Sometimes people think it’s a disaster when you have more predictors than data points, but I always point out that, no, it’s better to have 9 predictors than just 1 or 2. After all, if you really wanted just 1 or 2, you could just throw out most of your data!
Nate’s chart is excellent, especially the ordering of the candidates in order of the percent favoring resignation:
I also like the gratuitious exclamation marks which add fun value without actually making the graph any harder to read. The key reason this works is that Nate wisely did not fill in the blank squares with “No!”s.
My only comments are:
1. Maybe there would be a good way to order the rows.
2. I’d think a key, key, key predictor is the politician’s popularity before the scandal. In particular, I recall Spitzer having some problems even before all this came out, to the extent that fellow Democrats didn’t mind dumping him (in retrospect, maybe a big mistake, but that’s another story).
3. Why only 7 cases? Aren’t there lots, lots, lots more politicians who’ve been caught with their pants down? We all know about Barney Frank and Newt Gingrich, but there must be lots more with known extramarital affairs, illegality, etc. Although maybe these didn’t all hit the newspapers.
As you note, the columns have their own natural order, so they're not, to use Bertin's terminology, re-orderable. But the rows don't seem to have a natural order, so you can re-order them according to their values. This is usually best done by hand, but most people either haven't got the time to spend trying out different combinations, or, like tidying a room, they're paralysed knowing where to start.
One nice heuristic for starting with is the barycentre heuristic. You assign each "yes!" an imaginary weight, and calculate the centre of weight each row would have if you had to balance the weights on a seesaw. For instance, "hookers" would be evenly balanced, "gayness" and "criminal conduct" would lean slightly left, and "subordinate" would lean slightly right.
Then you can order the rows by their centre of weight. I find this is usually just a start, and I then begin to notice row categories that would go together better. But it is a good start, and if you're short of time, you can leave it at that.
You could also add, "Democrat?" Being a Republican appears to cancel out the impact of hypocrisy.
McGreevey also lacked the "believable stand by your man moment" trait.
This is fun and is a good starting point.
But it has a big deficiency if we truly think the point of the table is to evaluate predictors.
I fail to see which predictors are correlated with the outcome, not so with resignation and even less so with approval rating.
Take the first row… it would appear that hooker or not is completely unrelated to outcome! (okay there are two data points only…)
And then if we take Hypocrisy, it too appear unrelated to outcome since almost everyone had that characteristic.
So surely the rows ought to be arranged by predictiveness!
Thanks for the pointer though. Fun stuff.
Really, the best predictor seems to be whether the politician was a Democrat. Two out of the three Democrats resigned, and the third still paid a political price in impeachment. Being a Republican seems to allow you a free pass in these things.
I think the Republicans just have a more cohesive and disciplined party, allowing their politicians to get away with more stuff.
Of course, he'd already "resigned" when it hit the news. Nonetheless, I'm sure John Edwards is hugely relieved not to find his name on this chart.
Do mayors count? "Buddy" Cianci and Marion Barry come to mind. Cianci was forced to resign by law, and Barry did not resign (but chose not to run for mayor again) and continues in city politics. No doubt Cianci would not have resigned if possible. So, I would add "sense of invincibility" as a predictor. Psychologist readers of this blog can no doubt suggest a better term.
I don't think party acceptance has much to do with it per se. Maybe people who are good at being sneaky are more likely to be Republican and don't get caught until they are of high enough profile.
Some more that we could expand the data set with: Bob Livingston, Rudy Giuliani, David Paterson, Jim McGreevey…
Isn't using government money for trips to see your squeeze criminal conduct? Let's give Sanford another Yes!
Kwame Kilpatrick would be another good example. He checks off every box except for gayness. He resigned, but only after it became clear that he would've been kicked out of office. Only 43% thought he should resign in early 2008 (right about the time that the city council voted to ask the governor to kick him out of office… he resigned several months later).
I disagree that there's only 7 data points. I count 63 binomial data points.
Hey, where'd that pie chart come from?