Preamble: we have seen the UK CPI print this morning and we are not surprised. (for the record: the plan is to cover the UK, Canada, and some emerging markets going forward). What happened this morning in the UK is what we anticipated when talking about Spain (see here). The MoM saar of UK core CPI is now showing a “monster” reading for April (12%-18% depending on the filter). This is really unsurprising. The issue is that most series have now lost their pre-Covid seasonal pattern (the reader might want to check the NSA level). Therefore, the MoM saar can show large swings/revisions, given how filters work. Not only, but the YoY is not necessarily unbiased and the fact that it ticks up/down in a given month means little in the current environment. In our view, “consensus” is getting the forecast so wrong precisely because most forecasters are assuming that the series has a seasonal pattern at least similar to pre-Covid. But again, this is no longer the case. Finally, as a reminder: the same thing can happen with core HICP from now until (at least) September.
Some measures of inflation expectations have recently turned and point higher. We explain why, at the moment, the issue is less severe than it looks like for the Fed. Nevertheless, in our view and estimates it should prevent the Fed staff and the FOMC from turning dovish any time soon. The surge in long-term expectations in the survey of consumers of the University of Michigan in our experience has certainly caught the eyes of the Fed staff and FOMC members. We caution against a straight read for three reasons: (i), large movements in the preliminary reading are often reversed in the final reading, (ii) our nowcast of the Fed CIE (the index tracking several measures of expectations) points to a marginal decrease this quarter, and (iii) measures of trend inflation are only marginally affected.
Overall, we are getting conflicting signals in this moment. In our view (which we expect to be in line with the approach of the Fed staff) the best thing to do is to take time right now.
(For the record, last February we circulated a note titled “US: Expectations are Dropping, So What?” in which we suggested to take no signal from the back-then decline in expectations. Ex-post, it turned out to be the right thing. This time we argue the same, although for opposite reasons).
The evidence
Long-term expectations in the University of Michigan survey of consumers are at the highest level since 2011. Figure 1 shows the long-term inflation expectations from the Michigan survey. The preliminary reading in May came in at 3.2%, the highest reading since 2011. Not only, but the standard deviation (Figure 2, latest point March) indicates a high degree of dispersion with risks that remain skewed to the upside. In our experience, when the preliminary reading shows a large jump as it did in May, there is a high probability that part of the increase will be reversed in the final reading. In any case, in our experience the behavior of long-term expectations at this point is a headache for the Fed staff (and the FOMC). For the reasons we explain in the next section, a case can still be made not to panic. But at the same time, it certainly cannot be dismissed (from which our call that the Fed staff and the FOMC will not turn dovish any time soon).
Figure 1. Inflation expectations in the Michigan survey.
Note: horizontal lines indicates the 70% bands of the historical distribution.
Figure 2. Standard deviation of the long-term inflation expectations (Michigan).
How the drop in expectations will affect the Fed staff models and forecast
In our experience, the movement in long-term expectations will affect some of the staff models. It will also affect the judgmental forecast, although only indirectly. We review each step below.
The Fed CIE
The first model affected by the movement in expectations is the so-called “Fed-CIE” or “Common Inflation Expectations” model. We have updated our mixed-frequency Dynamic Factor Model (DFM) to nowcast the Fed-CIE, given the new readings in expectations. Our results are presented in Figure 3.
Our model estimates that the Fed-CIE has ticked down in 2023:Q2. According to our model, the level of the Fed CIE is estimated at 2.18 (C.I. 2.21-2.15) in Q2. This level of the Fed CIE is consistent with a reading of core PCE price inflation of about 2¼ percent.
Why is the CIE expected to tick down if Michigan ticked up? Answer: because the long-term expectations of the Michigan survey is only one of the series that feed into the model. Other series (some SPFs, some breakevens) have ticked down in Q2, at least so far. The net effect of the two is in favor of the latter and we currently expect the CIE to show a small drop in the current quarter.
Figure 3. Published Fed-CIE and our model-based nowcast
Note: the Figure shows the published Fed Common Inflation Expectations (CIE) index (the thick blue bar), and our Dynamic Factor Model (DFM) nowcasts (the black circles). Both series are displayed at quarterly frequency. The latest published quarter of the Fed-CIE is 2023:Q1. The latest circle in the Figure shows our nowcast for 2023:Q2. The dashed-blue lines are estimated confidence intervals. The Fed-CIE model is a DFM that includes a large number of measures of inflation expectations. The estimated common factor (which is specified as an autoregressive process of order 5) is projected on a chosen measure of long-term inflation expectations (the SPF PCE 10 years). By construction, the level of the common factor cannot be interpreted because the underlying series are normalized at the beginning of estimation. Instead, the level of the common factor can be compared to its own history.
The Trend models
The second set of models affected by the movement in expectations are the trend models. We have updated our set of trend inflation models which mimic Fed staff Rudd (2020). (As a reminder, some of the models use expectations as a proxy for the trend itself. What the econometric estimation does, one way or the other, is to provide a projection of a particular measure of expectation in core PCE price inflation space). Figure 4 shows the latest run of the models, including 2023:Q2 in sample. According to the latest run, the average across models is slightly higher in 2023:Q2 than the previous quarter, driven by an upward revision of the Phillips curve trend model which uses Michigan (long-term) and by a Time Varying Parameter VAR model.
Figure 4. Trend inflation models.
Note: the chart shows the estimated trend inflation from 11 econometric models. The models are split into three groups. The first group is a collection of Phillips-curve (PC) type of trend inflation models in which a measure of long-term inflation expectations is used as a proxy of trend inflation. The second group is a collection of state-space unobserved component models in which we have modelled trend inflation either as a smooth trend or as an augmented local level. Finally, the third group of models is a collection of Time Varying Parameters Vector AutoRegressive models (TVP-VAR) with different endogenous variables. This set of models follows the FEDS Note by Rudd (2020) “Underlying Inflation: Its Measurement and Significance”. The Fed staff assumption about the level of underlying inflation (set at 1.8 percent until end of 2021) is inferred from Laubach et al. (2014) “Long-term Inflation Expectations and Risks to the Inflation Outlook“. The Fed staff has revised its assumption of pi* in December 2021, as per the FOMC minutes.
Ok, but then… what? Ultimately, it’s about pi*.
In our view, the answer to the question “ok, but then… what?” is “not much for now” because the upward movement in Michigan is offset by other measures (so far). In our experience, each round the Fed staff has to make a call about pi* (or “underlying inflation”), the crucial variable for the dynamics of core PCE in the medium-term. The literature suggests that the estimate of pi* should rely on three pillars: (i) measures of inflation expectations, (ii) measures of trend inflation, and (iii) the general equilibrium of the economy considering the interaction between pi* and U*. Figure 5 shows the updated slide from our Pre-FOMC Meeting Package in which we summarize the signals from the three pillars, together with the Fed staff estimate (again, as inferred by reverse engineering the forecast from the minutes). In our estimate, the Fed staff is estimating pi* at around 2½ percent, below the average across the three pillars.
Figure 5. Updated slide #24 from our “Pre-FOMC Meeting Package”
Conclusion
For now, you can forget about expectations. Right now, we are getting conflicting signals about the economy. Inflation and expectations are no exception. We have shown that on the one hand the Fed staff has reasons to be worried. But on the other side the increase in long-term consumers’ expectations is mitigated by the drop in other measures. The net effect of the two is close to zero. In this circumstances, we do not expect the Fed staff to revise pi*. But the staff cannot minimize the upward movement of Michigan. If we put everything together, the best strategy seems to take time. In any case, there are good reasons to exclude that the staff (and the FOMC) can turn dovish any time soon.