Using Core Inflation to Predict Headline Inflation

November 28, 2023

In our previous On the Economy blog post, we discussed the importance of current food inflation as a predictor of future inflation more generally. We noted that while food inflation used to be a weak predictor, it has become considerably more useful. By contrast, current energy inflation was and remains a very poor predictor of future inflation. This is not to say that energy (i.e., gasoline) inflation is not important to our lives. Rather, if we are making economic plans, and those plans require a good forecast of future inflation, data suggest that keeping an eye on current food inflation instead could be much more useful.

In this blog post, we revisit our previous results from a related but more pragmatic perspective through some forecasting. The goal is to provide simple visual evidence of why Federal Reserve officials often reference the concept of core inflation and more recently have begun referencing the concept of core inflation excluding energy. The former is based on a price index of goods and services that excludes food and energy, while the latter excludes only energy.

Our approach to forecasting is standard. For any given month, we construct year-over-year inflation for several primary components of the consumer price index (CPI). We then use those values to forecast year-over-year headline CPI inflation 12 months into the future. We evaluate the accuracy of the forecast by averaging the squared forecast error over a five-year (60-month) period and taking its square root. This root mean squared error (RMSE), like Euclidean distance, is a common measure of closeness—or in the current context, accuracy. The higher the RMSE, the less accurate the forecast.

In the first figure, we plot the rolling five-year windows of RMSEs for the food, energy and transportation components of CPI.

Root Mean Squared Error for Forecasted Headline CPI Using Food, Energy and Transportation Components

A line chart shows the root mean squared error for forecasted headline CPI inflation using core CPI, core CPI excluding energy, and the CPI components of food, health care and shelter since 1988. The smallest RMSEs tend to be associated with core or core excluding energy.

SOURCES: Bureau of Labor Statistics (retrieved from FRED) and authors’ calculations.

NOTES: The higher the RMSE value, the less accurate the forecast. The values are estimated using a five-year rolling window over the Great Moderation sample (the period of relative macroeconomic stability from 1983 to the present).

At any point in time, food inflation has been a better predictor than transportation, which in turn has been a better predictor than energy inflation. Perhaps most striking is just how poorly energy inflation has predicted future inflation. There might have been a moment in the late 1990s when the component could have been considered, but by the 2000s, energy inflation was significantly less useful than food inflation as a predictor.

In the second figure, we focus on those components that tend to be more accurate (food, health care and shelter) and compare their accuracy with core inflation and core inflation excluding energy.

Root Mean Squared Error for Forecasted Headline CPI Using Core CPI, Core CPI Excluding Energy, and the CPI Components of Food, Health Care and Shelter

A line chart shows the root mean squared error for the CPI inflation components of food, energy and transportation since 1988. The smallest RMSE of the three, which indicates greater accuracy in predicting inflation 12 months in the future, was associated with food inflation.

SOURCES: Bureau of Labor Statistics (retrieved from FRED) and authors’ calculations.

NOTES: The higher the RMSE value, the less accurate the forecast. The values are estimated using a five-year rolling window over the Great Moderation sample (the period of relative macroeconomic stability from 1983 to the present).

There are several items to note. First, across the entire timeline, the smallest RMSEs tend to be associated with core or core excluding energy. Perhaps more importantly, even when core and core excluding energy aren’t the most accurate predictors, they never deviate much from being the best predictors.

In a nutshell, this is why Federal Reserve officials regularly reference these two measures of inflation. By contrast, while food has done well in a few periods, it also has periods like the one after the Great Recession, in which its accuracy diminished substantially. Interestingly, while health care was a poor predictor throughout the first part of the sample, by 2000 it had become competitive with food and shelter.

About the Authors
Michael McCracken
Michael McCracken

Michael W. McCracken is an economist and senior economic policy advisor at the Federal Reserve Bank of St. Louis. His research focuses on econometrics and macroeconomic forecasting. He joined the St. Louis Fed in 2008. Read more about his work.

Michael McCracken
Michael McCracken

Michael W. McCracken is an economist and senior economic policy advisor at the Federal Reserve Bank of St. Louis. His research focuses on econometrics and macroeconomic forecasting. He joined the St. Louis Fed in 2008. Read more about his work.

Trần Khánh Ngân

Trần Khánh Ngân is a research associate at the Federal Reserve Bank of St. Louis.

Trần Khánh Ngân

Trần Khánh Ngân is a research associate at the Federal Reserve Bank of St. Louis.

Related Topics

This blog offers commentary, analysis and data from our economists and experts. Views expressed are not necessarily those of the St. Louis Fed or Federal Reserve System.


Email Us

Media questions

All other blog-related questions

Back to Top