State-by-State

Survey USA's aggregation of 50 independent state polls sure does lead to some interesting results. Here's Obama versus McCain:



And here's Clinton versus McCain:




Provocative, but I don't buy it. Each of these polls has a sample size of 600, so the margin of error will come into play. What's more, there are 100 separate polls being aggregate here, so the odds are that several of these are just bad samples. On top of that you have all the usual problems with early polling. Were I to inject some pro-Obama spin into this, I would note that there are Democratic Senate pickup opportunities this year in Oregon, New Hampshire, Colorado, and Virginia where the polls show Obama winning and Hillary losing, but there are no such races in Arkansas, West Virginia, Pennsylvania, Michigan or New Jersey where the map shows Hillary winning and Obama losing. But as I say, while I think this method is clever, I don't really see it as methodologically sound -- Clinton's not going to lose Washington, Obama's not going to lose New Jersey, and what happens in Michigan will have more to do with whether or not McCain runs demagogic attacks on the Democrats' global warming plans than on who the Democrats nominate.