In the coming weeks, the U.S. Census Bureau will release the first set of results from the 2020 Census. Our goal for every census is to count everyone once, only once, and in the right place.

While we know that no census is perfect, how will we know if the 2020 Census results are good enough for their constitutional purpose and their many other uses? The answer to this question is perhaps more important than ever before because of the unprecedented challenges this census faced.

To answer this question, we are engaged in several efforts to assess the quality of the census data (as we do after each census). We’ve previously described these efforts in more detail.

For a quick summary, these efforts include:

• Comparing census results to the 2020 Demographic Analysis Estimates.
• Conducting a Post-Enumeration Survey to measure how many people and housing units were missed or counted erroneously in the census.
• Working with respected members of the scientific and statistical community to provide independent, external assessments of the census.
• Conducting a series of planned assessments and evaluations of the 2020 Census operations.
• Providing quality indicators in the form of operational metrics and preliminary response data.

Today, we’ll dive deeper into the operational metrics we plan to provide, particularly the host of metrics we plan to release along with the first results from the 2020 Census.

Operational metrics are data points related to how we manage census operations. For example, we can look at data about response characteristics, such as how we obtained a response for an address.

No singular number can definitively quantify the quality of the census. However, looking at these metrics and comparing them across geographies and with past censuses can give us insight into the quality of the data.

For example, we know from past research that the best quality information usually comes from a person within the household. As a result, understanding the percentage of self-responses (a response submitted online, by phone or by mail by a member of the household) is important. Similarly, understanding the percentage of population counts collected by talking to “proxies,” such as neighbors or building managers, is also important. 

We have already released a number of preliminary data quality metrics, including:

• Self-response rates down to the tract level.
• Total completion rates at the state level.
• Nonresponse Followup workload completion for area census offices.
• National administrative records and proxy respondent enumeration rates.

These metrics all gave a preliminary look at how we collected responses and accounted for each address.

We know many people are interested in more detailed metrics, so we’ve accelerated our schedule for releasing them. We plan to release the next set of metrics in April — on the same day we release the first 2020 Census results — followed by another set in May.

Each set will provide information for the nation, all 50 states, the District of Columbia, and Puerto Rico.

The operational metrics planned for April will provide information on the status of addresses in the census and how we resolved addresses in the Nonresponse Followup operation. (See the table shell.) The May set will provide similar information by average household size, and percentage of single-person housing units and two-person housing units.

For every address in the census, we need to know whether it is a real and livable residence and if so, whether it is vacant or occupied. The April metrics will show the percentage of addresses we determined to be occupied, vacant or nonexistent.

We’ll also show how that status was determined:

• Self-response: A household member responded online, by phone or by mail of their own accord.
• Nonresponse Followup: A census taker collected a response from a household member or a proxy respondent, such as a neighbor, building manager or landlord.
• Group quarters: We collected the response through the Group Quarters operation, which counts people living in places such as college dormitories, prisons and nursing homes.
• Unresolved: We used imputation to assign a status and population count for addresses that remained unresolved after we concluded our data collection operations. (Imputation is a statistical technique that fills in missing information using other available information. We’ll talk more about it in an upcoming blog.)

During the Nonresponse Followup operation, census takers knock on the doors of addresses that did not self-respond to the census to try to get a response or to verify whether the address is vacant.

The April operational metrics will show the percentage of occupied addresses resolved by:

• An interview with someone in the household.
• Administrative records: If a household didn’t respond on their own, a census taker would visit to collect a response. If they couldn’t reach the household, we attempted to use high-quality, existing records about the household. This new method for the 2020 Census provided a way to count people who otherwise wouldn’t respond.
• An interview with a proxy: If census takers couldn’t get a response directly from a household after three visits (except where high-quality administrative records were available), they tried to get information about the address from a proxy such as a neighbor, landlord or building manager.

The metrics will also provide:

• The percentage of vacant addresses.
• The percentage of deleted addresses resolved with a proxy interview or by administrative records, respectively. (We delete addresses we confirm don’t exist.) 

For decades, we have produced similar metrics to those we are releasing. What’s new is that we are releasing the metrics sooner than with past censuses, and we are releasing the metrics at the state-level in addition to the national level.

While there is interest in data at levels of geography below the state level, it is important to note that the Census Bureau is required by law to safeguard the privacy of respondents and the confidentiality of their responses. This requires that we balance the level of detail we provide, especially for smaller geographies, with that need for confidentiality.

We will apply privacy protections to the operational metrics we release to guard respondent privacy. We will also be limited in what detail we can publish at a local level for the operational metrics.

To make comparing the metrics among the nation and states easy, we plan to release the metrics through an interactive dashboard and downloadable tables. Alongside the 2020 Census numbers, we’ll also provide comparable metrics from the 2010 Census.

It’s important to note that we expect to see differences across geographies and the two decades. “Different” doesn’t necessarily mean “better” or “worse.” Many differences are a result of changes to the way we conducted the 2020 Census as compared with the 2010 Census, for example, implementing an internet response option.

A difference is just another data point as we consider the breadth of quality assessments in the works for the 2020 Census. 

The Census Bureau is committed to sharing what we know, when we know it, to help the nation understand the quality of the 2020 Census results. We are just a few weeks away from the first release of 2020 Census data, which represents the culmination of over a decade of hard work. We hope you will join us for the live news conference to hear these first results and stay tuned for more information over the coming months.