Predicting inflation: What have we learned?

What have we learned about inflation in the past few years—and how can it guide investor forecasting?

The pandemic-driven inflation shock is ending, as inflation rates in 2024 look likely to take a glidepath toward central bank targets. It’s a good time to reassess the lessons we learned, with the goal of becoming a better inflation forecaster.

It’s not merely an academic matter: Soaring inflation can damage both stock and bond performance, as investors were reminded in 2022.

Broadly speaking, economists agree on the basic theory of inflation. Yet they have disagreed intensely on how to read the recent data.

They’ve also differed on what historical analogy best captures the pandemic experience. One group of economists saw (to simplify the debate) a 1970s-style inflationary environment from 2021 in which rising prices were becoming so entrenched that only a recession could bring inflation down to the Federal Reserve’s 2% target. Another cohort found historical parallels in wartime inflation shocks, such as the Korean War of the early 1950s. During wartime periods, the supply and demand forces fueling inflation proved temporary, which allowed inflation to come down without the onset of an economic recession.

We are now close to settling the debate. Over the last year, the Korean War analog proved useful, while the 1970s analog mostly led investors astray.

To be sure, sometimes a recession is needed to bring down high inflation. But it’s now clear that the forces that drove inflation during the pandemic—fiscal stimulus, supply chain disruptions, consumer spending skewing toward goods, and low labor force participation—have mostly proved temporary.¹

In short, we believe inflation is moving steadily lower. Our base case macro outlook sees moderate economic growth and cooling inflation, which should continue to be supportive for both stock and bond returns.

But we acknowledge that resurgent inflation poses a key risk to that outlook. If inflation takes off again, investors could revise their expectations for central bank policy and, once again, increase the risk that it could take a recession to eradicate inflation. In such an environment, both stocks and bonds could struggle.

We don’t believe you need to excessively fear that inflation will re-accelerate in the next couple of years. But we do believe it’s important to have a framework for forecasting the future path of inflation. To that end, we’ve made some enhancements to our traditional forecasting models.

We hope they will help clients build investment portfolios not only for the next one or two years, but also for future inflationary regimes, should they change unexpectedly.

Here, we present five key takeaways from our latest analysis:

1. Unemployment is a less useful indicator

Wages, of course, are central to the inflation puzzle, closely watched by economists to detect any sign of the dreaded “wage-price” spiral. In traditional economic models, a lower unemployment rate—linked to higher wages—implies greater inflationary pressures. But when the inflation shock began in Q1 2021, the U.S. unemployment rate averaged 6.2%—relatively high. Then, in the second half of 2023, when inflation came down rapidly (and faster than consensus expectations), unemployment averaged 3.7%—relatively low. The unemployment rate told us little about the trajectory of inflation.

What data can we track to better understand if inflationary pressures are building in the labor market? To answer this question, we performed a statistical “horse race.” We tested 10 labor market variables in a regression framework to determine which had the best fit and predictive power for wage growth, a strong inflation indicator.

The winner is the so-called “pressure gauge,” defined as the quits rate divided by the layoff rate. It explains three-fourths of the variance of wage growth in the regression. The unemployment rate explains only half of the variance.

What makes the pressure gauge a useful inflation indicator? Essentially, it can help identify relative tightness in the labor market. More layoffs suggest a relatively loose labor market, while a higher quits rate (driven by workers’ enhanced bargaining power) signals a relatively tight labor market.

Around mid-2021, the layoff rate dropped below pre-pandemic levels, as labor was relatively scarce. Meanwhile the quits ratio soared in the context of the Great Resignation. As a result, the pressure gauge reached a boiling point. At the end of 2021, when the labor market was at its tightest, the pressure gauge was about 3 standard deviations above its long-term average.

During this period, wage growth accelerated, to about 6.5%, and overall price inflation reached 40-year highs.

Conditions then changed in 2023 as the labor market rebalanced. That is, labor participation surged,² layoff rates nudged slightly higher and the quits rate declined, marking the end of the Great Resignation. Wage growth then cooled to a rate of about 4%. Overall inflation also came down, so much so that investors are now pricing in Fed interest rate cuts this year.

The current signal from the pressure gauge finds a labor market only modestly tighter than it was in 2019: about 1 standard deviation above its long-term average versus 0.8 standard deviations above the average in 2019. We believe the labor market is currently consistent with underlying inflation rates in the range of 2%–2.5% (i.e., modestly above the inflation rates that persisted in 2018–2019).

Of course, we’re not planning to completely ignore the unemployment rate in our analysis of inflationary pressures. But we think the pressure gauge offers a valuable complementary measure of labor market tightness that we will now carefully consider in our inflation forecasts.

2. Shelter inflation moves with important time lags

Shelter inflation (inflation coming from the housing market) has long been a key component of U.S. inflation, accounting for a 34.4% weight in the CPI and a 15.5% weight in PCE inflation.³ (Shelter inflation is generally less important in economies outside the United States.⁴)

But the pandemic experience revealed the limitations of our traditional understanding of shelter inflation. Specifically, we learned that shelter inflation, which measures the average cost of housing in the economy, is slow moving and lags new housing rental contracts that are tracked on a monthly basis.

Bureau of Labor Statistics research suggests shelter inflation lags new tenant rents by about four quarters.⁵

Understanding the time lag allows us to make projections about where shelter inflation will trend over the next four quarters using data on new tenant rents.

In 2024, shelter inflation is expected to slow to a rate comparable to where it was in late 2019.⁶ As the weight of shelter in the index has grown since 2019, shelter’s contribution to CPI/PCE inflation by the second half of 2024 won’t be identical to what it was in 2019. But it will be close.

In short, post-pandemic, we better understand how rental inflation pressures can build well before they appear in the official data. And we can turn to reliable indicators for early warning signs of an uptick in this important inflation metric.

3. Old supply-and-demand inflation models failed spectacularly

Pre-COVID, economists typically used variables such as the U.S. dollar exchange rate, and food and energy prices to model the impact of supply shocks on inflation. But these variables failed completely to capture the magnitude and breadth of the COVID-era supply-side disruptions.

A commonly used core PCE inflation model that worked well pre-COVID missed the pandemic-driven inflation shock in the out-of-sample projection. But look what happens when we overlay the New York Fed’s Global Supply Chain Pressure Index (GSCPI), which was created during the pandemic and is now updated monthly.

Unlike the traditional model, the GSCPI did a better job in capturing the complex supply chain disruptions that emerged during the pandemic. These disruptions included semiconductor shortages, shipping backlogs at ports and lumber shortages caused by plant lockdowns.⁹

More recently, the GSCPI has normalized, which is consistent with continued supply-side inflation normalization in 2024. But this could change, and the data needs to be monitored closely.

If the GSCPI does especially well identifying supply-side inflation drivers, another approach (pioneered by Adam Shapiro at the San Francisco Fed) can excel at identifying demand-based drivers.

Shapiro’s approach is especially useful because supply and demand shocks can hit simultaneously, sparking inflation while making it very difficult to distinguish precisely which factors are generating the price spikes. Shapiro’s model disentangles the supply-demand nexus: For a given item in the inflation basket, if in a month prices and quantities of the item consumed move in the same direction (and with a magnitude that is statistically significant), that item is said to be influenced by demand. If prices and quantities consumed move in opposite directions, then supply is the driving force.

Helpfully for investors, Shapiro’s framework shines a light on a favored metric of the Federal Open Market Committee (FOMC): core services inflation excluding housing. As the FOMC sees it, this metric best identifies inflation driven by wage growth. But some components in this metric, for example, auto insurance and telecommunication services, are not especially labor-intensive, and thus likely not illustrative of wage growth pressure. Shapiro’s framework helps reconcile this confusion by isolating the demand component within core services inflation excluding housing.

Like the pressure gauge measuring labor market tightness, this metric shows a cooling over the last year, though it is not yet back to the pre-pandemic rate.

4. Short-term inflation expectations matter, too

Pre-COVID, economists modeling inflation focused on long-term inflation expectations and downplayed the significance of short-term expectations. No longer.

Economists had assumed that movements in short-term inflation expectations often reflect commodity price swings, and that central bankers would mostly ignore these measures when setting monetary policy. Long-term inflation expectations, on the other hand, signaled whether consumer, business and investor expectations were anchored at central banks’ 2% target.

But in 2022, central banks faced a tough test. Measures of long-term inflation expectations suggested that inflation would not spiral out of control. Yet measures of short-term inflation expectations soared to multiples of the 2% central bank target.

In response, central banks recalibrated. In a speech at the Jackson Hole Symposium in 2022,¹⁰ Fed Chair Jerome Powell argued that seemingly anchored long-term inflation expectations could start to shift higher if realized inflation stayed too high for too long. At the time, high realized inflation was driving uncomfortably high measures of short-term inflation expectations.¹¹

Lesson learned: Economists must now pay much closer attention to short-term inflation expectation measures.¹²

5. Energy shocks matter, especially outside the United States and Canada

That energy shocks matter much more in economies beyond the United States and Canada is a longstanding, well understood trend. But the COVID experience revealed how powerful those shocks can be in passing through to non-energy inflation.

We estimate this passthrough in the chart below, showing how the estimates jump when pandemic data is included in the sample.

The story in Germany is especially dramatic. The estimated passthrough excluding the pandemic is statistically indistinguishable from zero, but the measure jumps markedly when pandemic-era data is included. After Russia invaded Ukraine, Germany halted the Nord Stream pipeline project that brought Russian natural gas to Germany. Its energy sector was severely disrupted, and prices soared in an unprecedented way.

The episode reminds us that historical data can sometimes completely fail to characterize a country’s vulnerability to a novel shock.

We also note meaningful jumps in energy pass-through estimates for the United Kingdom, France and Italy when comparing pre-pandemic estimates to periods including the pandemic.¹³

Inflation outlook: The worst is behind us

Finally, post-COVID, we better understand how inflation captures people’s attention in a nonlinear fashion. Here’s what we mean by nonlinear: When inflation falls below a certain threshold, the general public and market participants do not perceive it to be a problem. But once inflation rises above that threshold, it quickly becomes alarming—and often a dominant factor in market performance.

We were aware of this dynamic before COVID, but fuzzy on the specifics. The pandemic has given us more reliable empirical estimates of the relevant thresholds that can trigger inflationary regimes versus non-inflationary regimes.

So what’s the threshold for U.S. inflation?

We illustrate it, from a markets perspective, in the chart below. Drawing on data back to 1965, the chart plots inflation against equity market price-to-earnings (P/E) multiples.

The fitted trend line is distinctly nonlinear. In other words, equity market multiples are typically high and stable when inflation is below a threshold of close to 3%, but multiples deteriorate rapidly once the threshold is crossed. Historically, when inflation was below 3%, the average equity multiple was 22.8x versus 14.9x when inflation was above 3%.

If inflation remains at 3% or below, we expect inflation to fade as a dominant driver of market returns. Some market participants worry that inflation could re-accelerate in 2025, moving past the 3% threshold in the wake of Fed rate cuts in 2024 and a possible pickup in shelter inflation. We acknowledge the risk, but it’s not our base case.

On the geopolitical front, the fluid and tense situation in the Middle East presents a risk to the inflation outlook, particularly from the perspective of defense spending. Yet absent another major military shock impacting the global economy, inflation looks set to remain below 3% through 2025. We’re encouraging clients not to become too fearful of the threat of inflation re-accelerating in the coming years.¹⁴ Multi-asset portfolios will likely be well supported in an environment of below 3% inflation.

Inflation is a critical economic variable, and forecasting its path is, necessarily, a challenge. Our approach to inflation forecasting is not radically new. But we know more about inflation than we did prior to COVID, and we think our analysis is now more refined. We hope this revised framework helps our clients in crafting investment portfolios for a variety of inflationary regimes.

¹We prefer the word “temporary” over “transitory” in describing the recent inflation shock because in 2021–22, the latter carried a specific meaning: that central banks wouldn’t hike aggressively to bring pandemic-driven inflation back down to target. In this context, the transitory narrative was wrong.

²Compared against the CBO’s pre-pandemic forecast (made in January 2020), the labor force participation rate is currently higher by 0.5%.

³The main reason shelter carries a higher weight in CPI than in PCE: PCE weights are determined by actual consumer spending, whereas CPI weights are determined by surveys that ask consumers to self-report what they regularly spend on a variety of items.

⁴While the United States, Canada and the United Kingdom include owner-occupied housing inflation estimates in their CPI baskets, most other countries do not, including Japan, the countries in the Euro Area and most emerging market economies. This is the main reason the shelter weights are higher for the United States, Canada and the United Kingdom. Owner-occupied housing is an odd concept, since it is measuring a cost that nobody actually pays: It is a measure of what homeowners implicitly pay to themselves in rent each month. Nevertheless, the argument for including it in the CPI basket is that homeowners, by choosing to live in the homes they have purchased, are effectively forgoing the opportunity to earn rental income from the home. This is treated as an opportunity cost and hence relevant for a holistic understanding of the cost of living.

⁵Brian Adams, Lara Lowenstein, Hugh Montag et al., “Disentangling Rent Index Differences: Data, Methods and Scope,” bls Working Paper 555, U.S. Bureau of Labor Statistics, October 6, 2022.

⁶One caveat: New tenant rent indexes are highly susceptible to end-of-sample revisions, due to the fact that housing units in the full sample are surveyed only every six months. As a result, shelter inflation projections based on new tenant rents need to be continually monitored and updated when revisions occur.

⁷Especially in the context of the recent downturn in multifamily housing starts, which could create the conditions again for a tight apartment market by 2025.

⁸We performed a Granger causality test confirming this leading relationship. Results available on request.

⁹That the GSCPI can capture these specific sector shocks effectively at the macro level is promising, as it would not make sense for macroeconomists to include a variety of sector-specific variables in inflation models.

¹⁰Jerome H. Powell, “Monetary Policy and Price Stability,” Speech at economic policy symposium, Jackson Hole, Wyoming, August 26, 2022.

¹¹Moreover, a circularity likely exists between short- and long-term inflation expectations: During a period of high inflation, long-term expectations could very well be anchored because consumers and businesses expect central banks to hike aggressively to bring inflation back to target.

¹²One nuance here: In 2011, amid a commodity price surge, the Fed ignored rising measures of short-term inflation expectations and didn’t hike interest rates. The likely key difference between 2011 and 2022: In 2011, the labor market was weak—far from conventional estimates of full employment. Thus we conclude that short-term inflation expectations are more or less important depending on other economic variables. If the labor market is at or near full employment, we believe it is important to closely track these expectations. Finally, we note that the European Central Bank (ECB), unlike the Fed, did hike interest rates twice in 2011. Yet those two rate hikes were very likely a policy mistake, one the ECB reversed in late 2011.

¹³From a level perspective, the United Kingdom exhibits the highest estimate when looking at the full sample period (from 2000 to Q3 2023), at 16 basis points, which means that for a 1% rise in energy inflation, the pass-through to non-energy inflation is expected to be 0.16%—0.16% may not sound very high. But consider that energy inflation is very volatile, with a 1 standard deviation YoY% change at about 12% when considering data back to 2000. During the pandemic, consumer energy prices in the United Kingdom rose by nearly 50% from the start of 2021 to the peak in Q3 2022. For a country like the United Kingdom, energy can quickly pass through and become an important feature of inflation during a shock such as the one induced by the pandemic and the war in Ukraine. In the United States, by contrast, the energy pass-through is estimated at only 0.02%. 

¹⁴Also, to be clear about the risk: It’s not so much about inflation “re-accelerating” but rather inflation getting “stuck” at too high a run rate that is incompatible with the Fed’s 2% inflation objective. Core PCE inflation rose at 2.9% over the last year. Our view is that it will be in the vicinity of 2-2.5% by the end of 2024. However, if it were to get “stuck” at around 3% for the foreseeable future, that would present a risk to the macroeconomic outlook and raise the odds that a recession would be needed to bring inflation down to 2% on a sustained basis.