Opinion: Daily Journal

Reading The North Carolina Polls

RALEIGH – With North Carolina’s filing deadline past us and the primary date fast approaching, news coverage of state elections is going to start ramping up. You’ll be able to find much of it summarized each day here at Carolina Journal Online, as well as other links and discussion in JLF’s “Locker Room” blog. Many of the horserace stories will reference public-opinion polls. I encourage you to read and use the poll results carefully, regardless of how the reporters themselves do.

Recent posts by bloggers Ryan Teague Beckwith at “Under the Dome,” Tom Jensen at Public Policy Polling, and Chris Hayes at Red Clay Citizen have all touched on the similarities and differences among the regular survey series. Rather than write yet-another cautionary piece about potential sources of error and how to read polls – you can glean a lot more insight about this general subject by reading Mark Blumenthal, for example – I’ll devote today’s column to presenting what I know about North Carolina’s polling outfits.

There are five such outfits conducting regular North Carolina polls in recent months: Elon University, the Civitas Institute, Public Policy Polling, Survey USA, and Rasmussen Research. Elon’s poll is designed by Hunter Bacot and other faculty there and is actually conducted by students using a specially equipped room on campus (similar to the one my fellow students and I used 20 years ago when conducting the Carolina Poll at UNC-Chapel Hill). As I noted in Thursday’s column on the governor’s race, the Elon University Poll is an interesting gauge of public sentiment on a host of issues. However, because it makes no effort to screen for voter registration or likelihood, it should never be used to frame articles or arguments about the political horserace.

The Civitas Institute is a conservative policy group in Raleigh. It used to be headed by Jack Hawke, the former leader of the state Republican Party, but he left to manage Pat McCrory’s gubernatorial campaign. The new president is Francis De Luca, who formerly headed the grassroots organization Americans for Prosperity in North Carolina and is a colonel in the U.S. Marine Corps. Civitas writes its poll questions but the actual live-operator survey is conducted by a contractor, Tel Opinion Research. Its monthly DecionMaker Poll samples registered voters who have participated in either the 2004 or 2006 general elections or have registered to vote since 2006.

While Civitas leans to the Right in political ideology, Public Policy Polling leans Left. It conducts monthly automated surveys (called Interactive Voice Response, or IVR) on its own while also collaborating with the North Carolina Justice Center (the Left’s able counterpart to JLF) on polling projects. Its founder, Dean Debnam, is an entrepreneur and longtime Democratic activist. PPP is conducting regular polling not just in North Carolina but also in many other states, and is properly gaining nationwide attention for its efforts.

PPP uses a tighter screen that Civitas in two ways. First, it samples registered voters who have participated in either the 2004 or 2006 primary elections. Second, when asking questions about Democratic or Republican primaries, it requires unaffiliated voters to choose one. Civitas asks both its Democratic and Republican primary questions to the same group of independent respondents.

As a result, while PPP’s recent Democratic match-ups for governor have averaged 89 percent Democrats and 11 percent unaffiliated in its samples, Civitas’ recent Democratic match-ups have averaged 79 percent Democrats and 21 percent unaffiliated in its samples. Similarly, PPP’s Republican-primary samples have averaged 87 percent Republican and 13 percent unaffiliated. The Civitas averages are 75 percent and 25 percent.

That’s probably one reason why Civitas polling tends to yield more undecideds than PPP does. You’ll find more undecideds among independents than among partisans. When I studied the two organization’s crosstabs, however, I was surprised to find few other significant differences in the samples. Civitas’ samples skew somewhat older than PPP’s, though not as much as I expected.

Far greater differences exist between the Civitas/PPP samples as those of Survey USA, a national firm that does IVR surveys for North Carolina television stations, including WTVD Channel 11 in Durham. Survey USA’s voter screen generates samples that are younger, more partisan, and have somewhat-lower percentages of white voters. For example, while the January and February samples for Civitas and PPP averaged about 20 percent young voters (by which I mean those 18 to the mid to high 40s, depending on how the pollster classifies them), Survey USA’s samples averaged 48 percent young. This is a huge difference. Also, Survey USA’s samples have averaged 93 percent partisan (either D or R) and 7 percent unaffiliated.

Finally, Rasmussen Research has been conducting brief monthly polls for WRAL Channel 5 in Raleigh. Like PPP and Survey USA, Rasmussen uses IVR technology, which lowers polling costs significantly but also draws some criticism from traditional pollsters. Unlike PPP and Survey USA, however, Rasmussen is not yet asking traditional horserace questions for WRAL, offering approval/disapproval data instead, so the poll isn’t really comparable to the others at the moment (and the February results are, well, simply odd).

By the way, I should say that Scott Rasmussen is an old friend of mine, a former chairman of JLF and Charlotte resident. JLF was one of the early customers of Scott’s first polling firm back in the mid-1990s (he and his father had previously founded the television network that became ESPN).

I’ll keep tracking the various surveys as we head towards the May primary. Perhaps the patterns will change. I assume that the major newspapers may also be commissioning polls as they have in the past. The News & Observer has recently used Research 2000, The Charlotte Observer has used in-house unit KPC Research (perhaps now the two papers will share a pollster), and the Winston-Salem Journal has used Mason-Dixon.

At this point, however, my conclusion is that PPP’s model is probably the closest analogue to the primary electorate, though if turnout trends in other states are replicated in North Carolina, Civitas’ general-election model may end up being a pretty good one. Either will do, though. The Survey USA samples, on the other hand, are strange enough that I won’t be putting much stock in the findings.

But that’s just me.

Hood is president of the John Locke Foundation.