Re:Looks like this was done by STAT 101 students (Score 2, Insightful)

For example, they did not account for: population growth between elections, the demographic makeup of such growth, median income change since the previous election, voter registration stats for each party and their changes since the previous election, reversals of traditional voting patterns (better-educated going Democratic, older going Republican), local issues that would draw certain voters to the polls, the involvement of and relative successes of get-out-the-vote organizations (the parties, unions, churches, etc), and the list goes on...

The paper took many variables into account. As many as could be expected on short notice with no budget. There are two reasons to take extra factors into account. First is because you believe they may correlate with the effect you are measuring, thus skewing the results. Second, you believe the factor is random and by accounting for it, you can reduce error and measure smaller effects.

Do you believe the change in demographics correlates with the county's decision to buy electronic voting machines (please note, the researchers took demographic makeup into account, just not the change in demographic makeup)? I ask the same question of the other factors you mentioned.

The whole point of statistics is to measure the likelihood that the effect was due to the various random factors that were not taken into account! The rho (p) values you see all over the paper represent the likelihood that the various factors you mentioned, along with countless other factors, could have caused the effect.

The accepted rho value for most academic journals is p<.05. The researchers had a rho value of p<.01. If they had accounted for the factors you mentioned, the value probably would have been even smaller. If they had accounted for every possible factor, the rho value would have been zero (they would know exactly what happened).

It is impossible to account for everything. The researchers accounted for those factors which were most likely to skew the results. No single study should be taken as absolute proof of anything. However, this study shows that the situation with electronic voting machines merits further investigation.

If you believe the factors you mentioned explain the correlation, then go ahead and do your own analysis.