Sinem Hacioglu Hoke, Diego Kaenzig and Paolo Surico
The response to the Covid-19 pandemic has included closure of retail outlets and social distancing. How large was the resulting consumption fall in the UK? In a new paper, we try to answer this question using a transaction-level dataset of over 8 million individual transactions. This gives a near-real time read on consumer spending, without the publication lags associated with national accounts consumption data. We find that the bulk of the fall had occurred before legally mandated lockdown started. The largest declines occurred in retail, restaurants and transport, but spending on some items such as online shopping, alcohol and tobacco rose. There is substantial variation in change in consumption across age, income group, housing tenure and local authority.
From individuals to a real time consumption measure
Our starting point is individuals’ transactions. Especially in the times of crisis, spending data present a valuable way of taking a peek at how consumer behaviour changes. We contribute to a burgeoning literature exploiting transaction level data to analyse consumption changes during the pandemic as in Baker et al (2020) for the US, Carvalho et al (2020) for Spain and Andersen et al (2020) for Denmark.
We use an anonymized transaction level dataset provided by Money Dashboard (MDB), a free online personal financial management company operating in the UK. Their main product is an app that gives its users the flexibility to link multiple accounts (current, savings or credit card accounts) and provides them with a set of tools for categorizing and keeping track of their spending. We focus on the time period between 1 January to 26 April 2020. This provides us with over 34,000 users who have consistently used the app over this period and overall more than 8 million transactions. The MDB user base is skewed towards higher income and younger individuals so the sample may not be representative of the whole UK population. Also analyses based on other data sources on the real time effects of pandemic in the UK might lead to different results due to different data coverage.
MDB has developed algorithms to classify transactions into over 280 spending categories. The granularity of the data allows us to construct weekly expenditure measures not only for non-durables, durables and services but also for a number of more detailed categories such as retail, restaurant, travel and transportation.
Our benchmark total expenditure measure combines different spending categories but excludes recurring expenses such as bills, council tax and more importantly rents so in that respect is different than the National Accounts’ consumption measure. Given that bills and rental expenditures are likely to not adjust much in the short run, our measure likely produces a larger fall. The actual total spending measure is almost real time as it is available with a few days lag and is of paramount importance to track household demand in times of crisis.
Abrupt changes in consumption even before lockdown
The UK government announced the nation-wide lockdown on March 23, 2020 although there were softer measures in place starting from March 15. For example: restaurants and pubs were already asked but not required to close before the lockdown, the public was advised to avoid public transportation, embrace social distancing and work from home whenever they can. All these measures affected consumption ahead of the legally mandated lockdown measures. But the effect of the pandemic on different kinds of spending has differed.
We compute the average weekly expenditure, and then normalize this measure relative to the 2nd week of January. The top panel of Figure 1 shows the average total expenditure in 2020 (blue) compared to 2019 (red, in 2020 prices). The second week of March 2020 records a similar relative level of spending to 2019. But in the third week of March total expenditure declines sharply, with a fall of almost 30%. By the end of April the cumulative falls exceeds 40%. Similarly, measures of nondurable and services consumption record significant falls in the course of March 2020. Most of the decline in consumption, however, occurs starting from the second week of March, before the lockdown. So, the bulk of the fall in consumption comes before legally mandated lockdown measures, suggesting voluntary social distancing makes a bigger marginal contribution than lockdown.
Figure 1: Indices of weekly total, nondurable and services expenditures excluding recurring expenses, second week of January is set to 100
Big differences across expenditure types
Figure 2 takes a more detailed look at different expense types. This time we report actual pound spent per week for different spending categories. On the left panel, retail, restaurant and transportation expenses all decline significantly in March. Although the declines stabilize in April, the levels at which these expenses settle are 25% to 30% lower than 2019 levels. Interestingly, travel expenses dip towards the end of March but return to 2019 levels in April perhaps as individuals preserve the hope of ‘going back to normal’ towards the end of the year and start booking holidays.
Figure 2: Average weekly spending by different categories
On the right panel of Figure 2, unsurprisingly, there is a big spike in online shopping in the second half of March and grocery expenses in the second week of March. However, the initial rise in online shopping and groceries purchases have been reverted to levels close to 2019. We also document the increase in alcohol and tobacco consumption, and DYI/home spending. Although the latter is seasonal and also increases around the same time in 2019, there is a stark difference between 2019 and 2020 levels. Overall, the most affected categories seem to be nondurable and services consumptions, spending in restaurants, retail, and transportation.
Variation across local authorities
Finally, the pandemic did not have the same effect across all parts of the UK despite the nation-wide lockdown. Figure 3 shows a heatmap for the percentage decline in total expenditure from the last week of February to the last week of April 2020 across different parts in the UK darker (lighter) shades represent larger (smaller) declines whereas the light grey denotes areas with insufficient observations. The heat map reveals pervasive heterogeneity in the economic costs both across the country and within broader local authorities, ranging in between −16% to −33% in a few locations in the Midlands, Wales, Belfast and the Scottish Highlands, and −56% to −65% in a handful of areas across Greater London, Hampshire, Berkshire, Surrey and Aberdeenshire.
Figure 3: The decline in total expenditure in the UK
Inequality and heterogeneity
The pandemic has not affected consumption uniformly across the income distribution. While the whole nation experienced a dramatic decline in consumption, the changes in consumption are far from unified across individuals. To explore the inequality the pandemic has given rise to, first, we classify individuals in our dataset according to their income and consumption at each point in time. We have three groups: 90th percentile, 50th percentile (median) and 10th percentile. We also condition on individuals’ February 2020 income and calculate the heterogeneity in savings rates in a similar way. Then in Figure 4, we report these groups’ weekly consumption (top left panel), monthly income (top right panel) and saving rates (bottom panel) in 2020.
There are two important points emerging about inequality. First, all three groups exhibit declines in consumption and income (top panel) over the months of March and April 2020. The 90th and 50th percentiles (shown in green and blue) are not as affected as the individuals at the 10th percentile of the distribution (shown in orange) although the overall decline in spending and income is apparent. Second, the percentage fall in consumption and income is far more pronounced among households in the bottom percentiles shown in orange lines. While there is an uptick in the savings of the richer individuals, the decline in the savings of poor is quite stark. So, poorer individuals are getting more constrained, especially in April.
Figure 4: Log weekly consumption (left) and log monthly income (right) and saving rate (bottom)
In the same spirit, we can explore which individuals’ consumption has taken a big hit with respect to their housing tenure (if they are renters or mortgagors); age (young, middle aged or older); income (high, medium and low income). Figure 5 documents a few interesting findings. First, on the top panel, mortgagors appear to have a higher level of consumption than renters during normal times. But they are also the ones experiencing a relatively more pronounced decline in their consumption. Second, in the middle panel, middle-aged and older individuals sustain a higher level of spending than young individuals. For middle-aged and older individuals, however, the fall in consumption is starker in March 2020. Third, at the bottom panel, while low and middle-income individuals (below £50K) sustain almost half of the consumption of the wealthier individuals during normal times, the wealthy reduce their consumption more than others. In summary, mortgagors, middle-aged to older individuals and higher earners exhibit the largest spending drop in terms of pound value.
Figure 5: Heterogeneity of weekly total pound consumption by housing tenure, age and income
A novel machinery to track consumption in real-time in the UK
Even in tranquil times, it is difficult for economists to track what is happening to consumption. In crisis times, timely indicators of consumer spending can be extremely important to policymakers for understanding the fallout from crises. Our headline result is that consumption fell substantially before the onset of formal lockdown measures, suggesting that the effects of voluntary social distancing or advice can be significant. Looking beneath the aggregate- there are substantial differences across income, housing tenure, geography and income level.
Sinem Hacioglu Hoke works in the Bank’s Monetary and Financial Conditions Division, Diego Kaenzig is a PhD student at the London Business School and Paolo Surico is a Professor at the London Business School.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
I don’t think this paper, nor it’s data; should be behind a paywall. This does not allow us to correctly read it, make use of it with technology or challenge any of the findings.
Is this data available anywhere?
Such an informative blog. Please give my blog a read too.