Reporting the results of polls

"Poll says less than half think Bush honest" This is the title of an Associated Press article in today's paper. The article cites a recent AP-Ipsos poll of 1000 adults. Forty eight percent of the respondees said Bush was honest while fifty percent said he was not honest. This was the extent of the explanation. Remember the title of the article?

What's wrong with the article? Well, for starters, no margin of error was given for the poll. Polls (whether done correctly or incorrectly) are intended to state the current sentiment of a population. In this case, the article only gave point estimates. These point estimates were used to state the current sentiment of the population (adults). But are these the true percentages that exist in America? That's where the margin of error comes in to play. With a sample size of one thousand, the margin of error (for a 95% confidence level) is approximately 3.1%. This helps set a boundary for the truths.

Taking the results of this poll and applying the margin of error: At a 95% confidence level, the true percentage of adults who think Bush is not honest could be as low as 46.9% and as high as 53.1%. The true percentage of adults who think Bush is honest could be as low as 44.9% and as high as 51.1%. So the "truths" overlap. You can argue that the title of the article is misleading.

In summary, based on the data in the sample, there is no significant difference between the percentage of adults that say Bush is honest and the percentage of adults that say Bush is dishonest.

It must be said that a similar AP-Ipsos poll in January 2005 did show a significant difference between the two percentages (53% honest, 45% not honest). It does appear that a shift is occurring in the population.