Estimating Sector Size With Payroll Taxes

The nonprofit sector must persistently track its programs and activities to measure societal benefits and communicate its impact to key stakeholders. This research note introduces an uncommon but useful approach to measuring the sector’s size using payroll tax information disclosed on the IRS form 990.

Researchers often measure the nonprofit sector size by counting the number of nonprofits or examining total program expenses. An alternative and complementary approach is to consider payroll data because they reflect the mobilization of human resources and can provide insights into the depth of community engagement.

Historically, payroll data have been difficult to acquire. Partnerships between nonprofit scholars and the Bureau of Labor Statistics have led to new methods for using the Quarterly Census of Employment and Wages (QCEW) to measure the size of the nonprofit workforce. But QCEW data on nonprofits are only updated every five years. Payroll tax information on 990 forms is more accessible and updated more frequently.

Although nonprofits are exempt from many types of taxes, they are required to pay federal payroll taxes. These payroll taxes include a 6.2% Social Security tax and 1.45% Medicare tax, and apply to part- and full-time employees. The total payroll taxes paid by the sector can serve as a proxy for estimating the number of full-time equivalent workers employed by the sector.

We estimate that nonprofits currently pay approximately $66 billion a year in payroll taxes. See below for details about how we arrive at this estimate.

Form 990 Data

We estimate the payroll taxes that the nonprofit sector contributes annually using IRS 990 data from 2019, the most recent full year of available data at the time of our analysis.

It’s important to note that there can be a two-year lag in returns, which makes it necessary to combine three years of IRS 990 data extracts to capture a single full year of data. This results in an idiosyncratic data aggregation challenge. We compiled our Form 990 data from the 2019, 2020, and 2021 Statistics of Income Extract files available through the IRS: https://www.irs.gov/statistics/soi-tax-stats-annual-extract-of-tax-exempt-organization-financial-data.

Sampling Considerations

Data availability varies greatly by the size of the nonprofit. Larger nonprofits are required to file the full Form 990, which involves significant disclosures on their programs and finances. Medium-sized nonprofits file a more compact version of the form called the 990EZ, and small nonprofits are only required to confirm their active status by filing Form 990N. Private foundations use the 990-PF. Payroll information disclosures vary by the type of form.

Larger Nonprofits: The Form 990 collects data on the amount of payroll taxes contributed by larger nonprofits (those with gross receipts >= $200,000 or total assets >= $500,000), which amounted to $55 billion in 2019.

Private foundations: As reported on the Form 990-PF, private foundations don’t provide data on the payroll taxes they contribute. Instead, they report the amount spent on salaries and benefits, which was $3 billion. To estimate the payroll taxes that private foundations contribute, we use a ratio calculated from larger nonprofits.

We divide the amount of payroll taxes larger nonprofits contribute by the amount they spend on payroll taxes to calculate a payroll tax/salaries and benefits ratio (5.75%). We then apply this ratio to the salaries and benefits data from the Form 990-PF to estimate the payroll taxes private foundations contribute.

There are some limitations, however. The IRS’s Form 990-PF data extract only includes the columns with data on “compensation of officers, directors, trustees, etc.” and “pension plans, employee benefits.” It excludes the column with data on “other employee salaries and wages.”

This means we will underestimate the payroll taxes that private foundations contribute. Additionally, the IRS does not provide a data extract for the 2019 Form 990-PF file, further contributing to the underestimation of the payroll taxes that private foundations contribute (estimated to be $153 million in 2019).

Smaller Nonprofits: The Form 990-EZ does not collect data on the amount of payroll taxes contributed by smaller nonprofits (those with gross receipts <$200,000 and total assets <$500,000). While it collects data on the amount that these smaller nonprofits spend on salaries and benefits, the IRS’s Form 990-EZ data extracts exclude these data. Therefore, we are unable to estimate the payroll taxes that smaller nonprofits contribute.

Invisible Nonprofits: Some nonprofits, such as churches, do not have to file annual returns. See https://www.irs.gov/charities-non-profits/annual-exempt-organization-return-who-must-file.

The Form 990-N also does not collect financial data from the smallest nonprofits (those with gross receipts <=$50,000). It also doesn’t collect data on the amount that these nonprofits spend on salaries and benefits. Therefore, we cannot estimate the amount of payroll taxes that the smallest nonprofits contribute.

Data Vignette

Packages

You can install R packages as follows:

install.packages( "knitr" )
install.packages( "tidyverse" )
install.packages( "pander" )
install.packages( "kableExtra" )

If you already have these packages, you only need to load the libraries.

library( tidyverse )  # data wrangling
library( pander )     # pretty formats
library( knitr )      # pretty formats
library( kableExtra ) # html formats for tables 

Data

Load the SOI Extract data obtained from the IRS:

Download all data and read locally:

URL <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/payroll.zip"
download.file( URL, destfile="payroll.zip" )
unzip( "payroll.zip" )

pc.2019 <- readr::read_csv( "payroll/19eoextract990.csv" )
pc.2020 <- readr::read_csv( "payroll/20eoextract990.csv" )
pc.2021 <- readr::read_csv( "payroll/21eoextract990.csv" )
pf.2020 <- readr::read_csv( "payroll/20eoextract990pf.csv" )
pf.2021 <- readr::read_csv( "payroll/21eoextract990pf.csv" )

Or read directly from the URLs:

url.pc.2019 <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/19eoextract990.csv"
url.pc.2020 <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/20eoextract990.csv"
url.pc.2021 <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/21eoextract990.csv"
url.pf.2020 <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/20eoextract990pf.csv"
url.pf.2021 <- "https://nccsdata.s3.us-east-1.amazonaws.com/replication/payroll/21eoextract990pf.csv"

pc.2019 <- readr::read_csv( url.pc.2019 )
pc.2020 <- readr::read_csv( url.pc.2020 )
pc.2021 <- readr::read_csv( url.pc.2021 )
pf.2020 <- readr::read_csv( url.pf.2020 )
pf.2021 <- readr::read_csv( url.pf.2021 )

Preview the data:

head( pc.2019[,1:6] ) %>% pander()

The data dictionaries are available in the zipped directory. Additional information can be obtained at the SOI Extracts page.

Data Preparation

Variable Cleanup

Convert six digit tax period dates (YYYYMM) to four digit years (YYYY):

pc.2021$year <- pc.2021$tax_pd  %>% substr( 1, 4 )
pc.2020$year <- pc.2020$tax_pd  %>% substr( 1, 4 )
pc.2019$year <- pc.2019$tax_pd  %>% substr( 1, 4 )
pf.2020$year <- pf.2020$TAX_PRD %>% substr( 1, 4 )
pf.2021$year <- pf.2021$TAX_PRD %>% substr( 1, 4 )

head(pc.2021$tax_pd)
head(pc.2021$year)

[1] 202005 202012 202104 202003 202006 202012
[1] "2020" "2020" "2021" "2020" "2020" "2020"

Isolate the Study Period

SOI Extract files are organized by filing dates, which represent when the IRS received the returns. The filing date is usually around six months after the end of an organization’s fiscal year. This fiscal year is also 12 months ahead of the calendar year corresponding to the nonprofit’s activities.

In addition, nonprofits can back-file late returns or submit amended returns, so the year of the SOI extract file is different from the periods of data contained within.

pc.2021 %>%
  group_by( year ) %>%
  summarise( n() ) %>%
  kable( col.names = c( "Tax Period", "Frequency" ),
         caption = "Tax Periods in the 2021 Form 990 Dataset", 
         align="c" ) 

Tax Periods in the 2021 Form 990 Dataset

Tax Periods in the 2021 Form 990 Dataset

pf.2021 %>%
  group_by( year ) %>%
  summarise( n() ) %>%
  kable( col.names = c( "Tax Period", "Frequency" ),
         caption = "Tax Periods in the 2021 Form 990-PF Dataset",
         align="c" ) 

Tax Periods in the 2021 Form 990-PF Dataset

Tax Periods in the 2021 Form 990-PF Dataset

We need data that corresponds to the 2019 calendar year, the most recent full year available at the time of writing.

pc.2021.subset <- filter( pc.2021, year=="2019" )
pf.2021.subset <- filter( pf.2021, year=="2019" )
pc.2020.subset <- filter( pc.2020, year=="2019" )
pf.2020.subset <- filter( pf.2020, year=="2019" )
pc.2019.subset <- filter( pc.2019, year=="2019" )

Inspect to ensure we have the correct data now:

pc.2021.subset %>%
  group_by( year ) %>%
  summarise( n() ) %>%
  kable( col.names = c( "Tax Period", "Frequency" ), align="c",
         caption = "Tax Periods in the Subsetted 2021 Form 990 Dataset" )

Tax Periods in the Subsetted 2021 Form 990 Dataset

pf.2021.subset %>%
  group_by( year ) %>%
  summarise( n() ) %>%
  kable( col.names = c( "Tax Period", "Frequency" ), align="c",
         caption = "Tax Periods in the Subsetted 2021 Form 990-PF Dataset" ) 

Tax Periods in the Subsetted 2021 Form 990-PF Dataset

Harmonizing the Data

In order to combine files we need to ensure that the variable names are the same.

R is case-sensitive; if one dataset uses uppercase letters for a column name and another uses lowercase letters for a column name, it will not recognize them as the same column. As a result, it will not combine the columns.

So we first make the names of all of the columns lowercase.

names( pc.2021.subset ) <- tolower( names( pc.2021.subset ) )
names( pc.2020.subset ) <- tolower( names( pc.2020.subset ) )
names( pc.2019.subset ) <- tolower( names( pc.2019.subset ) )
names( pf.2021.subset ) <- tolower( names( pf.2021.subset ) )
names( pf.2020.subset ) <- tolower( names( pf.2020.subset ) )

Then we filter the data to include only the columns we plan to use in the analysis. We use the following fields:

Form 990-PF Columns

Form 990-PF Columns

pc.2021.subset2 <- 
  pc.2021.subset %>%
  select( ein, payrolltx, compnsatncurrofcr, 
          compnsatnandothr, othrsalwages, 
          pensionplancontrb, othremplyeebenef ) 

pc.2020.subset2 <- 
  pc.2020.subset %>%
  select( ein, payrolltx, compnsatncurrofcr, 
          compnsatnandothr, othrsalwages, 
          pensionplancontrb, othremplyeebenef ) 

pc.2019.subset2 <- 
  pc.2019.subset %>%
  select( ein, payrolltx, compnsatncurrofcr, 
          compnsatnandothr, othrsalwages, 
          pensionplancontrb, othremplyeebenef ) 

pf.2021.subset2 <- 
  pf.2021.subset %>%
  select( ein, compofficers, pensplemplbenf ) 

pf.2020.subset2 <- 
  pf.2020.subset %>%
  select( ein, compofficers, pensplemplbenf ) 

Stack the data

pc <- 
  bind_rows( pc.2021.subset2, 
             pc.2020.subset2, 
             pc.2019.subset2 )
pf <- 
  bind_rows( pf.2021.subset2, 
             pf.2020.subset2 )

Remove duplicate EINs

There might be some duplicate EINs in the combined datasets because some organizations may have submitted amended returns.

# roughly 1,200 duplicates
n_distinct( pc$ein )
nrow( pc )

n_distinct( pf$ein )
nrow( pf )

[1] 310349
[1] 311145
[1] 98253
[1] 98337

Remove duplicate EINs from the combined datasets

pc <- pc %>% distinct( ein, .keep_all = TRUE )
pf <- pf %>% distinct( ein, .keep_all = TRUE )

# check to ensure it worked
n_distinct( pc$ein ) == nrow( pc )
n_distinct( pf$ein ) == nrow( pf )

[1] TRUE
[1] TRUE

Analysis

Step 1: Calculate Payroll Taxes Paid by Nonprofits

Calculate the amount of payroll taxes paid by Form 990 filers in 2019:

# helpful function for printing large numbers
dollarize <- function(x) 
{ paste0( "$", format( round(x,0), big.mark="," ) ) }

prtax <- sum( pc$payrolltx, na.rm = T )
prtax %>% dollarize()

[1] "$55,464,358,803"

Step 2: Estimate the Effective Payroll Tax Rate

Calculate a payroll tax/salaries and benefits ratio for Form 990 filers.

To calculate the ratio, we divide the amount of payroll taxes that Form 990 filers pay by the amount that Form 990 filers spend on salaries and benefits.

Calculate the amount that Form 990 filers spend on salaries and benefits.

salaries <- 
  sum( pc$compnsatncurrofcr,
       pc$compnsatnandothr,
       pc$othrsalwages,
       pc$pensionplancontrb,
       pc$othremplyeebenef, 
       na.rm = T )

The following function can be helpful for printing large numbers:

salaries %>% dollarize()

[1] "$965,155,601,055"

We estimate the effective tax rate by tabulating the amount of payroll taxes paid by nonprofits divided by total salary costs:

ratio <- prtax / salaries
paste0( 100*round( ratio, 4 ), "%" )

[1] "5.75%"

Step 3: Estimate PF Contributions

It’s not perfect, but it allows us to roughly estimate the payroll taxes paid by private foundations, even though those data are not included in the SOI PF Extracts:

salaries.pf <- 
  sum( pf$compofficers, 
       pf$pensplemplbenf, 
       na.rm = T )

salaries.pf %>% dollarize()

[1] "$2,670,373,181"

Estimated payroll taxes contributed by PFs:

prtax.pf <- ratio * salaries.pf

prtax.pf %>% dollarize()

[1] "$153,457,677"

Step 4: Tabulate Totals

Add the amount of payroll taxes that nonprofits pay to the estimated amount of payroll taxes contributed by private foundations:

tax.total <- prtax + prtax.pf
tax.total %>% dollarize()

[1] "$55,617,816,480"

Step 5: Adjust for Inflation

While inflation is usually minimal over a two-to-three-year period, inflation was high during the study’s time frame (2019 to 2023). To adjust for this, we apply a scaling factor, which can be obtained from a reliable online inflation calculator by entering the start and end years.

( tax.total * 1.1885 ) %>% dollarize()

[1] "$66,101,774,886"

The estimate increased from $55 billion to $66 billion, a significant amount.

We can interpret this number either as the value of 2019 nonprofit payroll tax contributions in today’s dollars, or as an estimate of what nonprofits are likely paying in 2023, assuming similar employment rates during both periods.

Conclusion

We estimate that nonprofits pay approximately $66 billion per year in payroll taxes. But this is a conservative estimate for several reasons:

• It does not include the payroll taxes that Form 990-EZ or Form 990-N filers pay.
• It underestimates the payroll taxes that private foundations pay.
• It excludes the payroll taxes from nonprofits that are not required to file annual 990 returns, such as churches.

Despite these limitations, this approach provides a useful way to demonstrate the size and economic contributions of the nonprofit sector using freely available IRS administrative tax data.