In 2014, national civil rights and human rights organizations endorsed the first unified Civil Rights Principles for the Era of Big Data, affirming the importance of data and privacy to historically disadvantaged groups. The principles emphasized both the potential for improvements in civil rights enforcement as well as the risks inherent in big data. The same potential and peril are applicable to the use of new technology in the decennial census.

The decennial census is far and away the most demanding of the many responsibilities of the Census Bureau, which range from the ongoing American Community Survey (ACS) and issue-specific and economy-focused surveys to the Survey of Income and Program Participation. Despite the herculean nature of a decennial census and the complexity of the U.S. population, Congress expects the 2020 Census to cost less than the 2010 operation. That requirement is driving efforts to implement a more efficient internet response option, automate key components of training and field operations, and develop cloud infrastructure to streamline data collection.

The operational changes planned and underway could reduce costs by up to $5.2 billion, and create a more efficient decennial census. Of course, to ensure an accurate and fully inclusive count, the Census Bureau must prepare for potential pitfalls. Technological failures could compromise data quality and cybersecurity. Populations at risk of being undercounted are also among the most likely to face tenuous and insecure access to the internet and other new technologies, making responding more burdensome and more open to danger from hackers and malware. Data quality is also at risk because vulnerable communities, especially those who are caught in the crosshairs of current political and
social tensions, may distrust the new methodologies.

To reduce the risks associated with new technology, Census Bureau staff have been conducting extensive research and testing for years. Now they are racing against the clock to be fully prepared to conduct a successful and affordable count. The last major test will be a large-scale “dress rehearsal” in 2018. Only operations included in the 2018 test are likely to be part of the 2020 Census,17 leaving a short window for stakeholders to ensure that the bureau’s procedures and technologies can be relied upon not to result in undercounting of socially and economically disadvantaged populations.