Polling the governor's race
Maryland political junkies were treated recently to a spate of governor's race polls that deserve a closer look.
The problem with political polls is that their results vary depending on differences in polling methodology. Even when pollsters seek accuracy by asking questions straightforwardly, they are faced with the daunting challenge of sampling.
Sampling means selecting a group of respondents that best resembles the demographics and political sentiments of the voters who actually show up on Election Day.
Pollsters don't just count the first 500 or 1,000 people they call on the telephone. If you've ever responded to a telephone poll you know that the initial questions were about your age, gender, race, party affiliation, voting history and so on. Based on this intel you either were included or excluded from the final poll results, depending on how many respondents of similar backgrounds the pollster needed to balance the sample accurately. In other words, if you're another white, middle-aged Republican male of which the sample was already at maximum, your response wasn't counted.
But this prompts the question: how does the pollster know what types of people are going to show up on Election Day? How does the pollster decide what percentages of women or blacks or Democrats to put into the sample so that it's weighted accurately? Well, it's an educated guess, but therein lies the art of polling.
When people vote, the only record they leave behind is their party affiliation and address. Pollsters must rely on so-called exit polls (people willing to be interviewed as they leave the polling place) to get a picture of voter demographics. But this is sketchy stuff and even if you know the voter demographics of the last election, it doesn't guarantee they'll be the same next election.
Within the polling community there is also debate about cell phones. As people increasingly use cell phones for their only phone service, do pollsters who exclusively call land lines unwittingly skew their surveys by missing cell phone users?
There's also a debate over robo-calls automated telephone polling. Some pollsters feel that people respond more candidly to a robot than to a live interviewer, especially about race and other touchy issues.
But, ultimately, it's sampling that mostly accounts for the difference in poll results. So let's look at the differences in the Maryland governor's race polls.
The most recent poll (July 10-12) was taken by Public Policy Polling showing Martin O'Malley leading Bob Ehrlich 45 percent to 42 percent.
Both PPP and The Washington Post poll measure a much broader sample (registered voters) than most polling services, which limit their samples to likely voters. This difference, alone, skews results.
There are 3,416,000 registered voters in Maryland, but for a typical governor's election only about 60 percent vote. Why some political pollsters measure people who don't intend to vote is a mystery. You can identify likely voters during telephone interviews by asking if they voted in the past few elections and if they intend to vote in the next election.
By definition, a poll of likely voters is more reliable than a survey of all registered voters. And, to underscore the difference in results, consider The Post's poll taken in early May. Among all registered voters, O'Malley was ahead of Ehrlich by eight points (49 percent to 41 percent). But when the same poll only counted voters who definitely intended to vote, the results were a dead heat O'Malley 47 percent and Ehrlich 47 percent.
O'Malley and his chief media booster, Washington Post reporter John Wagner, repeatedly refer to O'Malley's "eight-point lead" which is not only out of date but never was accurate in the first place.
Among pollsters who measure likely voters, Ehrlich is ahead by 46 percent to 43 percent (Magellan Strategies, June 29) and 47 percent to 46 percent (Rasmussen, July 8).
O'Malley lashed out at the Rasmussen results, calling it a "Republican poll," but the same poll, on the same day, using the same samples, gave Sen. Barbara Mikulski a 25-point lead over her Republican opponent, and in the 2008 presidential race Rasmussen nailed the final results.
However, the best way to detect political trends and momentum is to look at the same polling service's series of polls over time. For instance, the Rasmussen poll showed O'Malley ahead by 6 points in February, ahead by 3 points in April, tied in June and behind by 1 in July.
That's why this week Real Clear Politics, a national political handicapping service, reviewed the Maryland polls and said, "We still consider the race a tossup, but if forced to make a call today, we now put a thumb on the scale for the Republican (Bob Ehrlich)."
Thanks to the polls, that's a pretty good picture of how these two horses are running at the turn. But the finish line isn't until Nov. 2.
Blair Lee is CEO of the Lee Development Group in Silver Spring and a regular commentator for WBAL radio. His column appears Fridays in The Gazette. His e-mail address is email@example.com.