Grocery shopping in times of crisis: do income and wealth drops make consumers switch to cheaper products?

By Calogero Brancatelli, Adrian Fritzsche, Roman Inderst
Posted on June 03, 2021

How do households adjust their grocery shopping and consumption behaviour in times of economic crisis? The ongoing COVID-19 pandemic has reignited the interest in this question, especially as over the last year, households in many parts of the world have spent an increasing fraction of their income on food (e.g. Cavallo, 2020; Kouvavas et al., 2020). It’s a common hypothesis that households react to (unanticipated) changes in personal income and wealth by switching to cheaper products. Although it is difficult to establish plausible causality with time series data, corresponding analyses suggest large effects (e.g. Lamey, Deleersnyder, Dekimpe, and Steenkamp 2007).

In order to establish causality more rigorously, more granular microeconomic data has been used recently, specifically in form of purchases recorded directly by households through barcode scans and enriched with additional information, such as income and household size (e.g. Dubé, Hitsch & Rossi, 2018). However, these data typically pose three important measurement challenges. Firstly, income information in these data sources is often self-reported, making it subject to potential misreporting. Secondly, as marketing companies typically use information on income as one of several inputs to cluster shoppers or households – and timely updates of household-specific variables are costly – over-time variation of income may be measured less accurately. Thirdly, reported information on income and wealth typically relate only to a single (“main earner”) income source, not reflecting changes in a household’s total purchasing power.

Data innovation

In our current research paper (in cooperation with our co-author Thomas Otter from Goethe University), we revisit the question whether changes in income and wealth (notably during recessions) make households switch to private-label products, which are often substantially cheaper than national-brand products. To overcome the measurement challenges mentioned above, we introduce a novel data set from the Netherlands. As we obtained the explicit consent of panel members, we were able to match household purchases of consumer packaged goods (CPGs) as recorded in GfK’s consumer panel with data from the national statistical office (the Dutch Centraal Bureau voor de Statistiek or CBS). This includes income and wealth data, collected for tax purposes. This data set thus combines the high granularity of household scanner data with precise measures of household income and wealth in administrative data.

To answer our research question, we focus on the period after the financial crisis of 2008. In contrast to some other European countries (such as Germany), the Netherlands experienced a serious decline in house prices, a recession, and a surge in unemployment during the aftermath of the crisis. This translates into significant variation in income and wealth between 2011 and 2018, which is the time covered by our matched data.

Investigating causal effects of income and wealth changes on private-label consumption

Taking a panel perspective, we estimate the extent to which households adjust their spending towards cheaper private-label goods. As we rely on disposable household income as well as wealth variables collected for the entire household for tax purposes, we can minimize the measurement problems mentioned above. In order to isolate the effect of interest from other possibly confounding factors, we also enriched our dataset with information on precisely measured socio-demographic variables, such as household size and the number of income earner, as well as information on price and assortment changes made by retailers close to each household’s location.

As for income, we find that the effect on the private-label share is economically small (and much smaller compared to the literature using only aggregate information), even though it is substantially larger than found in previous studies with a focus on household-level variation. Interestingly, housing wealth alone has no significant effect in contrast to previous studies (e.g. Stroebl and Vavra, 2019).


Our results suggest that though households’ CPG-expenditures account on average for almost 10% of total income and though a switch to private-label products would realize considerable saving potentials, households react only little to changes in income and wealth, at least in the Netherlands. In ongoing work, we confirm this hypothesis, considering also households’ total CPG-expenditures – again after controlling for supply-side changes (Brancatelli and Inderst, 2021).

Our present study is of value for other researchers, as it documents the importance of data quality and measurement considerations, especially when the underlying variations are relatively small (like in the case of individual over-time variation of household income). It may also be of interest to policymakers and industry practitioners, as measurement errors may detrimentally affect conclusions about consumption dynamics of households for this segment of consumption.

Next steps

We are currently undertaking an additional match of the data, extending the covered time period. Analysing again the implications of changes in income and wealth, ongoing projects extend the consumption variable of interest to, for example, products with sustainability attributes. We have also enriched the dataset with nutritional information and plan to test related hypotheses, for instance on health outcomes.

Through our co-operation partner, the AiMark foundation, we plan to make this dataset available to other researchers interested in analysing the interaction between (precisely measured and representative) household consumption expenditures and other variables of interest, as available through CBS. This should also foster interaction between macroeconomics, especially household finance, and marketing science.