Phil: It’s worth comparing the candidates’ performance in the ARG polls among Republicans vs. independents. In New Hampshire, where Giuliani and Romney are in a statistical dead heat among both Republicans and independents, it might not matter too much whether the percentage of independents in ARG’s sample is representative. But in Iowa and South Carolina, a little bit of statistical noise on this front could make a big difference.
In Iowa, the sample of likely republican voters (those who “definitely plan to participate” in the caucuses) is 82% Republican and 18% independent. Romney is winning among the Republicans with 24% to Giuliani’s 21%; Romney registers only 6% among independents, who back Giuliani by 25% and McCain(!) by 37%.
In South Carolina, Giuliani leads F. Thompson 29%-25% among Republicans, but Thompson crushes Giuliani among independents, 43% to 17%. The South Carolina sample is 90% Republican and 10% independent.
In short, if ARG is oversampling independents in Iowa, or undersampling them in South Carolina, Giuliani’s leads in those states are illusory.
