Change theme
Change theme

Pricing behaviour and inflation during the COVID-19 pandemic: Insights from consumer prices microdata


As the Bank of Canada assesses whether inflation is firmly on the path back to the Bank’s 2% target, corporate pricing behaviour has been identified as a key issue. Using rich data on thousands of individual goods and services, we provide new insights into how firms’ price-setting behaviour has evolved and the role of this behaviour in recent inflation dynamics.

By tracking how often and by how much Canadian retailers have been changing their prices, we can characterize the price-setting behaviour underlying the initial surge and subsequent easing of inflation. Past research shows that the run-up in inflation was associated with firms passing rising costs through to consumers, thus maintaining healthy profit margins.1 Our findings reveal a significant change in firms’ price-setting behaviour, with retailers increasing prices much more often than in the past.

Over the past 12 months, pricing behaviour has been steadily returning to normal, which is consistent with insights from the Bank’s business surveys. However, pricing behaviour has yet to fully normalize. This is most evident among prices of goods excluding food and energy. For these items, the frequency of price increases relative to the frequency of price decreases remains high.

Data and methods

Our analysis is based on Statistics Canada’s publicly available consumer prices microdata, which include prices and metadata for a sample of goods and services of unchanged quantity and quality used to construct the consumer price index (CPI). The microdata file contains the majority of collected prices that contribute to the CPI calculation and that are available in the CPI’s main internal micro-database. While the consumer prices microdata file is similar to the database used to create the official CPI, it does not contain all the data from which the CPI is produced. It contains price quotes for around 100,000 goods and services each month and is intended to be used specifically for research purposes rather than to replicate the official CPI.

The consumer prices microdata file is governed by the Data Access Division of Statistics Canada and accessed through virtual research data centres. The dataset includes many variables, such as reference period, outlet identifiers, representative product names and price observations. For confidentiality reasons, the dataset withholds outlet names, addresses and details about the specific products selected (such as brand). Coverage begins in February 1998, and the latest available data at the time of our study are for January 2024.

Although the microdata file contains information for most products and services at a regional level, we conduct our analysis only at the national level. In addition, our analysis excludes shelter prices since those methodologies and data sources are unique, requiring separate specialized microdata processing systems. Shelter prices are also less relevant for examining corporate price-setting practices.

To study pricing behaviour, we generate two metrics that together comprise the rate of inflation:

  • the frequency of price changes, defined as the share (in percent) of prices that change each month2
  • the size of price changes, defined as the average non-zero monthly price change (in percent)

We further decompose each of these metrics into price increases and price decreases. We use all price changes in the data, including temporary sales.

The two metrics are affected not only by changes in pricing behaviour but also by changes in the underlying data and by the incorporation of alternative data sources over time. Most notably, the introduction of retail scanner data for food and other non-durables in 20183 raised the frequency of price changes while lowering their average size. Because of this, caution should be used when interpreting direct comparisons with the period before the COVID-19 pandemic, especially when examining more granular results.

In addition, the underlying data and derived metrics are highly seasonal and volatile. We therefore present our results as either 3-month moving averages of seasonally adjusted data, 12-month moving averages of unadjusted data, or both. Given that our pricing behaviour metrics are derived at a monthly frequency, 12-month moving averages can be used to interpret the dynamics of year-over-year inflation.

Frequency of price changes

In the years leading up to the pandemic, the frequency of price changes remained relatively stable. On average, about 14% of prices rose each month while 12% fell (Chart 1). However, this changed dramatically as the pandemic unfolded. In particular, the frequency of price increases began rising notably in early 2020, eventually peaking in early 2022 (Chart 1, light blue and dark blue lines). This period corresponds with some of the largest monthly increases in the history of the Canadian CPI.

The frequency of price decreases did not move nearly as much. It spiked briefly during the lockdowns in early 2020 and later eased back toward pre-pandemic levels as inflation began to rise in 2021 (Chart 1, red and pink lines). Thus, the rise in inflation was associated primarily with retailers raising prices much more often than in the past.

More recently, the frequency of price increases has come down substantially, but it remains elevated by historical standards. Interestingly, the frequency of price decreases has been trending up, reaching its highest level over the sample period. This may reflect that some prices became unsustainably high during the run-up in inflation, leading retailers to adjust price levels downward. Progress in achieving a better balance of price increases and decreases is more evident in the 3-month moving average than in the 12-month moving average, though the former is also extremely volatile.4

Chart 1: Frequency of price changes

Size of price changes

Our findings show that the size of the average price increase was getting smaller as inflation was rising during the pandemic (Chart 2, panel a). At first glance, this may seem counterintuitive. However, it simply reflects the fact that retailers were raising prices much more frequently than before, so each price increase did not need to be as large to still generate substantial inflation. The average price increase has ticked up since late 2023 but remains below historical norms.

We also see that price decreases became less pronounced as inflation was rising during the pandemic (Chart 2, panel b). This trend has continued as inflation has come down, partially offsetting the higher frequency of price decreases.

Chart 2: Average size of monthly price changes

Chart 2: Average size of monthly price changes

Average non-zero monthly price change (in percent)

* Monthly data, unadjusted
† Monthly data, seasonally adjusted
Sources: Statistics Canada and Bank of Canada calculations
Last observation: January 2024

Putting the pieces together

Having examined the frequency and size of price changes separately, we now investigate which of the two has quantitatively played a more meaningful role in inflation dynamics during the pandemic. To do this, we use the following simple equation, which expresses the inflation rate as a function of the frequency and size of price increases and decreases:

\(\displaystyle{Inflation}_t\) \(\displaystyle=\,{Frequency}_t^{Increase}× \,{Size}_t^{Increase}\) \(\displaystyle+\, {Frequency}_t^{Decrease}× \,{Size}_t^{Decrease} \)

We begin by generating a baseline inflation rate according to the above equation. Given that we are not working with the full CPI basket, this baseline inflation rate is not the same as the official inflation rate. However, the two are very highly correlated.

Next, we generate two counterfactual inflation rates:

  • For the first, we hold fixed the frequency of price increases and decreases.
  • For the second, we hold fixed the size of price increases and decreases.

As Chart 3 shows, holding frequency fixed generates a relatively flat inflation profile. In contrast, holding size fixed generates an inflation profile that closely mirrors the baseline rate. This reveals that inflation dynamics throughout the pandemic have mainly reflected changes in how often firms set their prices. However, since late 2023, the average size of price changes has become slightly more relevant.

Chart 3: Counterfactual exercise

How far from normal is pricing behaviour?

Given our findings on the importance of the frequency of price changes for inflation dynamics, the relative frequency of increases and decreases may be a good summary indicator of pricing behaviour. Chart 4 plots this relative frequency—defined as the percentage of prices that are increasing minus the percentage of prices that are decreasing—against the annual inflation rate of CPI excluding shelter.5 We see that the two are highly correlated, with a correlation coefficient of 0.89. This strong correlation also holds for major categories of the CPI.

Another advantage of this indicator is that it is better suited to historical comparisons than using either the frequency of price increases or the frequency of price decreases alone. This is because structural changes in the dataset, such as the adoption of retail scanner data, raised the frequency of price increases and decreases proportionally, leaving their relative frequency unaffected.

Chart 4: Relative frequency of price increases and decreases versus actual inflation

Chart 5 shows the relative frequency of price increases and decreases for the CPI overall and for select CPI components (the Appendix offers a more detailed look at the frequency of price changes for each of the three selected components). We compare the January 2024 values with the pandemic peaks and historical averages. Although the chart shows that corporate pricing behaviour is normalizing, it has not returned to pre-pandemic levels. This is most evident for the goods excluding food and energy component, which has the largest gap between its January 2024 value and its historical average. Note that for CPI overall, the gap between the January 2024 value and the historical average is very small. This is partly because the relative frequency of price increases and decreases for energy (not shown here explicitly) is deeply negative. Progress among non-energy categories has been notable but not as pronounced.

Chart 5: Relative frequency of price increases and decreases for CPI and select components

Concluding remarks

Our analysis shows that the surge in inflation during the COVID-19 pandemic was accompanied by a material change in corporate pricing behaviour. Specifically, retailers raised prices much more often than in the past. Over the past 12 months, price-setting practices have been normalizing steadily. While pricing behaviour is not fully back to normal, the broad trends in the data generally favour continued improvement in the months ahead. These findings are broadly consistent with the Bank’s business surveys, which show that fewer firms are planning unusually large or frequent price increases over the next year.

Our results also have important implications for the Bank’s broader understanding and modelling of inflation. For example, the Bank’s main economic models (and indeed most models used in the profession) assume that the frequency of price changes remains constant over time. But this assumption is becoming increasingly at odds with the data. Our results support the alternative models of price determination being explored as the Bank develops its fourth generation of policy models (Coletti 2023).

Going forward, Bank staff will continue to use microdata on consumer prices to track corporate price-setting behaviour and to enhance the Bank’s broader analysis of inflation in Canada.


  1. 1. This staff analytical note fits into a broader body of work at the Bank that seeks to better understand pricing behaviour and inflation during the COVID-19 pandemic. Past analyses have centred on firms’ costs and profits. See Asghar, Fudurich and Voll (2023); Bilyk, Grieder and Khan (2023); and Bouras et al. (2023). Also, see Montag and Villar (2023) for a similar study to ours using US data.[]
  2. 2. We are unable to account for multiple price changes within a given month.[]
  3. 3. In 2018, the Canadian CPI first introduced scanner data as another source of information on prices for grocery products. The use of scanner data allows the CPI to capture the prices paid by consumers at the time of the purchase, increasing the relevance of the data captured.[]
  4. 4. The fact that that the overall frequency of price changes remains elevated likely reflects some combination of behavioural factors and price collection practices. Examples of the latter, in addition to scanner data, include a greater share of prices being collected online since the start of the pandemic.[]
  5. 5. This relative frequency indicator is plotted only on a 12-month basis because the 3-month version is extremely noisy.[]


Asghar, R., J. Fudurich and J. Voll. 2023. “Firms’ Inflation Expectations and Price-Setting Behaviour in Canada: Evidence from a Business Survey." Bank of Canada Staff Analytical Note No. 2023-3.

Bilyk, O., T. Grieder and M. Khan. 2023. “Markups and Inflation During the COVID-19 Pandemic.” Bank of Canada Staff Analytical Note No. 2023-8.

Bouras, P., C. Bustamante, X. Guo and J. Short. 2023. “The Contribution of Firm Profits to the Recent Rise in Inflation.” Bank of Canada Staff Analytical Note No. 2023-12.

Coletti, D. 2023. “A Blueprint for the Fourth Generation of Bank of Canada Projection and Policy Analysis Models.” Bank of Canada Staff Discussion Paper No. 2023-23.

Montag, H. and D. Villar. 2023. “Price-Setting During the Covid Era.” FEDS Notes, August 29. Washington: Board of Governors of the Federal Reserve System.


Bank of Canada staff analytical notes are short articles that focus on topical issues relevant to the current economic and financial context, produced independently from the Bank’s Governing Council. This work may support or challenge prevailing policy orthodoxy. Therefore, the views expressed in this note are solely those of the authors and may differ from official Bank of Canada views. No responsibility for them should be attributed to the Bank.


We are grateful to colleagues in Statistics Canada’s Consumer Prices and Data Access Divisions, including Chris Li, Matthew MacDonald, Michael Henderson, Bin Hu, Shane Goodwin and Joshua Miller. This work also benefited from helpful discussions with Nicolas Vincent and Oleksiy Kryvtsov.


On this page
Table of contents