Recent data on used cars prices, and an upcoming shift in health insurance prices indicate some moderation in core CPI in the next few months. In this note, we explain the sources of this downward pressure, our model estimates, and we discuss how the Fed staff (and FOMC) might react.
Keep in mind
According to our models and estimates, all else equal, core (CPI) inflation could moderate from an average of 52bps (MoM, SA) in the last 6 months to 32bps in the next 4-6 months. The downward pressure is expected to come from used cars and health insurance prices, contributing roughly equally. Nevertheless, even considering these sources of disinflation, core prices are expected to growth at about 4% at annual rate, way higher than the tolerance level of the Fed.
For this reason, we do not expect the Fed to pivot now. At the same time, if the expected moderation will materialize, it is hard to imagine a more aggressive Fed than the September SEP dots plot going forward.
1. Used cars prices
Used cars prices are (finally) falling, both at the wholesale level (see Manheim used vehicle value index) and at the retail level (see used car price trends by CarGurus). Figure 1 shows the basic correlation between the Manheim used vehicle value index (the green line), and the CPI for used cars and trucks (the red line). Given the lagged passthrough from wholesale prices to consumers prices, the green line in Figure 1 is shifted forward by two months.
How much decline in the CPI used cars and trucks index (and when) should we expect?
We have set up a time series model to answer this question. Studying the passthrough from wholesale to CPI prices, our research shows that:
- The sum of the wholesale coefficients (in several models specifications) is around 0.6. This implies that when wholesale prices go up by 1 percent, CPI prices increase by about 0.6 percentage point.
- In most specifications, wholesale prices enter significantly up to the 3rd lag. This implies that it takes 3-4 months for the full passthrough to show up in the CPI.
- The fit of the models (R^2 up to 0.5) is remarkably good considering the simplicity of the specifications and the fact that the regressions are run on monthly (MoM) data. Figure 2 shows the left-hand-side variable (the red line), and the fitted values of our favorite specification.
Given the above, our models suggest that in September the CPI for used cars should drop about 1 percent. The models also suggest that, conditional on wholesale prices continuing their decline at the current rate, the CPI for used cars and truck could fall about 2% (MoM, SA) in the next 3-4 months.
Considering that the CPI for used cars increased at an average of -0.2% in the last 6 month and given its weight (4.2% in total CPI, 5.5% in core CPI), our models suggest that the current decline could subtract about 0.1 percentage point from the core CPI MoM in the next few months. In other words, all else equal, core CPI would move from an average of 52bps (MoM) to an average of 42bps (MoM).
Can the model be off for 1 month (or 2)?
Yes. As mentioned, the econometric analysis suggests that the passthrough is pretty slow (up to 4 months). Therefore, it can happen that the CPI does not react for 1 month or 2 after wholesale prices start moving. Having said so, as mentioned above we are already observing downward pressure not only at the wholesale level but also at the retail level. Therefore, we should (finally) see some relief in the CPI soon.
Figure 1. CPI used cars and trucks (index) and Manheim used vehicle value (index)
Note: the figure shows the level of the CPI used cars and trucks (red line) and the level of the Manheim used vehicle value index (green line). The green line is shifted 2 months forward.
Figure 2. CPI used cars and trucks (MoM, %) and model fitted values
Note: the figure shows the MoM (SA, %) of the CPI used cars and trucks (red line) and the fitted values of our preferred model (blue dashed line). The latest observation of the blue line (which is the model forecast) is September 2022.
2. Health insurance
Health insurance prices will put some downward pressure on consumer prices starting from the October CPI release. The reason has little to do with lower premia (in fact, premia will not be lower) but with a measurement issue.
How does the BLS measure health insurance prices in the CPI?
Even though insurance premia are an important part of consumers’ medical spending, the CPI does not directly price health insurance policies (see here for technical details). In a direct approach, the BLS would track the movement of insurance premia, holding constant the quality of insurance, and use these prices to build the Health Insurance index. However, the BLS has been unable to consistently control for changes in quality such as policy benefits and risk factors. Price change between health plans of varying quality cannot be compared, and any quality adjustment methods to facilitate price comparison would be difficult and subjective. As a result, the BLS developed an indirect approach called the “retained earnings method” (where “retained earnings” are defined as leftover premiums income after paying out benefits and rebates).
Without entering into the details, the general idea is that when the retained earnings, or profit margins, of health insurance companies are going up, that maps to a higher health insurance component in the CPI. When those retained earnings are falling, that works out to a lower CPI.
The data used to construct the CPI index for health insurance come from the National Association of Insurance Commissioners (NAIC) and they are released once a year, around October.
What happened during Covid and what should we expect now?
In 2020 utilization of healthcare services was low because people preferred postponing elective treatments. As a result, retained earnings increased substantially in 2020 and this was reflected in the CPI in the October 2021 release. Figure 3 shows the MoM of the CPI of health insurance. The yellow mark indicates the October 2021 data when health insurance jumped to 2% (from -0.95% the previous month) reflecting the 2020 drop in utilization.
In 2021 we got the opposite: utilization went up and retained earnings dropped. This effect will show up in the October 2022 CPI report. According to some estimates, the MoM of the CPI health insurance is expected to move from an average of +2% in the last 12 months to about -2.5% (the latest bar in Figure 3 shows our forecast for the October 2022 reading).
What is the implied effect on core CPI?
Health insurance has a weight of about 1% in the CPI (or 1.3% in core CPI). Therefore, given the expected drop in the MoM, health insurance could subtract up to 0.1 percentage point from the core CPI MoM going forward.
(Note: this effect is expected *only* in CPI space, as the BEA uses PPIs to construct the PCE medical services index)
Figure 3. CPI health insurance (MoM, %)
Note: the figure shows the MoM (SA, %) of the CPI health insurance. The yellow mark shows the October 2021 reading. The latest two observations are our forecast for September and October 2022, respectively.
Implications for the Fed staff
Our analysis shows that core CPI prices could moderate somewhat in the next few months. Nevertheless, the MoM of core CPI is expected to remain well above target even considering the drop in used cars and health insurance prices.
The Fed staff (and the FOMC) are both expecting core inflation to moderate in 2023 (latest SEP has core PCE to 3.1% in 2023). If our analysis is correct, the expected MoM of core inflation would be roughly consistent (although a bit above) the latest SEP and our “main” model forecast.
For this reason, in our view, the Fed cannot pivot in the near future. At the same time, conditional on our forecast being right, the most likely outcome for the FF rate going forward remains the September SEP dots plot. A real pivot will require much more than some idiosyncratic shocks and a genuine movement of the price changes distribution (which, at the moment, have shown no signs of disinflation).