The Hard Count: The Political and Social Challenges of Census Mobilization

Chapter Five: Conclusions and Consequences

114 Chapter Five EVERY DECENNIAL CENSUS differs in design and methodology from the one that preceded it.1 Our inquiry started by setting out the context for the 2000 census because every census is responsive to the inevitable changes in the social-political climate and the demographic context over the ten-year interval. Census design also changes because every census offers lessons for how to do the next one. Although the Census Bureau conducts a “test census” during the planning phase of every census, this is a weak substitute for the real thing—much as the Pentagon’s “war games” are a weak substitute for the real thing. Because the decennial is a learning opportunity for the Census Bureau, an extensive evaluation program is imbedded in each decennial on the basis of which the next census is shaped. It is in this spirit that we ask how our findings might be valuable in designing the 2010 census. The challenge of translating social science findings into policy recommendations has been a continuing theme for well over a century, dating to the nineteenth-century origins of the academic disciplines as we know them today. This discussion has often been framed as if there is a clear distinction between “basic” and “applied” social science , a distinction we find useless. Here we discard it in favor of a distinction between social science knowledge and policy knowledge. Metaphors are risky, but we might liken social science knowledge to an aquifer and policy knowledge as the ability to drill into that resource pool in a manner that facilitates making, defending, and changing policy. Social science knowledge tells us what is there—in our study, for example, it showed us how a civic mobilization campaign worked and how privacy concerns compromised that campaign. Policy knowledge tells us how to use these findings in real-world contexts. It requires going beyond the empirical findings to take into account known constraints and probable developments . Policy knowledge, so defined, blends social science findings with Conclusions and Consequences Conclusions and Consequences 115 sensitivity to what is reasonable to expect under current resource and policy conditions. Bearing this in mind, we start with the key findings in each of our three empirical chapters: 1. The mobilization campaign in 2000 was effective in improving the mail-back cooperation rate above what had been predicted, and it was particularly effective in reaching the population groups that are most often undercounted. 2. Privacy and confidentiality concerns depress cooperation with the census and, if inflamed during the census itself, as they were in 2000, compromise census coverage and increase selectivity in answering questions (item nonresponse), leading to a deterioration of data quality. 3. Household structure and composition, especially in light of rapid changes over the last few decades in the household composition of traditional and nontraditional families in the United States, influence census cooperation in ways that call into question the household as the appropriate unit of data collection and data reporting. Although our focus is on the decennial census, each of these findings has implications for the survey-based information system on which our government (and economy) depend. There are approximately seventy federal statistical programs and agencies that operate at a total annual budget of $5 billion and generate a steady flow of statistics about the nation’s population and economic establishments. In the end, the resulting national statistical system is based on millions upon millions of citizens voluntarily checking boxes, completing forms, and answering questions. Anything that erodes this compliance compromises the quality of the nation’s statistics— and of course, the policy and economic decisions based on them. In thinking about how the 2000 census experience can inform future censuses , we must recognize that the 2010 census is expected to be fundamentally changed by the American Community Survey (ACS), a major innovation now under way. If the ACS is fully implemented, the 2010 census will not include a long form. So, before drawing out the implications of what we have learned about public cooperation with the government’s statistical data collection , we briefly introduce the American Community Survey. THE AMERICAN COMMUNITY SURVEY The ACS is a large, continuous sample survey designed to replace the decennial census long form. It is administered to approximately three million households each year, and this summary provides yearly estimates of pop- ulation and household characteristics for communities of 65,000 or larger...