Q & A: Last Week’s Rasmussen Poll
Right now, Ballotopedia lists 12 declared or potential candidates for Governor of Wisconsin. There is still time for others to jump into the race. Unless someone with a lot of name recognition jumps into the race soon, however, it’s likely to be a two-way race between incumbent Scott Walker and Democratic challenger Mary Burke. Last week’s Rasmussen Reports poll showed Walker and Burke tied at 45%-45% (+/- 4.5 points) among those surveyed.
Nearly every release of poll data is coupled with the words “left-leaning poll” or “right-leaning poll”, so let’s find out a little more about this new poll from Rasmussen Reports.
Q: What’s Rasmussen Reports?
A: Rasmussen Reports was founded in 2003 by conservative activist Scott Rasmussen (he recently left the organization). There is now a subsidiary called Pulse Opinion Research that Fox News treats as a separate company, but which uses the same methodology as Rasmussen polls, according to Scott Rasmussen himself.
Q: Who, exactly, was surveyed for this poll?
A: They won’t tell us. Rasmussen will not make public their sampling methodology. This blogger wouldn’t pay for a one-year (@$20/month) subscription for pollsters with such crappy math, especially when it wasn’t entirely clear whether they would share all of their crappy math even with a subscription.
Q: What is Rasmussen Reports’ record on predicting election results?
A: Poor. Rasmussen had Mitt Romney and President Obama tied at 49-49 in Wisconsin. President Obama carried Wisconsin with a seven-point margin. Better pollsters predicted that President Obama would carry Colorado, Ohio and Wisconsin. He did. In the end, Rasmussen was right only one third of the time when it came to predicting election results in battleground states.
Q: What does Rasmussen do—or not do-- that makes its polling so inaccurate?
A: Several things, according to Nate Silver of New York Times’ FiveThirtyEight blog (probably the most accurate pollster alive today). First, they rely only on landlines for their robocalls, while other polling firms either call cell phones or adjust their numbers to estimate cellphone-only user results. Many younger voters rely solely on a cell phone and no longer have landlines. Second, at least in 2012, Rasmussen banked on Wisconsin’s people of color not turning out at the polls. Big mistake, as Milwaukee’s recent elections have seen all-time high rates of voter turnout among African American residents, who make up about 40% of the city’s population. Third, Rasmussen often predicts lower voter turnout for young voters, who tend to lean left, and higher voter turnout for older voters, who tend to lean right. (Neither of those things happened in 2012). Fourth through sixth, Rasmussen’s robocalls are made only during a four-hour period, they don’t do call-backs if they miss people, and they use computer scripts instead of live interviewers. These are cost-saving measures for Rasmussen that lead to less reliable results.
Q: So the results showed a +/- 4.5-point margin of error. What does that mean?
A: It means that the results for Walker or Burke could be as much as 4.5 points higher or lower, meaning either could be up to 49.5 or down to 40.5.
Q: So how much credibility can we even give this poll? What should we believe?
A: Nobody knows, though if you want to give more points in either direction, it’s much more likely you’ll need to give those to Burke. Put it this way: among older, whiter, landline-only voters, Walker is tied with Mary Burke. Across a wider age spread and an ethnically and racially diverse populace he’s faring poorly against Burke. Among young people who rely only on cell phones—a rapidly growing part of the state—he’s faring poorly against Burke.
Q: But they probably know who’s likely to vote which way, so why don’t they just use history as a guide, and extrapolate from those results?
A: Rasmussen weights surveys based on assumptions about how many people identify in which party in the state. This is a problem because many people do not identify with a specific party, they have political beliefs or leanings. Big difference.
Q: Maybe you’re just a Lefty who’s try to play with statistics, because we all know statistics can all be manipulated to say whatever people want.
A: No, not all statistics can be manipulated to say whatever people want. Rasmussen Reports tried that, essentially, and failed. Nate Silver’s FiveThirtyEight blog has tracked polls since 1998, and Rasmussen Reports had the most wildly off poll results in its 24-year history of poll-tracking (as of 2012): a whopping 40 POINTS off the mark in a Hawaii Senate race.
FiveThirtyEight puts the Rasmussen pollsters’ typical average error at 5.9 points, biased toward Republican candidates by about 4 points. So yes, this might mean that it’s more like 41% Walker and 49% Burke, but it’s likelier somewhere in between, with Burke LEADING.
Q: So you’re saying Burke is really ahead of Walker?
A: We don’t know for sure, but it’s likely she’s ahead, if only slightly. It’s best to wait until the real pollsters at Marquette University Law School publish their results later this month. When they do, they’ll also publish information on how they arrived at their results. For free!