Searching the internet for clues to consumer spending
admin | June 5, 2020
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors.
The sudden stop in economic activity in March and early April and the slow revival since has left plenty of investors guessing about the direction of consumer spending. But an inexpensive and accessible window on that spending is becoming clear: the Internet. Internet searches now point to major shifts in consumer spending. That is creating new potential winners and losers across providers of household goods and services and across lenders and investors that fund consumers or that fund businesses that rely on consumers.
Internet search traffic related to major categories of household spending has largely tracked the shifts in activity seen through the pandemic. Search activity has reflected changes in spending on healthcare and housing, transportation and recreation, financial services and spending on food and beverages for consumption in restaurants and at home. Search volume also tracks some key independent surveys of household activity, suggesting searches offer especially valuable information in areas where surveys or other measures do not exist. Some initial implications:
- Household spending in 2020 looks unlikely to rebound to levels of late 2019
- Spending on housing and groceries looks set to run above levels of late 2019
- Spending on transportation, recreation, restaurants, hotels and clothing looks likely to fall well below earlier levels.
A net drop in consumer spending should leave consumer balance sheets and the balance sheets of companies and lenders that serve consumers vulnerable at least through 2020.
The composition of consumer spending, and the impact of coronavirus
Household spending in the last quarter of 2019 accounted for 66% of US GDP. Households spent an annualized $14.4 trillion in a $21.7 trillion economy. The biggest shares of the household wallet went to healthcare (22.0%) and housing (19.0%) (Exhibit 1). Transportation (9.4%), recreation (9.0%) and financial services (8.2%) came next. Groceries (7.3%), restaurants and hotels (7.2%) and other goods and services (6.6%) followed, at least among categories with spending shares above 5%.
Exhibit 1: The composition of household spending in 4Q2019
Spending in the first quarter of 2020 shifted in areas most exposed to social distancing and work-from-home orders that spread across the US in March. Spending on health (-3%) fell as routine doctor and dentist visits and elective surgery stopped (Exhibit 2). Housing expenses rose (+1%) as more people stayed at home. Transportation fell (-8%) as people stopped flying and driving. Recreation fell (-4%) as people stopped going to gyms, movies, theater, sports and casinos, although video streaming rose. Spending on financial services rose (+4%) as people opened more retail investment accounts. Spending on groceries rose (+8%) as spending on restaurants and hotels fell (-8%). Spending on other goods and services also fell (-2%).
Exhibit 2: Spending shifted in 1Q2020 to reflect social distancing and WFH
Most of the shift in spending in the first quarter came in March. If March spending patterns persisted in April, May and June, then the shifts in spending from the last quarter of 2019 to the second quarter of 2020 will be three times as large as the shifts in the first quarter. The 8% drop in restaurants and hotels during the first quarter, for example, becomes a 24% drop from 4Q19 to 2Q20. Anything that can measure consumer spending relative to March can help indicate whether spending will come in above or below March levels.
Based on our review of Internet searches related to major categories of household spending, net household spending looks highly likely to lag levels of late last year through the balance of 2020 (Exhibit 3). Spending on housing looks likely to run slightly above the pace of late last year, and spending on groceries looks likely to run well above late last year. But spending on transportation, recreation, restaurants, hotels and clothing looks likely to fall well below earlier levels.
Exhibit 3: Weaker household spending looks set to outweigh stronger spending
Internet searches track the course of pandemic
Internet searches for terms related to these different spending categories reflect the sudden stop in economic activity in March and April. Google Trends reports search volume over time indexed to 100% at the peak level of activity. Interest in certain categories of consumer activity plummeted as the US responded to coronavirus. In some cases, searches reflect a clear rebound in May. In other cases, searches show little to no rebound.
Searches for healthcare showed a clear drop in March consistent with the drop in first quarter healthcare spending. Activity has started to rebound but remains below levels of late last year.
- Searches for “a doctor near me,” for example, paralleled and ran higher than activity for the year before through late 2019 and into 2020 (Exhibit 4). Searches in March, however, dropped from around 95% of peak to 50% through mid-April. It has since rebounded to around 80%.
Exhibit 4: Searches for doctors dropped in half March but have largely rebounded
- Searches for “a clinic near me,” as another example, also tracked and ran higher than activity the year before into 2020, including a rise around New Year’s, which could reflect holiday illnesses or a rush to get appointments before insurance deductions reset or the window for using flexible spending accounts closed (Exhibit 5). Searches in March then fell from 85% of peak to around 45% by early April. Searches for clinics have rebounded to 60% but is still run well below levels of late 2019 and early 2020.
Exhibit 5: Searches for clinics dropped in half March and have modestly revived
Searches for information about working from home showed a surge in March, consistent with the first quarter increase spending on housing. Interest has since returned to normal. Other searches suggest interest in buying a home has picked up dramatically since March.
- Searches for “work from home” hit their peak in mid-March, consistent with the spread of work-from-home requirements across the US (Exhibit 6). As those requirements have eased, search activity has almost returned to normal.
Exhibit 6: Searches for “work from home” track the wave of WFH policy
- Searches for “buy a house,” after closely tracking levels from a year earlier through 2019 and early 2020, have recently risen to a new peak (Exhibit 7). With the likely prospect of continuing to work at home for a wide range of employees, renters or owners of apartments may be thinking about buying houses. That interest also may partly explain the recent surge in the MBA Purchase Index.
Exhibit 7: Searches for “buy a house” have hit a new peak since March
Searches related to transportation suggest airplane passenger traffic will remain very depressed, but the auto sector may be in the midst of a full rebound.
- Searches for airplane flights using Google Flights ran parallel and slightly above levels from the year before through late 2019 and early 2020 (Exhibit 8). When coronavirus hit, searches spiked as people looked at deeply discounted airfares or searched for flights home and then sank. Before the virus, searches ran around 60% of peak. After the spike, searches have dropped to less than 20% and remain there. The pattern roughly parallels the daily volume of security checks reported by the TSA.
Exhibit 8: Searches for flights have plummeted with no sign of rebound yet
- Uses of Expedia to make airplane, car and hotel reservations also tracked prior year levels through late 2019 and early 2020 before spiking in early March and plummeting (Exhibit 9). Searches before the virus ran around 75% of peak. Interest dropped to 20% of peak in late April and has slowly climbed to 30% lately.
Exhibit 9: Searches for Expedia sank but have slowly climbed higher
- Searches for “car dealers near me” through 2019 and into 2020 followed the seasonal cycle of high interest in the summer, lower interest in the winter and rising interest in the spring (Exhibit 10). When the economy began to shut down in March, search activity sank from 90% of peak to 45% by early April. It has since climbed back above the same level of activity seen the year before and is at 92% of peak activity.
Exhibit 10: Searches for car dealers have rebounded above year-ago levels
Searches for help on savings, investing and trading showed a surge in March consistent with more household spending on financial services. But the interest has waned.
- Searches for “investing,” for example, tracked levels from a year earlier through 2019 and began to rise in early 2020 (Exhibit 11). They picked up even further in March, setting a peak, and returned to the peak in April. Since then, interest has dropped back below 80% of peak but still well above levels the same time last year
Exhibit 11: Interest in investing peaked in March and April before dropping back
Searches for movies, live music or even for things to do in Las Vegas point to a rough road ahead for spending on recreation. This could signal that activity drawing large numbers of people together in close proximity for long periods could take a long time to return.
- Searches for “movies near me” shows a pattern repeated in searches for “live music near me” and for “things to do in Las Vegas.” Search volume in late 2019 and early 2020 again tracked levels of a year earlier, showing familiar spikes around Thanksgiving, Christmas and New Year’s (Exhibit 12). From February to early April, activity dropped from roughly 40% of peak to 5% and has barely recovered to 10% at the end of May.
Exhibit 12: Searches for movies have barely lifted off the post-March bottom
Restaurants, bars and hotels
Searches for restaurants, bars and hotels suggest an ongoing but modest rebound from the lowest levels of activity in March and early April.
- Searches for “restaurants near me” in late 2019 and early 2020 paralleled and ran slightly higher than the year before (Exhibit 13). They spiked around Thanksgiving, Christmas and New Year’s. As pandemic set in, activity fell from roughly 65% of peak to 30% of peak. It has rebounded to 55% of peak. Search data shows more restaurant activity than the OpenTable Index, which tracks the volume of year-over-year seated diners, perhaps because search captures seated, curb-side and delivery business as well.
Exhibit 13: Searches for restaurants slumped as dine-in became take-out
- Searches for “bars near me” ran slightly above the levels of the year before up to March 9 (Exhibit 14). That included spikes in interest each year around Thanksgiving, Christmas and New Year’s. Interest then fell from near 70% of peak to less than 10% through April. It has rebounded through May to 40% but still remains well below its February levels.
Exhibit 14: Searches for bars fell sharply and are still below pre-March levels
- Uses of Hotels.com for hotel reservations in late 2019 tracked the prior’s cycle, dropping from peak summer months to begin rising early in 2020 (Exhibit 15). As parts of the US began to shut down in March, searches again spiked and dropped. From a level of roughly 55% of peak in February, interest dropped by the end of April to 15% and has since climbed back to 35%.
Exhibit 15: Searches for Hotels.com spiked, sank and have climbed partly back
Validating the usefulness of the search data
Where markets have other frequently reported sources of information about household activity, search data may not add much to the picture. But these other sources can help validate that all searches capture important information.
The Mortgage Bankers Association reports a weekly index of applications to refinance mortgages. Search activity for refinancing closely parallels the MBA index (Exhibit 16). Statistical analysis shows the two information series have an R2 of 0.70, meaning they share 70% of their information. In other words, searches for “Refinancing” closely approximate the MBA’s well established survey.
Exhibit 16: Searches for “Refinance” approximate the MBA Refinancing Index
The TSA started releasing information about daily security checkpoint volume as passenger air travel began to shut down in March. Search activity on Google Flights closely tracks the TSA releases (Exhibit 17). The two series have an R2 of 0.80, sharing 80% of their information. Search data captures much of the same thing as the TSA daily checkpoint reports.
Exhibit 17: Searches with Google Flights closely approximate TSA records
The close fit between search data and other independent sources of information suggests that where independent information does not exist, searches provide a valuable window on activity.
Search volume obviously does not directly reflect spending, so the link between search and spending in different areas can be noisy. Search could come well before or well after any purchases, or not result in a purchase at all. Also, every single search carries the same weight, so searches cannot reflect differences in individuals’ spending power. Nevertheless, in current markets where past spending patterns may have little value for estimating future patterns, searches provide accessible, inexpensive and timely indications of the new direction of consumer spending.
Implications for consumer markets
Search activity to date suggests net household spending will lag levels of late last year through the balance of 2020 (Exhibit 18). Spending on housing looks likely to run slightly above the pace of late last year, and spending on groceries looks likely to run well above late last year. But spending on transportation, recreation, restaurants, hotels and clothing looks likely to fall well below earlier levels.
Exhibit 18: Weaker household spending looks set to outweigh stronger spending
A likely decline in net consumer spending should put a drag on growth and employment, leaving consumer credit and the credit of companies serving consumers vulnerable through the balance of the year. But shifts in household spending will likely not fall equally across all businesses, their employees, their providers and funders. Businesses getting a higher share of the consumer wallet—grocery stores, realtors, home improvement providers and the like—should do better than late 2019 business. Businesses getting a lower share—airlines, movies, live entertainment, restaurants and hotels, clothing and related retail—should see much weaker results.