The COVID-19 pandemic brought both a public health crisis and a significant shift in economic dynamics. Consumer spending patterns changed rapidly, highlighting the way inflation is measured. Alberto Cavallo, Harvard Business School professor and co-director of D^3’s Pricing Lab described this issue in his research, “Inflation with Covid Consumption Baskets,” first published in 2020 in the National Bureau of Economic Research and updated in 2024 in the IMF Economic Review. The research focuses on the biases introduced in traditional inflation metrics, like the Consumer Price Index (CPI), during the pandemic and those biases’ varied effects on consumers in the US and around the world.
In his paper, Cavallo describes the “Covid basket,” an adjusted set of consumer spending weights he developed based on real-time debit and credit card data from the Opportunity Insights (OI) Economic Tracker, which he used to measure inflation during the COVID-19 pandemic. This basket reflects the shifts in consumption patterns that occurred as a result of the pandemic, such as increased spending on food and groceries and decreased spending on transportation, recreation, and other services. As part of his analysis, Cavallo acknowledged that OI consumer spending data has limitations and may not offer a complete picture of total consumer spending in the United States.
Key Insight: Traditional Inflation Measures Underestimated COVID-19 Inflation
“By May [2020], the annual inflation rate of the US Covid index was 1.02%, compared to only 0.13% of the official CPI (all-items, US city average, not seasonally adjusted).” [1]
Cavallo’s research reveals that the official CPI failed to capture the real inflationary pressures during the pandemic. With consumers spending more on food and other essentials, which experienced inflation, while cutting back on transportation and recreation, which experienced deflation, the basket of goods used to calculate inflation became outdated. The CPI basket weights, which determine the relative importance of different categories in the overall inflation calculation, were based on pre-pandemic expenditure data from 2017-2018.The result was a significant underestimation of inflation.
Key Insight: COVID Inflation Hit Low-Income Households Harder
“The difference between the bottom and top quintiles of the income distribution peaked in May [2020], when the low-income inflation rate was 1.34% compared to just 0.47% for high-income households.” [2]
The inflation experienced during the pandemic did not impact all income groups equally. Cavallo’s research shows that low-income households, which allocate a larger share of their spending to food and essential goods, faced a much higher inflation rate than wealthier households. In May 2020, the inflation rate for low-income households was nearly three times that of high-income groups.
Key Insight: Inflation Diverged Across Sectors Globally
“In 14 countries I find that the Covid-basket inflation rate was higher than that of the official CPIs in December 2020.” [3]
Cavallo’s analysis extended beyond the United States. He studied 19 other countries and, in 14 of those countries, found the COVID-basket inflation rate was higher than the official CPI, driven largely by the increase in spending on food and the drop in spending on transportation. Of the countries studied, Brazil showed the greatest divergence; the COVID inflation rate in December 2020 was 5.88%, compared to 4.81% in the official CPI. Beyond this common trend, Cavallo found variations in other countries driven by differences in sectoral inflation rates and noted that additional study of high-frequency transactions in other countries might help to explain the differences.
Key Insight: Expenditure Switching Not Noted During COVID
“With the updated COVID weights, I am explicitly allowing for [expenditure switching] to happen, so why do I find even higher inflation? In other words, why is this expenditure switching not happening during COVID?” [4]
In a further examination of the reasons behind the observed inflation discrepancies, Cavallo asked why he did not observe expenditure switching (or people moving spending to different categories) during the pandemic. He suggested three possible answers to this question:
- A combination of changing demand (people consuming more groceries and less transportation because they have to stay home) and supply disruptions
- Demand becoming less flexible due to lockdowns and social distancing, for example, people eating at home even when prices rise and not traveling even with lower fuel prices
- Consumers changing spending within categories, for example, buying cheaper types of food
Why This Matters
For C-suite executives and policy makers, accurate inflation data is crucial for making sound strategic decisions, such as pricing products, forecasting costs and revenues, allocating capital, and compensating employees. Especially in times of unforeseen and dramatic economic crisis, executives and policy makers need to have inflationary data they can trust. For an executive, underestimating inflation can lead to underpriced products, eroded profit margins, and inadequate compensation planning, potentially harming employee morale and retention. Meanwhile, for policy makers, inflation underestimation can and does lead to inadequate policy responses, potentially exacerbating the negative economic impacts of the crisis.
References
[1] Alberto Cavallo, “Inflation with Covid Consumption Baskets”, National Bureau of Economic Research (June 2020) and IMF Economic Review (June 2024) 1- 21, 1.
[2] Cavallo, “Inflation with Covid Consumption Baskets”, 2.
[3] Cavallo, “Inflation with Covid Consumption Baskets”, 2.
[4] Cavallo, “Inflation with Covid Consumption Baskets”, 14.
Meet the Author
Alberto Cavallo is the Thomas S. Murphy Professor of Business Administration at Harvard Business School, where he teaches in the Business, Government, and the International Economy (BGIE) unit. He is also co-director of D^3’s Pricing Lab. Cavallo’s research focuses on inflation. In particular, he studies the behavior of prices and its implications for macroeconomic measurement, models and policies.