COVID-19’s Impact on Small Businesses and State Policies
By Shreyas Hariharan, Sophie Lou, Varsha Madapoosi
Author’s Note: We were restricted by Medium to embed our interactive Tableau visualizations. However, we have included the links to all Tableau visualizations below each visual that was created using Tableau. Please utilize these links, as you will be able to directly interact with the metadata included in the visualizations and gain a more granular view as the reader. Thank you!
Overview of Small Businesses in the Pandemic
Accounting for 99.9 percent of all U.S. businesses, there are 30.7 million small businesses in the U.S (SBA, 2019). However, small businesses are no doubt one of the most affected sectors during the pandemic. In early June, once a one-stop shop for everything from stationery to cosmetics to home goods, Daiso in Berkeley closed after nine years of service. This was disheartening and shocking news for a lot of Berkeley students and local citizens. Daiso isn’t the only business that experienced closure. According to Facebook’s State of Small Business Report, as of May 2020, nearly one-third (31%) of small businesses in the U.S. are currently not operational. In our analysis, we wanted to understand how exactly small businesses were negatively being impacted by COVID-19 and find out whether state policies, such as mandating non-essential businesses closing their doors to customers, had any correlation with the negative impact, using data from US Census Small Business Pulse Survey and COVID-19 US State Policy Database.
Overview of Small Business Pulse Survey
The analysis on the impact of COVID-19 on small business used data from the Small Business Pulse Survey by the US Census. According to the US Census, the target population is all non-farm, single-location employer businesses with between 1 to 499 employees and receipts of $1,000 or more in the 50 states, District of Columbia, and Puerto Rico. The Census Bureau invites over 90,000 different businesses to respond weekly, reaching nearly one million small businesses across a 9-week rotation. The survey has 15 questions covering different topics designed to ask about the experience of small business over the last week and since the start of the COVID-19 pandemic.
Process
The US Census had different versions of the survey data broken down by sector, city, state, etc. Since we were interested in the relationship between state policy and negative impact, the 9 datasets broken down by state were used for analysis. After cleaning the data, we began to analyze each question of the survey to gain insights about certain types of negative impacts. We split the questions into five categories: overall impact, revenue, operation, employee and financial issues. We didn’t include every question, but instead listed several key findings from each category. After we finished analyzing the questions, the 9 datasets were reconstructed into a single aggregated dataset with state, date, and 13 metrics from the questions that gave us insights on certain negative impacts. It was then joined with the reconstructed policy dataset for further exploration.
Findings
- Overall impact
The above visualization showed the weekly change of COVID-19 impact on small businesses across 9 weeks. While negative impact (“large negative” + “moderate negative”) remained prominent throughout the 9 weeks and occupied the majority percent, the impact on small businesses gradually shifted from “large negative” to “moderate negative” and the percentage of small businesses experiencing “little or no,” “moderate positive,” and “large positive” impact began to increase. It is apparent that although the overall impact on small business was mostly negative, the overall negative level was gradually decreasing.
To better understand the negative impact, we created a function to calculate the negative impact level for each state as follows: Negative impact level (out of 100) = [large negative] + [moderate negative] — [large positive] — [moderate positive]. The Tableau map above uses the calculated score and displays the distribution of negative impact across 9 weeks. The visualization on the left (Week 1) is mostly blue with only a couple states in the central area colored red, suggesting most states have a higher negative impact level; however, the visualization on the right (Week 9) is mostly red with about 5 states colored blue, suggesting that the negative impact level eases over time. One thing to note is that although the situation appears to be slowly improving, the negative impact level remained above 70, suggesting COVID-19’s prominent negative impact on small businesses across the nation.
2. Revenue
The above line plot shows that a collective decrease in revenue of the percentage of small businesses in a certain state. We can see a decreasing trend, meaning a lower percent of small businesses were having decreases in revenue. In addition, most states remained above 40% throughout the 9 weeks, suggesting an overall negative impact of COVID-19 on revenue.
3. Operation
These two charts show the weekly comparison of small businesses’ temporary closure and disruptions in the supply chain. Both show a similar trend:
- The percentage of small businesses that have temporary closure or disruptions in the supply chain decreased over time.
- The percentage of small businesses that did not experience temporary closure or disruptions in the supply chain increased over time.
The majority (>50%) of small businesses answered no for both questions across 9 weeks, suggesting COVID-19 had a relatively lower impact on small businesses’ operation.
These two charts show the weekly comparison of small businesses’ shift in their production and services. Both display a similar trend:
- The percentage of small business shifting production or service and only doing pickup, carry-out, or delivery slightly decreased over time.
- The percentage of small businesses that didn’t shift production or service and only did pickup, carry-out, or delivery slightly increased.
Contrary to what we might expect, less than 20% of small businesses only do pickup, carry-out, or delivery. Above 80% of small businesses answered no for both questions across 9 weeks, suggesting that although COVID-19 had a severe negative impact, most businesses were able to stick to the same production or service and the same operation type.
4. Employee
In the scatter plot on the left, we can see a positive correlation between temporary closures and the number of lay-offs and furloughs. Additionally, the right scatter plot shows a positive linear relationship between temporary closures and a reduction in working hours (the less hours a business is open, the less hours are available for employees — which matches our intuition). From this, we can deduce that, unfortunately, many small businesses across the United States have struggled to transition to a virtual world. As a result, they have had to reduce their hours and reduce the number of employees on the payroll.
5. Financial issues
The Paycheck Protection Program is a loan established by the CARES Act designed to provide a direct incentive for small businesses to keep their workers on the payroll. The Small Business Administration will forgive loans if all employee retention criteria are met, and the funds are used for eligible expenses. From the visualization, we can see that the request rate stayed about the same, while the receive rate increased over time. This suggests that, in general, small businesses that experienced financial issues were getting more assistance from government issued support programs during the pandemic.
In conclusion, although during the pandemic small businesses have indeed been negatively affected in every aspect, from our analysis, we can sense the situation for small businesses has gradually become better with increasing support from the government. The situation was not as bad as we initially hypothesized, as small businesses were able to stick to their ways of operation and production of goods/services and less small businesses experienced temporary closure and disruptions in the supply chain. Although this is true, we cannot be sure if this positive trend will continue through the phase 2 of the survey. The US Census recently updated their phase 2 plan of the survey, for which the collection period will be from 8/9 to 10/10. Considering the situation of COVID-19 worsened even more after June, the results might be very different for phase 2.
State Policies Analysis
The state policies analysis determined if different policies that each state (and the District of Columbia) issued over the course of COVID-19 had any correlations with the impacts faced by small businesses. We used the COVID-19 US State Policy Database from the COVID-19 Data Repository of OpenICPSR.
Process
The June-updated dataset we used is comprehensive with over 70 policies and records of the dates when each state implemented them, making the cleaning process the most time-consuming. Since one of the main uses of this dataset was to be joined with our cleaned Small Business Pulse Survey data to identify correlations, the first step of the cleaning process was to determine which policies are relevant to small businesses. According to McKinsey’s “Which small businesses are most vulnerable to COVID-19 — and when”, the top categories of small businesses affected by the pandemic include food services, entertainment, and recreation. We settled on 10 policies that seemed to directly affect the broad term of “small businesses,” including 5 restrictive policies, such as“close nonessential businesses” and 5 easing policies, such as “ease shelter-in-place order,” as well as industry-specific policies like “close gyms” and “reopen restaurants.”
The cleaning process also included changing the records of dates to encompass the 6 weeks that were common with the survey previously analyzed. This called for creating and aggregating 6 datasets with binary values of whether a state enacted a policy or not. Also, since the majority of restrictive policies were issued before the survey took place, the multiple weeks of restrictive policies beginning in March were placed in the same category to match the first week of the survey.
Findings
- Revenue
Most of the analysis occurred after joining the restructured policies dataset with the cleaned survey dataset. The first question we explored was how certain policies correlated with the approximate mean revenue of small businesses across states. We calculated the approximate mean revenue metric in the survey analysis based on the percent of small businesses choosing each revenue group (e.g. $0-$500). We initially predicted that easing policies would show positive correlation with mean revenue, as roadblocks for small businesses would be mitigated. The graph shows a general trend of small business approximate mean revenues shooting upward after week 3 (5/10–5/16) with states like New York and California beginning to reopen businesses just before. Even for areas that show a different revenue pattern like D.C., it can be seen that soon after beginning to reopen businesses, mean revenue seems to increase.
Since all restrictive policies relevant to small businesses were implemented before the survey, this graph shows the policy to cover only the first week for all states. For both policies, we selected a few states to view, but check our Tableau Revenue Dashboard to choose the states you want to see.
2. Financial Issues
We explored financial issues that small businesses may have faced by mapping the percentage of small businesses requesting assistance from the PPP, as well as including metadata showing the number of new restrictive and easing policies issued by each state of the selected week (select different weeks to view in our Tableau Financial Issues & Operations Dashboard). It is interesting to see which regions seemed to have a larger proportion of small businesses requesting financial assistance. The southern and east-coast regions of America seem to have a higher proportion of small businesses requesting assistance in week 1 (4/26–5/2).
For identifying the correlation between states’s small businesses requesting for assistance and then actually receiving assistance, we created this scatterplot. The positive correlation by the trendline indicates that small businesses requesting financial help are fortunately likely to receive help.
3. Operations
Finally, we explored how the number of restrictive policies may correlate with the percent of small businesses experiencing disruptions in their supply chains. We predicted that these two metrics would be positively correlated, which can actually be seen as the darker circles (states with more restrictive policies) tend to be larger in size (have a higher proportion of small businesses experiencing supply chain disruptions).
Impact of policies on success of small businesses
Based upon our findings above, we decided to dive deeper into the relationship between state policy and negative impact on small business. We initially hypothesized that restrictive policies will positively correlate with the negative impact small businesses face.
From the above heatmaps, we can see that, in general, restrictive policies are more correlated with the overall negative impact on each state than easing policies. This makes sense because most restrictive policies restrict customers to go out and force small businesses to stop operating, whereas easing policies encourage them to operate close to normally.
However, there are still some unexpected findings from the visualization: first “began to reopen businesses” and “closed restaurants except take out” have the same correlation score. Second, “End/relax stay at home/shelter in place” correlates with negative impact more than “closed restaurants except take out.” These findings suggest that it would be inappropriate to simply conclude that restrictive policies yield a negative impact on small businesses. We can see that there is a nuanced relationship between policies and business, and to understand the relationship at a deeper level, we encourage YOU to explore how policies are impacting your favorite industry.
Unfortunately, one issue with the dataset was that it did not account for how long each policy was in place for. This limited some of our analysis since each week would account for the number of “new” policies issued rather than the total number of policies in effect that week. Therefore, these results should be interpreted as the immediate one-week impact, as opposed to long-term impact.
Effectiveness of policies restricting non-essential businesses on COVID-19 case count
People across the U.S. have been questioning the effectiveness of state and county-wide COVID-19 policies, specifically those regarding the closing of non-essential businesses. Non-essential businesses are typically organizations with a recreational purpose, such as indoor dining restaurants, salons, theaters, gyms, among others. Some believe that these restricting policies do not contribute to reducing the spread of COVID-19 in their community and that the economic implications and lifestyle changes are too important to forgo.
However, some research has shown that implementing restrictions on non-essential businesses is key to limiting the spread of the COVID-19. This analysis hopes to further explore this idea to observe the effect of policies that ease and restrict the opening of non-essential businesses and the percent increase in cases after 14 days, the upper end of the typical incubation period.
To begin the processes, we collected COVID-19 case and death data from the New York Times COVID-19 data repository and The University of Washington. The policy data was updated per county-wide policy implemented in a state, listing specific policy types, such as Emergency Declaration, School Closing, Gathering Restrictions etc, and associated descriptions. We used regex to search for policies that contained “non-essential businesses” in their description and split them further by searching for specific keywords (ie “ease”, “open”, “closure”, “suspend”). We joined the policy data set with the New York Times data set for both the policy enactment date and the date 14 days later, to see the change in COVID-19 cases and related deaths.
This graph shows a comparison of the percentage increase of COVID-19 cases from policies that eased the restrictions of non-essential businesses (blue) to those that implemented restrictions on non-essential businesses (orange). Although there is an increase in COVID-19 cases for all enacted policies, the average increase for restrictive policies (61%) is much lower than the average increase for easing policies (98%).
In addition, it is evident that up until the end of June, the rate of contracting COVID-19 had slowed down, as the percentage increases weren’t as high as those from late March, for both types of policies. This could indicate that as more research came out about the virus, people began to take additional precautions to ensure safety, regardless of the restrictions on non-essential businesses.
In conclusion, the effectiveness of policies that restrict the opening of non-essential businesses can be shown through its lower percentage increase of deaths. This supports existing research in the field on key ways that counties can limit the spread of the virus as well as sheds light onto the potential “flattening of the curve” that can be seen from mid-April to June. However, case count began to rapidly rise again, hitting peak levels in mid-July and August, which could be associated with stores, schools, and businesses beginning to open up again, indicating a need for more restrictive non-essential business policies.
Conclusion
COVID-19 has undoubtedly affected the entire nation. However, the negative impacts are not distributed evenly. Small businesses are more negatively affected by policies that restrict non-essential businesses than large corporations. Although small businesses are able to see growth with the slow implementation of easing policies, the public must ensure that more restrictive policies aren’t issued for small businesses to continue operating.
Follow us on Instagram @bigdata.berkeley and visit our website at bd.berkeley.edu !!!
