That sleepy #englishbulldog pictured in the previous retweet, is none other than the Trafalgar Group’s mascot “Horatio” Named of course after the Hero of the battle of Trafalgar, Lord Admiral Horatio Nelson. Since we are a polling firm, he goes by “Ratio”. pic.twitter.com/Kj5PhmqheH— The Trafalgar Group (@trafalgar_group) August 8, 2020
I spent 30 minutes on the phone this morning with Robert Cahaly, the Head of The Trafalgar Group which PBers will recall as the pollster who called Michigan, Pennyslvania and Wisconsin correct back in 2016.
I’m not going to deny, I’m intrigued by Trafalgar. Are they a group that correctly identifies and eliminates social desirability bias, therefore picking up a raft of shy Trump supporters? Or are they Republican operatives who are keen to push the narrative that this is a close race?
Traditional pollsters reach voters by either having an on-line panel (YouGov, for example) or by calling people, with either automated voice response or a real human. The former approach risks having politically active people who’ve lied about their affiliation, while the latter has very low response rates (especially of certain demographics). Plus, because there are often 30 questions (to help with weighting) you get very high drop out rates.
The Trafalgar model is different. They buy in a list of people from a consumer research agency, properly weighted by demographics (and because this data is bought in, rather than self reported, it is more accurate). They then reach out to these people using a combination of phone, email, and text.
The Trafalgar surveys are short – four or five questions long – and emphasise anonymity to try and get rid of social desirability bias. There is no post collection weighting of results, which seems odd, because given admittedly low response rates, it would seem necessary. There’s also no “spiral of silence” type adjustment.
So: bullshit or brilliance?
I don’t know.
I don’t even know if Trafalgar engages in real polling. You see, what Mr Cahaly told me about their data collection (i.e. a combination of email, texts, phone calls, and presumably repeated follow up to try and make sure you got a representative sample) would make it much more labour intensive than a traditional pollster. And yet Trafalgar Group appears to just be the one person.
Robert Cahaly was an engaging and interesting chap. But he kept going off topic to drop in what sounded awfully like Republican Party talking points. So, Hispanics are going desert the Democrats because they’ve seen what socialism did in Venezuela and they don’t want hard left Socialism in the US, for example. And Nevada was likely to flip because people there have seen what a Trump economy does for them. And if Pennsylvania were to go Democratic, it would be because of voter fraud.
Ultimately, though, while I enjoyed the chat, I wasn’t convinced. Mr Cahaly didn’t say the things I’d expect a pollster to say, about representative samples and the like. Instead, it was all about how to make it as anonymous as possible to ensure people told the truth. That seems pretty… crude… to me. And really, if social desirability was such a big issue, we’d see bigger differences between on-line and telephone pollsters.
I could, of course, be wrong. If I am, it will probably be because Trafalgar is conducting polls in a novel and brilliant way.