news article background image
2019-12-06Macro `n Cheese

Resurrecting the Phillips Curve

This week, Roger was set to keep it simple. He found one half-done graph on his desktop and ended up with an entirely new way of looking at labor markets, one which brings the old Phillips Curve back to life.

Wage growth may not be too low after all

Recycling and environmental issues are big here in the Nordics, and this week, as I was cleaning up the (computer’s) desktop of some old docs, I found a Macrobond chart containing data from the SF FED:s “Wage Rigidity Meter” and the unemployment rate (UNR) and the employment to population ratio (EPOP). Unfortunately, I don’t remember why or when I created it, but when I opened the graph, it immediately struck a chord. Hence, this week it was actually my own trash that turned into a treasure!

What was it that I saw? – With some light editing, it looked like this:

Macrobond moment: At times I have come across really interesting data, and one thing that I really like about our content team is that they don’t mind adding it to the application (if there are no obvious restrictions) – If you find something exciting, let us know!

 

What struck me was how well the number of people staying in their job’s correlate to the Unemployment rate (UNR). And, in a corollary to recent developments in the Phillips relationship, how despite historically low UNR-levels, many still stay in jobs where they have experienced no wage growth at all!

In addition, it is also a bit surprising as I almost daily read articles, speeches and other commentaries alluding to how the strong US labor market is pushing people into (better) jobs. Could this be one of the missing pieces to the low wage growth / inflation puzzle? – People simply don’t switch jobs as they used to; or should do!

 

Putting the pieces together

I am sure we all recognize the critique against using the UNR as a measure of capacity constraints due to low labor force participation etc., so let’s first check if the correlation remains when adding the Employment to population ratio, EPOP (and let’s also look at a more relevant time period, after the great moderation):

Note: Conveniently, the chosen time period is also for which SF FED produce a coherent, uninterrupted dataset.

 

Undeniably, the correlation improves and the EPOP also seems to better align with a still-high share of workers remaining in zero wage growth jobs. And if the share of job-stayers with no wage growth is indeed an explanation to the generally low wage growth and inflation, a simple extrapolation indicates that, at current rates, it is not until the 2030-something that the share will have decreased sufficiently to give us “normal” wage growth and inflation.

This is, of course, jumping to conclusions, but at the very least, the chart above underlines that we might need to look beyond the traditional Phillips curve to understand why we are not seeing much of wage or ‘proper’ inflation. Indeed, one of the most troublesome aspects of how we traditionally look at this relationship is that we assume labor markets to be competitive and that it is ‘pricing frictions’ (i.e., if people only knew what to charge/pay there would be no unemployment) that creates unemployment. However, this simply does not jibe with the canonical model for unemployment where it is instead ‘search frictions’ (matching problems) that is the crux of the matter.

Hence, maybe it is competition for the employed, rather than the unemployed, that should be the guiding light for monetary policy. It is only once the employed workers have been more efficiently allocated that cost pressures build (as it is only then that employers will need to pay relatively more to persuade workers to move). In such a model, unemployment is not a good indicator for wage growth (inflation) and the unemployed can therefore – somewhat cynically – be referred to as “workers of last resort”.

 

If this holds true, we should be able to construct a better measure of labor market slack than the pure unemployment rate by considering the flow of labor from unemployment to employment (UE), but also the flow of labor from employment to employment (EE), as it indicates that workers believe (and are believed to) provide higher utility elsewhere – “Revealed preferences”. Only when slack – misallocation of labor – is sufficiently reduced, should wages (inflation) start to increase as employers need to outbid each other to attract resources.

Subsequently, by putting EE relative to UE, we should be able to construct a measure of the relative probability of acceptance which varies only because already employed workers get pickier (while the unemployed will take any job offer). Let’s call it the ‘search friction rate’ (SFR). When this measure rises, it is in an indication of slack, and when it falls, it is a measure of tightness. Hence, like the unemployment rate, it should be expected to vary inversely to wage growth (and inflation).

Note: I had to use a number of exotic research series  that aren’t in the Macrobond database, but I embedded them into the document (Source: Federal Reserve). It’s quite easy to upload your own data into Macrobond, in this example I have used the xls-function but you can also connect with API:s, SQL etc etc. As some of these series are updated from time to time, here is a link to an excelfile which you can use to paste and update the data into Macrobond yourself.

 

Validating the model (this is where it gets technical)

Now, if we replace UNR with the SFR in a standard Phillips Curve specification, we can compare the leading properties of UNR and SFR. First UNR, then SFR:

Macrobond moment: It is almost ridiculously easy to test different model specifications in Macrobond – the combination of formula language and neat built-in statistical tools facilitates economic research of all kinds.

 

From the graph above it should stand quite clear that trying to capture search frictions, like with our SFR-measure, dominates UNR in a Phillips curve specification almost regardless of horizon (the standard errors are similar for both SFR and UNR). SFR has a stronger, negative impact on subsequent inflation outcomes than UNR and, as you can try for yourself when you download the Macrobond documents, this result holds for core-CPI, PCE, core-PCE, and wages as well.

– In a recent paper by Galí & Gambetti (2018) variants of simple Phillips specifications with UNR and lagged inflation (to control for both explicit and implicit indexation), but with all econometric bells and whistles, are explored in full. The aim is to try to specify a stable Phillips relationship also over more recent time periods. However, the authors come up quite empty handed as can be seen here (ff.) and they conclude their work by alluding to some kind of structural shift of the Phillips relationship in conjunction with the financial crisis.

That said, and when trying to replicate their findings, the size of the UNR coefficient is admittedly quite stable whether you estimate with the full sample (going back to the 1960s) or start around 1995 (when our Search Friction Rate comes into play). It is indeed the global financial crisis that wreaks havoc to its statistical properties. And yet again, exploring measures of misallocation/mismatching, search frictions seem to provide a way forward:

 

Admittedly, the post-crisis coefficient for SFR can only be said to be stable in a relative sense. The SFR-coefficient doubles while the UNR-coefficient even switches sign (underlining the difficulties many are having with forecasting inflation). Hence, using search frictions might provide us with a more stable (and forward-looking) Phillips relationship, but it hardly provides a panacea to the troubles central banks and other forecasters are having with forecasting inflation after the global financial crisis.

 

That said, when looking at simple estimations – as in the graph above – using measures of search frictions instead of price frictions does fit the facts much better and highlights the strong possibility that the global financial crisis displaced many workers – both outside and inside the labor market. As long as the employed workers are looking for and receiving new, better matched, offers there is slack on labor markets. Only when this relocation is over can we hope to see stronger wage growth.

 

Lessons learnt

By picking up some riff-raff from the desktop, I thought I would have a head start when writing this week’s blog. I was wrong. This blog has been a real nightmare (I’m writing this close to midnight, before deadline), with a lot of data sourced outside of Macrobond etc etc. But, thankfully I have also learnt a couple of things:

 

  • There might be better ways to model the Phillips relationship than what we are used to;
  • Today’s simple exercise signals that labor market slack has been far bigger than what standard analytical tools have indicated;
  • Which not only underlines how careful we should be with the aggregated measures we macro economists tend to be fond of (output gaps etc);
  • It also adds to the considerable pile of arguments in favor of central banks currently erring on the soft side
  • But the fact that employers are beginning to pay-up to lure already employed workers is a good sign, which normally precedes more general wage increases.

 

Finally, and perhaps more importantly, I have learned that recycling is the right thing to do. Good night!

 

Get these posts delivered to your inbox

Sign up now to get an email every Friday with the latest research posts from Macro n Cheese

Sign up

 

Disclaimer

We don’t usually have views and opinions about economic and financial states of affairs, (not ones that we express publicly as a company, anyway). We do believe, however, that people can and do appreciate a variety of perspectives. What you’ve just read is the perspective of our resident chief economist. While we think he’s very smart, Macrobond Financial does not expressly endorse the views he presents here. And, as the old adage goes, you shouldn’t believe everything you read (not without finding the data, performing a few analyses and presenting it in a nice chart). We want to make it clear that we are not offering this information as investment advice. That being said, if you have the application you can easily check everything that’s mentioned here, and decide for yourself. If you don’t have the application, now you have a great reason to get it.