The Covid-19 pandemic presents a unique problem to economic forecasters — one that’s uniquely urgent to solve. As consumers’ needs, attitudes, and behaviors continue to rapidly change, how can forecasts accurately adapt to reflect them? The key is to build consumer sentiment into economic forecasting models. To do this, economists must marry credit- and debit-transaction data with ongoing qualitative data on consumer sentiment. This article presents an effective example of a model like this, which accurately predicted the spending habits of consumers in hard-hit sectors between April and June of 2020.
In these difficult times, we’ve made a number of our coronavirus articles free for all readers. To get all of HBR’s content delivered to your inbox, sign up for the Daily Alert newsletter.
As the U.S. began to shut down in mid-March in response to the coronavirus pandemic, economic forecasts were thrust into the spotlight. Second only to public health concerns, the question of what happens next to our jobs, businesses, and the economy at large have become all-consuming for a country on edge.
The pandemic presents unique problems when it comes to forecasting. As noted by FocusEconomics, a provider of economic analysis that surveys hundreds of economic experts from the leading banks and think tanks, the spread among economic growth forecasts for Q2 in the U.S. grew from 3.5 percentage points in early February to a staggering 56.8 percentage points by late April. While most economists expect to see a GDP rebound in the second half of the year, how comfortable can policy makers be with Q3 and Q4 projections given the divergence in Q2 forecasting? In late April, the Fed’s economists offered a baseline projection that showed Q3 and Q4 improvement accompanied with the caveat that “a more pessimistic projection was no less plausible than the baseline forecast.” Statements like this don’t exactly inspire confidence.
There is a monumental need for timely and reliable economic forecasts. In lieu of them, policymakers and businesses are increasingly turning to real time alternative data sources, like mobility data from Apple and Google, or credit card transaction data. While valuable, these data sources lack important context and connectivity to larger economic trends. (For example, credit card data does not capture the impact that this pandemic is having on cash transactions, and mobility data does not explain whether consumers are spending more or less when they reach their destinations.)
Even so, reliable forecasting is possible right now. The missing ingredient is an enhanced understanding of consumers’ changing needs, attitudes, and behaviors. With this in mind, we’ve developed an approach that renders consistently accurate forecasts, even in uncertain times, by combining forward-looking consumer sentiment with real time transaction data.
Start by understanding consumer sentiment.
Personal consumer expenditures account for up to 70% of GDP. If consumers do not resume their pre-Covid spending, or if they change their purchasing patterns, the impact on the individual economic sectors and the overall recovery will be significant. The only way to improve the reliability of economic forecasts in this pandemic is to augment models with an ongoing evaluation of consumers’ past, current, and anticipated spending patterns.
With most of the country living under varying levels of restrictions for the last several months, the habits forged through social distancing and prolonged fears of infection will create long-lasting shifts in consumers’ buying behaviors. For instance:
- Consumers have had more than enough time to experiment and form new habits with remote work and digital commerce that will last long after the states reopen.
- As we emerge from lockdown, and as social distancing restrictions continue to change, consumers will continue to experiment with new digital commerce behaviors, so some habits will continue to form and evolve.
- Without a near term medical solution, even after the states reopen, some consumers will continue to self-impose social distancing to protect themselves and those around them.
There is no doubt that the combined impact of these changes will impact the shape and rate of the economic recovery.
Next, turn to transaction data.
In this constantly changing environment, economists and business leaders must gain clarity on consumers’ future behavior by analyzing credit and debit transaction data in concert with forward-looking consumer sentiment.
Leveraging a database of 500,000 credit and debit accounts with weekly access to transaction patterns, we are able to focus on expense categories that are most likely to be impacted by near- and longer-term changes in consumer behavior, such as Travel, Restaurants, Recreation, Retail, Groceries, Home Improvement, and Gas. Combined, these categories account for approximately $4 trillion in 2019 consumer expenditures or 30% of total Personal Consumer Expenditures.
To gauge future spending, we engage large and representative samples of U.S. consumers in continuous online interviews. We gather information on how consumers intend to buy under multiple realistic scenarios that include the timing and extent of reopening conditions. Since conditions are in a constant state of flux, these scenarios are updated every month to reflect new information, as it becomes available to the public.
Our forecasting establishes a connection between past transaction patterns and consumers’ anticipated spending in the coming months. From April to June we have conducted 9,000 consumer interviews and have used this approach to accurately predict the rate of recovery across the most severely impacted sectors like Travel and Restaurants.
One of the unique forecasting challenges in this environment is the impact that the pandemic has had on consumers reducing their cash expenditures. As a result, relying exclusively on credit and debit data is more than likely to paint an overly optimistic picture of the recovery, particularly in sectors where cash transactions are more prevalent, like Restaurants.
By combining past credit and debit transaction patterns with consumer feedback that captures anticipated cash, credit and debit expenditures, economists and businesses can vastly improve the quality of their forecasts. For example, in May, our forecast predicted a 13% improvement in Retail sales, while the rest of the industry anticipated only a 9% advance. In the end, U.S Commerce Department reported a record 17.7% improvement in May, placing our forecast within 5% of actual outcome.
So, what does the future look like right now?
With many states reporting a rise in Covid infections at the time of writing, no one can fully anticipate the long-term impact of this pandemic. However, by gauging consumer reactions to realistic and hypothetical reopening scenarios, we get a glimpse of the pent-up desire for a return to normalcy. Under a condition where all sectors are fully reopened, our data shows immense consumer appetite to resume spend even in the hardest hit categories, like Travel.
Continue measuring through the crisis.
Nothing about the current situation is static. As new information about the virus, constraints for reopening the various sectors of the economy, and the impact of Covid-19 on consumers’ finances come to light, our society will continue adjusting to the new normal. Given the unique nature of this crisis and the potential lasting impact of social distancing on consumption habits, continuous measurement of anticipated spend is an important part of reliable economic forecasts. By linking actual credit and debit transaction data with consumer sentiment that approximates future spend, analysts can create a sound foundation for forecasting what happens next.
If our content helps you to contend with coronavirus and other challenges, please consider subscribing to HBR. A subscription purchase is the best way to support the creation of these resources.