How reliable is the Atlanta sticky/flexi price framework?
And what does it predict about future inflation?
Last week the CPI numbers were released in the US. They weren’t great. It feels that markets are starting to consider the prospect of inflation continuing to rise.
One interesting feature of last week’s data release is the impact that it had on the Atlanta Fed’s sticky/flexi price CPI index. Here is the full series from the Atlanta Fed.
As we can see, until recently, inflation only seemed to be taking place in flexi price markets. These are mostly goods that have volatile prices and are priced on markets driven by supply and demand. Think: energy, used cars, food etc — all the stuff that took a big hit due to the lockdowns and the Russian invasion of Ukraine.
But in recent months, sticky prices have started to respond. These are the prices end consumer typically see at the checkout. These prices are typically ‘administered’ by corporations and set, not in line with supply and demand, but strategically in order to maximise sales and not alienate consumers. They are also the prices that spur consumers to demand wage hikes — and that can, as we noted before, trigger a dreaded wage-price spiral where inflation becomes entrenched.
In this post I’m going to do two things. First, I am going to look at the reliability of the Fed’s metric for forecasting purposes. Secondly, I’m going to use it to predict future sticky price inflation.
The easiest way to test if the metric is reliable is to look at the correlations between the sticky and flexi series. It is obvious eyeballing the chart that flexi price moves seem to predict sticky price moves with a lag. Is this a reliable relationship?
To see, we will look at the two key periods when the series would have been useful for forecasting: the beginning of the 1970s inflation between 1972 and 1975 and today’s inflation. Here are the correlations (r-squared) between the two series at various lags. Note that if the series is reliable for forecasting, the lag impact of the flexi price series on the sticky prices should be the same (or thereabouts) in both inflations.
This exercise has given us some interesting results. Up until mid-2021 the series looked highly unreliable. In the 1970s inflation, the flexi price series predicted the future sticky price series with increasing degrees of accuracy every month that passed. It peaked in its predictive power with a 9 month lag.
During the recent inflation, however, the flexi price series initially predicted the sticky price series. But as we moved through time and lagged the flexi price series (i.e. to use it for forecasting) it became less and less accurate for forecasting. This would typically indicate to me an overfitted series.
Around mid-2021, however, this changed. The correlations started to rise again. The flexi price series started to predict with greater and greater accuracy. Interestingly, it reached its peak forecasting efficacy at 9 months — the same as in the 1972-75 inflation!
I think that means that the flexi price series should reliably predict sticky prices moving forward — with a handy lag of 9 months. Okay, so what does this tell us in terms of where sticky prices are heading in the future?
Well, interestingly, sticky prices may be responding reliably to lagged flexi prices in recent months, but the magnitude by which they have been responding is much lower. Here are the slopes/beta for the full sample, the 1972-75 inflation and the recent sample.
We can generate forecasts using these slopes. I will leave the 1972-75 one to the side, as it is probably too extreme. Here are the forecasts for sticky prices using the recent slope and the full sample slope.
What does this tell us? A few things. Most strikingly that sticky prices have not been responding to flexi prices in the way that they typically have historically. This leads to the conclusion that if sticky prices at some point revert to responding to flexi prices the way they have historically we are in for some nasty inflation.
Why would they revert? Maybe after 15 years of deflation distributors and shops are trying much harder not to raise their prices in the face of rising flexi prices — but this could be a losing battle and at some point they might be forced to aggressively reprice their goods. At some point, administered prices have to respond to input costs unless firms want to take enormous hit to their profit margins.
But even if the new regime holds and flexi prices feed into sticky prices in a more benign fashion, we are still in for rising sticky prices in the next 9 months. This is already baked in. It has, in a sense, already happened. So, at the very least, sticky price growth — which averaged only 2.1% between 2010 and 2019 — is set to increase to around 5.4% in the coming nine months. We can say this with a high degree of confidence.
Now, perhaps flexi prices crash and these counterbalance the rise in sticky prices. I suppose this would be the ‘transitory inflation’ situation. But then the question is if the sticky prices start feeding on themselves via a wage-price spiral.
In summary, 2022 is set to be an inflationary year no matter what happens. There is also some risk that inflation could get really out of control. The two most likely scenarios for that development would be:
Sticky prices stop responding so tepidly to flexi prices, we see the correlation between the two revert to the historical trend and sticky prices explode.
Flexi prices continue to climb on the back of continued supply side disruptions (say, something happens around the Ukraine situation or the Shanghai lockdowns start to bite).