C-19: Where is the economy headed? Studying the role of small businesses in the recovery

Updated: Jul 7

There have been several initiatives in the US around collating relevant and critical social, economic and healthcare data to help stakeholders manage the COVID-19 crisis. One such area in need of close attention is that of economic recovery from the destructive effects of harsh lockdowns that were put in place around mid-March. The hope is that with relevant data in hand, decision makers, the general public and other relevant parties would not only be able to closely track the recovery process, but also better formulate the right set of actions necessary to ensure optimal recovery.


One initiative example is tracktherecovery.org. It provides a wide range of datasets particularly around small businesses; such as store openings, store revenues, unemployment stats, and consumer spending data. The data is often provided on a daily cadence and broken down to the city and county level, which is very useful to study impact to a granular level.


In this article we explore the data they provide in detail. With the hope to uncover useful insights about the current and future state of the US economy, as the country pursues their recovery. In doing so, we will hopefully establish a credible baseline that might be helpful to other countries whom may be exploring the merits of data collection and analysis around their own economies as they chart their own pathways to avert a further economic crisis from the pandemic.


#COVID #analytics #economy


Background & introduction

Our previous article looked at macro level spending patterns across major sectors in the US, during and shortly after lockdowns. We go deeper this time with the data supplied from tracktherecovery.org, by looking specifically at data around small businesses and low income jobs; to quantify the impact that business closures and reduced mobility due to COVID-19 has had on them. While we will focus primarily on national US data in this article with some focus at state level; almost all datasets covered also provide city and county levels too. We will leave that study for another time.

Reference legend for terms used in the charts that follow;

  • Spend_: spend data from Affinity Solutions. Legend: acf-accom. & food service; aer-arts, entertainment & recreation; apg-general merchandise & apparel; grf-groceries & foods; hcs-healthcare & social; tws-transport & warehousing; inc_ prefix indicates zip code distribution by income levels*.

  • GPS_: mobility data sourced from Google. Legend: self explanatory.

  • Merchants_*: small business openings/closings sourced from Womply.

  • Revenue_*: small business revenues sourced from Womply.

  • Emp_*: employment level for low income workers sourced from Earnin and Homebase.

  • Pay_*: earnings by low-income workers sourced from Earnin and Homebase.

  • by_posts: average level of job postings relative to baseline sourced from Burnglass.

  • _ss30: manufacturing sector (NAICS)

  • _ss40: transportation sector (NAICS)

  • _ss60: professional and business services

  • _ss70: leisure and hospitality

  • _pay/emp44_45: pay or employment in accommodation and food services

  • _pay/emp48_49: pay or employment in transportation and warehousing

  • _pay/emp72: pay or employment in accommodation and food services

  • Note (*): inc_ indicates location of store by US zip code, i.e. high income indicates high income zip code.

Overall statistics sourced for US from before lockdown to present day. Click here for an interactive version of this chart.

As above but with focus on consumer spending trends

As above but with deeper focus on store opening ("merchants"), store revenue, employment & jobs, and pay / salaries. Categorisation here refers to select sectors of the marketplace.

% deviation in overall spending from baseline across US states, compared to number of total cumulative deaths from COVID-19 [snapshot 10 June]

% deviation in overall spending from baseline (y-axis) compared to % deviation in retail mobility from baseline across US states (x-axis) [snapshot ~week 2 June]

Weekly datasets on job postings for manufacturing sector (NAICS: ss30), professional services sector (ss60), and leisure & hospitality sector (ss70) vs. initial unemployment claims per 100 people.


General observations

  • Introductory note: We split these observations across the supply side and the demand side. The former refers to business/store openings, business revenue receipts, (un)employment, and pay; while the latter, the demand side, refers to consumer spending and physical mobility.

  • On the demand side, consumer spending across sectors has varied widely through- and after- the lockdown period; consistent with our previous findings. Entertainment, hotels & food service/restaurants, and transport spending still remain low today; despite the fact that overall spending has risen in tandem with mobility since reopening. Particular segments that have gained strong momentum include general merchandise and apparel, as well as groceries. In fact grocery spending remained high throughout lockdown and peaked during the panic buying period of the first 1-2 weeks.

  • During the lockdown overall consumer spending in high income zip codes bottomed out at -36%, and is now still at -17% from baseline. This is lower than the equivalent levels of middle and low income zip codes, which are down only -10% and -4% respectively from baseline today. The inability to 'shop physically' seems to have inhibited the rich from spending more (or equivalent to previous levels), possibly due to restrictions on travel & tourism, dining out and an overall negative sentiment affecting purchasing of luxury goods and services.

  • Spending decline does not seem well correlated to COVID-19 spread, at least in terms of total deaths from the virus - take for example, Oregon with almost no deaths but a spending decline at -18% from baseline. There is also a slight skew towards Republican states having incurred less of a spending decline compared to Democrat ones, consistent with the attitude of the party towards faster and more aggressive reopening of state economies championed by president Trump. Regardless, there is an extremely wide variation of spending decline from state to state; while spending levels are some 4% higher than pre-lockdown baseline for Indiana and Tennessee, they are more than -20% down for California and Rhode Island. A primary driver of the variation between states is the mobility levels, which is expected. Tennessee retail mobility is now +5% from baseline, while California is some -21% below; an almost 1 to 1 correlation between the two factors.

  • Mobility outside of the home has generally increased steadily since gradual reopening around early April, with park mobility increasing the most from initial baseline, followed by grocery & pharmacy, then by retail & recreation. Transit station mobility still remains the lowest at some -40% from baseline at the national level.

  • On the supply side, some 15-20% of small businesses are still shuttered as at mid-June, affecting more distinctly, businesses in higher income areas compared to lower income ones. Accordingly, revenues of those businesses in high income areas are still some -30% from baseline levels, while those in low income areas only some -13% down. Again, we observe spending among the rich is struggling to bounce back.

  • Pay and employment for lower income workers do not seem perfectly in sync with the trend of mobility levels - instead, they show a distinct lag both in decline and recovery. This could well be a reporting issue. However we believe it is indicative of the fact that businesses simply take some time to recalibrate their operations to the reality of their operating environment. Employment reduced to its lowest levels some 6 weeks after mobility reached its lowest point (early May vs. mid March), indicative of how employers may be taking a relatively lagging view towards making changes to their business after an initial gauge of the impact from the environment.

  • Presently, national level pay for lower income workers is -36% from baseline, while employment is down -38%. Again, workers in high income zip codes are more deeply affected compared to low income ones, reflecting the same trend as revenue generation and openings on the business side. These levels are quite shocking considering that overall spending today is only -15% from baseline.

  • Weekly unemployment claims peaked at 4% of the total workforce at initial stage after the lock down, equating to some 6mn workers. This is now at ~1% but showing suggestive signs of a plateau at this level. The reader should note that this figure represents the initial claim during that particular week, and the cumulative number would be far higher as it compounds with time. By early June the total insured unemployment claims stood at a total of some 21mn people according to the Dept of Labour.


Highlighted key macro trends across factors

Cross correlation from mid January to 9 March (before lockdown)

Figure: cross correlation from 19 March to mid June (reopening period)

As above, but with alternative categorisation in store openings, revenue receipts, employment and pay


Additional observations from these figures:

  • Note: Cross correlation was computed across four major groups of data: (1) consumer spending data, (2) small business (merchant/store) opening and revenues, (3) employment and pay for low income workers, and (4) mobility/people movement trends.

  • Before lockdown (up to 9 March): Cross correlation data seems more erratic, we believe due to; (1) the limited span of data available, and (2) relatively lower levels of perturbation from baseline levels. Regardless, some mild relationships can be gleaned from the pre-lockdown period, although most are as expected. For example, consumer spending patterns are seen to be well correlated to mobilities in retail and groceries. Also, small business store openings and revenue receipts are also positively correlated to consumer spending.

  • After lockdown (from 19 March): as mobilities in retail, transit stations and workplaces have gradually increased, so too has overall spending. In particular, spending in general merchandise & apparel, and accommodation & food service have improved significantly coinciding with increases in mobility within the retail & recreation, groceries & pharmacy, as well as transit stations categories.

  • It turns out that the particular mode of transport has a well defined contribution on consumer purchase behaviour. There is strong correlation between transit hub mobility and overall spending, especially in the restaurant, general retail and groceries segments with ~80% cross correlation. And so as the country reopens, the choice of mode of transport consumers take plays a role in how quickly and where the economy gets reinvigorated.

  • On the other hand, store opening and revenue receipts are moving in lockstep with consumer spending (correlation ~80%); suggesting that accelerating physical store openings is key to encourage more spending to take place (vs. alternative 'online' based channels of retail spending where shops could in theory remained closed to walk-in trade activity).

  • As we saw earlier, employment and pay for low income workers has lagged behind changes in spending and store opening - in fact these have yet to show as distinct an upswing since reopening. Correlation between spending and employment & pay is in fact relatively low at <50%, and may be a worrying sign of things to come as jobs could in general continue to lag in spite of spending picking up.


Describing the economy during this recovery process; an abridged narrative

How would we best describe what has happened across the economy? A short narrative could read as follows;

  • Restricted movement through social distancing and lockdowns put in place to stem COVID-19 spread have effectively reduced the ability (and to an extent, muted the demand) for consumers to purchase products and services from small businesses. Hardest hit have been those operating in the hoteling, restaurant, and entertainment sectors.

  • Due to reduced footfall, and instruction from authorities for temporary closures, many businesses have seen their revenues decline significantly; and some may have already been driven out of business. Their workers may have been furloughed, but many too have also lost their jobs entirely; raising unemployment claims significantly.

  • As the economy reopens, the hope is that much of this will be reversed. However the question of if and when still remain, and clearly applies differently from sector to sector. Some are showing stronger signs of recovery, for example casual dining and fast food, retail merchandise and apparel, however others continue to struggle and are unlike perhaps to recover in the next few quarters, e.g. the hotel, fine dining, travel sectors.

  • It is important to distinguish between tactical observations and more structural changes when assessing what is happening. Closure of some retail businesses and a sharp decline in public transportation use is well expected through a lockdown. These are expected to be temporary and should return to normal levels in time. This is therefore a tactical response.

  • A structural shift is perhaps best described by what is happening to the hotel and hospitality sector where it will take far longer for tourism and travel to return to previous levels due to heightened apprehension among the consumer base, in addition to restrictions on travel and social interactions that have remained in place.

  • The impact from state to state has been devastating with nationwide spending now trailing at -15% to baseline levels. While it is assumed that the impact is consistent with the extent of COVID-19 spread, that has not been the case. The stringency of lockdowns and the reduced mobility that comes with it has been the single biggest driving factor of economic decline.


Discussion, and some general ideas to boost the economy

  • Spending by higher income consumers has to increase if the economy is to recover. Crucially, spending on accommodation/hotels and food service/restaurant dining need to be unlocked further and faster, as these could be key areas of spending for these consumers. As it stands, spending here seems to be slow to recover. Additionally, travel spending, including air travel and cruises, as well as arts & entertainment have highest contribution historically from these consumers in these (higher spending) zip codes - which has dipped greatly due to restrictions on travel and movement.

  • Boosting transit station mobility could bring about greater spending in the small business economy, given strong observed correlation. Notably, for restaurants & food services, as well as transportation as presumably most commuters would also spend on food on the go, and last mile ride hailing services.

  • Store (re)opening is just as important as mobility increase amongst the population. It is observed through the lockdown period that mobility changes down to lowest levels preceded that of spending and store revenues, followed later by store closures. More stores opened would mean more ability for the general public to shop.

  • Job losses will be structural, and therefore require major initiatives to solve for. Sectors such as travel, hospitality and food service (restaurants) will not recover to previous levels any time soon. The job losses have and will continue to be devastating.

  • Better balanced coordination between federal and state level is needed. The issue is that while the pandemic is indeed a national crisis, the challenge of balancing COVID-19 infections and deaths, with the devastating economic impacts are a localised i.e. state reality. For large countries where the differences could be great from state to state, too much broad sweeping policy without accounting for the granular differences in regional level realities risks over engineering the response in some states, and under delivering in others. Take for example the national level lockdown in the US - while hard hit New York and New Jersey clearly benefitted from it, one could argue states like Oregon, South Dakota, Nevada, New Mexico, Alaska, Delaware, Montana, who were hit far less badly, suffered dearly from the stunted economic activity arising from it. For the US, the bipartisan divide made matters even worse.

  • From country to country we see visible tension between national level policy and directives, and state level implementation, owing to this challenge. One possible demarcation could be for the federal level to hold policies around the criticality and availability of the healthcare system, while providing tools and guides for other matters which are under the control of the state to determine the best course of action. Preparation should start now for the next crisis - while COVID-19 has been the first global pandemic over the past century, it may not be the last.

  • Highly targeted support to affected industries and workers is needed to temporarily buoy the hardest hit areas until major structural issues are resolved. Additionally, discrimination with respect to states and regions is likely also needed given what we have observed. Many countries have indeed announced and started implement recovery and stimulus plans, ranging from cash handouts, loan moratoriums, business bailouts, credit extensions and the like. While the details are wide ranging from country to country, we duly hope that sharp and implementable discrimination is a key tenet within each.


This article is still a work in progress.

32 views

© 2020 Agility Plus 

Pushing Boundaries Together