Exposed: The American Research Group’s Anti-Giuliani, Pro-McCain Bias

The other day, The Bij emailed to ask why the American Research Group’s state-by-state primary polling seems so out of whack. In each of the states they survey, McCain remains the prohibitive frontrunner with Giuliani struggling to break out of low double-digits. Do they have a special insight into the race that all other pollsters don’t? I decided to do some digging.

For the longest time, New Hampshire-based ARG refused to even add Giuliani to its polls, using the flimsy excuse that he hadn’t visited New Hampshire. Huh? By this standard, Hillary Clinton should be excluded from ARG’s polls!

Then, when did another round of polling of other early primary states, they continued to exclude Giuliani, to the obvious benefit of McCain. I’m not sure what his not visiting New Hampshire has to do with a poll of South Carolina (where he has visited) or Maine.

In their latest rounds of polls, ARG seemingly bit the bullet and included him – but on a second question, after a first non-Giuliani ballot test. Odd, I thought. Most other polls included Rudy straight-away, and if they wanted to gauge the impact of his departure from the race, would ask a second non-Rudy question. And a bell went off: studies have shown that question order matters in polls. Ask the Bush approval question early in the poll, as opposed to last, after twenty or thirty leading questions, and you get a more favorable result. I wondered if the same could be true of ARG’s unorthodox practice of including Rudy on the second rather than the first ballot. More importantly, how to prove it?

And then a second bell went off: Condi.

Numerous pollsters have asked the 2008 primary question with and without Condi included. Was there any impact in those polls that added Condi to a second ballot, as opposed to taking her off after the first?

There was. On average, Condi’s support was 40% lower when she was added as an afterthought and not included right away.

Condi on First Ballot		Condi Added to Ballot
WNBC/Marist 2/06	22%	SV Wisconsin 6/06	10%
WNBC/Marist 10/05	21%	SV Florida 5/06	10%
WNBC/Marist 2/05	14%	SV Michigan 5/06	10%
CNN/USAT/Gall 12/05	18%	SV Washington 5/06	12%
Zogby* 12/05	12%	SV New Jersey 5/06	18%
Diageo/Hotline 11/05	22%	SV Pennsylvania 5/06	10%
Fox News/OD 9/05	19%	SV Georgia 5/06	10%
		SV New York 4/06	11%
Average with Condi on 1st Ballot	18.3%	Average with Condi Added on 2nd Ballot	11.4%
* Low numbers for candidates across the board, high undecideds

This isn’t a perfect comparison of course. All the polls to the left are national polls. And all the polls to the right are state polls from Strategic Vision – because they were the only other ones I could find who added candidates to a second ballot. No national pollster did this – perhaps because a more experienced pollster would know that this technique raises a huge red flag. A one-to-one comparison assumes the race has remained fairly static (a safe assumption given the Giuliani-McCain numbers haven’t moved) and that SV’s states are representative overall (eyeballing it, they sure seem to be).

Looking at the evidence, it’s fairly clear that adding Rudy to the ballot is depressing his numbers. What happens if you give him the same 60% boost Condi gets with first ballot treatment? His numbers across the eight ARG states rise to 26% -- right in line with his national numbers. Mystery solved.

Is ARG doing this intentionally? A long line of evidence suggests a disturbing answer. They initially refused to add the national frontrunner to their poll, even outside their New Hampshire backyard. When they did, they chose a technique that gave him numbers that were abnormally low. Beyond the early states, their selection of states also tends to be weighted to the most heavily RINO/McCain states from 2000 – Vermont? Rhode Island? Maine? Massachusetts?

ARG's polling on the 2008 Republican primary is extremely unreliable. Here at Giuliani Blog, we will not report on the result of another ARG poll until ARG responds adequately to questions raised about its methodology and impartiality.

Find this post interesting? Get posts from this blog in your inbox once a day: