The diagonal line is the linear best-fit line. The two variables have a correlation coefficient of 0.779.
Romney ended up paying the most per vote: $238. Obama and Clinton both spent about $103 per vote. The biggest bargains were for Edwards ($48/vote) and Huckabee ($35/vote).
One of the things John didn't point out is that, generally speaking, the politicians got what they paid for. John McCain and Fred Thompson hardly spent any money and, as a result, got hardly any votes. Hillary Clinton and Barack Obama spent the most money and got the most votes. Of course, there are two primary exceptions, namely Romney and Edwards, but otherwise the model here works fairly well: ya get what ya pay for.
[Note: For those of you who are statistically-challenged, the correlation coefficient measures how well the data fits to the "best-fit line," as John calls it. The number ranges from +1 (perfect fit) to -1 (perfect "anti-fit"). If the correlation coefficient had been +1, each dot would have lined up exactly on the best-fit line. A correlation coefficient of 0.779, then, shows a strong correlation. This is a good thing. ;) ]
HT: Advertising is Good for You