Essentially, only two ingredients are required to calculate the inflation rate: a consumer basket containing the goods and services consumed by a representative household and the change in the price of these goods and services.²⁷ However, the problem of isolating changes in the quality of goods from the changes in prices is well known. So is the difficulty of pricing goods and services that, for some periods in 2020, could not be sold because of the pandemic’s restrictions. We will therefore focus our attention on an additional complication: changes in the composition of the representative consumer basket. This basket consists of a wide range of goods and services divided into 12 major groups,²⁸ each with a weighting that reflects its relative importance in the average consumer basket.

Prior to the pandemic, the annual review of these relative weights went largely unnoticed due to the stability in consumption patterns. However, the advent of COVID-19 and the consequent imposition of restrictions completely have altered our consumption habits. The official agencies in charge of calculating inflation found themselves tied to a distribution of relative weights based on a basket that was representative in 2019 but had ceased to be so as of March 2020. So the question we must ask ourselves is: if the official bodies had been able to modify the relative weights of the representative basket to adapt it to the new situation, what would have been the inflation rate for the Spanish economy in 2020?

To answer this question, we have used CaixaBank’s internal, fully anonymised data on card payments at point-of-sale (POS) terminals.Given that the HICP includes consumer spending by residents and non-residents in Spain, we have extracted, from the internal data, the spending with Spanish and foreign cards at all the bank’s POS terminals. First, to ensure the consumer basket resulting from our internal data is representative, we compared the relative weights according to our 2019 data with the official data for the groups of goods and services we can capture using POS card transactions (i.e. those whose payment is not mostly made via transfer or direct debit).³⁰ The result of this comparison clearly shows that the consumption observed via our internal data fits quite well with the consumption patterns of the average consumer, as shown in the following chart.

One year later, in 2020, the pieces of the puzzle have fallen apart and a disparity has emerged between the official relative weights (calculated in early 2020 based on 2019 consumption patterns) and the relative weight distribution shown by our internal data. The goods and services most affected by the restrictions (transport, hotels, cafés and restaurants, and leisure and culture) are the ones whose relative weight decreased the most, being offset by the group of food and non-alcoholic beverages. This redistribution of weights is due to the fact that we are spending much more time at home than in pre-pandemic times, which also explains the slight increase in household expenses. The role played by e-commerce as a cushion against the slump in in person sales in the clothing and footwear group is also evident, as shown in the chart below.³¹

As the basket used by the official agencies to calculate inflation in 2020 gave too much weight to the groups in which prices grew less or even fell sharply,³² and not enough weight to the food and non-alcoholic beverages group (with a 2.4% price increase in 2020), all the evidence suggests that the official inflation series suffered from a downward bias. Consequently, from April onwards and according to our internal data, inflation was higher (and deflation lower in those months when both series show a fall in prices) than the figures published by the National Statistics Institute, as shown in the first chart below. In contrast, this gap between the two series is non-existent during 2019 and much smaller in 2021 once the official agencies updated the relative weights of the consumer basket at the beginning of the year, adapting it to the consumption patterns observed during the first year of the pandemic (second chart below).³³

So far we have analysed inflation in relation to the average consumer but the basket of goods and services obviously varies depending on the demographic group that consumes it. After breaking down our total clients into age group and income level, we have concluded that the inflation rate was higher for the over-60s and lower for 16-29 year olds during 2020 (see the chart below).³⁴ This difference between generations can be explained by their differing consumption patterns: seniors spend a higher proportion of their expenditure on basic necessities (inflationary during 2020) and less on transport or leisure and culture (deflationary), as opposed to younger people.³⁵

Likewise, the chart below shows how the consumer basket became more expensive for consumers with a salary in the lower part of the distribution (10th percentile) compared with the higher income group (90th percentile), something that can be explained by the fact that low incomes concentrate a greater part of their consumption on basic necessities and less on goods and services related to transport, tourism and leisure.³⁶

To conclude, having verified that inflation estimates for Spain in 2020 would have been higher if the data on consumption patterns had included real-time changes, it is important to gauge the potential of high-frequency data in a changing environment such as the one we are experiencing. It is true that events as extreme as the current health crisis, which transformed the structure of spending overnight, do not happen often. However, our world is becoming more and more dynamic, to the extent that having this type of information will become increasingly important, especially for a variable such as the inflation rate.