that we should see lines of constant slope on a log graph possibly interrupted by "shocks" $\sigma_{i}$. Using the same methodology as my paper, this is in fact what can be seen:

The model is generally good, except for a bit of overshooting in the Great Recession. And compared to some other purported leading indicators I've looked at on this blog, it's not too bad! It definitely seems to lead the early 90s recession, and roughly tied with conceptions for the Great Recession [1] (click to enlarge for all images):

However, $C/I$ has a somewhat inconsistent relationship over time, sometimes leading and sometimes lagging (which is part of Steve's point). But it also falls apart if we look at earlier data:

This was interesting to me because the transition is where I've also posited a qualitative change in the behavior of the economy — in the wake of the end of the demographic shift of women into the workforce:

In the 90s and 2000s, women's labor force participation stops generally rising and becomes more correlated with the business cycle (as well as men's labor force participation). The $C/I$ ratio does this as well. In fact, Bar-Yam et al also note a transition from an exponential to a cyclical behavior around the same time for another time series (their Figure 5). This also matches up with the transition from the "Phillips curve" economy to the "asset bubble" economy I've described before.

It's that latter part that makes me doubt the cyclic nature of the indicator and these recessions in the paper. The asset booms and busts since the 90s correspond to the dot-com and housing bubbles — these involved entirely different causes and mechanisms making it exceedingly unlikely that they represent a first and second oscillation of one "cycle" that is supposed to continue [2].

More likely, the $C/I$ ratio will just continue to decline until it is hit by another "shock" (possibly a recession in the 2019-2020 time frame based on other indicators) with random timing (see this discussion of the "linear with random shocks" approach versus "nonlinear/chaotic dynamics").

...

Update 4 June 2018

Another ratio came up today (Justin Fox via Noah Smith) that I labeled $S/L$: service sector payrolls over total non-farm payrolls (FRED SRVPRD/PAYEMS). The growth rate (dynamic equilibrium $\alpha$) is about −0.0007/y (−0.07% per year) which is close enough to zero.

Overall, this appears to be the flip side of the loss of manufacturing employment (I didn't resolve the individual recessions in this one):

...

Footnotes:

[1] The different measures are:

C/I = Consumption over investment ratio
Cons = Conceptions
JOR = JOLTS Job Opening Rate
U = Unemployment
EPOP M = Prime age employment population ratio (men)
EPOP ratio = Prime age employment population ratio
Wage growth (ATL Fed) = Atlanta Fed's wage growth data

[2] Notably, Bar-Yam et al leave out data after 2015 (which would have been available in December of 2017) which would show the bump up in 2016 (possibly associated with the mini-boom of the mid-2010s which included a bump up in wages as well as bump down in unemployment).

Unemployment rate time series is on trend

The latest unemployment data came out last Friday, and despite the current president wanting to take credit (read Justin Wolfer's twitter thread) it's really just a continuation of the dynamic information equilibrium model trend.

It's true it is a bit below the 90% confidence region, but we should expect at about two of the 17 post-forecast points to fall out side it (which is roughly what's happened). If fewer than 10% of the points fell outside the region, we've likely estimated our errors too conservatively (or there is a model that could do better). Plus, there are the annual data revisions from BLS that come with the January numbers.

Overall, the continued decline in the unemployment rate is expected and the possible turnaround with the next recession will be first seen in e.g. JOLTS data.

Latest PCE inflation data

Here is how the forecast of core PCE inflation is doing (both log derivative/"continuously compounded annual rate of change" and year-over-year). Click to embiggen ...

Vacancy yield and labor market analysis

Nick Bunker's labor market analysis are the go-to for detailed, nuanced — and yet "mainstream" — views. By this I mean that if you're not all that into listening to crackpots, he's definitely a good way to go. But then, I'm a crackpot with my own theory of how labor markets work (paper here).

Bunker is attempting to explain why wages haven't risen as much as they have in the past, and halfway down he notes the "vacancy yield" (hires per vacancy, or H/V):

Hiring has not been particularly strong during this recovery in the U.S. labor market, particularly when measured against the number of vacant jobs. Part of the decline in hires per job vacancy—a metric known as the vacancy yield or the job-fill rate—is due to the tightening of the labor market, but even accounting for the low unemployment-to-vacancy ratio hiring is down. (See Figure 2.)

Here's that Figure 2:

Of course, back in September 2017 I modeled H/V and made a forecast that shows the decline in H/V is about what we should expect (and the decline is actually a bit less than we'd expect which might be due to the leading edge of an upcoming recession):

I show a counterfactual recession as a dashed line and the post-forecast data as a black line. However, Bunker also notes this:

But only certain kinds of hiring are down. Hiring of workers who were previously unemployed or out of the labor market is in-line with the previous labor market recovery. The hiring that is down is the hiring of already-employed workers.

You could read this as a lower risk tolerance: fewer people with jobs out looking for new, better jobs — a kind of hunkering down against the uncertain future. A similar sentiment may be measured in the conceptions data.

In any case, this made me want to look at separations in the JOLTS data (which I had for some reason neglected). It tells a similar story to job openings (vacancies) — a possible leading edge of a turnaround (click to enlarge):

*  *  *

PS I also wanted to note that wage growth is also about where it should be based on this model:

Speculation: is this lack of wage growth evidence of sexism? As male labor force participation falls, the rate of wage growth falls with it:

Is wage growth falling because men get better raises than women, but are becoming less and less of the workforce?

Inflation and the labor force in Japan

Noah Smith has called Japan the place where macro models go to die, and recently tweeted about high employment and low inflation creating a definite puzzle. However, in the dynamic information equilibrium model, Japan seems mostly like a normal economy just will really low equilibrium inflation (0.1% per year). 

In the data below, I removed the effect of the VAT increase in 2014, but not the 1997 one (I actually modeled it as a shock to the price level). The reason I did this is that the 1997 shock comes roughly when the CPI is at equilibrium (0.1% inflation) after the previous transition is ending, while the 2014 comes right in the middle of a transition/shock that follows the Great Recession.

I also left off the data from the 70s and 80s because it contains at best half a transition which makes it problematic for the parameter estimation algorithms especially when including multiple shocks. It's not that I wouldn't be able to find a fit, it's just that I'd have to carefully adjust intial conditions — a process that is tedious and wouldn't add any useful information besides saying inflation surged in the 70s. (Which we know; it's the recent lack of inflation that makes Japan a difficult case for macro models.)

Also note that I switch randomly between references to the labor force, labor force participation, and the employment rate. It should all be employment rate (per the data). But as Japan's unemployment rate is low and fairly constant, there isn't a lot of difference between the basic structure of these different measures of employment. I therefore elected to write this paragraph instead of editing and correcting the references in this post.

Here are the model results (click to embiggen):

The basic story is as follows:

• CPI surge follows the surge in women’s labor force participation in the 70s and 80s. This increase in the employment rate is cut off by early 90s recession.
• Early 1990s recession shows loss of employment for both men and women, and cuts off the inflation surge.
• The 1997/8 Asian financial crisis negatively impacts employment (for both men and women), but not inflation (the 97 blip is due to VAT tax changes).
• Early 2000s recession impacts both employment and inflation in a normal fashion.
• The Great Recession impacts inflation and men’s employment, but just cuts off the surge in women’s employment that began after the early 2000s recession. After the Great Recession, women’s employment rate begins to surge again (while men’s just grows at its equilibrium rate). The recent inflation "surge" (to just 1%) closely follows this rise.

This is all to say Japan has a pretty normal relationship between employment and inflation in terms of dynamic equilibrium (i.e. rates of change): when employment falls, so does inflation. The overall employment rate increased from the 70s until the 90s (inflation was higher), fell though the 90s (Japan's "lost decade" where inflation fell), and began to increase again in the 2000s (with a pause at the Great Recession). Inflation caught up a small amount, and it is possible it will continue to increase with the increasing size of the labor force (inflation lags labor force increases).

However, Japan does not have a normal relationship between unemployment and inflation in terms of numerical values. Equilibrium inflation is almost zero, so the labor force can increase at its equilibrium rate and inflation will be almost zero. A surge in the employment rate causes inflation to rise ... to 1%. But that 1% is 1% above equilibrium; it would be analogous to US inflation rising to 2.7% (PCE), 3.5% (CPI), or 3.4% (DEF). That would be major news in macro! But since we expect Japan to have 2% inflation for some reason (I guess the BoJ said it wanted 2% inflation at one time), we see 1% inflation as a "failure" of monetary and/or fiscal policy [1].

Japan seems to have low inflation because it has a very low employment rate dynamic equilibrium: it is about 1/3 the rate of the US (see graph below), so if US core inflation is 1.7%, we might expect Japan to have an equilibrium inflation rate of 0.6%. In that context, 0.1% isn't that far off from this back of the envelope estimate. This is probably directly related to low population growth. Australia, with its higher population growth, has a higher equilibrium inflation rate (deflator inflation of 2.8% compared to US of 2.4%).

...

Update 26 May 2018

Here's the picture with both employment rates from women and men in Japan at the same scale (click to embiggen):

...

Footnotes:

[1] If "Abenomics" is responsible for the increase in women in the workforce, then Abenomics "worked". If we just think of Abenomics as fiscal and monetary policy, those appear to have done nothing.

The price level in Germany

I fit the model for the CPI in Germany and came up with the same general story as most of the rest of the countries I've been looking at. The only thing I don't really understand is a sudden drop in inflation (deflation) the mid-80s. Otherwise most of the shocks are associated with recessions. Here is CPI, year-over-year inflation, as well as two zoomed-in graphs of the post Great Recession period. As always, click to embiggen.

One thing to note is that I used the US model code which ignores data after the forecast date (used to compare the latest data with forecasts, and usually shown in black). However, I technically had all the data so this is really more of a reserved subset out-of-sample forecast than a genuine forecast. Therefore I changed the color to dark blue.

This was going to be another post about the relationship between inflation and the labor force, but alas I was only able to find data organized by gender going back to 2005:

So instead, there will be another post on Japan and the relationship between the labor force and inflation.

Women in the workforce and investment

Sri Thiruvadanthai‏ questioned the "quantity theory of labor" model on Twitter, showing some relationships between the labor force and investment [1] with the latter being causal. However in my "Twitter talk" (also available as pdf from a link here), the general causal structure of the 60s-70s period is lead by impacts on women's participation in the labor force:

However, this did not look at investment; so I've added two measures (Gross Private Domestic Investment, as well as the (nominal) Capital Stock). Women entering the labor force (as well as the general increase in the labor force) also precede [2] the shocks to investment and the capital stock (click to expand):

I will put up a "seismograph" version when I get a chance.

...

Update 21 May 2018

Here it is (click to expand):

One modification I did make was to decrease the scale of the "Great Recession" shock in GPDI because it made the 70s expansion difficult to make out (low contrast). This should actually be telling; the size of the expansion in GPDI relative to its typical growth rate is small, making it one of the smallest shocks.

...

Footnotes:

[1] I was unable to figure out exactly which measure of investment he was using, though it was in a ratio with GDP. One issue with dividing measures that might have independent temporal structures is that it can produce a result with a much different temporal structure as an artifact:

Combined with using 10 year moving averages and e.g. 20 quarter changes, the exact timing and causal structure can get confusing. I tried to show this in some graphs where I show both the 20 quarter percent change compared to the instantaneous (continuously compounded rate of change) for the CLF total and for women:

[2] By precedes, I am using the "2-sigma" shock duration (middle 95%) as demarcation lines for the beginning and end. The "Great Recession" shock does peak in investment first. However, the shock to inflation (which is small) still lags the shock to the labor force (the former in 2013-2014 (PCE) or even 2015 (CPI), the latter in 2011):

Here's the Great Recession shock to investment preceding the shock to the labor force:

Update: That was the 1-sigma width above. However, the 2-sigma width does show the shock to CLF preceding the shock to investment. Incorporating uncertainty in the estimate of the width does not completely eliminate the possibility that it was a change in labor force participation that preceded the Great Recession (!)

Market & business cycle forecasts: update

Checking in on my forecasts of the S&P 500 and the 10-year interest rate (click to expand):

The 10 year rate has increased its deviation from the model, but the S&P 500 is tracking the forecast fairly well despite heading towards a deviation in early 2018.

Also, I shared this set of counterfactual recessions using the JOLTS job opening rate on Twitter. Each frame is a different assumption for the center of a possible recession between 2018.5 (~ July 2nd) and 2020 (December 31st) in steps of 0.1 year (36.524 days) because metric system is best:

A center of 2019.8 produces a shock with amplitude parameter a₀ = 1.4 ± 0.6 and width parameter b₀ = 0.9 ± 0.2 year. That's somewhat wider and larger than the 2008 recession (a₀ = 0.84 ± 0.01 and  b₀ = 0.37 ± 0.03 year), but largely consistent with it. A center of 2018.8 produces a smaller shock of comparable width (a₀ = 0.6 ± 0.1 and  b₀ = 0.8 ± 0.2 year). I chose a year + 0.8 because that puts us in October which has a history (actually exactly at October 19th which was the date of 1987's "Black Monday", close to 1929's "Black Tuesday", as well as around the time of the biggest losses of the 2008 recession). The silver lining of a 2018.8 recession would be potential amplification of a "blue wave" in the midterm elections. Such a recession would likely also send the interest rate data closer to the model as well.

The only signs of a recession (in the information equilibrium framework) are the abnormally high interest rates and the negative deviation in the job openings data. If those evaporate, then so does any evidence of a possible recession. There are other more traditional signs out there as well, such as yield curve inversion.

A list of macro meta-narratives

In my macro critique, I mentioned "meta-narratives" — what did I mean by that? Noah Smith has a nice concise description of one of them today in Bloomberg that helps illustrate what I mean: the wage-price spiral. The narrative of the 1960s and 70s was that the government fiscal and monetary policy started pushing unemployment below the "Non-Accelerating Inflation Rate of Unemployment" (NAIRU), causing inflation to explode. The meta-narrative is the wage-price spiral: unemployment that is "too low" causes wages to rise (because of scarce labor), which causes prices to rise (because of scarce goods for all the employed people to buy). In a sense, the meta-narrative is the mechanism behind specific stories (narratives). But given that these stories are often just-so stories, the "mechanism" behind them (despite often being mathematically precise) is frequently a one-off model that doesn't really deserve the moniker "mechanism". That's why I called it a "meta-narrative" (it's the generalization of a just-so story for a specific macro event).

Now just because I call them meta-narratives doesn't mean they are wrong. Eventually some meta-narratives become a true models. In a sense, the "non-equilibrium shock causality" (i.e macro seismograms) is a meta-narrative I've developed to capture the narrative of women entering the workforce and 70s inflation simultaneously with the lack of inflation today.

Below, I will give a (non-exhaustive) list of meta-narratives and example narratives that are instances of them. I will also list some problems with each of them. This is not to say these problems can't be overcome in some way (and usually are via additional just-so story elements). None have yielded a theory that describes macro observables with any degree of empirical accuracy, so that's a common problem I'll just state here at the top.

Macro meta-narratives

Meta-narrative: Wage-price spiral
Narrative: e.g. Exploding inflation in the 70s/"stagflation"
Problems: Doesn't seem to apply to today

Meta-narrative: Human decisions impacting macro observables
Narrative: e.g. Rational expectations and 70s inflation
Problems: Leads to theories that do worse than VARs

Meta-narrative: Monetary policy primacy
Narrative: e.g. Volcker disinflation
Problems: Monetary policy seems ineffective today

Meta-narrative: The Phillips curve
Narrative: e.g. Observed inflation/employment trade-off in the 50s and 60s
Problems: Flattening to the point of non-existence

Meta-narrative: Boom-bust cycles
(von Mises/Minksy investment/credit cycle, Fisher debt-deflation)
Narrative: e.g. The Great Depression, the Great Recession
Problems: Post hoc ergo propter hoc reasoning; recessions aren't cyclical making each investment boom a just-so story of a particular length and critical point ("Minsky moment")

Meta-narrative: Money as a relevant variable
Narrative: e.g. 70s inflation, Friedman-Schwartz account of the Great Depression
Problems: No specific measure of money makes sense of multiple periods of inflation or deflation; extrapolated willy-nilly from hyperinflation episodes to low inflation; lack of inflation with QE