Radical
Statistics
Communication, politics, and the census
Martha Farnsworth Riche
By their nature, population censuses are people-based activities: people counting people. Every population census involves three main groups of people:
- the professionals who plan, undertake, and evaluate the census;
- the politicians and policy makers who determine the census environment and apply the census results; and
- the public, whose participation is vital.
Each of these groups enters the census process at different times, generally in the above sequence. That makes it imperative for the first group, census planners, to plan for the inclusion of the other two groups. What might British census planners learn from the current U.S. experience?
The point of departure: the 1990 census
From the point of view of U.S. census professionals, the nation's 21st decennial census was probably its most successful. Post-census measures showed that the 1990 census counted 98.4 percent of the population. (That is a net measure: the census actually missed over 8 million people, while double-counting over 4 million). However, the public and most policy-makers repeatedly assailed the census as perhaps the worst ever. This conflicting census evaluation was the result of an almost complete failure of communication between census professionals and the public, and census professionals and policy-makers.
The public had expected the census to count 100 percent of the population, in part because the Census Bureau and census stakeholders had enlisted so many people in volunteer activities designed to get a complete count. The failure to meet this unrealistic expectation was laid directly at the door of the Census Bureau: 'They did a bad job'. The results seemed to bear this assessment out, since the 1990 census was the first to have a larger undercount than its predecessor. (The 1980 census had the lowest net undercount: 1.2 percent.) Moreover, the differential undercount of racial minoritiesCthe prime reason for measures to eliminate the undercountCactually widened in 1990.
The politicians, who disagreed among themselves over the merits of statistically correcting the census results for the under- and over-counts, agreed that the 1990 census cost far too much. Again, they attributed this failing to the Census Bureau, which had not been able to communicate the connection between census costs and various census operations that policy-makers either favored or opposed during the planning process.
The real reason for higher-than-expected costs was the unprecedented low public response to the census. This reflected the interaction of census operations and demographic changes.
The U.S. moved to a mail census in 1970 because the traditional source of temporary census employeesChousewivesCdried up with the widespread movement of women into the paid labor force. In 1970, 85 percent of the public that received the form by mail (people living in urban areas) returned it to the Census Bureau. In 1980, the mail return rate dropped to 70 percent (all but very remote places now were enumerated by mail). In 1990, the rate dropped to 65 percent. As a result, in 1990 the door-to-door follow-up, the most costly part of census activities, ballooned from the 6 weeks originally scheduled to 13 weeks.
The same population trend that led the U.S. to adopt a mail censusCworking womenCmeant that fewer people were at home to the census taker. The people who were at home tended to be those who had been willing to take the time to fill out the form and mail it in. The people who did not mail in their form said they were too busy or did not receive it (i.e. did not recognize it among the mass of solicitations that fill their mailboxes). Busy people whose interest is focused elsewhere are very hard to reach, whether by mail or in person.
Since both public and policy-makers considered the 1990 census a failure, it was important for census professionals to take their views into account in planning Census 2000. The two Congressional committees with oversight of the censusCone responsible for its outcome, the other for its costCmade sure this would happen by creating a panel of experts at the National Academy of Sciences, charged with recommending new ways of taking the census. The panel=s report was issued in November of 1994, in time for the Census Bureau to incorporate its recommendations in the design for Census 2000.
The design for Census 2000 offers considerable innovation compared to the 1970-model that was revised for 1980 and 1990:
- It addresses the issue of prolonged and costly non-response follow-up by sampling the last 10 percent of nonresponding households in each neighborhood. This saves both time and money to enable the following innovation.
- It tackles the perpetual and persistently differential net undercount through a giant (750,000 household) post-enumeration sample survey, integrated with the results of traditional enumeration methods. (This is the One-Number Census.)
- It addresses the changes in the population that limit the public=s willingness to mail in census questionnaires: increasing reluctance to provide personal information, increased labor force participation by women (who usually fill out questionnaires), and decreased respect for government, particularly the federal government.
All three of these innovations are controversial for one reason or another, and it is difficult at this point to know whether they will be implemented. The first two innovations are being debated in the political arena because one political party fears that a corrected and therefore complete count will result in Congressional re-districting that would be harmful to its electoral prospects. Since Congressional leaders cannot acknowledge this motive, they are attacking the use of statistical sampling as unconstitutional and unreliable. The third set of innovations, designed to increase the public=s response to the census mailing, has been lauded by both the policy-makers and the public. However, it is not supported by the professional census community, which is generally resistant to complicating the census process by efforts that only deliver acceptability, not improved data quality.
The plan for Census 2000: Census professionals and the public
The public campaign for Census 2000 was developed the way any marketing campaign is: by going to the public and finding out how people react to the product to be marketed. Two major activities conducted in the first part of the planning decade laid the groundwork. The first was a series of tests of census mailing strategies, some testing new ideas census professionals had gathered from the direct marketing community, others testing ideas offered by the politicians and public stakeholders who criticized the 1990 effort. The second was a series of focus groups.
The mail tests produced a number of important results. The most important from an operational standpoint was the demonstration in census terms of a key principle of direct marketing: 'Repeated reminders pay big dividends'. Bracketing the mailed questionnaire with pre- and post-mailing reminders increased the response rate, as did the mailing of a second, replacement questionnaire. The Census Bureau estimates that this 'full mail strategy' will stem the continued decline in response rates, and, ceteris paribus, deliver a 66.9 percent mail response.
The most important result of these tests, from a communications standpoint, was the stunning refutation of repeated recommendations from policy makers and the public:
- to remove the message, 'Your response is required by law', because it was seen as threatening and consequently off-putting, and
- to substitute a patriotic message.
Although both groups continue to deplore the 'mandatory' message, they now acknowledge the 12-percent boost in response rates that it produced in the tests. (The clincher: each additional percentage point of mail response saves $25 million in follow-up costs).
The focus groups also produced several important results. From a communications standpoint, they reinforced the policy-makers' concerns that the public experienced the census as too burdensome and bureaucratic. Although census professionals repeatedly pointed out that these concerns were not reflected in the census outcome, and therefore were not worth scarce resources, Congressional policy-makers did not accept this argument. Consequently, the planners made several changes, including hiring professional designers to turn the questionnaire and mailing package into what is essentially a direct marketing campaign.
From an operational standpoint, the focus groups helped shape the development of the statistical sampling innovations, particularly the plan to sample for non-response follow-up. The National Academy of Sciences panel that recommended this innovation acknowledged that there was no scientific basis for determining the point where follow-up should be truncated. It thought that 70 percent might be effective from both a cost and an accuracy standpoint, but invited further research. The focus groups, however, showed extreme discomfort with sampling the last 30 percent of a community. They showed shallow support for sampling the last 10 percent once they were assured that every household would have multiple opportunities for answering the census. They agreed that there would always be some people who would not participate, no matter how hard the Census Bureau worked. However, they were very concerned about fairnessCeverybody should have a 'fair' chance of being included. There was also a sense that 'the last 10 percent' meant 'someone other than me' while 'the last 30 percent' could mean 'somebody like me'. Since six of the seven extra weeks of follow-up in 1990 were devoted to the last 10 percent, and responses for many were eventually gathered indirectly, it was easy to communicate the costs and relative inaccuracy of repeated follow-up visits for the last 10 percent.
In short, though the focus groups revealed a great deal of public dissatisfaction with the way the census has been conducted, they also showed that people very much wanted to be counted. But they wanted to be counted on their terms, not the Census Bureau's. They wanted census professionals to recognize the reality of their lives, and to make it simple and fast for them to participate, as well as communicating clearly why it was important to them to do so.
The plan for Census 2000: Census professionals and policy-makers
Although Congressional policy-makers had commissioned the expert report that is the basis of the design for Census 2000, they did not accept the report when it was completed. In part that was because the political makeup of the Congress had changed, and the new leadership eliminated the committee which would have received and digested the recommendations of the expert panel. In part that was because the report rejected remedies some of its sponsors had expected it to proffer.
Nevertheless, in its subsequent reports and actions, particularly the annual funding appropriations, the Congress has continued to ask the Census Bureau to meet the goals it gave the panel: a census that is both more accurate and costs less. Consequently, the census professionals accepted the recommendations the policy-makers had asked for but refused to hear, since it was the only way they saw to meet these seemingly conflicting goals. This landed them with the burden of communicating the experts' views to the policy-makers. (In the United States, the Congress has responsibility for the census, though it delegates the implementation to the Secretary of Commerce.)
This burden was unusually heavy because the new Congressional leaders had privately rejected the experts' recommendation to use statistical sampling to both increase accuracy and reduce costs. (The new Speaker of the House reportedly told his party caucus, 'Sampling is a dagger pointed at the heart of the Republican party'.) And, by not giving the expert panel an opportunity to report, Congress also had not had to hear the experts reject the remedies the members continued to propose instead of sampling:
- have the Postal Service conduct follow-up ('they know everyone on their routes'),
- eliminate the bulk of the questions ('the length and intrusiveness of the questionnaire is what depresses response rates'), and
- 'creative and innovative' outreach and promotion will eliminate the undercount.
This meant that the census professionals had to spend a great deal of their scarce time and money developing ways to communicate the ineffectiveness and relative costliness of these methods, and the relative effectiveness and cost-saving nature of using statistical sampling.
These communications had to be air-tight conceptually and materially to be convincing, since the audience had already made up its mind to the contrary. Since this is a new, and unwanted, activity for census professionals, they did not use the same standard of quality in producing these communications as they do for measuring the population or the economy, and made some embarrassing mistakes in their calculations. These mistakes were particularly damaging because the audience used them as arguments against letting the Census Bureau use the new methods. ('The Census Bureau keeps making mistakes; therefore we can't trust them to get sampling right'.) This has kept advocacy for less effective methods alive, and forced the Census Bureau to continue to devote scarce resources to them.
On the other hand, extensive and geographically diverse town meetings created a public voice for the census professionals' plan. The Census Bureau's regional offices hosted a dozen or more town meetings in major cities around the country for three years, amounting to more than 40 by 1997. Local members of national stakeholder organizations (such as the Association of Public Data Users, the League of Cities, and the Chamber of Commerce) were invited, and the Bureau's director and the senior staff member in charge of building relationships to outside organizations presented the census plan and asked for comments. They also showed video film clips of two national meetings where representatives of these and similar organizations had participated in finalizing important aspects of the census. This avoided the contentiousness that would have resulted had the audiences felt they 'had not been at the table' when decisions were made; instead, they were being asked to ratify and extend a process in which they had been represented.
A professional press effort ensured widespread coverage of the town meetings, and editorials in many local newspapers followed the meetings. As a result, as the debate between the planners and the policy-makers continues, the public is fully informed. An added benefit is an edited videotape of many of the meetings so census staff can understand the public's expectations for the census.
Conclusion? To be written in 2000
The author, who can be contacted through Radical Statistics, was the Director of the U.S. Bureau of the Census, 1994-98. She gave an earlier version of this talk to the One Number Census research workshop in Leeds, May 12th 1998, funded by the ESRC.
