A Framework for Forecasting US Inflation

Summary

• A new IMF working paper revisits the causes of the dramatic rise in US inflation during the Covid era.
• The authors find the ongoing labour market tightness, plus shocks to headline inflation from energy prices and supply chain issues, can explain the acceleration in core CPI.
• Looking ahead, unemployment at 4.4% is unlikely to return inflation to target by 2025, meaning the Fed will need to slow the economy by more than it currently forecasts.

Introduction

Inflation has become the central economic problem of the Covid era. In 2022 alone, our Deep Dive series has covered its origins, level, persistence and impact. We also explained how shocks to food prices and supply chains filter into headline CPI and showed how developments in the housing and labour markets help determine the trajectory of inflation.

With 2022 ending, a new IMF working paper looks back at the origins of the US inflation surge. In doing so, it creates a framework for forecasting 2023’s inflation dynamics. According to the authors, the outlook hinges on inflation expectations and the Beveridge curve, but core CPI will likely be in a range of 5.0-7.2% by the end of next year. Moreover, the economy will need to slow by more than the Fed is currently forecasting to return inflation to target by 2025.

Inflation Identity

The paper’s framework hinges on the following common decomposition:

Headline inflation = Core inflation + Headline shocks

Core inflation, or underlying inflation, is a relatively slow-moving component that, the authors argue, depends on inflation expectations and labour market slackness. Headline shocks refers to deviations of headline inflation from core inflation. These are high-frequency movements arising from large price changes in particular sectors of the economy.

An Alternative Measure of Core Inflation

Developed in the 1970s, the traditional measure of core CPI, XFE inflation, excludes food and energy prices. It is the one the Fed focuses on and currently stands at 6.3% (October 2022). Instead, the authors use the weighted median CPI inflation rate, published by the Federal Reserve Bank of Cleveland, to measure core CPI. In October, the annual rate was 7%.

The reason for the difference is that the traditional measure includes large price swings in industries besides food and energy. Measures like weighted medians or trimmed means, on the other hand, filter large price changes in any industry, thereby more accurately reflecting the underlying prices.

As a result, headline shocks are equivalent to any deviation of headline inflation from the weighted median CPI inflation rate (Chart 1). These deviations turned positive in March 2021 and peaked 12 months later at 3.6pp. That is, when underlying prices were increasing at 4.9% YoY in March 2022, temporary headline shocks were pushing headline CPI to 8.6%.

The recent rise in core CPI reflects stronger growth in underlying prices currently. However, median inflation actually drifted down in the first part of the pandemic and, as late as September 2021, was still below its level in January 2020. Given the common view that non-core inflation movements are transitory, this chart may explain the Fed’s nonchalance about inflation in 2021.

Decomposing Core Inflation

The authors’ framework allows core inflation to be influenced by three variables: expected inflation, labour market tightness, and past headline inflation shocks.

Mainstream economists argue that the inflation rate depends strongly on expected inflation. This concept is simple – if individuals act on their beliefs regarding future prices, current prices will follow. This argument has many pitfalls, but it is more likely to be true when looking at an economy’s long-run price level, as here. The authors capture long-run inflation expectations with the median 10-year-ahead CPI inflation forecast from the Survey of Professional Forecasters (SPF).

The link between core CPI and the labour market is less debatable. Since 2013, the correlation between annual wage growth and core inflation is 0.86 (Chart 2). Moreover, wages are closely connected to the tightness of the labour market, which can be proxied by vacancy rates (V), the unemployment rate (U), or a combination of the two (V/U). The authors focus on the average ratio of job vacancies to unemployment (V/U).

Finally, large movements in headline CPI are likely to feed back into core CPI through wages or other costs of producing output. These ‘second order effects’ can, in theory, cause wage-price spirals but are more likely to keep the underlying price level higher for longer, as now.

The Inflation Gap

What moves core inflation? And, in particular, what moves core inflation away from the long-term expectations of professional forecasters? The two generally move together, as they should in equilibrium, but they occasionally deviate, as now (Chart 3). When they do, and if the three-way decomposition of core inflation above is correct, the drivers will be due to developments in the labour market or headline shocks.

To answer this, the authors define the ‘inflation gap’ as deviations of the underlying price level from 10-year-ahead inflation forecasts:

Inflation gap = Core Inflation – Inflation Expectations

Next, they look at the historical relationship between (i) the inflation gap and V/U and (ii) the inflation gap and headline shocks.

For 1985-2019, the relationships turn out to be strikingly non-linear and asymmetric (Chart 4). Negative headline shocks – occasions where headline inflation rates are below median inflation – have negligible impacts on core inflation. Meanwhile, positive headline shocks raise core inflation, as we see today. The same is true for labour market tightness.

We can use these relationships to make predictions, or fitted values, of core inflation during the pandemic. These fitted values (Chart 5, blue dotted line) are remarkably close to the dynamics of actual core inflation (Chart 5, solid blue line), showing just how important these relationships are in determining the pathway of inflation. Looking at labour market tightness alone (Chart 5, red dotted line) is also very informative.

No Room for Fiscal Story?

Many argue government support programmes caused the rapid acceleration of inflation in 2021 and 2022. While this framework does not explicitly include a fiscal channel into core inflation, it still plays a role.

Estimates show the American Rescue Plan (ARP) increased V/U by approximately 0.6 at the end of 2021 and 0.5 at the end of 2022 (currently 1.8). It means the ARP influenced inflation by increasing labour market tightness, either by reducing unemployment and/or increasing vacancies.

Without the support, the authors estimate that core inflation in September 2022 would have been 5.1% rather than 7%. In total, they expect government support added as much as 40% to core inflation since the end of 2020.

These assumptions even exclude the possibility that government support could have entered headline shocks. It could have done so by, for example, helping intensify supply chain disruptions by supporting demand.

Headline Inflation

Headline shocks are monthly deviations of headline from core inflation, which affect inflation both directly and through their pass-through to core.

It comes as no surprise that the authors find three variables – energy-price shocks, backlogs of work, and auto-price shocks – to explain over 90% of the changes in headline shocks during the pandemic.

Combined, headline shocks account for about 4.6pp of the 6.9pp rise in headline inflation between December 2020 and September 2022 (Chart 6). Most of this reflects energy-price shocks and supply chain issues. Two thirds of the contribution is the effect on current headline inflation and one third is the pass-through into core.

There is also a significant pass-through from past auto-price shocks, reflecting the run-up in auto prices in Summer 2021. But the direct effect on headline inflation has turned negative as these price increases have been partly reversed.

Meanwhile, the contribution of V/U to the rise in headline inflation is 2pp, nearly a third of the total inflation increase. However, the rise in V/U explains more – nearly half – the rise in core inflation, and the effect of V/U is rising over time.

Looking Ahead

With tightness in the labour market clearly important for the direction of inflation, the future trajectory of prices will depend on the unemployment and vacancy rates. The Beveridge curve captures this relationship best.

We recently documented the outward shift in the Beveridge curve. This is equivalent to an increase in the natural rate of unemployment. In other words, it has made the trade-off between unemployment and inflation steeper. To curb inflation now requires a relatively larger shift in unemployment than pre-pandemic.

To show this, the authors simulate a modest unemployment rate rise to 4.4% by the end of 2023 – a figure projected by the Fed in September. For a given Beveridge curve and assumptions about inflation expectations, the authors can predict the likely path of inflation, excluding any additional shocks from energy prices etc.

A range of scenarios are tested. If the Beveridge curve stays where it is, which is likely if we believe that the shift outwards is because of a smaller employment pool post-Covid, inflation will take longer to subside (Chart 7, red lines). If inflation expectations quickly revert to their pre-pandemic level of 2.2%, inflation under this scenario would fall to 5.8% by the end of 2023 and 4% a year later.

Only under the most optimistic scenario, where expectations fall back in line and the Beveridge curve returns to pre-pandemic levels, will a 4.4% unemployment rate be enough to drop inflation back to 2% by 2025.

Bottom Line

The paper creates a simple framework to explain the drivers of headline and core inflation. Anything that influences the labour market, including government support programmes, or inflation expectations can directly affect headline inflation through its core component. Anything else, like food or energy shocks, can influence headline inflation directly through headline shocks, or indirectly by passing through into core inflation.

You could tell any story with this approach. This makes it valuable for not only appraising the current period of inflation, but future ones too. It also allows you to forecast inflation with just three parameters – unemployment, vacancies and inflation expectations. And in doing so with the Fed’s forecast of the unemployment rate, the authors find inflation does not return to 2% anytime soon. Instead, the Fed will either need to accept higher inflation for longer or slow the economy down by more.