Thursday, September 29, 2016

Waiting for a post-debate Clinton bounce

I've gotten this one from a bunch of students, so here we go.  Clinton trounced Trump on Monday, so when, if at all, will this show up in the polls?  Well, the latest PPP poll has Clinton up by four points over Trump.  That was done over Tuesday and Wednesday.  Now, PPP is a partisan organization, Democratic-leaning, but that doesn't necessarily mean they are wrong.  But, it is all we have.  It will take a few days before we have enough polls to confirm or disconfirm a bounce.  The question is whether or not that gets sustained, and that really depends on Trump.  If Trump's gain in the polls over the last month was from the fact that he mostly refrained from the kind of inflammatory statements that got him in trouble in the past, then his post-debate attacks on Alicia Machado suggest that his debate performance was not an isolated incident.  And, as I argued yesterday, if Trump performs similarly in the next two debates, then Clinton's bounce, should it occur, won't just be transitory.

But, we don't have enough data yet to know whether or not Clinton really will gain in the polls.  The PPP sample was 933, and Clinton's lead wasn't that much smaller in the national polls before the debate.  So, the evidence of a polling bounce so far is actually weak factoring in PPP's leanings.

Right now, the Democrats shouldn't be celebrating that much...  This thing is still close.


  1. Nate Silver favors comparing a pollster to their last poll.....
    In which case, nothing about this race has changed all summer, as PPP has said either +4 or +5 Clinton ever since May.

    Your students just have to wait.

    Problem is that if you're looking at the aggregators, Clinton has been trending up this week....and that DOESN'T include any polls this week! So, it's going to be VERY possible that Clinton was already improving slightly and we can't distinguish a debate effect from normal bumps and wiggles.

    1. Note: I don't like Silver's logic on comparing apples to apples, as it seems to kinda ignore that whole sampling variation thing.

    2. The irony is that Silver's method can only apply to the kind of rolling cross-section that the LA Times uses, but then it would be circular reasoning anyway. Have I mentioned that I'm not a fan?