Estimated reading time: 7 minutes

Earlier this year, we released two sets of 2020 Census operational quality metrics. Operational quality metrics can provide important information about the quality of the census by looking at how we obtained a response for each address. We released these earlier results for the nation and for each state.

It’s been suggested that we release some of the metrics at lower levels of geography, such as counties or census tracts. To safeguard the privacy of respondents and the confidentiality of their responses, and in keeping with a recommendation from the JASON group, we are not planning to release metrics for specific jurisdictions below the state level, but we are providing sub-state summaries for each state to show the variation and spread in the range of operational quality metrics between localities within each state.

In this blog, we provide highlights from today’s release as we did with the first and second releases. 

The first release of the operational quality metrics focused on how we resolved whether each address, including housing units and group quarters, was occupied, vacant, delete (for example, uninhabitable or nonexistent) or remained unresolved.

Now, we are providing national and state-level summaries of the county and tract distributions for selected metrics:

• Percentage of all addresses unresolved (went to count imputation).
  
• Percentage of all addresses resolved by self-response (responding on their own online, by phone or by mail).
• Percentage of all addresses resolved by online self-response.
• Percentage of addresses resolved by self-response that were from online responses.
• Percentage of occupied households in Nonresponse Followup that were enumerated with a household member (a census taker interviewed someone living at the address).
• Percentage of occupied households in Nonresponse Followup that were enumerated with a proxy respondent such as a neighbor, building manager or landlord.
• Percentage of occupied households in Nonresponse Followup that were enumerated using high-quality administrative records.
• Percentage of occupied households in Nonresponse Followup, excluding those enumerated by administrative records, that provided only the population count and no other characteristics such as age, sex, race, Hispanic origin or household tenure.

We calculated each of these metrics for every county and census tract in the nation. Then we calculated the mean, standard deviation and median values for each state. 

The mean is the sum of all the values divided by the number of counties or tracts in each state — what most people call the “average.”

The standard deviation tells us how much the average measurement for a specific geography differs from the mean value. It is a good way of measuring how spread out the numbers are because it shows the amount of variation. If there were no variation and all counties and tracts had the exact same number, the standard deviation would be 0. Typically, in statistics, about 95% of the values will be within two standard deviations of the mean.

Lastly, the median is the middle point of the data. Half of the values will be below the median and half will be above the median.

Note that for each of these calculations, we excluded from the calculations very small tracts — the 336 tracts across the nation (not including Puerto Rico) with 20 housing units or less. We did this to better protect confidentiality and minimize influence of too-small geographies on the statistics. For this same reason, we are providing only the mean and standard deviation at the county level for the District of Columbia and states (Delaware, Hawaii and Rhode Island) with fewer than six counties.

Together these numbers provide important information about the distribution of the operational quality metrics across the counties and tracts within each state and across the country. Naturally, as with any population or housing characteristics, there are some differences from one geography to another. Some counties or tracts, for example, will be higher and others lower than the average.

We know for example that nationally 65.3% of census addresses were resolved by self-response. However, this varied throughout the country, with some localities much more likely and others less likely to self-respond.  

Out of all counties in the country, the mean resolved from self-response was 58.1%, with a standard deviation of 12.2%, and the median was 59.7%. Collectively this tells us a lot. The mean and median are both lower than the national average of 65.3%, which indicates that many less-populated counties generally have lower rates of self-response. Some of that was due to the nature of the census operational design, in which different data collection methods were used such as Update Enumerate or Update Leave, which both have lower self-response than traditional mailout areas and tend to be in more rural areas.

Another example is the metric for the percentage of occupied households in Nonresponse Followup that were enumerated with administrative records, which was 18.4% nationally. Enumerating households with high-quality administrative records was a planned innovation for the 2020 Census. In our sub-state summary, out of all counties in the United States, the mean was 16.5%, with a standard deviation of 6.0%, and the median was 17.2%. Here, the mean and median are both close to the national average of 18.4%, which suggests that the county-level variation around the country is statistically normal, as expected. There is some variation, as evidenced by the standard deviation and the state-level county averages, but no obvious outlier patterns that drive the center too far from the national average.

We see similar patterns and variation with the other sub-state summary metrics.

The metrics are available here in a downloadable table on the 2020 Census Data Quality webpage. The numbers from today’s release (as well as all previously released operational quality metrics) are also available on the 2020 Census Data Quality webpage.

The 2020 Census results so far have shown once again that the United States is a very diverse country, and this diversity varies from state to state, county to county, and place to place. The census operational quality metrics are no different, in that there is natural variation all over the country. The sub-state summary metrics released today show patterns similar to what we would expect from a normally distributed population. Some counties and tracts are higher on some metrics and some are lower on other metrics, but no signs point to anything unexpected in the results.

The decennial census is a very complex operation, and no single number can definitively quantify its quality. But the various operational quality metrics we released provide insights into how our operations and processes have a hand in accurately counting our nation’s population.

Throughout this year, we have been sharing as much information as we can about the 2020 Census results. As a reminder, some of the other ways that we are measuring the quality of the 2020 Census include:

• Comparing census results to 2020 Demographic Analysis estimates and other population benchmarks. More information is available in our Comparisons to Benchmarks as a Measure of Quality blog.
  
• Releasing information on the improvements to how we measure race and ethnicity.
• 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.

Additionally, on August 25, the Census Bureau will provide item nonresponse rates for the population count, age or date of birth, race, and Hispanic origin questions. These rates will be available for the nation, 50 states, the District of Columbia and Puerto Rico.

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 will continue to update the 2020 Census Data Quality webpage as new information becomes available.