Skip Navigation
Analysis

Major Census Quality Check Spotlights Persistent Undercounts

The results of the Census Bureau’s Post-Enumeration Survey reveal inequities that require concerted action to fix.

Last Updated: May 24, 2022
Published: March 15, 2022

Over the past few months, the Census Bureau has been publish­ing the results of its major qual­ity check on the 2020 Census. While the bureau’s initial find­ings from the Post-Enumer­­a­­tion Survey suggest that the census very accur­ately coun­­ted the total number of people living in the coun­­try, they also show that this decade’s data suffered from the same seri­ous prob­lem as past censuses: signi­fic­ant under­­­counts in many communit­ies of color. The bureau’s subsequent find­ings, moreover, reveal that not all states were coun­ted equally well.

These data prob­lems complic­ate the census’s abil­ity to perform its basic func­­tions and should spur action for future fixes.

What is the Post-Enumer­a­tion Survey?

The Post-Enumer­­a­­tion Survey is a sample survey the bureau conducts shortly after the decen­nial census. Its goal is to meas­ure how well the Census Bureau coun­­ted people and captured basic demo­­graphic inform­a­­tion, specific­ally the age, sex, race, and ethni­­city of each person in the coun­­try. This data is used to produce two estim­ates that help eval­­u­ate the coun­­t’s accur­acy. The first is how many people were under­­­coun­­ted or over­­­coun­­ted in the census, known as “net cover­­age error.” The second is how many people were correctly coun­­ted, erro­neously coun­­ted, or missed alto­­gether, known as “compon­ents of cover­­age.” This data gives the public a sense of how accur­ate the 2020 Census ulti­mately was compared to prior censuses, and it gives the bureau a sense of what it needs to improve before the next one.

Why is it import­ant to have an accur­ate census?

The stakes of the census are high. Decen­nial census data is used to divide seats in the House of Repres­ent­at­ives and Elect­oral College votes among the states. It is also used to redraw district maps used for voting and to distrib­ute over $1 tril­lion in fund­ing every year for services like food, schools, and health­­care. Getting the count right is essen­­tial to ensur­ing that every­one receives their fair share of demo­cratic repres­ent­a­tion and govern­­ment support.

The stakes of this decade’s census are espe­­­cially high for communit­ies of color. The nation’s popu­la­­­tion growth over the past 10 years came entirely from nonwhite communit­ies. For these communit­ies to receive polit­ical repres­ent­a­­­tion and crit­ical invest­­­ments commen­sur­ate with their size, they must be coun­­­ted fully.

Inac­­curacies in census data that fall along racial lines, in turn, perpetu­ate systemic discrim­in­a­­tion against and disen­­­fran­chise­­ment of communit­ies of color by depriving those groups of polit­ical power and public services.

What are the major takeaways from the Post-Enumer­a­tion Survey?

The pandemic and unpre­ced­en­ted meddling by the Trump admin­is­tra­­tion led to worries about a disastrous 2020 count. The Post-Enumer­a­tion Survey results indic­ate that the national popu­la­­tion count was unaf­­fec­ted by these chal­lenges — the 2020 Census did not have a stat­ist­ic­ally signi­fic­ant under­­­count rate over­­all. However, that national stat­istic hides the more alarm­ing trend of racially inequit­­able under­­­counts, an issue the census has consist­ently faced. The top-level results also mask dispar­it­ies in the coun­t’s accur­acy among the states.

The Post-Enumer­­a­­tion Survey shows that in 2020, people who identify as Black, Hispanic/Latino, Native Amer­­ican and Alaskan Natives living on reser­­va­­tions, or Some Other Race were signi­fic­antly under­­­coun­­ted, while white non-Hispanic and Asian people were over­­­coun­­ted, and Native Hawaiian and Pacific Islander people were neither over­­­coun­­ted nor under­­­coun­­ted. The Hispanic popu­la­­tion, moreover, went under­­­coun­­ted at over three times the rate it was in 2010.

Mean­while, six states were under­coun­ted at a stat­ist­ic­ally signi­fic­ant rate in 2020: Arkan­sas, Tennessee, Missis­sippi, Flor­ida, Illinois, and Texas. Eight states, on the other hand, were over­coun­ted: Ohio, Massachu­setts, Utah, New York, Minnesota, Rhode Island, Delaware, and Hawaii.

The Census Bureau states that the over­counts and under­counts in the 2020 Census were not outside the range of expec­ted vari­ab­il­ity based on prior decades. The 2010 PES sugges­ted that no states had stat­ist­ic­ally signi­fic­ant under­counts or over­counts, while the 2000 PES estim­ated 22 states and D.C. had stat­ist­ic­ally signi­fic­ant prob­lems of that kind. The Bureau cautions, however, that uncer­tain­ties asso­ci­ated with the meth­ods it used in those earlier qual­ity checks make it diffi­cult to make histor­ical compar­is­ons of these qual­ity stat­ist­ics.

What should be done to improve future censuses?

Plan­n­ing for the 2030 Census is already under­­way, and the results of the Post-Enumer­­a­­tion Survey show that the bureau needs to prior­it­­ize redu­­cing racial and ethnic differ­­en­­tial under­­­counts the next time around. The bureau should take a hard look at its current coun­t­ing prac­­tices and research how it can improve all levels of the census process to ensure fairer counts. Inter­net responses, for example, helped insu­late the 2020 count from the COVID-19 pandem­ic’s disrupt­ive effects, but the digital divide — that is, unequal access to broad­band services across racial and ethnic groups — means that moving the census to the inter­net can’t ensure an accur­ate count. Congress should fully fund bureau efforts to study and improve its count­ing meth­od­o­lo­gies, as well as consider reforms that would safe­­guard the bureau’s inde­pend­ence and enhance the accur­acy, inclus­iv­ity, and equity of the count. Lawmakers should also consider chan­­ging the formu­las used to distrib­ut­e federal fund­ing to compensate state and local govern­­ments for under­­­counts in their juris­dic­­tions.

The differ­ence in state-level accur­acy also suggests that robust get-out-the-count efforts led by states, local­it­ies, and civil soci­ety groups may enhance the qual­ity of the count. States that inves­ted in census oper­a­tions and get-out-the-count efforts — such as Cali­for­nia with its $187 million invest­ment — faired better than those that refused to invest in the census like Texas and Flor­ida.

An equit­­able census that counts all groups equally well would not only produce more accur­ate numbers, it would result in a distri­bu­­tion of polit­ical power and fund­ing where every community gets its fair share. The 2020 Post-Enumer­­a­­tion Survey results under­­score the need for the Census Bureau, Congress, and other stake­hold­ers to work together to ensure that the next decen­nial census does better.