Internet Access and Self-Response Rates

Posted October 26, 2017

For the 2020 Census, the Census Bureau will urge most households to fill out the census form online rather than use a paper questionnaire.  This is an effort to boost self-response rates and reduce costs.  (See Census Bureau’s Self-Response Research and 2020 Census Operational Plan.)

But in some areas of the country, especially in disadvantaged communities that have proven more vulnerable to being undercounted in the past, relatively low internet access may hinder the effort to increase online response.¹  This could make it even more challenging to conduct a fair and accurate census in the nation’s hardest-to-count areas, and, as explained below, counterintuitively may also adversely impact some areas that would be considered more likely to self-respond during the 2020 Census based on historical trends.

In other words, while the 2020 Census will deploy new technologies and procedures designed to reduce paperwork and save money, those changes will not by themselves overcome the difficulties of reaching hard-to-count areas — and could compound the challenges.

Recommendations for increasing self-response during the nation’s most technologically advanced Census in 2020 are discussed in detail in the Counting Everyone in the Digital Age report (PDF) co-authored by the Leadership Conference Education Fund and the Georgetown Center on Poverty and Inequality (GCPI).

Worrisome trend? Low 2010 mail return rates, low shares of household internet access

The Center for Urban Research at the CUNY Graduate Center has developed a Hard to Count 2020 Census Map that displays 2010 mail return rates and the share of households in each congressional district with inadequate internet access, as measured by households that reported either no or dial up-only internet subscriptions based on the latest estimates from the 2016 American Community Survey.²

An analysis of the 2010 mail return rates and the percent of households with no or inadequate internet subscriptions revealed that congressional districts with lower mail return rates also tend to have higher percentages of households with no or inadequate internet subscriptions³.

This means that many districts that are already considered hard-to-count based on how many households did not self-respond during the 2010 Census, also have relatively low rates of household internet subscriptions.  The mail return rates are based on households mailing back their census questionnaires in 2010 (before the internet was an option for census self-response).  Therefore, if these districts already tend to have low self-response via mail, and households in these districts also tend to be less connected to the internet, then the preferred online option for self-response in 2020 may not help boost response as much as it would in other districts with better digital access.

Figure 1 below shows the patterns of 2010 census mail return rates by congressional district, with an overlay highlighting the districts that have the greatest shares of households with either no internet subscription or dial-up-only subscriptions. Districts with the worst mail return rates nationwide are shaded in dark red. Districts with the most households with inadequate internet access are marked with a green diamond. Most districts with the worst rates of 2010 census self-response are in the South and West, as well as in urban areas (such as New York City, Detroit, New Orleans, Phoenix, and other cities that are obscured on the map because their districts are geographically small).

Figure 1
(click to open high-resolution map image in new browser window)

 

 

 

Table 1 below lists the 39 congressional districts that had the worst mail return rates in the country in 2010 (a district-wide mail return rate of less than 76%) and also have the highest shares of households with no or poor internet access (23% or more households having no broadband or satellite subscriptions).

Table 1
(full spreadsheet in Excel format accessible here)

The relationship between congressional district mail return rates and shares of household internet access spans a continuum. In addition to the 39 districts listed above, there are several dozen other districts with slightly better (but still relatively low) mail return rates and also slightly higher (but still relatively low) shares of household internet access. These additional districts are also at risk of having the self-response rate in 2020 hindered by these two characteristics.

Some high mail return rate districts also have low share of households with adequate internet access

At the same time, other congressional districts had high mail return rates, but also have relatively low rates of internet access. In these districts, we would expect households to have a greater likelihood of self-responding in the 2020 Census. But because of poor internet access, they may have a harder time self-responding than in 2010. To be sure, the paper and mail option remains available. However, most households initially will be mailed information about submitting a questionnaire online, rather than a paper form.

Figure 2 below shows the patterns of congressional districts and household internet access. Districts with greatest share of households that have either no internet subscription or dial-up-only are shaded in dark red. Districts with the fewest households with inadequate internet access are shaded in light orange or yellow.

Figure 2
(click to open high-resolution map image in new browser window)

 

 

 

Table 2 lists the 13 districts in this "best/worst" category.

Table 2
(full spreadsheet in Excel format accessible here)

Footnotes

1. For information on census accuracy and disproportionate undercounting, see Census Accuracy and the Undercount: Why It Matters; How It's Measured
2. American Community Survey data is from 2016 1-year estimates in table B28011  . Households with adequate internet subscriptions, as described in the Census Bureau’s latest report on internet access Computer and internet Use in the United States: 2015, have either satellite subscriptions or broadband subscriptions including cable, fiber optic, or DSL. Some people without household internet access may be able to complete the forms on public computers or smartphones.
3. The relationship was statistically significant, with a moderate Pearson correlation statistic of -.232.