| Get Jonathan Bernstein's newsletter every morning in your inbox. Click here to subscribe. The 2020 elections aren't over. No, I don't mean President Donald Trump's ongoing efforts to undermine democracy. Nor do I mean New York's embarrassingly long count — hey, New York, get your act together! No, I'm talking about the two Senate election runoffs coming in Georgia on Jan. 5, the ones that will determine whether there will be a 52-48 Republican majority, a 51-49 Republican majority or a 50-50 Democratic majority with Vice President-elect Kamala Harris breaking the tie to organize the Senate once she takes office. And Thursday brought us polls from SurveyUSA, a well-regarded company, showing toss-up races for both seats, with Democrats Jon Ossoff and Raphael Warnock slightly leading incumbent Republicans David Perdue and Kelly Loeffler. I know what you're thinking: More polls? After they got everything wrong in November? Well … yes, more polls. For one thing, the polls didn't actually get everything wrong (see the link below). And polls — properly understood — are still the best way we have of finding out what's actually going on with voters. The trick is that proper understanding. All public-opinion surveys are limited in numerous ways. Election polls have a particular set of challenges, beginning with the inescapable fact that they're attempting to measure something, the actual voting electorate, that doesn't exist until the election. Pollsters are constantly having to update their practices as reality changes — who responds and how. The same mathematics that make calculating the views of many people by surveying only a fraction of them also tell us not to take the specific estimates of any one poll too literally. And while averaging the results of many surveys can help with the problem of random error, those averages have their own potential problems, since they're only as helpful as the underlying polls are accurate. So we go back to the basics. Election polls give us information, but that's not the same thing as saying that they perfectly predict outcomes. Single polls don't get us very far, and should be used with extreme caution. Averages are better than single polls, but they may not be available in many elections, and they're still subject to potential errors. If we want to know what will happen in an election, however, surveys are still among the best evidence available, and certainly better than counting lawn signs or interviewing people at diners (we can learn interesting things from both of those methods, but not who will win an election). Talking to campaign insiders or seasoned local observers is likewise problematic, not least because such people will themselves be using polling in forming their opinions. We could, of course, give up on predicting what will happen, or give up on horse-race coverage of elections altogether. But wanting to know outcomes in advance seems to be an inherent human desire, widely shared, and so horse-race coverage is always going to be a major part of how the news media reports on elections. Getting rid of polling won't fix that. But we should try to underline the uncertainty inherent in the polls, and to read polls with that uncertainty in mind. 1. Rick Hasen on the latest with Trump's hapless election lawsuits. 2. Robert Y. Shapiro at the Monkey Cage on election polling in 2020 and other public-opinion surveys. 3. My Bloomberg Opinion colleague Shannon K. O'Neil on President-elect Joe Biden and Latin America. 4. Carrie Cordero on Trump's pardons. 5. And Lydia DePillis on Trump's longer-term effects on the bureaucracy. Get Early Returns every morning in your inbox. Click here to subscribe. Also subscribe to Bloomberg All Access and get much, much more. You'll receive our unmatched global news coverage and two in-depth daily newsletters, the Bloomberg Open and the Bloomberg Close. | 
Post a Comment