The COVID-19 pandemic has led many countries to implement social distancing, lockdowns and travel restrictions, which have resulted in a collapse in the world economy unprecedented in peacetime. The real-time effects of the ‘Great Lockdown’ on employment and consumer expenditure have been widely documented (e.g. Bartik et al. 2020, Chetty et al. 2020, Villas-Boas et al. 2020). However, much less is known about how the crisis is impacting inflation.
In a new paper (Jaravel and O’Connell (2020), we use live scanner data to document patterns of inflation. We hope our approach can serve as a template to facilitate rapid diagnosis of inflation risks during economic crises, leveraging scanner data and appropriate price indices in real time.
The Great Lockdown entails a combination of substantial shocks to both demand and supply (e.g. Brinca et al. 2020, Guerrieri et al. 2020). It is therefore plausible that the crisis may lead to deflation, disinflation, or higher inflation. Falling aggregate demand, due to heightened uncertainty and reductions in incomes and liquid wealth, may lead to deflationary pressures. Conversely, inflationary pressures may arise from increases in production costs, due to interrupted supply chains and to the impact of social distancing restrictions on labour supply. By shutting down some sectors of the economy, the Great Lockdown may lead to changing patterns of demand that translate into shifts in the degree of market power firms exercise, which will affect equilibrium inflation. These pressures will differ across sectors, and therefore it is likely inflation also will. Sectoral inflation heterogeneity in turn is likely to feed through to heterogeneous inflation experiences across households.
Accurate and timely measurement of inflation is key for the design of policies aimed at paving the way for the recovery – both for central bankers in charge of maintaining price stability and for policymakers in charge of the design of transfer programmes aimed at mitigating the effects of the economic shock on vulnerable groups.
Scanner data, in which a large sample of households record purchases of fast-moving consumer goods, provide a way of tracking what is happening to prices in an important sector of the economy. Such data have a number of key advantages for tracking inflation over the crisis, including that they contain millions of transaction prices, up to date expenditures weight, and information on households’ socio-demographic characteristics.
In Jaravel and O’Connell (2020), we leverage real-time scanner data for the UK from Kantar FMCG Purchase Panel to track how prices have changed during the Great Lockdown among fast-moving consumer goods (i.e. grocery products, including food, alcohol and non-foods). We use information on over 30,000 households and 13 million transaction in 2020 up until mid-May, as well as data over the same period in the years 2013 to 2019. We compute price indices which are grounded in economic theory and allow us to measure inflation across product categories and households. We establish four findings.
First, we document a large spike in inflation. In Figure 1 we show cumulative monthly inflation over the five months running to 17 May for all years from 2013 to 2020. In the first three months of 2020, month-to-month inflation is close to zero and similar to previous years. However, in the month 18 March to 17 April, there is a large increase in inflation of 2.4 percentage points. In the month 18 April to 17 May, there is modest deflation, though prices remain well above their pre-lockdown level.
Figure 1 Cumulative monthly inflation
Notes: Computed based with a chained Fisher price index using Kantar FMCG Purchase Panel. Source: Jaravel and O’Connell (2020).
This inflationary spike is unprecedented across all comparison years and constitutes more inflation than normally occurs in a year. We show that the increase in prices mainly happened in the first week of the UK’s lockdown (which began on 23 March 2020), and that a key driver was a reduction in the fraction of promotional transactions as retailers cut back on both price promotions and quantity discounts. This fall in promotions contrasts with the Great Recession, during which consumers purchased more on sale (see Griffith et al. 2016 for evidence in the UK, and Nevo and Wong 2019 for the US).
Second, we show that declining product variety strengthens inflation. Typically, inflation between two successive periods is computed by comparing the prices of products available during both periods. However, consumers’ effective cost-of-living is also impacted by the removal or entry of new products; all else equal, if less products are available consumers will be worse off. In Figure 2 we show the evolution in the number of unique products purchased per week in 2020 and in preceding years. Prior to the start of lockdown, and similar to previous years, the number of products sold in each week is stable. However, from the beginning of lockdown, there is a fall of around 8% in the number of products we observe purchased. This points towards a reduction in product variety, which erodes consumers’ effective purchasing power.
Following Feenstra (1994), we show, based on CES preferences, that this reduction in product variety is equivalent to 85 basis points of additional inflation, compared with prior years where product variety was expanding instead of shrinking. Overall, once we take account of reduced product variety, month-to-month inflation in the first month of lockdown increased by over 3 percentage points relative to the same month in prior years.
Figure 2 Product variety
Notes: Number of unique UPCs purchase per week, normalized by mean value in first four weeks. The red vertical line denotes the first week of lockdown. Computed using Kantar FMCG Purchase Panel. Source: Jaravel and O’Connell (2020).
Third, we document heterogeneity in inflation rates across households. Individual households will experience price changes differently depending on the products they choose to buy. As we observe the same households through time, we are able to compute household-specific inflation rates. In Figure 3, we show the distribution of household-specific inflation rates over the period 18 December 2019 to 17 May 2020. We also plot the distribution over the same period in the preceding year. In both 2019 and 2020, there is substantial variation in household-specific inflation: the standard deviation in 2020 is 1.7 percentage points, rising from 1.5 in the preceding year. However, in 2020, almost all households experience inflation; in contrast, in 2019, fewer than half of households had positive inflation rates.
Figure 3 Household-specific inflation rates
Notes: Distribution of household-specific inflation rates computed with a fixed base Fisher price index using Kantar FMCG Purchase Panel.. Source: Jaravel and O’Connell (2020).
Fourth, we investigate inflation heterogeneity across product categories. The scanner data also allow us to measure product category specific inflation rates. In Figure 4 we show the distribution of inflation rates in the first month of lockdown and how they compare with the same month in 2019; the distribution of inflation rates across product categories has shifted rightwards compared with 2019. In the first month of lockdown 13% of product categories experienced deflation, while over half of categories did over the same period in the preceding year. In addition, the variance in category specific inflation rates has increased, consistent with the fact that different sectors were impacted by different shocks. A natural hypothesis is that increased inflation may be driven by a few categories for which there has been a large increase in demand. We show, however, that there is increased inflation across many categories, including those for which output has fallen. The category-level average inflation rate is 3.2% both for categories with increases and decreases in output.
Figure 4 Product category inflation rate, 18 March to 17 April 2019 and 2020
Notes: Distribution product category specific inflation rates computed with a chained Fisher price index using Kantar FMCG Purchase Panel. Source: Jaravel and O’Connell (2020).
What lessons about the dynamics of inflation can be drawn from these findings? Lockdown coincided with unusually high inflation, which was experienced by almost all households and in almost all product categories. This finding is noteworthy given financial markets expect the COVID-19 pandemic to be a disinflationary shock (Broeders et al. 2020). The pervasive nature of the inflation, along with the fact that it is observed even in product categories with declines in output, point toward a risk of stagflation.
It is naturally too early to say for sure whether persistent stagflation will materialise. While the higher price level has persisted for several weeks, the inflation spike coincided with a one-time event, the beginning of lockdown; in addition, we do not observe the entirety of households’ consumption baskets (e.g. rents and services are not included). Nonetheless, it is crucial for central banks, fiscal authorities, and statistical agencies to closely monitor inflation risks going forward. Our work highlights the advantages of real-time scanner data for this purpose. One can track changes in spending patterns for disaggregate products in real-time and observe changes in promotion activity and product variety, all of which are important drivers of inflation and are typically overlooked by statistical agencies.
Bartik, A W, M Bertrand, Z B Cullen, E L Glaeser, M Luca, and C T Stanton (2020), “How are small businesses adjusting to covid-19? early evidence from a Survey”, NBER Working Paper 26989.