Much of the work of the Census Bureau is embedded in a culture of science. The notions of quality of our work are derived from scientific principles. Our language to describe our products uses scientific and statistical terms. Inherent in this perspective is the ethic that the scientist must reveal all the weakness, qualifications, and alternative possibilities of a piece of work. In short, scientists tend to emphasize the negative qualities of their work before revealing their conclusions. Bertrand Russell once said that the more scientists know, the more forcefully they articulate what they do not know.

Interestingly to me, the census and survey statistics given to the public are delivered into a very different culture. This is the culture of public media and politics that rarely can permit extended descriptions of facts and their attributes. This is the world where single phrases are often used to describe the result of large works. In such a context, qualifications about findings and descriptions of alternative conclusions are rarely revealed.

With the growing desire on the Census Bureau’s part to be transparent in all its work, this situation poses an interesting problem. If we give a full picture of the results of the decennial census, with full intent of transparency, little of what we distribute may make its way to the public. If we provide only our selection of the key facts, we fail to provide the full picture, and violate our scientific principles.

I fully believe that the ultimate measure of quality of government statistics is the extent to which the public judges they are credible. However, for many statistics there is no single feature that defines their utility to the user; that is, the extent that they deserve to be credible.

Further, the “quality” of a census or survey estimate varies across different statistics. For example, the US population count, the 50 state population counts, and the roughly 8 million census block counts face different challenges. We need to find an effective way to communicate the multiple features of the multiple statistics we distribute, despite the need for simplicity in our communications. Without that, users cannot adequately judge the statistical information’s credibility.

In short, we have our work cut out for us. We have to communicate simply and clearly to the public about our professional evaluation of the census statistics. In addition, we must describe multiple indicators of survey and census quality.

We’ll attempt to do so in clear and terse language.

Please submit any questions pertaining to this post to ask.census.gov