In the Two Minute Drill, we explain complex issues in politics in 500 words or less (roughly the amount of words it takes the average adult two minutes to read on a monitor). Politics just isn’t always that complicated. Without the fluff and partisan bias, even the most complex of our political differences can be explained succinctly. This week: we are constantly being inundated with polls, but when are they worth taking seriously? This is The Two Minute Drill for May 20, 2016.
As he’s slipped farther behind in the primary and delegate math, Bernie Sanders has made a lot out of general election polling as one of the justifications, if not the main justification, for him staying in the Democratic race. “The Democrats want to see the strongest candidate possible take on Trump or some other Republican,” Sanders told NBC’s Andrea Mitchell in April. “At this point, according to the polls, that is me.”
In terms of raw polling numbers, Sanders has a point. According to the RealClear Politics polling averages, in a hypothetical Sanders-Trump matchup, Sanders holds a 13.0 point lead (Sanders 51.8%, Trump 38.8%). In contrast, Hillary Clinton only holds a 5.2 point lead in a hypothetical Clinton-Trump matchup (Clinton 47%, Trump 41.8%).
Sanders’ poll numbers do not necessarily mean that he is the best candidate to take on the Republican nominee (almost assuredly Trump) in November. But they beg the question: When do general election polls become meaningful?
The Explanation (500 or Bust)
General election polls become more reliable as the months and weeks go by. Polls taken well ahead of voting day have a history of being wrong. According to data compiled by FiveThirtyEight, polls taken one year in advance have been inaccurate by more than 5 percentage points in the last 10 out of 14 elections. Even polls six months out – roughly where we are in this election – have been inaccurate. At this point in the 2000 election, for example, George W. Bush had a 5-percentage point advantage over Al Gore. As the scrutiny over Bush’s record as Texas governor grew, however, his net favorability dropped considerably – from +43 percentage points in December 1999 to +16 percentage points on the eve of the election. And despite winning in the electoral college, Bush lost the popular vote to Gore by half a percentage point that November.
This does not mean, however, that the general election polls become more reliable in a steady, gradual arc. Rather, after starting off as essentially meaningless, they ratchet up sharply in two defined steps.
As you can see from the graph above, the first step occurs toward the beginning of the primary process (between 200-300 days before Election Day). At this point in the nominating process, primary and caucus voters have been introduced to the party candidates, have formed preliminary opinions about them, and, in some states, have even cast primary ballots. In 2016, the first primary – historically the Iowa Caucus – occurred 281 days before Election Day on February 1.
The second step occurs roughly 100 days before Election Day, corresponding to the parties’ conventions. Once the nominees are known and the general election campaigns begin, the polls start to get very good at predicting the outcome.
So, what can we learn from the current polls?
As of this writing, we are 172 days away from Election Day. While general-election polls will become more predictive as we get closer to the nominating conventions, they can tell us quite a bit already. And according to Sam Wang at the Princeton Election Consortium, they “give a way to estimate the probability of a Hillary Clinton [or Bernie Sanders] victory.”
Going off Wleziena and Erikson’s research, Wang measured how far general election polls move over time, including data for 16 elections to estimate the standard deviation of poll-outcome differences. His findings look like this:
The poll variation, which is quantified by the standard deviation in the chart above, is what matters here. For mid-May, the standard deviation is around 8 percentage points. Hillary Clinton’s polling margin over Trump is currently +5.2% – within the standard deviation. In contrast, Bernie Sanders’ polling margin over Trump is +13.0 – 5 percentage points above the standard deviation. What does that mean? It means that Sanders has a legitimate argument about the polls. Given the standard deviation at this point in the election, a Clinton victory is almost completely uncertain. In contrast, the likelihood of a Sanders victory in November vs. Trump is quite high.
Of course, the polls could change, and this exercise hardly means that Clinton will doom Democrats in November. But now that the polls are beginning to be predictive, we have a more direct measure of what will happen.
Word Count: 544 (bust)
Issue Brief: A Comprehensive Guide to Hillary Clinton’s Email “Scandal”
In case you missed it, check out The Weekly Column. By the end of it, you should have a pretty good idea of the factual and legal issues present in the Clinton’s email controversy. Read the Column for May 17, 2016.