Weighting bias and inflation in the time of Covid-19
Weighting bias and inflation in the time of Covid-19: Evidence from Swiss transaction data
The Covid-19 pandemic and the measures enacted to contain it have led to a standstill of public life and a severe downturn of economic activity in many countries, including Switzerland. The measures implemented by the Swiss Federal Council – including lockdowns, mobility restrictions, and social distancing rules – have greatly affected consumer expenditure. In the rapidly evolving crisis, there has been a surge of research on tracking consumer response with high-frequency data. Transactions data from banks and other financial institutions have proved to be particularly fruitful sources and have been used for the analysis of consumer spending by, among others, Baker et al. (2020) and Cavallo (2020) for the US, Andersen et al. (2020) for Denmark, Bounie et al. (2020) for France, and Carvalho et al. (2020) for Spain.
Using similar data on debit card expenditure,1 Figure 1 depicts changes in Swiss consumption patterns since January 2020.
Figure 1 Changes in Swiss consumer spending as measured by debit card transactions
Notes: Cumulative expenditure change across categories of goods and services in Switzerland since January 2020 (normalization month). The vertical dotted line marks the declaration of the “extraordinary situation” by the Swiss Federal Council and start of the lock-down on 16 March 2020.
Two observations stand out. First, during the roughly two-month lockdown period from mid-March to mid-May, consumer spending was severely restricted and different consumption categories were affected differently by the official measures. Non-essential retail outlets and many service industries such as restaurants, bars as well as entertainment and leisure facilities were temporarily closed. Only grocery stores, pharmacies, banks and post offices were allowed to remain open (Eichenauer and Sturm 2020). As a result, while it was virtually impossible to spend money on entertainment or personal services, consumption expenditure for groceries has increased significantly before and during the lockdown period.
Second, expenditure categories recover fairly heterogeneously after the lockdown. Expenditure on personal and professional services spiked at the end of April, when a first opening step was taken. “Other Retail” (including garden stores, clothing and furniture) is stepwise retracing the two relaxations steps of 27 April and 11 May. While most categories are at least partially recouping their losses, “Accommodation & Food” and “Transport” are recovering only very slowly, and currently still well below their levels at the beginning of the year.
Inflation with Covid consumption
These sudden and profound changes in consumption can introduce significant bias in the consumer price index (CPI) used to measure inflation. The CPI is compiled on the basis of expenditure weights that are kept constant within a given year, reflecting the purpose of the index to measure changes in prices only without accounting for adjustments in consumption patterns. Most national statistical offices update their CPI expenditure weights once a year, often with lagged expenditure data. While this practice is reasonable in normal times, it makes inflation indices much harder to interpret during the Covid-19 pandemic (Lane 2020, Tenreyro 2020), as the underlying weighting scheme is no longer representative of what is being consumed or what can be consumed at all in the lockdown period, thus introducing a weighting bias in inflation.
In a recent paper (Seiler 2020), I study the effect that biases induced by such changes in spending patterns have on the measurement of inflation in Switzerland during the Covid-19 crisis. For this purpose, I combine estimates of consumer spending based on debit card transactions with official CPI data. In particular, I match debit card transaction categories with CPI expenditure items, update CPI basket weights with transaction-based expenditure changes and compute an alternative price index based on such “Covid weights”.
Figure 2 compares the annual inflation rates of the resulting Covid price index with the Swiss consumer price index.
Figure 2 Swiss consumer price index (CPI, not seasonally adjusted) and the Covid index construed using estimates of the consumption expenditure shares during the Covid-19 crisis
Notes: The vertical dotted line marks the declaration of the “extraordinary situation” by the Swiss Federal Council and start of the lock-down on 16 March 2020.
CPI inflation was low but relatively stable in the beginning of the year, before it contracted strongly during the lockdown period. It fell by -0.4% in April 2020 and was at -1.1% compared with the same month of the previous year. The Covid price index, by contrast, was consistently higher during the crisis. It contracted by -0.1% only in April 2020, and Covid inflation stood at -0.4%. Due to the Covid-induced weighting bias, the CPI hence underestimated inflation during the lockdown period. This result is consistent with various contributions2 to the current debate on inflation and potential biases of it during the Covid-19 crisis. It is a consequence of the relative weight shifts, and particularly driven by the relative increase in consumption of “Food & Non-Alcoholic Beverages”, which are more inflationary than other spending categories.
Inflation with low-touch consumption
Beyond, persistent changes in consumer behaviour are likely to keep underestimating inflation in the short to medium term. While most of the changes in consumer spending during the crisis are obviously due to the containment measures and forced closures of many retail sectors, it is still likely to be a long and difficult task to return to pre-Covid consumption patterns. Consumer behaviour may have changed persistently during the lockdown experience, and for quite some time be driven by ‘low-touch’ considerations. This ‘low-touch’ consumption is characterised by the continuing uncertainty about the spread of the virus, changed working habits and the lockdown lifestyle. If teleworking, for example, is widely maintained, it will create more opportunities for eating at home and reduces eating out, which in turn increases spending on food and beverages and reduces expenditure in restaurants and canteens. Similarly, spending on hospitality, tourism and traditional entertainment activities may remain subdued for fear of contagion, because of the great uncertainties surrounding the spread of the virus and because of newly acquired leisure activities that are compatible with the maxim of social distancing.
In Seiler (2020), I examine the effects of such persistent changes in consumer spending on inflation in the short to medium term. I do so by producing inflation forecasts based on disaggregated CPI data, which I aggregate using two different weighting schemes: the official CPI expenditure weights, and the expenditure weights implied by the ‘low-touch’ consumption scenario described above.
Figure 3 Comparison of inflation forecasts with official CPI weights and with ‘low-touch’ consumption expenditure
As depicted in Figure 3, CPI inflation is projected to increase gradually from its low in May 2020 (-1.3%) but will remain considerably below 0% well into 2021. For 2020 an average inflation rate of -0.7% is predicted. In contrast, inflation under the ‘low-touch’ scenario is persistently higher, and the average annual inflation in 2020 is -0.4%. Without accounting for potentially lasting changes in consumer behaviour driven by low-touch considerations, there is thus a risk that the CPI underestimates inflation by more than a quarter of a percentage point through to the end of 2020.
It is difficult to say to what extent the presumed ‘low-touch’ consumer behaviour will be permanent. The extent to which consumer behaviour will change in the medium to long term is likely to depend to a large degree on the further course of the pandemic, as well as on how the new work experiences are integrated into existing work habits and consumers’ assessment of their future prospects.
Nevertheless, in light of the fact that the CPI is an essential tool for economic policymaking, my results have important implications for the crisis period and beyond. They provide evidence that conventional price measures have underestimated inflation during the crisis. By unveiling this, I hope that my results contribute to the assessment of inflation in times of economic turmoil. Beyond the crisis, they raise conceptual issues concerning adequate price measurement. While most of the changes in prices and consumption expenditure can be expected to reverse once the crisis is over, some may be more persistent. Considering current calculation methodologies, this can make official CPI measures less informative, in particular during the transition period. In response to this challenge, the use of high-frequency and alternative data sources on both prices and consumer spending may therefore become key to development of more robust and informative price measurement tools in the future.
Andersen, A L, E T Hansen, N Johannesen and A Sheridan (2020), “Consumer responses to the COVID-19 crisis: Evidence from bank account transaction data”, Covid Economics 7: 88-114.
Baker, S R, R A Farrokhnia, S Meyer M Pagel and C Yannelis (2020), “How does household spending respond to an epidemic? Consumption during the 2020 COVID-19 pandemic”, NBER Working Paper 26949.
Bounie, D C Y and J W Galbraith (2020), “The COVID-19 containment seen through French consumer transaction data: Expenditures, mobility and online substitution”, VoxEU.org, 26 May.
Carvalho, V, J R García, S Hansen, A Ortiz, T Rodrigo, J V Mora and P Ruiz (2020), “Tracking the COVID-19 crisis through the lens of 1.4 billion transactions”, VoxEU.org, 27 April.
Cavallo, A (2020), “Inflation with Covid consumption baskets”, NBER Working Paper 27352.
Diewert, E E and K J Fox (2020), “Measuring real consumption and CPI bias under lockdown conditions”, NBER Working Paper 27144.
Eichenauer, Vand J-E Sturm (2020), “Die wirtschaftspolitischen Massnahmen der Schweiz zu Beginn der Covid-19-Pandemie”, Perspektiven der Wirtschaftspolitik.
Jaravel, X and M O’Connell (2020), “Inflation spike and falling product variety during the Great Lockdown”, CEPR Discussion Paper 14880.
Lane, T (2020), “Policies for the great global shutdown and beyond”.
Seiler, P (2020), “Weighting bias and inflation in the time of Covid-19: Evidence from Swiss transaction data”, Covid Economics 35: 96-115.
Tenreyro, S (2020), “Monetary policy during pandemics: inflation before, during, and after Covid-19”.
1 These data are based on weekly transaction volumes of debit cards issues by banks to their customers in Switzerland and include debit card payments at points of sale such as grocery stores or service providers (e.g. hairdressers, restaurants, or petrol stations). They are publicly available as part of the Consumption Monitoring for Switzerland, which is a project powered the University of St. Gallen (Prof. Martin Brown, Prof. Matthias Fengler) and Novalytica together with Dr. Robert Rohrkemper (Distinguished Expert, Senior Data Scientist at Worldline) and Prof. Rafael Lalive (University of Lausanne). See http://monitoringconsumption.org/switzerland.
2 These results are consistent with the analytical argument of Diewert & Fox (2020) who show that a downward bias in consumer price indices may result from current calculation methodologies. Using scanner data of fast-moving consumer goods in the UK, Jaravel & O’Connell (2020) empirically document a spike in inflation in the first month of lock-down. Using official price indices and updating CPI weights in a similar way to this study, Cavallo (2020) finds comparable results for the US but overall mixed international evidence.