Final Self-Response Analysis

Posted May 3, 2010

On April 28, the U.S. Census Bureau announced that the nation’s final mail-in rate for the 2010 Census was 72%, equaling the rate from a decade earlier. Now that the first phase of the census is largely over (Mail-out/Mail-back and Update/Leave operations) and the second phase has begun (Nonresponse Follow-up, or door-to-door enumeration), the Center for Urban Research at The Graduate Center, CUNY has analyzed the final participation rates.

Download the full report

Our analysis focuses on two areas:

1. Impact of the Census Bureau’s replacement mailing strategy. We examine the Census Bureau’s determination that census tracts receiving replacement forms had higher mail-in rates. Our analysis supplements their information with more detailed findings.
2. Characteristics of communities that will require the most door-to-door follow-up work. We examine demographic characteristics of tracts with relatively low response rates, which will likely present the greatest challenges during the Nonresponse Follow-up (NRFU) operation that began May 1.

One new factor in 2010 compared with 2000 was the Bureau’s strategy of sending replacement census forms to selected neighborhoods nationwide. In early April, a “blanketed” replacement mailing was sent to all households in tracts with historically low response rates. A week later, a “targeted” replacement mailing was sent to households in tracts with historically mid-range response rates that had not yet returned a census form. The Census Bureau notes that tracts that were in the replacement mailing “universe” improved substantially on their 2000 mail-in rates.

More than 600,000 census takers are now in the process of visiting homes that did not mail back their census questionnaires. This work will be far more difficult in communities with low participation rates in the first phase of the census. Therefore, our analysis updates our earlier findings for tracts that had relatively low participation rates. This does not capture the full “NRFU” universe – even tracts with high mail-in rates will have households that need to be enumerated in person – but it provides neighborhood-level context for areas that will experience a substantial level of door-to-door activity.

The Center for Urban Research will maintain its Census 2010 Hard to Count mapping site while the NRFU operation continues. The site also promises to be a platform for the display of American Community Survey data and the results of the 2010 Census.

Key Findings as of April 28

1. Replacement mailings had a strong impact in census tracts nationwide, across metropolitan/non-metropolitan regions, and across hard-to-count categories.

Our analysis concurs with the Census Bureau’s finding that tracts that received a replacement mailing had a high likelihood of improving their mail-in rates over 2000. Almost three-quarters of tracts nationwide that received a targeted or blanketed mailing improved their mail-in rates over 2000 (74 and 71%, respectively).

• Only 21% of tracts receiving a single questionnaire improved their participation rates. More than 70% of these tracts had mail-in rates worse than 2000.

2. Stronger effects of replacement mailings in areas outside cities.

Replacement mailings were more extensive in cities than in other parts of the country. Just under half (47%) of census tracts in cities were in the replacement mailing “universe” (compared with 44% nationwide), while just over a third of tracts in metro suburbs and non-metro areas were in this universe.

Although replacement mailings were more prevalent in cities, it appears that the replacement mailings had a greater impact in boosting response in metropolitan suburbs and non-metro areas than in cities.

• Almost 70% of tracts in cities that received a targeted or blanketed mailing improved their mail-in rates over 2000.
• In metropolitan areas outside cities, 85% of tracts receiving a targeted replacement mailing (and 81% of tracts receiving a blanketed replacement mailing) improved their participation rates over 2000.
• In non-metropolitan areas, more than four-fifths of census tracts that received a replacement mailing improved their participation rates over Census 2000.

3. Replacement mailings were extensive in several major cities.

For example, 89 of the 90 tracts in Newark, NJ received a blanketed replacement mailing, and the remaining tract was in the targeted mailing universe.

In New York City and Baltimore, MD, more than 90% of those city’s tracts received replacement mailings.

4. Stronger effects of replacement mailings in easier-to-count tracts.

Participation rates improved across hard-to-count categories, with some differences.

• In all of the HTC-score categories, fewer than 25% of census tracts that received a single census mailing (just over 30,000 tracts nationwide) improved their mail-in rates over 2000. As many as nine in ten of these tracts had lower mail-in rates than 2000.
• Relatively easy to count tracts that received a targeted replacement mailing were highly likely to improve their participation rates over 2000 (as many as 85% of 14,000 tracts). For harder to count tracts (HTC scores of 61 or higher) in this category, between 53% (for tracts with HTC scores 76 or more) and 59% (for tracts with HTC scores of 61 to 70) improved their mail-in rates.
• Relatively easy to count tracts that received a blanketed replacement mailing were also highly likely to improve their participation rates over 2000 (as many as 80% of 15,000 tracts).
• For the harder to count tracts that received blanketed replacement mailings, between 66% (for tracts with HTC scores of 61 to 70) and 68% (for tracts with HTC scores 76 or more) improved their mail-in rates. Therefore, in harder to count tracts, it appears that the blanketed replacement mailing had a somewhat stronger positive effect than the targeted mailing.

5. Census tracts with relatively low participation rates (<= 60%) generally were more racially and ethnically diverse, especially in cities.

As of April 28, almost 10,000 tracts (9,546) had mail-in rates of 60% or less. More than two-thirds (~6,500 or 69%) are in cities, while 14% are in metro suburbs and 17% are outside metro areas.

The race/ethnicity characteristics (based on 2000 census data) are strikingly different, however, in low participation tracts where census takers will visit a higher percentage of households during the Nonresponse Follow-up operation than in neighborhoods that will have a lower proportion of door-to-door visits.

In cities:

• the 60%-and-under tracts on average are diverse: 30% Black, 17% Hispanic, and 46% White.
• the higher mail-in rate tracts are much less diverse: 73% White, 11% Black, and 11% Hispanic.