In this note we discuss the near-term and medium-term implications of yesterday’s CPI.
Keep in mind
There are now some promising signs of disinflation but shelter continues to surprise to the upside masking those signs. Based on this, we do expect improvements but only starting in 2023:Q2. Our models continue to signal significant upside risks in the medium-term, especially if we include our 2022:Q4 nowcast in the sample. For this reason, we expect a marginally more hawkish Fed from now till the end of the year.
What was the message of the CPI?
Our models and analysis suggest that several (in fact, the majority of) items in core CPI have stabilized:
- In September, the left tail of the MoM changes distribution was thicker than the previous month, despite the aggregate weighted mean (0.6% MoM) was the same.
- The distribution of the last 3 months is similar (in fact, even more benign because it has a thicker left tail) that the distribution of 3-6 months ago and 6-9 months ago.
- The estimated common component across items has dropped and it is now comparable to a strong month in the pre Covid period.
Another, more direct way of seeing the above points is to look at Figure 1 below which shows the 3m/3m (a.r.), the 6m/6m (a.r.), the YoY, and the 2Yo2Y (a.r.) of core CPI excluding shelter. In September, the 3m/3m (a.r.) was 5.6%, the lowest level since November 2021.
Put it differently, net of shelter, there are now visible signs not only of stabilization but also disinflation.
Figure 1. CPI excluding food, energy, and shelter items
Note: the figure shows several metrics of the CPI excluding food, energy, and shelter items. The red line shows the 2 year over 2 year changes at annual rate. The thick blue line shows the year-over-year at annual rate. the light blue line shows the six months over six months changes at annual rate. The light dashed blue line shows the 3 months over 3 months changes at annual rate.
Why was core CPI strong? (0.6% MoM)
The answer is simple and unpleasant: because of rents/OER and their weight. Figure 2 shows the CPI of rent of primary residency. In September, the 3m/3m (a.r.) reached 9.2%, the highest value since November 1981. Not only but in the last 2-3 months rents and OER have been much stronger than predicted by the models (see Dolmas and Zhou (2022)), hinting that the strong readings will continue in the near-term.
Figure 2. CPI of rent of primary residency
Note: the figure shows several metrics of the CPI of rent of primary residency. The red line shows the 2 year over 2 year changes at annual rate. The thick blue line shows the year-over-year at annual rate. the light blue line shows the six months over six months changes at annual rate. The light dashed blue line shows the 3 months over 3 months changes at annual rate.
Will the Fed (staff) and the FOMC care about the details just discussed?
The Fed staff monitors a long list of indicators; therefore, we are sure they are well aware of the details. But in this environment, in our view and experience, the only possible message remains the same “inflation is too high”. The details will matter at some point in the future, but not now.
What are the implications for the near-term and medium-term forecast?
The near-term forecast
Given the strength of shelter, the last two reports imply revising the near-term forecast (next 3 months) to an average of about 40bps (MoM, SA) in core CPI space and 35bps (MoM, SA) in core PCE space. In our view, the Fed staff has also taken signal from the incoming data and revised its near-term forecast accordingly.
When will core CPI finally show significant signs of disinflation?
Given the evidence in Figure 1 and Figure 2, the answer to this question is determined by the disinflation of shelter. Unfortunately, model-based evidence suggests that shelter inflation will remain very elevated until 2023:Q1. Therefore, at the moment it is hard to see core CPI averaging 30bps before 2023:Q2.
How did the “main” model react to the incoming data?
This is the most worrying part of the story. The “main” model expected softer readings in Q3. Because the starting point of the model is now higher, the model delivers a forecast which is about 3 tenths higher at the end of the medium-term compared to the forecast at the time of the September FOMC. Figure 3 shows the “main” model forecast stopping the estimation in 2022:Q3, conditioning on the path of the FF rate and the unemployment rate of the September SEP. The main takeaway remains the same: without some help from the supply side (or a more aggressive monetary policy), inflation does not converge back to target in 2025.
The most problematic evidence, however, is implied by the quarterly arithmetic of Q4. Because August and September enter into the Q4 (QoQ) growth rate calculation, the strong readings of the last two months not only pushed up significantly the Q3 core PCE figure but also the Q4 nowcast.
(Note: yes, August and September do influence the Q4 (QoQ) figure because the quarterly numbers are chained).
Assuming that core PCE prices will grow at an average of 35bps in Oct-Nov-Dec, the Q4 figure is 5.0% (a.r.) which is much higher than the model forecast for Q4. For this reason, we have run the model including the Q4 5.0% nowcast in the sample. The results are shown in Figure 4. The model revises the forecast at the end of the sample by almost 1 percentage point. The reason for the upward revision is not only the higher starting point but most importantly the much higher estimated persistency (the coefficient of pi* drops close to zero, implying that the model becomes almost unanchored (!)).
As usual, we never take too much signal from a single CPI print and the model forecast can change in the future, should we get some moderation as discussed in the first part of the note. In any case, for the time being the message remains the same: conditional on the latest SEP path of the FF rate (and the unemployment rate), the risks around the FOMC forecast are skewed to the upside.
Figure 3. “Main” model forecast – end of sample 2022:Q3
Note: the confidence intervals (C.I.) are estimated using quasi-out-of-sample methods (estimate the model over a sub-sample, forecast, and calculate the root mean squared forecast errors).
First quarter of forecast: 2022:Q4.
Figure 4. “Main” model forecast – end of sample 2022:Q4
Note: the confidence intervals (C.I.) are estimated using quasi-out-of-sample methods (estimate the model over a sub-sample, forecast, and calculate the root mean squared forecast errors).
First quarter of forecast: 2023:Q1.
Implications for the Fed staff
The Fed only cares about the medium-term (forecast). Therefore, we do not expect neither the Fed staff nor the FOMC to take any signal from the deceleration in core prices net of shelter. Rather, we expect the Fed to signal a higher terminal rate (north of 5%) with no cut in 2023. Having said so, as mentioned yesterday, in our view, it is unclear what a more aggressive monetary policy can achieve now, given that most of its effects are still in the pipeline.