by Aleks Kajstura, August 22, 2014
The next decennial census is still 6 years away, but the Census Bureau is already running tests and exploring areas for methodological improvement. As part of this process, the Bureau commissioned and released an ethnographic study of incarcerated populations that analyzed how the Census Bureau could count incarcerated people more efficiently while improving accuracy. Although technically not about tabulating incarcerated people at home, we think this ethnographic study fulfils an important prerequisite to the Bureau’s exploration of counting incarcerated people at their home addresses.
The authors describe how, unlike most Americans who fill out a Census form for themselves or their family, incarcerated people are often counted using facilities’ administrative records. This method is sometimes quicker and easier but often results in less accurate data.
But how well do these records serve Census purposes? Ultimately that is the question the authors sought to answer:
Does the use of administrative records meet the goals of the census of “counting everyone in the right place and the right time”? …As has been asserted repeatedly, administrative rosters in correctional systems meet the first half of the census requirements by counting everyone present in these group quarters on the day the rosters are obtained. They are less successful in meeting the next goal of everyone in the right place”….
…these administrative records are least successful in answering valuable questions about race and ethnicity and home addresses. Addressing these questions will provide specific directions for the planning of Census 2020 in correctional group quarters.
The study presented many examples of the shortcomings of administrative records, for example:
Misalignment between official census categories of race and ethnicity and the administrative rosters resulted in subjective, and often arbitrary, [Census form] completions.
While names, birthdays and gender are highly reliable in correctional administrative records, the mismatch between federal labels for race and ethnicity and widely variant labels used in these correctional systems illustrate the need to examine these important categories to improve census coverage.
Given some advanced planning, however, the use of administrative records could be a great time-saver for the Census Bureau (and correctional staff). Relying on administrative data could save the Census Bureau from having to visit each prison individually. Instead, the Bureau could simply have a single contact in a state’s Department of Corrections to gather census data for every facility under their jurisdiction. The authors explain how such an agency-level approach can benefit the Census Bureau:
Investing time in an agency approach presents additional benefits for accuracy and cost-savings. Working closely with agency programmers, preparation for the 2020 Census could result in more consistent definitions of the race and ethnic categories both across the system and with the established census categories. Gaining knowledge of the structure and content of these databases could also result in the development of “census subroutines” in electronic form, eliminating the need to code data from hard copies into census machine-readable forms.
Such an approach would not be appropriate for the over 3,000 county jails, with one exception. The large, urban jails are likely to maintain databases similar to these prisons and could be included in a modified agency approach.
Although technically outside the scope of the inquiry, the authors return to the issue of home addresses several times, recommending that “A separate study of the availability and accuracy of these records would have to be designed and implemented.” A recent report, published after the ethnographic study, by a New York Law School professor Erika L. Wood and Dēmos, Implementing Reform: How Maryland & New York Ended Prison Gerrymandering, tackles some of those questions.
The study also points out the internal dissonance in the way the Census Bureau interprets its residence rules for incarcerated people, especially those in jail:
For example, a person arrested at 11:00 p.m. on March 31 may appear on the roster on April 1 but released back into the community later in the day on April 1 and thus not meet any census definition of residence in the jail.
And as the authors point out, this incongruity also leads to a potential overcount of people incarcerated in jails:
Jails, with a high level of population turnover, may present a challenge as their inmates may be counted in both housing unit and GQ [correctional population] enumerations.
So far in this post I have described the parts of the report that are most relevant to our prison gerrymandering work, but I cannot stress enough the wealth of information found in this report on a wide range of issues relating to the enumeration of incarcerated populations. In addition to identifying a few main problems and making recommendations, the report includes a 10-page detailed documentation of the enumeration process in the facilities under study. These pages (section 4 and its subparts) shed light on the census process from the facilities’ perspective and provide great insight into possible approaches for improving future censuses.
The full report is certainly worth a read, and represents a great foundation for the Census Bureau’s potential progress toward more accurate enumeration of America’s incarcerated populations.