Whether inflation is short-lived or persistent, concentrated in a few sectors or broad-based, is of deep relevance to policymakers. We estimate a dynamic factor model on monthly data for the major sectors of the personal consumption expenditures (PCE) price index to assess the extent of inflation persistence and its broadness. The results give a measure of trend inflation and shed light on whether inflation dynamics are dominated by a trend common across sectors or are sector-specific.
Multivariate Core Trend (MCT) inflation was 2.9 percent in September, a 0.3 percentage point increase from August (which was revised up from 2.5 percent). The 68 percent probability band is (2.4, 3.3).
Services ex-housing accounted for 0.54 percentage point (ppt) of the increase in the MCT estimate relative to its pre-pandemic average, while housing accounted for 0.50 ppt. Core goods had the smallest contribution, 0.03 ppt.
A large part of the persistence in housing and services ex-housing is explained by the sector-specific component of the trend.
What is the goal of the Multivariate Core Trend (MCT) analysis?
The New York Fed aims to provide a measure of inflation’s trend, or “persistence,” and identify where the persistence is coming from.
What data are reported?
The New York Fed’s interactive charts report monthly MCT estimates from 1960 to the present. The New York Fed also provides estimates of how much three broad sectors (core goods, core services excluding housing, and housing) are contributing to overall trend inflation over the same time span. The New York Fed further distinguishes whether the persistence owes to common or sector-specific components. Data are available for download.
A dynamic factor model with time-varying parameters is estimated on monthly data for the seventeen major sectors of the PCE price index. The model decomposes each sector’s inflation as the sum of a common trend, a sector-specific trend, a common transitory shock, and a sector-specific transitory shock. The trend in PCEinflation is constructed as the sum of the common and the sector-specific trends weighted by the expenditure shares.
How does the MCT measure differ from the core personal consumption expenditures (PCE) inflation measure?
The core inflation measure simply removes the volatile food and energy components. The MCT model seeks to further remove the transitory variation from the core sectoralinflation rates. This has been key in understanding inflation developments in recent years because, during the pandemic, many core sectors (motor vehicles and furniture, for example) were hit by unusually large transitory shocks. An ideal measure of inflation persistence should filter those out.
PCE data are subject to revision by the Bureau of Economic Analysis (BEA). How does that affect MCT estimates?
Historical estimates in our MCT data series back to 1960 are based on the latest vintage of data available and incorporate all prior revisions.
How does the MCT Inflation measure relate to other inflation measures?
The MCT model adds to the set of tools that aim at measuring the persistent component of PCE price inflation. Some approaches, such as the Cleveland Fed’s Median PCE and the Dallas Fed’s Trimmed Mean, rely on the cross-sectional distribution of price changes in each period. Other approaches, such as the New York Fed’s Underlying Inflation Gauge (UIG), rely on frequency-domain time series smoothing methods. The MCT approach shares some features with them, namely: exploiting the cross-sectional distribution of price changes and using time series smoothing techniques. But the MCT model also has some unique features that are relevant to inflation data. For example, it allows for outliers and for the noisiness of the data and for the relation with the common component to change over time.
How useful can MCT data be for policymakers?
The MCT model provides a timely measure of inflationary pressure and provides insights on how much price changes comove across sectors.
GDPNow is not an official forecast of the Atlanta Fed. Rather, it is best viewed as a running estimate of real GDP growth based on available economic data for the current measured quarter. There are no subjective adjustments made to GDPNow—the estimate is based solely on the mathematical results of the model. In particular, it does not capture the impact of COVID-19 and social mobility beyond their impact on GDP source data and relevant economic reports that have already been released. It does not anticipate their impact on forthcoming economic reports beyond the standard internal dynamics of the model.
“We’ve had a time of red-hot housing market all over the country… Shelter inflation is going to remain high for some time. We’re looking for it to come down, but it’s not exactly clear when that will happen. Hope for the best, plan for the worst.”
The rapid run-up of shelter costs—both house prices and rents—during the recovery from the pandemic has raised questions about how inflation pressures might affect housing affordability. Since March 2022, the Federal Reserve has rapidly lifted its federal funds rate target from near zero to over 4%, and policymakers have signaled that they are open to keeping the monetary policy stance sufficiently restrictive to return inflation to the longer-run goal of 2% on average. The tightened financial conditions following those policy changes, especially the surge in mortgageinterest rates, have helped cool house price growth. However, rentinflation remains elevated.
This Economic Letter examines the effectiveness of monetary policy tightening for reducing rentinflation. We estimate that, during the period from 1988 to 2019, a policy tightening equivalent to a 1 percentage point increase in the federal funds rate can reduce rentinflation—measured by 12-month percentage changes in the personal consumption expenditures (PCE)housing price index—by about 3.2 percentage points, but the full impact takes about 2½ years to materialize. Based on housing costs’ share in total PCE, this translates to a reduction in headline PCEinflation of about 0.5 percentage point over the same time horizon.
Rentinflation also accelerated during the pandemic period. Figure 1 shows that rentinflation—measured using 12-month changes in the PCEhousing price index and including rents for tenant-occupied housing and imputed rents for owner-occupied housing—rose from a low point of about 2% in early 2021 to 7.7% by December 2022, the highest level since 1986. During the same period, rentinflation measured by 12-month changes in the shelter component of the consumer price index (CPI) experienced a similar increase. Thus, following the tightening of monetary policy, house price growth has slowed but rentinflation continues to rise.
Economic theory suggests that some common forces such as changes in housing demand can drive both rents and house prices. For example, the expansion of remote work since the COVID-19 pandemic has increased demand for housing, raising both house prices and rents (Kmetz, Mondragon, and Wieland 2022). To the extent that the stream of current and future rents reflects the fundamental value of a house, house prices can be a leading indicator of future rentinflation (Lansing, Oliveira, and Shapiro 2022). Thus, monetary policy can affect both house prices and rents by cooling housing demand.
Housing demand responds to changes in financial conditions, such as increases in mortgageinterest rates. However, theory suggests that house prices are more sensitive than rental prices to changes in financial conditions, because home purchases typically require longer-term mortgage financing. In addition, unlike rents, house prices can be partly driven by investor sentiments or beliefs, which explains the observed larger swings in house prices than in rents over business cycles (Dong et al. 2022). Long-term rental contracts can also contribute to slow adjustments in rentinflation.
Rentinflation is an important contributor to overall inflation because housing costs are an important component of total consumption expenditures. On average, housing expenditures represent about 15% of total PCE and 25% of the services component of PCE. In CPI, shelter costs represent an even larger share, accounting for about 30% of total consumption of all urban consumers and about 40% of core consumption expenditures excluding volatile food and energy components.
The contribution of rentinflation to overall PCEinflation has increased since early 2021. As Figure 2 shows, in the first quarter of 2021, rentinflation accounted for about 22% of the four-quarter change in the PCE services price index, excluding energy: 0.5 of the 2.3 percentage points increase in service prices was attributable to rentinflation. By the third quarter of 2022, the contribution of rentinflation had climbed to about one-third, or 1.5 of the 4.7 percentage point increase in service prices.
Figure 2: Rising contribution of rent inflation to services inflation
For our analysis, we use a measure of monetary policy surprises constructed by Bauer and Swanson (2022). Their measure focuses on high-frequency changes in financial marketindicators within a short period surrounding the Federal Open Market Committee (FOMC) policy announcements. If the public fully anticipates a policy change, then the financial market would not react to new policy announcements. However, if the market does react to an announcement, then the policy change must contain a surprise element. Thus, changes in financial marketindicators, such as the price of Eurodollar futures, in a narrow window around an FOMC announcement can capture policy surprises. In practice, however, the data constructed this way are not complete surprises because they can be predicted by some macro and financial variables shortly before FOMC announcements. We follow the approach of Bauer and Swanson (2022) to purge the influences of those macro and financial variables from the measure of policy surprises. We use the resulting quarterly time series to measure monetary policy shocks, with a sample period from 1988 to 2019.
We then use a local projections model—a statistical tool proposed by Jordà (2005)—to project how rentinflation responds over time to a tightening of monetary policy equivalent to a 1 percentage point unanticipated increase in the federal funds rate in a given quarter. The model takes into account how monetary policy shocks interact with other macroeconomic variables, including lags of rentinflation, real GDP growth, and core PCEinflation.
In the final step, we compute the responses of rentinflation relative to its preshock level over a period up to 20 quarters after the initial increase in the federal funds rate.
Gradual impact of policy tightening on rent inflation
Figure 3 shows the response of rentinflation during the first 20 quarters after an unanticipated tightening of monetary policy (solid blue line). The shaded area shows the confidence band, indicating the statistical uncertainty in estimating the responses. Under the assumption that the model is correct, the shaded area contains the actual value of the rentinflation responses to the monetary policy shock roughly two-thirds of the time. The policy shock is normalized such that it is equivalent to a 1 percentage point unanticipated increase in the federal funds rate.
Figure 3: Response of rent inflation to monetary policy tightening
Source: Bureau of Economic Analysis, Bauer and Swanson (2022), and authors’ calculations. Note: Response of rentinflation to a monetary policy shock equivalent to a 1 percentage point surprise increase in the federal funds rate. Shaded region shows 68% confidence band around the estimate.
The figure shows that monetary policy tightening has significant and gradual effects on rentinflation. On impact, a 1 percentage point increase in the federal funds rate reduces rentinflation about 0.6 percentage point relative to its preshock level. Over time, rentinflation declines gradually, falling about 3.2 percentage points in the 10 quarters following the impact. The slow adjustment in rentinflation partly reflects the stickiness in nominal rents due to long-term rental contracts. Since housing expenditures account for about 15% of total PCE, this estimate translates to a reduction in headline PCEinflation of about 0.5 percentage point, stemming from the decline in rentinflation over a period of 2½ years.
The rent component of PCE is measured based on average rents, including those locked in long-term rental contracts, which are slow to adjust to changes in economic and financial conditions. Rents on new leases, however, are more flexible. For example, the 12-month growth in Zillow’s observed rent index, which measures changes in asking rents on new leases, has slowed significantly since March 2022 (see Figure 4). Asking rents are typically a leading indicator of future average rents. Thus, the slowdown in asking rent growth could portend lower future rentinflation.
Figure 4: Year-over-year observed rent growth starting to slow
Source: Zillow and Haver Analytics. Note: Twelve-month percentage changes in Zillow’s observed rent index. Gray shading indicates NBER recession dates.
Conclusion
Rents are an important component of consumer expenditures. Recent surges in rentinflation have led to concerns that overall inflation might stay persistently high despite tightening of monetary policy. We present evidence that monetary policy tightening is effective for reducing rentinflation, although the full impact takes time to materialize. A policy tightening equivalent to a 1 percentage point increase in the federal funds rate can reduce rentinflation up to 3.2 percentage points over the course of 2½ years. This translates to a maximum reduction in headline PCEinflation of about 0.5 percentage point over the same time horizon. Although average rents are slow to respond to policy changes, growth of asking rents on new leases has started to slow following recent monetary policy tightening. Our finding suggests that this tightening will gradually bring rentinflation down over time, thereby helping to reduce overall inflation.
