If a published report of poll results contains a reference to other polls on the same subject, then it probably also contains statements about how opinions have changed over time. Provided with enough relevant information of the kind we have been discussing, a poll consumer can evaluate the previous results. This is the basis for assessing whether attitudes, opinions, or behaviors have really "changed."
The meaning of change is a difference in "comparables," that is, comparing apples to apples and not to oranges. Sometimes, a discussion of change over time really reflects differences in measurement over time (completely different results) rather than change in the same measurement. To make direct and accurate assessments of change requires a comparison of the responses to the same question, asked of similar individuals, and with the results coded in the same fashion.
Poll consumers should look to see that the same question was asked of respondents in two similar samples, if not of the same respondents interviewed at more than one point in time. The same question wording means that the same concept, such as presidential approval, was measured. If the same polling organization conducted both surveys, it is more likely that this is the case.
It is also important to note whether the same kinds of people were asked the question. Often the findings from two polls will be reported as presenting conflicting results In the first poll, the question was asked of a representative sample of the entire population; but in the later one, it was asked only of partisans of one kind or another (Democrats or Republicans). Another common problem is that results from the trial-heat question appear to change because it was asked of a sample of "adults" or "registered voters" in an early poll, but it was asked of "likely voters" as Election Day neared.
ASHER, HERB. 1995. Polling and the Public: What Every Citizen Should Know, 3d ed. Washington, D.C.: CQ Press.