Census of Governments Employment and Payroll Data for March 2017 (1997 and 2007)

The U.S Census Bureau conducts a Census of Governments every five years on the employment, payroll, revenues, expenditures, assets and debts of every state and local government in the country. (State-level estimates for state and local government are released in other years).   The data is reported by government function (police, schools, parks, etc.)  The Bureau recently released its employment and payroll data for March 2017, and I have spent a considerable number of hours tabulating it to get it into usable form.

I now have spreadsheets that show local government full-time equivalent employment (full time workers plus part time workers converted into full time workers based on hours worked), by function, for the United States, New York City, other areas of the state, every county in New York State and New Jersey, and other states and counties elsewhere in the country selected for comparison.  This is expressed per 100,000 residents of each area.  Selected private sector industries that are either substantially government-funded (health and social services and infrastructure construction) or a substitute for government services (private schools and automobile-related industries) are shown on the same basis.  And local government payroll per worker of each area, expressed as a percent higher or lower than the U.S average for that function, and compared with the extent that an area’s private sector earnings per worker are higher and lower than the U.S. average.  Similar data is provided for state government in for U.S. as a whole, compared with New York State, New Jersey, and other states I consider relevant.  This is a process I will repeat for the finance phase of the 2017 Census of Governments, when it is released.

As is my custom, this post will simply provide the data in tables and spreadsheets, explain where it comes from and how I tabulated it, and provide a brief overview of trends in total state and local government employment and payroll.  Function-by-function posts, with the data reorganized and charts included, will follow.  Read this post to fully understand what you are seeing.  Download the spreadsheets and look at the numbers to decide for yourself what they mean, before getting my take on them.  I’ll write more when I can.  But anyone else is free to use this information right now.

The Census of Governments data was downloaded from the Bureau’s website, here for individual governments (a spreadsheet with 300,000-plus records)…


And here for state and local government totals by state, summarized in tables.


To standardize this data at the county level, I downloaded data on state and county population, private employment and earnings, and metropolitan area private employment and earnings, from the U.S. Bureau of Economic Analysis, Local Area Personal Income data.


The employment data for the relevant government-funded private sector industries was downloaded from the Quarterly Census of Employment and Wages from the U.S. Bureau of Labor Statistics.


After a long series of steps to provide fair comparisons between areas, which will be described below, I ended up with this spreadsheet for local government in March 2017.

Revised 2017 Census of Governments Employment by County NY NJ Selected Others

Including this table for employment.


And another table for mean payroll per employee.

What you are looking at as text in the table is formulas in the spreadsheet, not numbers.  In the spreadsheets, the original data that is used to calculate the tables is located below.  And information for additional areas can be seen by moving right.  The spreadsheet is set up to keep the titles and U.S. average in place as data for additional areas is viewed.

I am compiling this data from the point of view of New York City, where I live.  Other areas of the state are summarized for comparison.  In this tabulation, the “Downstate Suburbs” are Nassau, Suffolk, Westchester, Rockland, and Putnam.  The “Upstate Urban Counties” are Albany, Broome (Binghamton), Dutchess, Erie (Buffalo), Monroe (Rochester), Niagara, Oneida (Utica), Onondaga (Syracuse), Orange, Rensselaer (Troy), Saratoga, and Schenectady.  The other area of the state includes rural areas, small towns, and small cities, and is designated “Rest of New York State.”  Of course the data for all the individual counties of New York State is off to the right, for others to use.

Data is provided for counties elsewhere that are comparable with New York City because they include, or are comprised of, other major cities.  – Los Angeles, Cook County (Chicago), Harris County (Houston), Suffolk County (Boston), San Francisco, Philadelphia, Washington DC., etc.

For comparison with the Downstate Suburbs there is Fairfield County, Connecticut, Middlesex County, Mass, Montgomery Counties in Pennsylvania and Maryland, Palm Beach county in Florida, Orange County and San Mateo counties in California, Oakland County in Michigan, etc.

For comparison with the Upstate Urban Counties, there are such “Rustbelt” counties as Alleghany (Pittsburgh), Cuyahoga (Cleveland), Wayne (Detroit), Milwaukee, St. Louis, and Hennepin (Minneapolis).

For comparison with more rural and snowy areas of New York State, there is state level data on local government in Vermont, New Hampshire, and Maine.

Counties that include booming Sunbelt and Western cities are includes as well, such as Mecklenburg (Charlotte), Wake (Raleigh), King (Seattle), Denver, Mariacopa (Phoenix), Dallas, Travis (Austin) and Bexar (San Antonio).  State level data is included for other areas as well.

Identical spreadsheets, with everything in the exact same place, are provided for March 2007 and March 1997.

2007 Census of Governments Employment by County NY NJ Selected Others

1997 Census of Governments Employment by County NY NJ Selected Others

The structure of local government varies from place to place, with the same person often provided with, and paying for, services provided by counties, townships, municipalities, and special districts of different kinds, including school districts.  The City of New York, in contrast, is the only local government within its boundaries, save for (part of the) Port Authority of New York and New Jersey and the bi-state Waterfront Commission.  Data for the City of Los Angeles, on the other hand, does not even include its school system, which is in a separate special district.

The Census Bureau has traditionally tried to enable fair comparisons across areas by producing “County Area” files, which add up all the local governments within each county and produce one figure for all “local governments” in the county as a whole.

The downside is that some government agencies and districts provide services in more than one county, and yet all the employment, payroll, expenditures, and revenues are reported exclusively for the county where the agency is headquartered.  This makes the headquarters county’s local government seem very large, relative to its population, for that function.

In the areas I’m interested in, and for which data is provided in this tabulation, that is mostly an issue for mass transit (although, for example, the New York City Water Board also provides water to parts of Westchester).   More on that below.  In other parts of the country, however, particularly in the South, counties are smaller, and some cities are actually located in more than one county.  Taking a national view, the Bureau decided to scrap the County Area local government file.


Yes the removal of the County Area file is a change since the last Census of Governments. There are many geographical concerns with providing data in that manner since far too many governments do not fall directly within county geographies. This is an issue we are hopeful to solve through some form of shapefiles in the future.

As we have been working to prepare individual unit files however, we have actually found that we will be able to provide all units even if not reported (these units are imputed as indicated in the flags). So I believe what you are looking for will be available in the Census of Governments individual unit files by taking all individual units and looking at them in whatever combinations are necessary for your needs. We’ll be doing our best to recreate the same layout of individual unit files previously published. This means for a Census of Governments year you should find every possible unit, and we will also be ensuring all units have at least one line of data even if the government does not have any employment. We won’t have every possible function listed for every government as we’ve worked in recent years to clean up our data and only show the relevant functions each respective government actually has. If a government does not have a specific function listed then it’s effectively zero employment for that function.

So the first thing I had to do was re-create the County Area data myself, adding up each function (say police, or schools, or parks) for all the governments within all the counties I decided to tabulate, using the “sumif” function in Excel.

In most of the areas I included, I believe, adding up all local government within a county still provides the best comparison with other counties, on an employment per 100,000 residents basis, Transit aside.  Although one needs to bear in mind that (for example) the extremely high Public Hospital employment in Mecklenburg County, NC (Charlotte), relative to the population of that county, is likely the result of that hospital providing services to residents of a broader multi-county area with larger population.  (The same may be said of NYC’s relatively large private hospital employment).

To deal with multi-county (and in some cases multi-state) mass transit agencies, such as the MBTA in metro Boston, SEPTA in metro Philadelphia, etc., I divided the Transit employment for those counties not by the county population, but instead by the agency’s “service area” population, as reported the Federal Transit Administration, National Transit Database.  You’ll see “Transit Population” as a line of data separate from “Population” in the spreadsheet, in some cases.  The National Transit Database may be found here.


I also divided up the employment of the Port Authority of New York and New Jersey between NYC and New Jersey based on simple assumptions:  the Seaport and the PATH are 2/3 New Jersey, the airports are 2/3 New York, and the rest is 50-50.

The division of responsibility between local government and state government presents other issues, especially for Transit and Elementary and Secondary Education.

Here in New York, while New York City Transit and the Triboro Bridge and Tunnel Authority are reported as New York City local government, for historical reasons, the rest of the MTA – Metro North and the LIRR — is reported as state government.  And combined with other large transit agencies in the state such as the regional transit agencies for the Capital District, metro Syracuse, metro Rochester, metro Syracuse, and metro Buffalo.  With just one number for state Transit employment and one for payroll for the entire state.

I used National Transit Database employment and payroll data to allocate Census Bureau’s  State of New York Transit employment and payroll around the state – and shifted some of Metro North to Connecticut based on the number of stations in each state.  (I would have used ridership by county, but MetroNorth responded to that request by suggesting that I submit a freedom of information act request).

But New Jersey Transit really is just one statewide agency, with no separate FTA-NTD data for different parts of that state. I asked the agency to provide ridership data by county to allow an allocation of its Census of Governments employment and payroll, but it did not respond.  So while I reassigned Transit employment to Local Government at state level data, you won’t see much of anything for Transit in individual New Jersey counties.  If you are from New Jersey and can come with a basis of allocation, you can drop it in the individual counties and calculate FTEs per 100,000 residents yourself to compare with other parts of the state/region.

I also reassigned all “State Government” Elementary and Secondary School employment and payroll to “Local Government,” for the U.S. average and statewide data for individual states. Hawaii has one state government-run school district.  In some other states the state government has sometimes taken over local school districts.  That is true in New Jersey, where the state had 10,517 FTE elementary and secondary instructional employee and 5,576 non-instructional employees to go along with 11,077 transit employees.

For New Jersey as a whole, the state level Elementary and Secondary Education data is added to the local government total to be comparable with other places.  But I have no basis to divide it up the state elementary and secondary school data and assign it to individual New Jersey counties.  So in those counties where the state is operating schools, Local Government Elementary and Secondary School employment appears lower than it actually is.

In the spreadsheets, I highlight data I’ve adjusted by putting it in bold.

For public school employment, moreover, the more relevant denominator is not the overall population of a county, but the number of public school children.  I obtained enrollment data from the separately collected Education Finances data from the Census Bureau, once again adding up individual school districts to the county totals.


I had previously used this data for a FY 1997, FY 2007, and FY 2017 for a comparison of public school revenues and expenditures in New York State and around the country, which started with this post.


That dataset is very detailed on expenditures and revenues, but does not include school employment data.  In the education finance data, “instructional” expenditures refers only to teachers and a few related categories of workers, while expenditures on school administrators, principals, and others are tabulated separately.  But in the Census of Governments employment and payroll data there are only two categories for schools, “instructional” and “non-instructional.”  So in this data, all “pedagogical” employees are counted in “instructional,” including administrators.

The related private sector employment data has another issue to overcome – data suppression to ensure confidentiality. While investors can see what their stocks and bonds are worth on the market any minute of the day, employers consider data on what workers are paid to be confidential.  So unless there are enough business establishments in a particular industry in a particular county to ensure that no one could figure out the employment and pay level of an individual business, the data is suppressed.

If you want to know private hospital employment in New York City, therefore, you have to overcome the fact that Richmond County (Staten Island) data for the Hospitals industry is suppressed. Though we know that Staten Island University Hospital is there.  If you want to know private college and university employment in Upstate New York, you have to overcome the fact that private colleges and universities data for Onondaga County, Saratoga County, and Schenectady County are suppressed, even though we know that Syracuse University, Skidmore College, and Union College are there.

To overcome suppression, and at least provide data for broad areas of New York State, I’ve had to interpolate, bring in other sources, and fudge.   In the spreadsheets, if a number is in bold in the original data down below, it is massaged and estimated rather than simply added up.

Often data for a particular industry in a given county is not available for the Census of Governments year, but is available for another year(s).  If so, I used the data for the nearest year available for 1997, 2007 or 2017 as well, or perhaps modified it based on the statewide employment growth/decline trend in that industry, assuming that is the closest estimate I could get.   Een though a business must have closed to put it under the number that requires suppression). Otherwise, I had to get data from other sources, such as county “largest employers” lists.

Private sector data for the “Rest of New York State,” rural and with small towns and cities, is obtained by subtracting New York City, the Downstate Suburbs, and the Upstate Urban Counties from the New York State totals.  Due to inaccuracies in the data for those other areas due to suppression, some employment actually located elsewhere ends up counted in “Rest of New York State.”  I know that there is at least one business in the “Parking Lots and Garages” industry in Rockland County, but the number of people employed in that industry is suppressed.  That employment ends up in Rest of New York State.  That doesn’t mean someone built a parking garage in Yates County.

In addition to the state vs. local issues in education and transit, some states just have more activity managed at the state level than others.  Particularly in small states, where the state capital is never far away. Connecticut has eliminated county government, and Massachusetts is doing so.  So workers who in bigger states such as New York or California would employees of county police, county jails, county social services, county courts, and county community colleges, are state government workers in these states smaller states.  Connecticut’s local government Corrections and Higher Education employment total?  Zero.  Vermont’s government employment in Public Welfare? Ten FTEs statewide at the local government level – compared with 1,675 for the State of Vermont.  There is no way to overcome this other than to start the discussion of each function with statewide state and local government totals.

State government data, with selected local government functions included as well, may be found in this spreadsheet.

Census of Gov Employment 1997-2017 Selected States

Here is table for the U.S., New York State and New Jersey.

Data for many other states is located to the right in the spreadsheet.

Here is a chart for March 2017, comparing total state and local employment for selected states.


The second table in each spreadsheet shows average pay per employee (March payroll divided by full time equivalent employment) for each function in each area.  An example, for local government in March 2017 for the U.S., different parts of New York State, New Jersey and Fairfield County, Connecticut, is below.

This is cash pay only, and does not include the considerable (and in New York City particularly massive) compensation (many) public employees receive in the form of non-wage benefits, notably early retirement on rich terms.

Those reviewing payroll per FTE employee data need to keep in mind factors that sometimes cause these figures to mis-represent how much a type of government worker is paid in one area compared with others.

First, since public employees tend to earn progressively more during their careers, places where employment is increasing will show lower average pay, because there will be relatively more recently-hired, lower-paid workers as a share of the total.  And places where government employment is shrinking will have fewer new hires to pull down the average, making average pay appear higher.

Second, a particular group of unionized workers may be working without a contract, and may in fact eventually be paid more for March 2017 than the Census of Governments for that month shows, later when that contract is settled.

This is a frequent situation in New York City. The mean payroll per FTE for New York City police officers, in particular, seems lower than it really is because NYC police officers had been working without a contract for years in March 2017 – and March 2007 and March 1997 (and many of the years in between).   They were paid more for that time after the contracts were settled.  To further confound my attempt to provide a fair comparison, NYC police officers were paid 50.9% more than the U.S. average in March 2012, but only temporarily due to 9/11-related overtime.  In March 2012, the mean payroll per FTE NYC police officer was 23.6% above the U.S. average, probably a true measure of the difference.

Third, in some functions with very high pay per employee, this may be an artifact of not having the most actual workers included in the totals.  Take the Air Transportation function in metro New York, run by the Port Authority of New York and New Jersey.  The management and operation of those airports is contracted out.  The employment and payroll data only shows a very small number of very highly paid contract managers.  (The expenditures data from the finance phase of the Census of Governments will provide a fuller picture).

Although in many cases public employees in the NY Metro Area earn more than the U.S. average, so do private sector workers, and the cost of living (and ability to pay taxes) are higher here as well. To make the comparison fair, I show not only the extent to which the mean payroll (cash only) per local government worker exceeds the U.S. average, but also the extent to which the mean earnings per private sector worker exceeds the U.S. average.  For places where private-sector earnings are inflated by Wall Street (Manhattan and Fairfield County), I show mean private earnings per worker without the financial sector as well.

In the past, I have used Employment and Wages data from the BLS for Downstate New York for this comparison.  That data shows that excluding finance, the mean payroll per worker for private wage and salary workers in Downstate New York was 28.4% higher than the U.S. average in 1997, 32.3% higher than average in 2007, and 27.3% higher than average in 2017.

But this time I have decided to use the BEA private sector earnings data, which includes the self-employed (a growing share of the workforce in the so-called “gig” economy) and includes non-wage benefits (for those who still have them).  Because it is for a year rather than a month, and includes benefit income, the BEA data I present for the private sector is not comparable with the Census of Governments data for the public sector, which is monthly wages and salaries only, and full-time equivalent.  But the percent above the U.S. average iscomparable, and that’s the figure to focus on.

The BEA data (and school enrollment data) for 1997, 2007, and 2017, which is interesting in its own right (including the self-employed and looking at those working – rather than living – in the city makes the city of Philadelphia’s economy seem better off than I had known), is in this spreadsheet.

BEA County and State3

For NYC and most other counties in metro New York, I’ve used mean earnings per private sector worker for the metro area as a whole, excluding Wall Street, as my comparison.  In general, I use metro area private earnings data for other areas for other areas as well.  But I use other figures for Orange and Dutchess counties in New York, and Sussex and Ocean Counties in NJ.  These are in the NY Metro because many workers commute from there to closer-in suburbs, but they are far from the extremely well-paid jobs in Manhattan.  The area used for comparable private-sector data is noted in the table.  For the other counties in New Jersey, the comparison is with the portion of that state that is located outside metro New York, where mean earnings per worker was 8.2% above the U.S. average in 1997 but 0.6% below that average in 2017.

In general, private sector workers are earning less, relative to the U.S. average, in the NY metro area, New Jersey, Connecticut and Upstate NY than they used to, particularly if Wall Street is included (but Wall Street earnings are going down too).  But Bureau of Economic Analysis data shows that private sector workers are earning less, including benefit costs, relative to state and local government workers across the country.  And median cash pay, less affected by the one-percenters, has been going down for some time, with Millennials paid 25 percent less than Baby Boomers had been at the same point in their lives, despite higher educational attainment.

For the U.S., New York City, other areas of New York State, New Jersey and Fairfield County, this spreadsheet shows local government employment and payroll data for March 1997, 2007 and 2017 side by side so one can see the trend.

Revised census-of-governments-employment-ny-nj-1997-2017

Here is a PDF, for the U.S., New York City, and the Downstate Suburbs on employment…

Revised census-of-governments-employment-ny-nj-1997-2017

And mean payroll per worker.

The three years of data for Upstate New York, statewide data for New Jersey, and Fairfield County can be viewed in the spreadsheet above.  And data for still other areas can be dropped into that spreadsheet (“paste special,” “values”) to examine their trends over time.  Since everything is in the exact same row in each of the local government spreadsheets.

I’m only including these JPEG “picture” files to show people what is in the spreadsheets, and tempt them to download.  It is the spreadsheets that should be looked at.


So in an overall sense, what are the trends?

From 2007 to 2017, local government employment fell relative to population in most states, and most of the area around NYC, after having risen from 1997 to 2007.


The biggest factor is the exit of the large millennial population from childhood, and from the public schools, and therefore a relative decrease in local government Elementary and Secondary Education employment.  For the U.S. as a whole, instructional employment in the public schools fell from 1,559 per 100,000 U.S. residents in 2007 to 1,463 in 2017.  The number of students per instructional FTE, however, fell from 10.3 in 2007 to 10.2 in 2017.

A relatively smaller number of public school children is not the entire explanation, however, as for many other functions local government employment fell relative to population as well.

New York City’s local government employment increased from 431,540 full time equivalent in 2007 to 437,982 in 2017, but the city’s population increased faster, and local government employment fell per 100,000 city residents, in total and in most categories.

For the “Rest of New York State,” however, population and school enrollment decline is leading to a choice of drastic local government employment decreases or an increasing tax burden (on someone). There, local government FTE employment increased from 122,167 in 1997 to 135,145 in 2007, but then fell to 133,066 in 2017.  But the total population in the Rest of New York State fell from 2007 to 2017, and the number of public school children fell 12.8%, causing local government employment to rise compared with that population.

These are the sort of declines that were seen in New York City (and many other older central cities) in the 1970s.

Today, no doubt many similar rural areas in the Northeast and Midwest have similar issues.  Governor Cuomo has attempted to wean Upstate New York off state-funded local government employment as an economic base, but thus far private sector success has been limited, and the richer generations that have retirement income from the unionized manufacturing jobs of the past are dying off and/or moving away.


Taxes Increase 17% For Elmira Property Owners, Kinda By Mistake

In contrast with local government, state government employment increased relative to population in most states from 2007 to 2017.

Here again the movement of the Millennials through life may be the biggest factor, as the biggest increases are in state Higher Education employment.  But at least at the national level, instructional employment seems to have increased faster than enrollment, and non-instructional employment appears to have increased faster than instructional employment. And in any event, state government higher education may have peaked and started to fall, now that the Millennials are aging of out that government function.

While state government employment, in total, increased per 100,000 residents in most states from 2007 to 2017, inflation adjusted payroll per state worker fell.

No doubt rising employment was a factor. There are more new, and thus low paid, workers pulling down the average. But that is not the only factor. Pension costs have soared across the country, either due to retroactive pension increases beyond what public workers have been promised, the past failure of taxpayers to fund the pensions workers were promised to start with, or a combination.   As result total state and local government wages and salaries are down as a share of the personal income of all state workers, as discussed here, while benefit costs are up.


On the other hand, local government payroll per FTE employee increased when adjusted for inflation, in the U.S. and the various parts of New York State.

No doubt the falling number of local government workers was a factor, as there were fewer low-paid newbies bringing down the averages.  Meanwhile, local government pension costs have also soared, notably in New York City.

For cash income, mean New York City local government worker payroll averaged $6,271 per FTE in March 2017.  Multiplied by 12, that is a mean payroll per worker of $75,252 per year.

To put this in perspective according to the American Community Survey data, adjusted for inflation the median cash earnings of all workers residing in New York City fell 0.5% over a decade to $38,430 in 2017.   The median earnings of men working full-time year-round, as most local government workers do, fell 0.2% to $52,210, but the median earnings of women working full time year-round rose 3.9% to $50,735.


Finally, the Census Bureau would like you to know its data isn’t perfect, and has asked that this statement be included.

2017 State & Local Government Employment & Payroll

Individual Unit File Disclaimer

The individual unit data are reviewed and edited by the U.S. Census Bureau solely with the purpose of creating the most accurate estimates at the state and type of government level, and should not be viewed as an accurate time series for any individual unit.

If these data are released to others, you may cite the U.S. Census Bureau as the source of the original data only; however, the following reliability and cautionary statements should be included with the release:

  • The Census Bureau has not reviewed the individual unit data as separate time series, therefore caution must be exercised in their use and interpretation.
  • Data users who create their own estimates using data from these individual unit files should cite the U.S. Census Bureau as the source of the original data only. The Census Bureau has not sanctioned, conducted, or reviewed any analysis using these estimates. Conclusions drawn from any analysis of this data are the sole responsibility of the performing party.
  • Although every effort is made in all phases of collection, processing, and tabulation/ estimation to minimize errors, the sample data are subject to non-sampling errors (such as inability to obtain data for every variable from all units in the sample, inaccuracies in classification, response errors, misinterpretation of questions, mistakes in keying and coding, and coverage errors). These same errors may be evident in census collections and may affect the Census of Governments data (years ending in a ‘2’ or ‘7’), which are used to adjust the sample during the estimation phase and used in the imputation process.
  • A discussion of the methodology and processing used for the Survey of Public Employment & Payroll can be found at:


Tables explaining the contents of the text-based files and the excel worksheet are contained in the individual unit file technical documentation (2017 COG-E Individual Unit File Tech Doc.pdf).

The Census Bureau recommends that any results based on these data should include cautionary statement concerning the data limitations and the potential influences of sampling and non- sampling errors, as noted above.

There is that old data game I learned long ago. If the data shows what you want it to, you use it, because it is the best available, and better than public policy being based on nothing.  But if the data doesn’t show what you want it to, you point out that it is imperfect, and bring your own “alternate facts” to support your own self interest.

This data is in fact the Bureau’s best estimate to measure state and local government as part of the Economic Census program, so I’m putting it out there.  And in Census of Governments years it is supposed to include the full data for all state and local governments, not just a sample used to create state-level estimates.

As for non-sampling error, over the years when I’ve found errors for New York, I’ve pointed them out to the Bureau.  But for other places I don’t know as well, I’m not in a position to do so.  All I can say about the data there is that it is the best available, and huge changes between 1997, 2007 and 2017 that go in a different direction from the statewide totals and similar places may not be real.

1 thought on “Census of Governments Employment and Payroll Data for March 2017 (1997 and 2007)

  1. Pingback: Public Elementary and Secondary Schools: Census of Governments Employment and Payroll Data for 2017 | Saying the Unsaid in New York

Comments are closed.