In a previous post I noted that the recent ARG poll (and the similar Rassmussen poll) showing a tight race between Lieberman and Lamont was actually not that close. I had intended to demonstrate the challenge for Lamont with some basic mathematics, but never found the time. Well, now I have the time – and the interest.

First we need to recall basic numbers in the ARG poll. Leiberman leads with 44%, Lamont comes in second at 42% and Schlessinger is at 3%. 11% remained ‘undecided”. In the Qinnipiac poll (53-41 Lieberman) the undecideds were obviously pushed to reduce the undecideds to only 2%, the 9% basically all going to Leiberman. I think this is the most optimistic possible outcome for the race for Leiberman – his potential high water mark. The fact that Lamont and Schlessinger are at the same level in all 3 polls means we need to make some predictions on what the undecideds will end up doing in order to bound the realistic outcomes – as of today.

The first prediction can be made assuming undecideds are no different from the 89% who did make a choice in the poll and we simply treat the results of indicative of the general population. To do this we add 44 + 42 + 3 to reach the 89% of those who did chose a candidate and compute the new percentages of the total. This gives us 44/89 = 49.5% for Leiberman, 42/89 for Lamont 47% for Lamont and 3.5% for Schlessinger. This is actually the best Lamont can hope for at this moment – that the undecideds follow the general population.

Unfortunately for Lamont this is not how the undecideds were broken down in the ARG poll. In that poll 5% of dems were undecided, 14% of Reps and 13% of Independents. That means of the 11% still undecided, 17% are democrats, 40% independents and 43% Reps. The sad truth of the numbers are the undecideds are 4 to 1 coming from the voter pools now leaning heavily towards Leiberman. If we assume each subgroup of undecideds will follow their group’s overall pattern we see a different story. It is not unrealistic to assume undecided dems will break down just like all dems, etc. If we assume these undecideds will trend with their general group, then the dems will favor Lamont 65-30, the Reps Lieberman 57-18, and Independents Lieberman 48-38.

So let’s recompute based on this assumption alone. Reps: 43% of 11% is 4.75%. Indies: 40% of 11% is roughly 4.4%. Dems: 17% of 11% rounds to roughly 1.85%. If Leiberman gets 57% of the Republican undecideds we can add 2.7% to his 44%. We had .85% to Lamonts 42% and .5% to Schlessinger’s 3%. The new subtotal is Leiberman 46.7%, Lamont 42.85% and Schlessinger at 3.5%.

Next we do the Independents. Leiberman will get 48% of their vote which turns into an additional 2.1%. Lamont will get 38% or an additional 1.7%. And Schlessinger will get basically nothing (only 1% of independents are for Schlessinger). The new totals then are Leiberman now at 48.8%, Lamont at 44.2% and Schlessinger basically at 3.5%.

Now we do the Democrats which are the smallest group in the undecideds. Lamont gets an additional 1.2% from this group. Leiberman adds 0.5 and basically nothing for Schlessinger. The final tallies are Leiberman at 49.3%, Lamont at 45.5% and Schlessinger with 3.5%. Those paying attention will note that these will not add to 100, but to accurately compute the share of votes for each group required recalculating the share without undecideds. For example, Dems favor Lamont 65-30 with 5% undecided. The 65-30 split with the undecideds removed translates to 66.3-33.6 for Lamont. The extra step is adding precision without any benefit. Because what we want to see is where the race would end up if the undecideds broke along the tendencies of the three voting blocks. And we see Joe Lieberman within striking distance of the magic 50% mark still – and that is in a three way race! No matter how you slice it Lamont will come up short.

I still doubt any poll estimating the dems will break 65-35 for Lamont when a recent primary measured the split to actually be 52-48. As we can see from this example, how you allocate the votes across your sample can change the results dramatically. A 2% difference turns into a 4% gap depending on the model and assumptions one uses. No matter how people spin the numbers, Lamont is anchored in all three polls in the low 40%, and that could be unreasonably high because it assumes those Democrats who shunned Lamont once for Leiberman will not do so again. Highly unlikely IMHO.