Nielsen’s Senior VP of data science on why making sure the U.S. census is based on solid science should not be a debate, but a given

What are the key issues and implications for researchers should the Supreme Court approve the inclusion of the administration’s proposed question on U.S. citizenship following the open letter of Nielsen’s CEO to the New York Times.

The United States Constitution requires that every resident of the U.S. be counted in the decennial census regardless of whether or not they are citizens. The decennial census serves as the benchmark for the datasets that underpin critical government decisions (e.g. Congressional representatives, Electoral College votes, allocation of federal dollars to local areas for programmes). However, U.S. businesses also rely heavily on U.S. Census data to size markets and make critical commercial decisions such as where to build new homes, where to place stores or what programmes to make available on television.

The advertising industry and many other businesses depend on U.S. Census data to make calculated and operational decisions, as they plan what to make, who to make it for, where to market it, where to sell it and how to adapt to the nation’s changing demographics.

The advertising industry and many other businesses depend on U.S. Census data to make calculated and operational decisions, as they plan what to make, who to make it for, where to market it, where to sell it and how to adapt to the nation’s changing demographics.

But what would happen if you introduced an untested question that could lead to underrepresentation of certain persons and potentially flawed data?

When the U.S. Commerce Secretary announced the addition of a citizenship question to the U.S. Census in March 2018, many of the nation’s leading social scientists and scientific organizations expressed concern. The addition of this question — late in the census planning cycle — would introduce the risk of an undercount, particularly for persons of color. An undercount would not only negatively impact the persons who did not participate but all people who live in these areas. Further, an undercount could lead to incorrect estimation of market size for key economic and business investments, and the data could be compromised for the entire decade. The science suggests that the addition of this question is likely to impact both the participation rate and the cost of the 2020 census.  

The likely impact

While the lack of testing makes it hard to precisely quantify the expected impact of the citizenship question, scientists have provided research to demonstrate that a negative impact is likely, particularly among minority racial and ethnic groups. Citizenship is a sensitive topic for groups who are concerned about privacy. Threat of disclosure is one reason why survey topics may be considered sensitive.  (Tourangeau and Yan, 2007). Sensitive questions often lead to higher levels of item nonresponse (failure to answer a question) and unit nonresponse (failure to return the entire survey). It is also well documented that high levels of nonresponse can significantly increase survey costs (Groves 1989). 

The cost impact is especially concerning for the decennial census due to the constitutional requirement to count all residents which necessitates exhaustive nonresponse follow-up operations — including expensive in-person enumeration efforts for respondents who do not return forms.

Prior to the announcement of the citizenship question, the Census Bureau documented concerns regarding 2020 Census participation levels as they began to experience an increased number of respondents spontaneously sharing concerns related to privacy and the potential use of Census data to enforce immigration policies. These findings were presented in the fall of 2017.

A 2018 Census Bureau survey designed to inform communications efforts reported that “the citizenship question may be a major barrier” to an accurate population count. Georgetown Law’s Center on Poverty and Inequality published a report that further illustrates the increased risk of nonresponse issues due to the presence of the citizenship question. The research, based on an analysis of the American Community Survey (ACS), shows that item nonresponse levels for the citizenship question on the ACS are much higher than they are for any other decennial census question and that the rates vary considerably for different groups and geographies. For example, there are higher rates of non-response in places such as Arizona where immigration law enforcement efforts have been more prevalent. Five Former Census Bureau Directors, including those who served under Republican administrations, filed a brief citing concerns about lowering participation rates as well as other unpredictable consequences given the timing and lack of testing.

What does this mean for businesses?

Census data serves as critical inputs to, for example, Nielsen’s estimates of television, radio and digital audiences, as well as many of the products our advertising and marketing clients use to make important business decisions. This data is also used as the baseline for our universe estimates and for sample design, selection and recruitment procedures. 

Universe estimates are used to understand demographic trends and estimate market size; manage the proper demographic representation and balance Nielsen panel data; and, project audience estimates for key demographics across its products. For example, national marketing campaigns are often run across multiple markets. The rank for a given Demographic Market Area (DMA) plays a role in which DMAs are selected. The DMA ranks also help determine prices for media buying, with higher ranked/higher population markets commanding higher dollars. This is particularly evident during election years when marketing dollars are allocated to important voter market areas.

Census measurement also powers decision-making by the country’s biggest retailers and manufacturers that influence more than $1 trillion. They make critical decisions based upon Census data, and it’s essential that they can continue to rely heavily on the census data as a foundational truth set.

Census measurement also powers decision-making by the country’s biggest retailers and manufacturers that influence more than $1 trillion.

If the citizenship question is added to the 2020 Census, the results will contain flawed data, which will have vast negative consequences for commerce and the economy, among other things like infrastructure, job creation, social programmes, educational funding, etc. The most heavily impacted will be diverse citizens, who, by 2044, will be the majority of the U.S. population. American businesses require the best possible data to make the right decisions to properly serve diverse populations and grow their businesses.

However, if that were to occur, Nielsen would need to assess the impact​ on the data quality first,​ in order to apply any necessary modifications to the estimates, as well we would work with U.S. Census and other data sources to quantify any type of undercount at a total population level and for specific demographic groups.

What can businesses do to plan for the future?

In the era of big data, an accurate Census is more critical than ever. If the citizenship question is included in the 2020 Census and there is an undercount and potentially inaccurate data, businesses and government will have to work harder to overcome inherent structural biases and errors in the census that could be amplified over and over again as the data is used in new and ever evolving ways. Businesses and advertisers will need to find different ways to measure and account for the population. For retailers and manufacturers, strategic business decisions around consumer spend forecasting, innovation planning, marketing projections and shelf assortment needs still need to be made. But if the Census becomes unreliable, they may have to turn elsewhere.