Download experimental Business Dynamics Statistics of SBA COVID Response recipients (BDS-SBA COVID Response) data tables below. A detailed description of each loan program can be found in the BDS-SBA COVID Response Definitions dropdown below. For a description of the BDS variables refer to the BDS Codebook and Glossary.

The BDS-SBA COVID Response is an experimental product of the U.S. Census Bureau and was developed by the Center for Economic Studies (CES). The published statistics were tabulated from the Longitudinal Business Database, an internal Census data product that tracks firms over time, and loan-level data from the four pandemic loan programs administered by the Small Business Administration (SBA). As with the main BDS, all statistics are presented at the annual frequency. Statistics cover the 2007-22 period.

The Census Bureau has reviewed this data product to ensure appropriate access, use, and disclosure avoidance protection of the confidential source data used to produce this product (Data Management System (DMS) number: 7513031, Disclosure Review Board (DRB) approval number: CBDRB-FY24-0487,CBDRB-FY24-0446).

The Paycheck Protection Program (PPP) was an SBA-backed loan program that provided uncollateralized loans to small businesses (firms) with fewer than 500 employees to help blunt the negative effects of the pandemic. Alternatively, if a business had 500 or more workers and thus didn’t meet the SBA’s core definition of “small”, it could have qualified for PPP by using one of the SBA’s industry-specific size standards. Another important detail is that the size threshold was generally firm-specific. However, businesses operating in the NAICS 72 sector, which includes hotels and restaurants, were able to meet the size threshold at the establishment level, meaning firms with more than 500 total employees could qualify if individual establishments were small.

Importantly, PPP loans were fully forgivable if recipient businesses maintained employment and wages at pre-pandemic levels in the months after receiving payment. Recipient businesses were also able to qualify for forgiveness through two “safe harbor” options which either extended the timeframe to restore employment and wages to pre-pandemic levels or completely absolved businesses of this requirement if they could document in “good faith” that lockdowns and local restrictions made it impossible to return to full business activity. The vast majority of all PPP loans (92 percent) were forgiven.

In total, there were three tranches of funds appropriated by Congress which totaled roughly $800 billion. Starting in late March 2020 eligible businesses were able to apply for a “first draw” loan of up to $10 million from the first two tranches of funding. The third tranche helped distribute loans to first-time recipients in early 2021 as well as businesses that qualified for a “second draw” loan. To promote greater access to funding during the first several weeks of 2021, loan applications were only accepted from community financial institutions and first-time borrowers in low- or moderate-income neighborhoods with no more than 10 employees. Businesses were eligible for a second PPP loan of up to $2 million if they met three conditions. These included having already received and appropriately depleted a first draw loan, being able to show a reduction in gross receipts of at least 25 percent between comparable quarters in 2019 and 2020, and finally meeting a more stringent definition of “small” which was fewer than 300 employees. Lasting for roughly a year and a half, PPP ended at the end of May 2021. 

Although smaller in scale, the COVID Economic Injury Disaster Loans (COVID-EIDL) program provided an alternative source of funding to businesses during the pandemic starting around the same time as PPP. Unlike PPP, COVID-EIDL loans did not have a forgiveness condition. Each loan came with a term of 30 years and fixed interest rate between 2.75 percent for private non-profit organizations and 3.75 percent for businesses with payments being deferred for the first two years. COVID-EIDL loans were eventually distributed up to $2 million and were to be used for operating expenses, such as payroll, rent, mortgage payments, debt, and other ordinary business expenses. Also in departure from PPP, COVID-EIDL loan recipients were required to provide collateral for loans in excess of $25,000. And while some businesses only received COVID-EIDL funds, many businesses also received PPP.

The SBA also backed two additional and more targeted loan programs. The first was the Restaurant Revitalization Fund (RRF), which launched in May 2021. RRF was designed to provide emergency funds to eligible restaurants, bars, and other qualified businesses that were impacted by the pandemic. Other qualified businesses included food trucks, caterers, and snack bars, as well as bakeries, breweries, wineries, and inns that had on-site sales to customers amounting to at least one third of gross receipts. Eligible businesses received funding up to $10 million per business. Similar to the loan forgiveness nature of PPP, recipients of RRF were not required to repay the funding so long as the funds were put toward eligible uses. And while being smaller in scale than PPP, nearly thirteen times as many businesses received both PPP and RRF than RRF alone. 

The final program that operated during the pandemic was the Shuttered Venues Operator Grants (SVOG) program, which launched in April 2021. This program was aimed at providing pandemic assistance to operators of live venues, theatres, live performing arts organizations, museums, and motion picture theatres, as well as talent representatives. Businesses that were born after February 2020 were not eligible for SVOG. Additionally, venues or promoters that received a PPP loan after December 27, 2020 were still eligible, but their SVOG funds would be reduced by the amount of their PPP loan. The maximum grant amount for eligible businesses was $10 million and similar to PPP, these funds were forgivable.

There are a total of 38 tables available to download. We present one set of tables for each loan program, where tables conform to the format and methodology laid out by the main BDS, with several key additions. The first is a program status variable that delineates between several types of loan recipients and non-recipients. Recipiency status is defined based on the specifics of how each program operated and what types of administrative data were collected.

Firms often received funding through more than one SBA pandemic relief program. We highlight this feature of the SBA Covid response by creating a combined table that examines dual recipients, or businesses that participated in two or more programs.

The first matching step was to link the loan-level SBA data to the BR, which provides a comprehensive database of all business establishments and firms in the United States. This was done in two phases. In the first phase, a deterministic matching process used tax identifiers to match records in the loan-level data to the BR. Specifically we used the Employer Identification Number (EIN) from the loan application to match to tax filing data in the BR, including annual revenue returns and quarterly payroll returns. In addition, if the loan application provided information that enabled the Census Bureau to assign a Personal Identification Key (PIK) to the business owner, we matched the PIK to 1040 Schedule C filings of sole proprietorships with employees and to self-employed individuals. If records were not matched using a tax identifier, we used a probabilistic linking procedure to find matches by comparing the name and street address of the business from the loan application to names and addresses on the BR. This process was done separately for each of the four programs.  

Table 1 shows the high-level outcomes of the PPP loan matching process. The PPP loan database shared with the Census Bureau by SBA contained 12.5 million loans. Approximately 1 million of these had a status of “Voluntary Termination” or “Fully Canceled” and were dropped. Of the remaining 11.5 million loans, 90% of them matched to the Business Register in the time period of 2019-2021. Table 2 breaks down the types of PPP-BR matches. Most of these loans (70%) have both an associated EIN (of the business) and PIK (of the business owner) and either the EIN, the PIK, or both matched to the BR. Among the remaining matched loans, 24% only matched by PIK, because they were non-employer businesses (i.e., self-employment) which were also allowed to apply for PPP loans, and 6% only matched by EIN. Less than one percent of loans matched only by name and address but did not identify an EIN or PIK.

We require an EIN in order to match a loan to an employer business, and thus drop all loans that matched to the BR solely by PIK and did not have an EIN before matching to the LBD. Because of the dual-draw nature of PPP, EINs may appear more than once in the loan-level data. Collapsing to one record per EIN, we were left with roughly 5.6 million unique EINs belonging to businesses that received at least one draw from PPP.

Next we matched the list of unique EINs that were found in both the BR and PPP loan data to the LBD for the year 2020 in order to identify employers. These results are shown in Table 3. We found approximately 70% of these EINs in the 2020 LBD. Most, but not all, matched EINs are in scope for BDS tabulation. It is notable that a very small share of these EINs also match to the non-employer universe. These EINs belong to businesses that operated as employers in 2019, but in 2020 they both received PPP funds and transitioned into a non-employer business.

Approximately 30% of EINs do not match to the LBD, and among these, 65% matched to the universe of 2020 non-employer (self-employment) businesses also found in the BR. A little over half of the non-employer matches reported having only one or zero employees on the PPP form, suggesting that these applicants were indeed self-employed. However, the remaining 47% of non-employer matches reported more than one employee on their PPP application. It is likely that this reported employment is a noisy measure of the true employment of the business but it is also possible that some non-employer records in the BR are incorrectly missing employment and are misclassified. The BDS-SBA COVID product includes only employers and all non-employer matches are dropped prior to tabulation.

The remaining 35% of non-LBD-matches do not link to either a non-employer or employer record for 2020. The group contains 578,000 EINs in total and makes up approximately 10% of all unique PPP EINs that were found in the BR. These are cases which matched to an EIN found in an alternate BR year (2019 or 2021). On the PPP loan application, 58% of these cases reported employment of one job or less. These could have been non-employers in an earlier year and either ceased economic activity in 2020 or have a missing tax return. The remaining 42% of records report having employees. There are a few possible explanations for this. First, the business could have two EINs, one used for payroll and one used for filing business income tax returns. If the business used its income tax EIN instead of its payroll tax EIN to file for a PPP loan, it is difficult to link the PPP loan to the employer business. Second, the business might have used a Professional Employer Organization (PEO) to handle its payroll. In this case, the business does not show up as a separate entity in the LBD, but instead has its employment reported as part of the PEO in NAICS 56133. If the business used an EIN of its own (maybe one used to file IRS income tax) to apply for the PPP loan, this record will not match to the LBD.

Finally, we attempted to link non-matched EINs to the LBD in 2021 and find 75,000 additional matches. This could be explained by businesses switching back and forth between using a PEO and managing their payroll internally; or tax filing issues in 2020 that create missing BR data for these businesses in the year of PPP application; or new businesses that formed early in 2021 and received very late first draw loans.

In the PPP tables, establishments and their parent firms are binned into one of seven groups: four recipient and three non-recipient bins. The recipient groups include establishments belonging to firms that only received an early first-draw loan, firms that only received a later first-draw loan, firms that received both an early first-draw and an eventual second-draw loan, and finally firms that received both a later first-draw loan and an eventual second-draw loan. For non-recipients, we distinguish between those that were in scope in 2020, either as active firms with employment in March 2020 or exiting firms that had employment in March 2019 but not 2020, those that died early enough that they were not in scope in 2020, and those that were born after March 2020. Note that establishments are counted as exits in the year they first have zero March employment. Establishments with no March 2019 employment will be counted as exits in 2019 and are classified in the “died before 2020” group in the BDS tables. This group contains any establishment exit recorded prior to 2020. This provides data users more flexibility in choosing an appropriate comparison group for any of the recipient categories. However, given that we cannot match all PPP loans to the LBD, we caution that the non-recipient bins likely include some PPP recipients, although this percentage is probably small. EINs that do not match to either a non-employer or employer record in 2020 make up only 10% of all the unique PPP EINs that were identified in the BR and of these, less than half, or about 4% of all EINs, report having more than one employee on their PPP application.

In addition to preserving the first- versus second-draw nature of PPP, we draw from the recent economic literature and add the concept of timing, which has been shown to be an important factor in businesses’ ability to retain employment and avoid shutdown (Autor et al., 2022). To distinguish between early versus later first-draw loans, we utilize the dates in which the first two tranches of program funds were signed into law. The Coronavirus Aid, Relief, and Economic Security (CARES) Act was signed into law on March 27, 2020 and provided $350 billion in appropriations. The Paycheck Protection Program and Health Care Enhancement Act became law on April 24, 2020 and added an additional $320 billion in appropriations which served as the second tranche of PPP funding. For the purposes of this experimental data product, businesses that received an early first-draw loan were approved for funding after CARES but before April 24, 2020. Any business that received a later first-draw loan had their funding approved on or after April 24, 2020. The final tranche of funding was approved by Congress at the end of December 2020 as part of the annual appropriations process. Some applicants continued to receive first draw loans from this new funding in early 2021. In addition, all second draw loans followed this approval and were distributed between January and May of 2021. In summary, we label first-time PPP loans made between March 27 and April 23 as early first draw loans; first-time loans made between April 24, 2020 and May 2021 as later first draw loans; and second-time loans made between January and May 2021 as second draw loans.

The remaining three programs, COVID-EIDL, RRF, and SVOG, do not have the same timing element as PPP and so there is only one recipient category. However for these programs, data on applicants who were denied loans are also available. Hence tables for these programs report program status in three bins: applicant recipients, applicant non-recipients, and non-applicant non-recipients. Importantly, this last group includes both businesses that never applied to one of these programs and businesses that applied but did not use an EIN that matched to an employer record in the BR.

Because this experimental data product reports annual statistics for employer businesses starting in 2007, recipient and non-recipient establishments in 2020 need to be followed longitudinally backward in time. This process begins by first identifying which establishments in the LBD use an EIN that is also found in the loan program data. For single-unit firms this process is straightforward as there is a one-to-one correspondence between the EIN and the establishment. For multi-unit firms, there are instances where some, but not all, of the firm’s establishments use an EIN that matches to the loan program data. After a firm received pandemic relief through PPP or any of the other loan programs, there were no rules that prohibited the firm from distributing those funds to all of its establishments. Hence we paint a broad brush and classify all of the firm’s establishments as having been treated if at least one uses an EIN that matches with the loan program data. This method converts seemingly “partially” treated firms into fully treated.

For multi-units with multiple recipient EINs, the early versus later designation of first-draw PPP loans is determined by the EIN with the earliest loan approval date. In addition to program status, each set of tables includes a loan size variable. For all four loan programs, the program-specific loan size is calculated as the sum of loans received by all of the firm’s EINs.

Importantly, as establishments are longitudinally linked backward in time, their treatment status and loan size follows them. Figure 4 provides a simple two-firm example across two years. Starting in the upper left, Firm A has three establishments in 2020, each with the program status of “early first draw only” and loan size of “a) up to $24,999.” For the year 2020, Firm A would be tabulated in one program status bin and one loan size bin. Tracking this firm's three establishments backward in time reveals that in 2019, establishment 1 belonged to Firm B. Since treatment status follows each establishment throughout time (both forward and backward), Firm B in 2019 would be tabulated in two program status bins and two loan size bins: early first draw only and non-recipient. Although this approach does not preserve the ability to vertically sum firm counts in the BDS tables, it more cleanly identifies the establishments that were treated and those that were not, an approach that does not rely on choosing an arbitrary dosage of treatment required to be classified as “treated.” Thus, in every year of the BDS tables, firms will be tabulated in any bin for which they have at least one establishment present. Establishment counts will sum vertically as each establishment is tabulated in a treatment status bin that is constant over time.

The BDS-SBA COVID data product contains 38 tables with 10 specific to PPP, 9 each for COVID-EIDL, RRF, and SVOG, and one for recipients of multiple loan programs. For each program, we create separate tables that stratify the statistics by firm age, firm size, loan size, loan size and state, metro/non-metro, rural/urban, industry sector, and state. Due to the size of the PPP program, we are able to include an additional PPP table stratified by both sector and firm size. Finally, since firms often received funding through more than one SBA pandemic relief program, the product also includes a combined table that examines recipients who participated in two or more programs.

In alignment with the core BDS, this experimental data product uses the standard coarse definition for firm size. Although generally these programs were limited to firms with less than 500 employees, there are recipients in the 500+ category bins for each program. There are several reasons for this. First, coarse firm size is determined by the average of firm-level employment between the prior year and current year during the March 12 reference pay period. It is not specific to one period in time. For example, if a firm received pandemic assistance in 2020 and had an average employment level of 500 in that year, it could have employed 510 workers in March 2019 and only 490 in March 2020. Based on the SBA size requirement in 2020, this firm would have been eligible for pandemic relief, but its coarse firm size bin would be 500+. Second, larger businesses in some industries qualified for pandemic relief using the SBA's industry-specific size standards, which, for example, allowed restaurants to apply on the basis of establishment not firm size.

Because the RRF program was much more targeted to restaurants and businesses operating in the NAICS 72 sector, we include a detailed industry breakout by program status. This includes the following NAICS industries:

• 721 (accommodation)
• 7223 (special food services)
• 7224 (alcoholic drinking places)
• 722511 (full-service restaurants)
• 722513 (limited-service restaurants)
• 722514 (cafeterias, grill buffets, and buffets)
• 722515 (snack and non-alcoholic beverage bars)

This breakout can be particularly useful for data users looking to compare, for example, full-service restaurants with limited-service restaurants or bars, all of which were impacted by stay-at-home requirements during the beginning parts of the pandemic. The “other” bin in these tables includes several industries that were eligible for RRF but were not classified under NAICS 72. These include the following NAICS industries:

• 312120 (breweries)
• 311811 (retail bakeries)
• 312130 (wineries)
• 311812 (commercial bakeries)
• 312140 (distilleries)
• 445110 (supermarkets and other grocery stores, except convenience)
• 445299 (all other specialty food stores)
• 445310 (beer, wine, and liquor stores)
• 445120 (convenience stores)
• 445291 (baked goods stores)
• 447110 (gasoline stations with convenience stores)
• 445292 (confectionary and nut stores)
• 445210 (meat markets)
• 445220 (fish and seafood markets)
• 445230 (fruit and vegetable markets)
• 447190 (other gasoline stations)
• 454210 (vending machine operators)
• 713950 (bowling centers)
• 713990 (all other amusement and recreation industries)
• 713120 (amusement arcades)

In contrast, the SVOG tables report an “other” bin that includes all two-digit NAICS sectors other than NAICS 51, 71, and 72, which were the most-common recipients.

New with the release of this experimental data product are tables that present annual statistics across an urban-rural continuum. Using population data from the 2020 Decennial Census, counties are classified into four categories: predominantly urban counties have at least 80 percent of their population living in urban areas, mostly urban counties have between 50 and 80 percent of their population living in urban areas, mostly rural counties have between 20 and 50 percent of their population living in urban areas, and predominantly rural counties have less than 20 percent of their population living in urban areas. It is important to note that establishments are tabulated according to the urban-rural attribute of the county in which they reside. In some instances, businesses are considered “statewide” if they are not assigned to a specific county or “unclassified” if it is not possible to determine where the business operates. Both of these represent a very small share of the data.

It is important to note one way in which we do not stratify our BDS tables. Since 92 percent of all PPP loans were forgiven, our data lack meaningful variation along this dimension. Hence, we do not split firms by whether their PPP loans were forgiven.

Below are links to selected publications related to BDS-SBA COVID Response.