The ABS is the primary source of information on research and development among for-profit businesses operating in the United States with one to nine employees. The ABS also collects data on innovation, technology, intellectual property, and financing from U.S.-based companies of all sizes.
The ABS is cosponsored by the Census Bureau in partnership with the National Center for Science and Engineering Statistics (NCSES) within the National Science Foundation. The ABS collects data on research and development (for businesses with one to nine employees), innovation, technology, intellectual property, and business owner characteristics, with additional rotating content that changes from year to year.
The ABS is cosponsored by NCSES and the Census Bureau.
Status | Active |
---|---|
Frequency | Annual |
Reference Period | Calendar year 2020 |
Next Release Date | October 2024 |
The Annual Business Survey (ABS) provides information on selected economic and demographic characteristics for businesses and business owners by sex, ethnicity, race, and veteran status. Further, the survey measures research and development (R&D) for microbusinesses, new business topics such as innovation and technology, as well as other business characteristics.
Not applicable.
Annual.
Calendar year 2017.
Calendar year 2020, for innovation, the reference period is 2018–20.
Firm.
Sample survey of for-profit employer companies with a U.S. presence engaged in the mining, utilities, construction, manufacturing, wholesale trade, retail trade, or services industries.
Approximately 4,900,000 employer companies.
Approximately 300,000 employer companies.
Key variables of interest are listed below.
The target population consists of all companies that allocated most of their annual domestic payroll to for-profit entities and had at least one establishment that had paid employees based on the company's Internal Revenue Service Form 941 that covers 12 March 2021, was classified in an in-scope NAICS industry, was physically located in the United States, and was in business at the end of calendar year 2021.
The Business Register (BR) serves as the primary input to the sampling frame. The BR is the Census Bureau’s comprehensive database of U.S. businesses. BR data are compiled from a combination of business tax returns, data collected from the economic census, and data from other Census Bureau surveys.
The ABS has a stratified systematic sample design. Stratification is based on demographic characteristics of the company owners and a NAICS industry code assigned to the company based on its constituent establishments. Companies were selected with certainty if they are known R&D performers with 1–9 employees, 500 or more employees, or meet sampling stratum specific cutoffs for receipts and payroll.
The ABS collects data electronically.
All data submitted by respondent companies are reviewed to ensure that data fields are complete and that data are internally consistent. Survey responses often include errors that require correction or unusual patterns that require validation. Automated edit checks are applied to improve the efficiency of data review and correction. Edit checks are designed to catch arithmetic errors and logically inconsistent responses (balance edits). The remaining automated edit checks are designed to flag outliers for further review (analytical edits). If additional information or data corrections are needed, respondents may be contacted to clarify or correct data. If additional information or corrected data cannot be obtained from respondents, data are imputed.
Estimates are produced from sums of weighted data (reported or imputed). Exceptions are described in detail in the annual reports containing detailed statistical tables.
Estimates of sampling errors associated with the detailed statistical tables will be available by request.
There can be both undercoverage error, where units are not included in the frame, and overcoverage error, where units included in the frame are actually out of scope for the population of interest. The ABS uses the prior-year Business Register to construct the frame, so any changes in businesses that would change the inclusion or exclusion of the business to the survey scope could be sources of coverage error. Prior to tabulation, survey units’ information is updated with the most recent available Business Register data to mitigate this source of error.
Unit nonresponse will be treated by adjusting weighted reported and imputed data by multiplying each company's sampling weight by a nonresponse adjustment factor. Detailed descriptions of the adjustments for nonresponse are available in the Technical Notes of the ABS data tables (https://www.nsf.gov/statistics/srvyabs/#tabs-2).
Expected sources of measurement error include differences in respondent interpretations of the definitions of R&D activities and differences in how companies count and report numbers of employees in various categories, including whether they work on R&D full time or part time. Little public information exists for most of the small businesses surveyed by the ABS, so it is difficult to determine whether companies are reporting R&D that satisfies the survey's definitions, particularly where the development of software and internet applications are concerned. Although quantitative metrics of measurement error are not available, there are ongoing efforts to minimize measurement error, including questionnaire pretesting, improvement of questionnaire wording and format, inclusion of more cues and examples in the questionnaire instructions, in-person and telephone interviews and consultations with respondents, and post-survey evaluations.
Data produced from the ABS will be available at https://www.nsf.gov/statistics/srvyabs/.
Some estimates from the 2021 ABS may not be directly comparable to prior ABS results due to changes in survey methodology. Sources of possible incomparability include changes to questionnaire wording and instructions, changes to data editing and tabulation, and changes to imputation and nonresponse adjustments.
ABS data will be published in NCSES InfoBriefs and data tables in the Business and Industry R&D series.
Results from this survey will be available on the NCSES Web site https://ncses.nsf.gov/. The ABS contains confidential data that are protected under Title 13 and Title 26 of the U.S. Code. Restricted microdata will be available at any of the 15 secure Research Data Centers administered by the Center for Economic Studies (CES) at the Census Bureau. Researchers interested in accessing microdata can apply for a restricted-use license by submitting a proposal to the CES, which evaluates proposals based on their benefit to the Census Bureau, scientific merit, feasibility, and risk of disclosure. To learn more about the Research Data Centers and how to apply, please visit the CES page on research with restricted-use data. For additional information about the application process, including how to initiate a project, please contact the administrator at the primary site where the research will be conducted. Per the Federal Cybersecurity Enhancement Act of 2015, the data are protected from cybersecurity risks through screening of the systems that transmit the data.
Purpose. The Annual Business Survey (ABS) provides information on selected economic and demographic characteristics for businesses and business owners by sex, ethnicity, race, and veteran status. Further, the survey measures research and development for microbusinesses, business topics such as innovation and technology, and other business characteristics. The ABS is conducted jointly by the U.S. Census Bureau and the National Center for Science and Engineering Statistics (NCSES) within the National Science Foundation (NSF). The ABS replaces the quinquennial Survey of Business Owners (SBO) for employer businesses, the Annual Survey of Entrepreneurs (ASE), the Business R&D and Innovation for Microbusinesses Survey (BRDI-M), and the Innovation section of the Business R&D and Innovation Survey (BRDIS).
The ABS is designed to incorporate new content each survey year based on topics of relevance. For the 2021 ABS, new content includes technology & innovation, financing, and management practices. The ABS also collects various data regarding business owners, such as sex, ethnicity, race, veteran status, and other characteristics. R&D data are collected on the ABS for businesses with a W-2 employment range between 1 employee and 9 employees.
Data collection authority. Title 13, U.S. Code, Sections 8(b), 131, and 182; Title 42, U.S. Code, Sections 1861–76 (National Science Foundation Act of 1950, as amended); and Section 505 within the America COMPETES Reauthorization Act of 2010, authorize this collection. Sections 224 and 225 of Title 13 require mandatory response. Office of Management and Budget (OMB) No. 0607-1004.
Survey sponsor. NCSES within NSF.
Survey collection and tabulation agent. The survey is conducted annually by the U.S. Census Bureau in accordance with an interagency agreement with NCSES within NSF.
Frequency. Annual.
Initial survey year. 2017; for innovation, the reference period was 2015–17.
Reference period. 2020; for innovation, the reference period is 2018–20.
Response unit. Firm.
Sample or census. Sample.
Population size. 4.9 million employer companies; approximately 511,000 companies with 1–9 employees and in North American Industry Classification System (NAICS) industries 31–33 (manufacturing), 42 (wholesale trade), 51 (information), 5413 (architectural, engineering, and related services), 5415 (computer systems design and related services), or 5417 (scientific research and development services) were in scope for the R&D module.
Sample size. 300,000 employer companies.
Target population. The target population consists of all companies that allocated most of their annual domestic payroll to for-profit entities and had at least one establishment that had paid employees based on the company's Internal Revenue Service Form 941 that covers 12 March 2021, was classified in an in-scope NAICS industry, was physically located in the United States, and was in business at the end of calendar year 2021.
Sampling frame. The sampling universe was constructed from the final 2019 Business Register. The Business Register is the Census Bureau’s comprehensive database of U.S. businesses. Business Register data are compiled from a combination of business tax returns, data collected from the economic census, and data from other Census Bureau surveys. The Business Register includes sole proprietorships, partnerships, and corporations reporting business activity to the IRS on any one of the following IRS tax forms: 1040 (Schedule C), “Profit or Loss from Business” (Sole Proprietorship); 1065, “U.S. Return of Partnership Income”; 941, “Employer’s Quarterly Federal Tax Return”; 944, “Employer’s Annual Federal Tax Return”; or any one of the 1120 corporate tax forms.
The Business Register contains establishments that are out of scope for the ABS sample. These establishments are removed from the sampling universe. They include the following:
Information on industry classification, receipts, payroll, and employment was extracted from the Business Register during the frame construction.
The sample is also designed to estimate demographic characteristics of the business owners. To efficiently sample for demographic characteristics, a variety of sources of information are used to estimate the likelihood that a business is a woman- or minority-owned business. Administrative sources include the Decennial Census, the American Community Survey (ACS), and the Numident file, which is the Social Security Administration’s (SSA) comprehensive database of information from Social Security applications. Individual business owners are identified through IRS K-1 filings for partnerships and corporations and from the Business Register for sole proprietorships. The owners are matched to the 2000 and 2010 Decennial Censuses to get race, sex, and ethnicity data; to the 2000–19 ACS; and to the Numident file (in that order) through a Protected Identification Key. Country of birth is also identified through the linkages to the ACS or Numident data. Each firm is then placed in one of the following eight strata for sampling: American Indian or Alaska Native, Asian, Black or African American, Hispanic, non-Hispanic White men, Native Hawaiian and Other Pacific Islander, publicly owned, and women-owned. Businesses are assigned to one and only one stratum, with priority given to the less common race or ethnicity groups.
Sample design. The ABS frame is stratified by state, expected owner race or ethnicity, and industry and is systematically sampled within each stratum. A standard type of estimation for stratified systematic sampling is used. Large companies were selected with certainty based on volume of sales, payroll, or number of paid employees. Specifically for the 2020 ABS, firms were selected with certainty based on the following criteria:
The certainty cutoffs vary by sampling stratum, and each stratum is sampled at varying rates, depending on the number and size of firms in a particular stratum.
The remaining frame is subjected to stratified systematic random sampling. Sampling rates vary by strata and are determined by the following calculations. Note that a systematic sample uses take rate to sample, and the sampling rate is 1/take rate.
The ABS sample consisted of 300,000 employer companies. There were 43,000 selected with certainty. The certainty portion of the sample consisted of
The remaining 254,000 noncertainty cases were selected using the systematic stratified random sample selection. The maximum sample weight was 40.
Data collection. Prior to mailing the survey, certain businesses selected were determined to be out of scope and were not contacted. The survey was mailed to 300,000 employer companies in July 2021. Businesses were sent a letter informing them of their requirement to report. The letter also provided instructions on how to access the survey and submit online. There were three mail follow-ups conducted to increase response. Additionally, the Census Bureau conducted e-mail follow-ups to respondents who entered the electronic system but did not submit the questionnaire. The collection period closed in January 2022.
Mode. The 2021 ABS was collected using the electronic instrument.
Check-in rate. The check-in rate is defined as the unweighted number of surveys that were submitted online by in-scope companies, divided by the unweighted total number of all in-scope companies in the sample. Response to individual questions did not factor into this metric. At the close of the collection period in January 2022, there were 190,000 responses submitted (63% of the sample).
There were an additional 1,800 businesses that contacted the Census Bureau via the call center to indicate that the company was no longer in operation or had been sold for the reference year.
Companies selected to report R&D represented 37.2% of those mailed. Of the companies mailed and selected to report R&D, 76,000 companies (or 68%) submitted responses.
Unit response rate (URR). The URR is the unweighted number of responding companies for the survey. For the ABS, response is defined as a company providing the number of owners, number of paid owners, and number of employees; or a company responding that it ceased operations prior to 2020.
For the ABS, the URR was 68.8%. The URR for businesses eligible to report the R&D module was 73.1%
Item response rates (IRR). The 2021 ABS collected data on approximately 650 variables, and the distribution of values reported by sample companies is highly skewed. Thus, rather than report unweighted item response rates, total quantity response rates are calculated, which are based on weighted data. The survey skip patterns vary for respondents; therefore, it is not possible to know an exact denominator for item response calculations.
Total quantity response rate (TQRR). For a given published estimate other than count or ratio estimates, TQRR is the percentage of the weighted estimate based on data that were reported by units in the sample or on data that were obtained from other sources and were determined to be equivalent in quality to reported data and weighted only by sampling but not nonresponse weights. The TQRR for total sales in the United States in 2020 was 72.9%.
Total quantity nonresponse rate (TQNR). For a given published estimate, TQNR, defined as 100% minus TQRR, is calculated for each tabulation cell from the ABS, except for cells that contain count or ratio estimates. TQNR measures the combined effect of the procedures used to handle unit and item nonresponse on the weighted ABS estimate. Detailed imputation rates are available on request from the survey manager.
Data editing. Prior to tabulating the data, response data were reviewed and edited to correct reporting errors. R&D data were tabulated for records reporting $50,000 or more in R&D expenditures.
Additionally, R&D data were only tabbed for records classified in the following NAICS industries:
Survey analysts reviewed the R&D reported by the survey respondents. Research was done by evaluating the reported business descriptions, reported R&D-to-sales ratio, and company website information. The majority of corrections involved false-positive reports or data reported using incorrect units (such as whole dollars instead of thousands of dollars). For NAICS industries 5415 and 5417, it is difficult to differentiate R&D from other technical work based solely on company website information. Due to this difficulty and the large number of companies sampled in these industries, it was not feasible to review each case individually; thus, relatively few corrections were made for false-positive reports. Unlike years past, we were able to review all cases in these industries that had greater than $50,000 in R&D expenses.
Additional data errors were detected and corrected through mass corrections and an automated data edit system designed to review the data for reasonableness and consistency. The editing process interactively performed corrections by using standard procedures to fix detectable errors. Quality control techniques were used to verify that operating procedures were carried out as specified.
Unit nonresponse. Unit nonresponse is handled by adjusting weighted reported data as follows. Each company’s sampling weight is multiplied by a nonresponse adjustment factor. To calculate the adjustment factors, each company in the sample that is eligible for tabulation is assigned to one (and only one) adjustment cell. The adjustment cells are based on employment size and NAICS sector. For employment size, there are five categories: 1–4 employees, 5–9 employees, 10–49 employees, 50–249 employees, and 250 or more employees. For a given adjustment cell, the nonresponse adjustment factor is the ratio of the sum of the sampling weights for all companies in the cell to the sum of the sampling weights for all companies in the cell with reported data. The nonresponse adjustment factor for adjustment cell h was calculated as:
where
r |
is the partial and complete respondents (RESP in (“CR,” “PR”)), |
nr |
is the complete nonrespondents (RESP = “NR”), |
wgt |
is the TAB_WGT value, |
h |
is the adjustment cell (SECTOR × EMPSIZE_COLLAPSE), and |
NAFh |
is the nonresponse adjust factor for cell h. |
For the nonresponse adjustment, a company is considered a respondent if it responded to either the R&D module, innovation module, or technology module of the survey.
Tabulation. Though as many firms as possible were identified as out of scope during sampling, additional out-of-scope firms were identified with either response or updated administrative data not available at the time of sampling. These 21,000 firms were removed for tabulations and include
Industry classification. The industry classifications of firms are based on the 2017 NAICS (https://www.census.gov/naics/). Firms with more than one domestic establishment are assigned a single industry classification using a hierarchal system based on the largest payroll sector, largest payroll 3-digit NAICS (within the largest sector), largest payroll 4-digit NAICS (within the largest 3-digit), and largest payroll 6-digit NAICS (within the largest 4-digit). For tabulation, industry classification was based on administrative data for 2017.
Geography. Firms with establishments operating in more than one state are tabulated as unclassified geography and counted only once in state and national totals.
Variance estimation. The ABS uses the delete-a-group jackknife variance estimator. Note that certainty cases do not contribute to the sampling variance. The delete-a-group jackknife variance estimator requires that every sampling stratum contains at least two sampled firms. Sampling strata that do not meet this requirement are collapsed as needed to create a new set of variance estimation strata that satisfies this requirement.
Detailed relative standard errors are available on request from the survey manager.
The estimates produced from the ABS are subject to both sampling and nonsampling errors.
Sampling error. The sampling error is described above in the variance estimation section.
Coverage error. Coverage error occurs when the frame fails to completely enumerate the population of interest. There can be both undercoverage error, where units are not included in the frame, and overcoverage error, where units included in the frame are actually out of scope for the population of interest. The ABS uses the prior-year Business Register to construct the frame, so any changes in businesses that would change the inclusion or exclusion of the business to the survey scope could be sources of coverage error. Prior to tabulation, survey units’ information is updated with the most recent available Business Register data to mitigate this source of error.
Measurement error. Expected sources of measurement error include differences in respondent interpretations of the definitions of R&D activities and differences in how companies count and report numbers of employees in various categories, including whether they work on R&D full time or part time. Little public information exists for most of the small businesses surveyed by the ABS, so it is difficult to determine whether companies are reporting R&D that satisfies the survey's definitions, particularly where the development of software and internet applications are concerned. Although quantitative metrics of measurement error are not available, there are ongoing efforts to minimize measurement error, including questionnaire pretesting, improvement of questionnaire wording and format, inclusion of more cues and examples in the questionnaire instructions, in-person and telephone interviews and consultations with respondents, and post-survey evaluations. The most common source of measurement error was reporting in different units (e.g., reporting whole dollars rather than thousands of dollars).
Some estimates from the 2021 ABS may not be directly comparable to prior ABS results due to changes in survey methodology. Sources of possible incomparability include changes to questionnaire wording and instructions, changes to data editing and tabulation, and changes to imputation and nonresponse adjustments.
Changes in survey coverage and population. There were no changes to survey coverage and population for the 2021 ABS.
Changes to questionnaire. The survey section that collected information on R&D from businesses (section D) had nearly identical question wording and instructions for both the 2020 ABS and the 2019 ABS questionnaires. The 2021 ABS collected additional detail from businesses on the location of R&D performance and explicitly highlighted the concept of “domestic R&D performance.” For the 2020 ABS, detailed questions about R&D expenditures (types of costs, funding sources, and R&D categories) were tied to domestic R&D performance (a subset of total R&D costs), whereas the 2019 ABS asked similar questions tied to total R&D costs.
The following are survey changes in the questionnaire from 2020 ABS to 2021 ABS; question numbers reference the 2021 ABS:
Changes in reporting procedures or classification. There were no changes to reporting procedures or classification.
Changes to data editing and tabulation. For data collected in the R&D section of the ABS, every response with $50,000 or more in R&D costs was reviewed for reasonableness. All positive R&D cases were reviewed in industries with high rates of false-positive reports. Outside of the review for potential false-positive R&D cases, the editing of R&D data was performed similarly in previous years’ ABS. There have been no major changes to the edits from year to year.
Changes in imputation methods. In 2019, a change was made in how item nonresponse is treated in ABS. Item nonresponse occurs when a company responds to the survey but leaves some items blank. In 2018, item nonresponse was treated using a combination of mode and donor imputation. For 2019 and future years (including 2021), no imputation was performed to treat item nonresponse, and missing items were included as no or zero in the estimates. For most of the estimates in 2019 (91%), this change in methods did not yield a change in the estimates. For an additional 9% of estimates, there was a difference in the estimate; the difference, however, was not statistically significant. In 2021, the imputation method was updated to exclude out-of-scope firms as donors and recipients. In previous ABS years, all sampled firms were included in imputation and out-of-scope firms were later removed.
Changes in collection methods. During the 2021 ABS collection period, all initial and reminder letters were mailed to respondents using U.S. Postal Service. A paper questionnaire was not mailed to respondents during the 2021 ABS, which collected data electronically.
Recommended data tables
This report provides data from the 2021 Annual Business Survey (ABS) (data year 2020). The ABS is the primary source of data on research and development of for-profit, nonfarm businesses with one to nine employees operating in the 50 U.S. states and the District of Columbia. The ABS also collects data on innovation, technology, intellectual property, and business owner characteristics of businesses of all sizes. The ABS is designed to incorporate new content each survey year based on topics of relevance. The 2021 ABS is the fourth year of the ABS. The survey is conducted annually by the Census Bureau for the National Center for Science and Engineering Statistics within the National Science Foundation.
The content and subsequent tables from this report have been reviewed by the Census Bureau’s Disclosure Review Board (DRB) and approved. The DRB approval number is CBDRB-FY22-ESMD010-017. This report is initially published with a selection of tables containing the core data of the survey. The report is updated with additional tables, as more detailed data becomes available. In some instances the initial selection of tables may also be updated from a preliminary version to add further data.
The first set of tables in this report, 1–20, released in December 2022.
The second and final set of tables in this report, 21–70, released in May 2023.
Audrey E. Kindlon of the National Center for Science and Engineering Statistics (NCSES) developed and coordinated this report under the guidance of Gary Anderson, Acting NCSES Program Director, and under the leadership of Emilda B. Rivers, NCSES Director; Vipin Arora, NCSES Deputy Director; and John Finamore, NCSES Chief Statistician. In partnership with NCSES, the Census Bureau conducted the survey and prepared the tables. NCSES staff members who made significant contributions include Jock Black, Jennifer Beck, Rebecca Morrison, and Timothy Wojan.
The Census Bureau, under National Science Foundation interagency agreement number NCSE-1748418, collected, processed, evaluated, and tabulated the data for this report. The Annual Business Survey is conducted within the Economic Directorate of the Census Bureau under the direction of Nick Orsini, Associate Director for Economic Programs, and Stephanie Studds, Assistant Director for Economic Programs.
The data were prepared in the Economic Reimbursable Surveys Division under the direction of Kevin Deardorff, Division Chief, and Aneta Erdie, Assistant Division Chief. This work was performed under the supervision of Patrice Hall, assisted by John Clark, Lakitquana Leal, and Gail White, with staff assistance from Concepcion Arenas Alvarez, Ahmad Bakhshi, Elaine Emanuel, Mary Frauenfelder, Aaron Finkle, Samantha Hernandez, James Jarzabkowski, Jessica Welch, and Tesfay Weldu. Additional support, including table creation and subject matter expertise was provided by Brandon Shackelford.
Mathematical and statistical techniques were provided by the Economic Statistical Methods Division under the direction of James Hunt, Assistant Division Chief. This work was performed under the supervision of Roberta Kurec, assisted by Sandra Peterson, and Dhanapati Khatiwoda, with staff support from Taylor Beebe, Alexandra Abzún Cadenas, Charles Champion, and Daniel Cordes.
Data collection procedures and operations were provided by the Economic Management Division under the direction of Lisa Donaldson, Division Chief, and Michelle Karlsson, Assistant Division Chief. The staff of the National Processing Center performed mailout preparation, respondent assistance, and correspondence processing. Project management support was provided by Laura Hardesty.
Development and coordination of the computer processing system was provided by the Economic Application Division under the direction of Sumit Khaneja, Division Chief and Olajumoke Oyewole, Assistant Division Chief. This work was performed by Marilyn Balogh, Michael Feldman, David Gonzalez, Chakravarthy Sharad, and Joseph Talbot.
National Center for Science and Engineering Statistics (NCSES). 2022. Annual Business Survey: 2021 (Data Year 2020). NSF 23-313. Alexandria, VA: National Science Foundation. Available at https://ncses.nsf.gov/pubs/nsf23313/.
For additional information about this survey or the methodology, contact