The Fiscal Impact of the Census Undercount on Cities  A 34-City Survey
January 1999



The Extent of the Undercount

The number of persons estimated to have been undercounted in the survey cities represents an average of four percent of those cities’ populations. The number undercounted ranges from 159,000 in Chicago and 138,878 in Los Angeles, to 21,600 in San Francisco and 21,500 in Long Beach, to zero or an insignificant number in Cuyahoga Falls, Green Bay, Pembroke Pines and Scottsdale. The following table shows the total population and the undercount for the cities able to estimate the number undercounted.

City Population Number Undercounted Percent
Anaheim, CA 282,133 7,334 2.6
Charleston, SC 80, 414 2,500 3.1
Chicago, IL 2,731,743 159,000 5.8
Cincinnati, OH 358,170 11,000 3.1
Covington, KY 43,264 3,800 8.8
Cuyahoga Falls, OH 48,950 0 0
Dayton, OH 178,540 4,187 2.3
Fayetteville, AR 52,662 1,477 2.8
Green Bay, WI 102,708 insignificant 0
Henderson, NV 122,339 4,000 3.3
Kansas City, KS 149,000 na 2.0
Lafayette, IN 48,764 858 1.8
Long Beach, CA 433,852 21,500 5.0
Los Angeles, CA 3,553,638 138,878 3.8
McAllen, TX 84,021 2,216 2.6
Minneapolis, MN 354,590 6,041 1.7
Pembroke Pines, FL 101,500 insignificant 0
Pinellas Park, FL 45,000 200 0.004
Plano, TX 200,000 1,000 0.5
St. Joseph, MO 71,7111 5,000 7.0
St. Petersburg, FL 240,000 3,497 1.5
San Francisco, CA 734,676 21,600 2.9
Scottsdale, AZ 197,012 0 0
Spokane, WA 197,044 2,699 1.4
Youngstown, OH 95,732 4,424 4.6

Among the cities’ comments on the undercount:

Charleston: Charleston has had to contend with the inaccurate 1990 census count and annual population estimates for years now. Even the census count of 80,414 persons resulted only after a very extensive and time-consuming challenge of the 1990 census undercount. As the City felt its population was 83,000 persons in 1990, this low number has caused the City considerable loss of revenue over time.

Covington: The City of Covington has an estimated 3,000 homeless individuals who were not counted during the last census. We had one entire apartment complex which was not included in our initial count. This was adjusted, however, in the raw count later. City Housing Department staff have indicated that the number of persons per household in our 800 Section 8 apartments should be between three and four, rather than the reported 2.3 individuals. Based upon the undercount of the homeless, our corrected count of 19,117 dwelling units and a conservative estimate of 3.3 persons per Section 8 household, we estimate an undercount of about 3,800 individuals.

Dayton: In 1990, the City of Dayton worked very closely with the U.S. Census Bureau to assure an accurate count for our city. Due to extensive "Were You Counted?" efforts, Dayton’s final 1990 population count increased by as many as a few thousand persons. However, due to our high poverty rate – both African-Americans and Appalachians – it is quite likely Dayton’s population was undercounted....Assuming we were undercounted to the degree Milwaukee was believed to be undercounted – 2.3 percent – Dayton could have been undercounted by 4,187 people.

Green Bay: We do not believe that there was a significant undercount for our city.

Henderson: In the 1990 census the City of Henderson initially suffered an undercount estimated at approximately 4,000. In a city of 60,000 at that time, you can appreciate that a discrepancy of this nature would have been quite substantial. However, after filing a petition for adjustment – and, we might add, with not much assistance from the Census Bureau – our official population was adjusted to include the 4,000.

Hendersonville: Our city was able to get undercounts corrected during the federal census post-review period.

Kansas City: Kansas City, KS is an older urban city with a significant minority population, nearly 35 percent. The Census Bureau has estimated that population undercounts are more likely in communities with a comparatively higher percentage of low-income residents and minority concentrations. A review of 1990 census data and school enrollment totals indicate an approximate four to five percent undercount of the minority population. Assuming this percentage for the city’s minority population only results in an estimated total undercount of two percent.

Kokomo: While Kokomo did lose a small amount of population between the 1980 and 1990 counts, we never believed it to be because of an undercount. We know the population did decrease, and we note that since then it has been rising once again. We are working at the local level to assure the 2000 count is as accurate as possible.

Lafayette: One of our biggest problems for the census is the undercount of students, many of whom do not consider themselves residents and prefer to be counted at home, and who may have left for the summer by the time of census follow-up.

Long Beach: Records for the Census Bureau’s corrected 1990 count, based on data from the initial post-enumeration survey, showed that the unadjusted census missed at least 21,500 people, or about four percent of the city’s official 1990 population of 429,433 persons....If adjusted, the City’s actual population as of April 1990 would, therefore, have been 451,000....While a subsequent analysis of the post enumeration survey by the Census Bureau in 1992 revised the estimated undercount for Long Beach downward 3.7 percent, this figure still represents a serious undercount for the City. In fact, post-census studies conducted as part of the City’s efforts to have the 1990 undercount officially recognized showed that the population of Long Beach in 1990 was, in fact, significantly larger than either PES. A sample survey conducted in several blocks having a significant concentration of Cambodians, for example, revealed an undercount of about 18 percent. The personal estimate of the City’s Advance Planning Officer is that the undercount in Long Beach was more on the order of 30,000 to 40,000 persons.

A large portion of the uncounted Hispanic and Asian residents were also in their teens and 20s, have remained in Long Beach and subsequently married and formed families – families that are generally larger....It is the recent immigrants and their children that have driven population growth in Long Beach during the 1990s. Because of the sizable undercount in 1990, it is therefore likely that Long Beach’s population is significantly larger than any of the current estimates based on the official 1990 census count. Indirect evidence of this may be found in the fact that the Long Beach Unified School District’s current academic year student population of 89,408 persons is nearly 3,000 students greater than the 1999-2000 enrollment projection established in 1996.

Pembroke Pines: Pembroke Pines, FL is a rapidly growing upper-middle-income city that in 1990 had approximately 65,000 people and in 1998 had 120,000 people. Any undercount that we experienced in 1990 was primarily because the "window" of the count caught only a brief picture of our population. When we are increasing at 7,000 to 9,000 people per year, any one month doesn’t accurately reflect the population for very long.

Pinellas Park: The City of Pinellas Park was impacted minimally. The City’s Planning Director was part of the local team participating in the preliminary counts and addressed the identification process leading up to the 1990 census. The difference between the actual census count and the population count estimated by the local team was determined to be about 200 persons.

Plano: We estimated that the 1990 census count for the City of Plano was approximately 1,000 less than our actual population. Based on funding formulas, it is unlikely that Plano lost federal and State funding as a result.

The Financial Loss to the Cities

Across the 34 cities, the estimated total loss in federal and state funds during the1990s resulting from the 1990 undercount was $536 million. The loss ranged from $184 million in Chicago and $120 million in Los Angeles; to $20.5 million in San Francisco, $13.3 million Spokane, $13 million in Covington and $10.6 million in McAllen; to $1 million in Charleston, $950,000 in Kansas City, $581,401 (in State funds only) in Fayetteville and $40,000 in Lafayette.

The amount lost to the cities during the 1990s averaged $1,230 for each person not counted in the city, or $56 per person for each person in the city who was included in the city’s 1990 population count. The table on the following page shows the estimated total and per capita loss in federal and state funds for each city.


City Total Federal and State Funds Lost in the 1990s Per Capita Loss per
Total per Population
Per Capita Loss Undercounted Popuplation
Anaheim, CA $ 1,587,619 $               6 $            216
Charleston, SC 1,000,000 12 400
Chicago, IL 1,84,440,000 68 1,160
Cincinnati, OH 40,000,000 11 3,636
Covington, KY 13,000,000 300 3,421
Cuyahoga Falls, OH 0 0 0
Fayetteville, AR 581,401 11 394
Green Bay, WI 0 0 0
Henderson, NV 4,000,000 33 1,000
Hendersonville, TN 0 0 0
Kansas City, KS 950,000 6 na
Lafayette, IN 40,000 1 47
Long Beach, CA 10-15,000,000 581 29
Los Angeles, CA 120,000,000 34 820
McAllen, TX 10,6000,000 126 4,810
Minneapolis, MN 8,800,000 25 1,460
Pembroke Pines, FL 0 0 0
Pinellas Park, FL insignificant insignificant insignificant
Plano, TX 0 0 0
Rantoul, IL 0 0 0
Roswell, NM 4,000,000 100 na
St. Joseph, MO 40,000,000 558 8,000
St. Petersburg, FL 1,300,000 5 364
San Francisco, CA 20,520,000 28 950
Scottsdale, AZ 0 0 0
Spokane, WA 13,339,000 68 4,942
Youngstown, OH 3,230,000 34 730

Among the city officials’ comments on loss of federal and state funds:

Anaheim: City officials estimate the following program revenue losses for the decade:

CDBG                                                              $ 845,551
JTPA                                                                    675,800
Local Law Enforcement Block Grant                       44,219
Emergency Shelter Grant Program                           22,050

TOTAL                                                           $1,587,619

Charleston: We have used the figure of $4,000 as the amount of revenue generated per person per year. The 2,500 per person deficit could mean a loss of $100,000 each year for the City through State and federal funding grants, etc.

Chicago: The City’s estimate of lost federal aid is based on a study which used formulas for five major programs and estimated the impact for 96 other programs. This study is conservative because it used 1989 funding levels and does not reflect increases in entitlement funding during the 1990s. It is also important to remember that the cost of the undercount does not directly correspond to the impact of sampling since sampling will not capture the entire undercount. Among the factors in determining financial impact:

  • Funding formulas are very complex. It is not accurate to simply add up the amount of grants allocated based on population and determine the dollar amount lost per person. Formulas provide different amounts of funding for different types of people. Typically, the most frequently undercounted groups, such as low income individuals and children, receive more than the standard funding allocation.

  • Studies do not include all the grants which consider population, especially those passed through state and regional agencies.

  • Studies do not include competitive grants, where increased documented need from census data can increase the ability to obtain grants.

  • Even with statistical sampling, there will not be a 100 percent count. Therefore, we will never be able to capture all of the lost revenue.

  • The amount of funding available is limited by the amount by which other municipalities increase their populations.

  • The amount of funding available may increase based on documented increased need.

  • The amount of funding available may increase based on increased representation.

Covington: Not only does the undercount impact funding based upon total city population, it also impacts those programs which are based upon need and income level. The latter is important because the undercount in our city was predominately skewed toward the low income segment of the population.

  • CDBG funds are based upon need and population counts. Our City is attempting to forge a coalition with four other Northern Kentucky river cities to apply for HUD block grant funds. If the designation is approved, it could be worth approximately $2.7 million per year in funding.

  • Empowerment zone designation is based upon need demonstrated by each city as indicated by census tract income levels. We have applied for the $100 million in the second round of designations currently being made which HUD will disclose in January 1999. (Editor’s Note: When these designations were announced, Covington was not included.)

  • EDA has a program for economic development which we were told we did not qualify for because our per capita income level was too high – a loss of $10 million in potential funding.

  • We did not receive State CDBG residual funds, which were redistributed based on demonstrated need as reflected in census tract and city income levels. This amounted to a loss of $3 million which we had requested to build an access road to a development site.

Deerfield Beach: The City of Deerfield Beach only recently secured a successful challenge and convinced the U.S. Bureau of the Census that we are a community of more than 50,000, rather than 47,000, as was originally included in the 1990 census. What does this mean to the City of Deerfield Beach in terms of dollars? We had to secure a Washington, D.C. lobbyist and pay over $60,000 to cut through the bureaucracy and fight for a real population count, which was finally recognized by the Bureau of the Census. Unfortunately, we found out that was only a quarter of the battle, and we have now spent additional monies, time and effort to convince the Department of Housing and Urban Development that we are now eligible to be a CDBG entitlement community. We will not receive funds as an entitlement community until 1999, so it is fair to say that for eight years our community has been unable to secure federal monies under the HUD CDBG entitlement program because of an undercount in population in 1990. A conservative estimate of monies lost is $2.5 million, plus another $200,000 in staff time and consulting fees required to convince the federal government of the facts that should have been recognized objectively back in 1990.

Kansas City: In Kansas, cities receive three revenues directly from the State based on their relative share of population. The revenues include monies from 1) a local ad-valorem tax reduction fund, 2) a city-county revenue sharing fund, and 3) a city street and highway fund. During this decade the City’s annual share of these funds, distributed on the basis of population, is $4.7 million. A two percent increase in these funds generates an additional $95,000 per year, or nearly $1 million for the decade. Nearly three-fourths of the monies are received from the highway distribution. It is more difficult to estimate losses from federal programs. CDBG appears to be less impacted. The Kansas City CDBG entitlement is based on a formula that considers age of housing units, poverty rate, and population growth lag. An overall loss in population could increase the CDBG allocation. It could be argued, however, that the population undercount results in a lower poverty rate, and reduces CDBG funding.

Lafayette: Very few taxes in Indiana are distributed solely on the basis of population, as are "sin" taxes on alcohol and cigarettes. This translates into lost revenue of about $4,000 annually. Population is a factor in State road, street and highway distribution, along with highway mileage. We cannot estimate the dollars lost here associated with population. Population is only one factor in many federal programs as well, making it difficult to estimate loss. Our understanding is that declining population is actually a plus in the Community Development Block Grant program as an indicator of "distress."

Long Beach: The failure to include even the recognized population undercount present in the 1990 census has had serious financial repercussions for the City. According to a study of federal funding in 1990, it is estimated that each miscounted person represented a loss of $56 per year. The Los Angeles County Urban Research Unit currently estimates that the value of each "census person" is more on the order of $100 per person per year. The 21,500 missed residents, therefore, represent a likely loss in funding to the City of between $10 million and $15 million (perhaps more) during the period 1990-2000. If the recognized undercount was, in fact, only about half of the true figure, then the loss to the City could be as high as $30 million to $50 million over the decade.

Pembroke Pines: As an upper-middle-income city, there are very few residents in groups that are often undercounted....We don’t feel that the failure to reach the traditional undercount groups greatly affected us in 1990.

Pinellas Park: The impact upon federal and State funding received by the City was considered insignificant.

St. Joseph: It would be fair to say that the programs hit hardest would be State funding (gas tax revenues, etc.), followed by State/federal transportation dollars, followed by the CDBG program.

San Francisco: The amount is a very conservative estimate of the adverse financial impact of the undercount on the City because we are unable to determine the total per capita amount the City receives from the State and federal governments. It is known, however, that a number of the formulas used by the federal government to allocate funds in various programs include the number of people who are part of a socioeconomic group – for example, those living in poverty. Since such groups are the ones that historically are the most likely to be undercounted, the loss of federal funds in a city with large portions of such populations is particularly profound.

Youngstown: The estimate is based on CDBG, HOME and ESG dollars and an estimated undercount based on a comparison of housing units and population.

Estimated Future Financial Loss to the Cities

Twenty of the cities were able to estimate the amount they would lose in federal and state funds during the first decade of the 21st century if the 2000 census maintains the same level of inaccuracy as that of the 1990 census. Their estimates total $677 million. The loss would range from $227 million in Los Angeles, $184 million in Chicago and $80 million in St. Joseph; to $24.6 million in McAllen and $22.7 million in Covington; to $5 million in Roswell, $1.3 million in Saint Petersburg and $950,000 (State funds only) in Kansas City.

This would average $2,263 for each person not counted in the city, or $129 per person for each person in the city who is included in the city’s 2000 census count. The table below shows the expected total and per capita loss during the 2000s in federal and state funds for each city able to make an estimate:

City Total Federal and State Funds Lost in the 2000s Per Capita Loss per
Total per Population
Per Capita Loss Undercounted Popuplation
Anaheim, CA $             2,84,915 $            8 $       382
Chicago, IL 184,440,000 68 1,160
Covington, KY 22,700,000 525 5,974
Cuyahoga Falls, OH 0 0 0
Green Bay, WI 0 0 0
Henderson, NV 12,000,000 99 3,000
Kansas City, KS 950,000 6 na
Los Angeles, CA 227,000,000 59 1,550
McAllen, TX 24,526,342 226 8,578
Minneapolis, MN 8,800,000 25 1,460
Pembroke Pines, FL 7,612,500 75 na
Pinellas Park, FL insignificant insignificant insignificant
Plano, TX 0 0 0
Rantoul, IL 0 0 0
Roswell, NM 5,000,000 125 na
St. Joseph, MO 80,000,000 1,116 8,000
St. Petersburg, FL 1,300,000 5 364
Scottsdale, AZ 0 0 0
Spokane, WA 15,375,000 78 5,336
Youngstown, OH 3,230,000 34 730

Among the city officials’ comments on future losses due to an inaccurate 2000 census:

Anaheim: City officials estimate the following program revenue losses for the first decade of the 21st century:

CDBG                                                           $1,385,122
JTPA                                                               1,144,231
Local Law Enforcement Block Grant                   220,458
Emergency Shelter Grant Program                         55,104

TOTAL                                                         $2,804,915

Augusta: Based on recent estimates, we believe that the 2000 census will be 2.5 percent low. We have no idea what the financial impact would be, however.

Caguas: For Census 2000, the Municipality of Caguas is collaborating in the preparation and review of census maps to ensure that new housing developments (built between 1990 and 2000) are included. The statistical area subdivisions (census tracts, block groups) are being reviewed to assure that they are representative of the real socioeconomic characteristics of the different communities in Caguas. This aspect is of special importance in requesting State and federal assistance for low income families.

Charleston: An accurate count can definitely mean more revenue for the City. Every person added to the population count increases the opportunities for non-profits and public entities to qualify for grants and federal assistance. The health of the economy is also affected. The attraction of new business to the area in anticipation of higher volumes of customers or of an accessible employee pool begins with a market strategy based on population counts.

Detroit: The 1990 census final population total for Detroit was 1,027,974. If less than one million Detroit residents are counted in the 2000 census, the following are the three largest negative impacts on City revenues that will come into play:

  • The Utility Tax – the City of Detroit will not be able to levy this tax if the 2000 census finds less than one million residents. In FY96-97, the Utility Tax brought in $54.7 million. This revenue is dedicated to public protection and enables the City to hire additional police officers.

  • The Income Tax – Michigan cities with over one million residents are allowed to tax residents’ incomes at three percent and the incomes of non-residents who work in the city at one-half of the resident rate (1.5 percent). Both rates will be reduced by one-third if Detroit’s population does not reach one million....In FY96-97, the reductions from these revenue sources would have been $103,000,000. Income tax revenues go into the general fund and finance those departments that are not supported by grants or other revenues.

  • State Revenue Sharing – The State of Michigan returns a portion of the funds it collects to local units of government. The State uses a three-part formula to determine how much each unit of government is to get back. Parts of the formula are based on that unit’s total population and on the State’s total population. Assuming the population of Michigan to be 9,786,700 in 2000 (a five percent increase), a decline in Detroit’s 2000 census population to 1,000,001 would result in a reduction of $16.5 million in State revenue sharing funds. If Detroit’s 2000 census population were 990,000 in 2000, Detroit’s share would be reduced to 18.4 million. Thus for each decline of 10,000 people within the current population ranges, Detroit’s share of funds will be reduced by nearly $2 million annually.

If Detroit’s total population drops from 1,027,974 to 990,000 in the 2000 census, the impact from these three primary revenue sources will be significant. In terms of revenues collected at the FY96-97 levels, the losses identified would be:

Utility Tax                                                    $ 54,700,000

One-third Reduction in Income Taxes            103,000,000

Reductions in State Revenue Sharing              18,400,000

TOTAL Annual Reduction                           176,100,000

In FY96-97, appropriations totaling $1,244,373,075 were authorized to operate tax-supported general fund agencies. In FY96-97 dollars, the immediate annual impact from loss of the above resources would amount to at least a 14.2 percent reduction in general fund revenues.

Henderson: If the same level of inaccuracy were to exist in the 2000 census, we estimate that the impact would be at least three times as great. This estimate is due in part to our rapidly expanding populations. In our initial discussions with the Census Bureau representatives it seems they are attempting to accommodate some of our unique challenges as they craft their count strategy.

Hendersonville: The pre-review for the 2000 census insures that the federal government has a complete address listing. But without post-review, there is no guarantee that all areas will be counted.

Long Beach: The conditions that resulted in Long Beach suffering a significant undercount of its residents in 1990 have not only continued, but have intensified. The City, therefore, remains more likely than many other areas of the country to be adversely affected if sampling is not used for the 2000 census. Test have consistently shown that communities having a large, relatively recent immigrant population, as well as those with a relatively large proportion of its households living in rental units, are especially prone to undercounts. Several indicators support the contention that there is an increased risk of a significant undercount if statistical sampling is not used in 2000. The City, therefore, considers it essential that the 2000 census be as accurate as possible, and has joined in various efforts to ensure this outcome.

Omaha: Electronic sampling will cause Omaha and Nebraska to risk lower capitation reimbursements.

Pembroke Pines: As our population becomes more diverse, an undercount will be an increasing concern....We believe that an undercount in 2000 will cost us at least $7.50 per resident per year in federal funds.

Pinellas Park: If the 2000 census maintains the same level of accuracy as was realized in the 1990 census for Pinellas Park, again we expect the impact upon the community would be considered insignificant.

Plano: Given a proportional undercount in the year 2000, Plano would probably not lose funding in the next decade. This does not mean that the City of Plano is unconcerned about significant census undercounts in the year 2000. This is because rapid growth and development in Plano is adding more than 4,000 new housing units each year. New subdivisions, new streets and new houses and apartments will be constructed almost simultaneously with the 2000 census count. In addition, workers moving to Plano to support the new construction may complicate the census process. For these reasons, we are extremely concerned about the potential for increased inaccuracies in the 2000 census. It is essential, therefore, for the Census Bureau to develop an accurate, expandable database to support the census process.

San Francisco: The impact of the undercount will be greater in the next decade if the Census 2000 reflects the same inaccuracy. The City is more likely than many other areas of the United States to be adversely affected if sampling is not used in Census 2000. Studies have shown that communities having a large, relatively recent immigrant population, as well as those with a relatively large proportion of their households living in rental units, are especially prone to undercounts.

Recent Immigrants: The 1990 census reported that 40.3 percent of the City’s population over the age of five spoke a language other than English at home, an increase from 34.2 percent of the population in 1980 for whom English was not the primary language. The City’s immigrant population continues to increase by even higher percentages. Between the 1980 and 1990 censuses, 54,000 immigrants came to the City. According to the California State Department of Finance, over just seven years (1991-1997), the City’s net immigration has been an additional 66,000 people. Clearly, the population who may not be counted because of language barriers is growing.

Homeless: The City’s significant homeless population also continues to increase. The Census Bureau’s efforts in 1990 to count this group resulted in about 2,000 people being found. According to other studies, this number represents an undercount of at least 4,000. Since 1990, the homeless population has grown so that, according to some estimates, on any given night, 6,000 to 8,000 San Franciscans may be homeless. Further, over the course of a year, an estimated 11,000 to 16,000 San Franciscans experience at least one episode of homelessness. To make matters worse, families with children make up roughly 25-30 percent of the homeless population. This means that many children are not being counted. In addition, many homeless San Franciscans suffer from disabilities such as substance abuse, mental illness, or HIV/AIDS, thus being an especially vulnerable segment of a population that is already difficult to enumerate.

Rental Units: In the City, 65 percent of residents rent their homes and 35 percent own their homes, which is the reverse of the percentages nationwide.