The chance of a Fed pivot in March is low, but it is higher at the June meeting. One of the reasons is that the YoY of CPI rents is expected to be on a downward trajectory only by then. The Cleveland Fed has published a paper titled “Disentangling Rent Index Differences: Data, Methods, and Scope” by Adams et al (2022). The paper has received a lot of attention. We discuss why we disagree with some of the authors’ claims, why we agree with them that the YoY of CPI rents is expected to peak in 2023:Q2, and what is still unclear in the data.
What the paper does
The authors have created rents indexes using the same source of the BLS CPI rents, the so-called BLS rent microdata. They show that the discrepancy between published statistics and their own indexes is almost entirely explained by differences in rent growth for new tenants relative to the average rent growth for all tenants. The authors also show that the new tenants rents index has peaked in 2022:Q2, similar to other private rents measures. Finally, the authors show that the new tenants and average tenants rents indexes lead the CPI rents index by 4 quarters and 1 quarter, respectively.
The paper is interesting because if the authors’ claims are correct, the behavior of the new tenants and (especially) the average rent growth in their dataset are informative of the future behavior of the published CPI rents.
Our research using the authors’ dataset
Our first result is that the average tenants’ rents index does not lead the CPI rents index by 1 quarter but it is coincident. Figure 1 shows the YoY of the average tenants’ rents index (“ATRR” yellow line) and the CPI rents. For the latter, Figure 1 plots the YoY using monthly data, that is the YoY in the last month of each quarter (as opposed to the YoY calculated on quarterly data as the authors do). In the Figure, the red dot indicates the peak of the YoY, and a green dot indicates the bottom. The main takeaway is that the “lead” of ATRR disappears. While a hard conclusion would require getting access to the monthly dataset of the authors (not available), the suspect is that what the authors claim is nothing but the difference between the YoY of monthly data and the YoY of quarterly data. By construction, the former anticipates the latter (for the record, in the last few turning points, the YoY of monthly CPI rents anticipated the quarterly YoY by about 3 months, precisely one quarter). Conditional on this result, we are not sure how the authors’ ATRR can be more useful than looking at the YoY of published rents using monthly data. For completeness, we have also tested the hypothesis that ATRR leads CPI rents using a Granger causality test. The result is that, no matter how the YoY is calculated, both series Granger cause the other. In other words, there is no strong evidence that ATRR leads CPI rents and the relationship appears, again, to be coincident.
Figure 1. YoY of CPI rents of primary residencies and authors’ ATRR
Our second result is that a much simpler approach yields the same conclusion: for any practical matter, we suggest to look at the 3m/3m of published CPI rents. Figure 2 shows the metrics (3m/3m ar, 6m/6m ar, the YoY, and the 2Yo2Y) of published CPI rents. The evidence is unsurprising: the 3m/3m leads the YoY by about 3/4 months (for instance, in 2021 the 3m/3m bottomed in January and the YoY in April). Put it differently, there is no need for the ATRR because the 3m/3m of BLS rents has the same (or even better) properties. In this sense, if one is interested in forecasting the “peak” of the YoY of BLS rents, it seems enough to look at the 3m/3m ar. Should the next MoM (ar) be below the latest 3m/3m (9.6% ar) as we suspect, then the “peak” of the 3m/3m will be in. Consequently, the YoY should peak in 2023:Q2. This result is in line with the authors’ claim that the new tenants rents index (“NTRR”), not discussed here for brevity, has peaked in 2022:Q2 and leads the CPI rents by 4 quarters. It is also in line with Phillips curve model-based forecasts that we have run.
Figure 2. CPI rents of primary residencies – metrics
Our final result is that the authors’ dataset does not clarify why the MoM of CPI rents is still so elevated. Nevertheless, we estimate that the level of NTRR is within 3% of ATRR, confirming that moderation in CPI rents should be coming soon. The dataset of the paper contains only the YoY of the new tenants (NTRR) and average tenants (ATRR) rents indexes. Because the NTRR and ATRR levels can be informative about the current readings of CPI rents, we have reverse engineered the NTRR and ATRR levels (technically, in order to reverse engineer the levels of the series we had to make an assumption about the first 3 quarters. The safest option is to impute the observed CPI rents growth rates in those quarters, assuming that NTRR and ATRR were growing at the same rate in a period of rents growth stability).
Figure 3 shows our estimates of the NTRR and ATRR levels. As expected, the level of both NTRR and ATRR takes off in mid-2021. Consistent with the author’s results, the level of NTRR drops in the most recent quarter, while the level of the ATRR continues to growth. This brings one good news and one puzzle. The good news is that, no matter how we tried to construct the levels, the gap between NTRR and ATRR at the end of the sample is around 3% (and closing). This suggests that the MoM of the BLS rents should start moderating soon, as the marginal increase should already be below the average. However, the recent MoM readings of the BLS rents are way above what it is implied by the NTRR and ATRR levels, which instead seem to suggest a much lower reading in the current quarter. Unfortunately, without access to the raw series, the puzzle will continue.
Figure 3. Estimated levels of NTRR and ATRR – index 2019Q4=100
Implications for the Fed
The implication for the Fed is, in our view, straightforward: the Fed will be very careful, and a real pivot will require a significant drop in the 3m/3m (ar) of BLS CPI rents. In our experience, the paper has been heavily scrutinized by the Fed staff and discussed in meetings. The issue, as shown, is that the contribution does not seem to add much to the existing information set. The Fed knows that marginal rents have already moderated. But the timing (and magnitude) of the passthrough to CPI rents remain vague. For most practical matters, the easiest approach (look at the 3m/3m of CPI rents) continues to be timely and uncontroversial. The Fed staff is probably relying on a similar approach.