Those of us who obsess about politics and midterm elections are laser-focused on two variables right now—the “stickability” of President Trump’s uptick in approval ratings since the beginning of the year, and the partial closure of the Democratic advantage in the national generic-ballot test.
We know that midterm elections are almost always referenda on the incumbent president, and that Trump’s approval and favorability numbers have been at record lows. We also know that those low numbers had contributed to a depression in overall GOP numbers, measured by the generic-ballot test and party-favorability ratings. Unfortunately, the generic-ballot and party-favorability questions are not asked by some pollsters, at least not regularly. Many surveys were last in the field just before the State of the Union speech and don’t have fresh numbers yet. Hence the uncertainty about whether the recent uptick of a couple of points is the beginning of a trend, or something less durable.
The new Gallup presidential approval numbers came out Monday, showing Trump’s approval rating at 40 percent for the second week in a row; his disapproval also remained constant at 57 percent. These are terrible numbers, to be sure—the lowest of any elected president at this point since modern polling began just after World War II. But it’s also worth noting that the last time Trump had a weekly Gallup approval this high was in late May, when he hit 41 percent. Among Republicans, Trump’s approval was 86 percent, with independents it was 36 percent (his best since July), and among Democrats it was 7 percent.
Pollsters, political analysts, and strategists watch Gallup’s numbers in the same way that a physician monitors patients’ vital signs. While there are a half dozen or so other fine nonpartisan national polls conducted periodically, the value of Gallup is the regularity of its weekly numbers (a change from its previous daily release of three-day rolling averages), and the constancy of its methodology.
My strong preference is for live telephone interviewing, but in the online realm, SurveyMonkey does a fine job as well, and its numbers also come out weekly, on Thursdays. Trump is improving in both the telephone and online interviewing averages, though he tends to run several points higher in the latter. For the last couple of months of 2017, the president’s SurveyMonkey approval ratings stayed in the 39-to-41-percent range, with disapprovals between 57 and 60 percent. For Feb. 1-7, Trump’s approval rating was 44 percent, same as the previous week, and his disapproval rating was 54 percent, down 4 points from the previous week.
Mark Blumenthal, the head of election polling for SurveyMonkey, wrote last week that Trump’s “improvement owes mostly to better numbers among Republicans. This week, 89 percent of Republicans and 88 percent of independents who lean Republican approve of Trump’s performance, representing 4 to 6 percentage point gains from the average in each category from August through November.” The new SurveyMonkey poll also showed that Trump’s strong approval clicked up to 26 percent, his highest since March.
Because Trump’s numbers tend to be so stable, whenever there is even a small amount of movement it gets noticed—perhaps too much. Indeed, some argue that there has been little if any movement at all. Emory University’s esteemed political scientist Alan Abramowitz pointed out to me an interesting blog post by a bright young University of Texas student, G. Elliott Morris, who argues that the change in the generic-ballot test is less significant when you look at surveys over a broader time frame. Morris makes a good case, and the fact is that many political analysts are looking at a lot of data beyond the generic-ballot test in public polls.
A warning for those cherry-picking polls to make the case either way: Too often, partisans will point to a poll from a few weeks or a month ago from one pollster, then compare it to a newer poll from a different pollster to make an argument. Remember that it’s always better to compare apples to apples, using the same pollster’s numbers for the before and after or the same organization’s poll averages (either RealClearPolitics.com or FiveThirtyEight.com). Also keep in mind that the two entities (RCP and FiveThirtyEight) that we rely on for averages have slightly different criteria for which polls are included, but both tend to reflect the most recent mix of polls. Given that some polls tend to run higher than others, during times when there aren’t many surveys being released a small number of them can affect the averages.
Once we have a fresh supply of ABC News/Washington Post, CBS News, CNN, Fox News, Pew Research, and NBC News/Wall Street Journal polls, we will be in a better position to determine just how meaningful this trend is, particularly with the election still nine months away.