One galling approach that some people in the media have is to present polling data as indisputable facts. In many cases, the polling data are inferential statistics. The conclusions are, at best, conjecture.
By way of illustration, it is possible to find out what per centage of the population like broccoli. Taking a representative sample of the population, it is possible to be forthright and ask “do you like broccoli?“. There would be a quantifiable answer and the data can be supported by asking how long ago broccoli was served as part of the family meal. The answers would be fairly truthful, as there is no great social stigma to liking or not liking broccoli.
Polling data are less clear cut when examining intervening variables such as racism. It would be useless to ask a question such as: “do you consider yourself a racist?“. Of course, the answer would be negative. The term itself has a wide range of meaning. Further, there are social expectations involved and the validity of the answer may be questionable.
The best that polling data can do when measuring such abstract constructs is to ask questions that may imply a racial element in decision making. The questions are not direct in order to mask the variable that is being measured. The results, at best, are only suggestive – and they are subject to a wide range of interpretations.
For media people not to state a disclaimer on data purporting to measure abstract concepts is misleading. It either shows a poor understanding of the limits of statistical measure and / or a quick acceptance of polling data. Unfortunately, much of these so statistical ‘facts’ are educated guesses – and they should be presented as conjecture.