It’s morning again in America. You grab your first cup of coffee, click to your favorite news site and are greeted by a new poll with a huge generic ballot spike in the congressional vote. What should your reaction be? Is it time to freak out, or calm down and assume the poll is an outlier?

The answer is neither. When a particular survey suddenly shows a significant shift in one direction or the other, political and media analysts and the public need to approach the data with caution. Before assuming there was a change in voter preference, we need to ask whether party identification in the survey also changed significantly, and if so, why. 

Did voters suddenly have a big change of heart, or did the survey’s methodology reflect a pollster’s wishful thinking or even partisan bias? Important questions to answer. One way to do that is to understand the party ID patterns of the last three decades, which give context and reference points to help evaluate the validity of any survey data.

New lows (and highs)

Here is a condensed review of party ID results based on national congressional exit poll data from 1984 forward, covering the last 17 national elections. From 1984 through 2000, the electorate’s party self-identification was remarkably consistent. Democrats ranged from 36 to 40 percent of the electorate, while Republicans made up 34 to 36 percent of voters.

In fact, in seven of the nine elections over this 16-year period, Republicans came in at a reliable 36 percent of the electorate. Independents were almost as consistent, with party ID between 25 percent and 27 percent of the electorate (excluding 1990).

The data also show that presidential election years were generally better for Democrats throughout the 1990s. They had clear advantages in 1992, 1996 and 2000, while the two parties were basically at parity in 1994 and 1998. 

The 2002 election became a watershed moment by disrupting what had been a fairly consistent pattern of party ID. That year saw Republicans finally break out, jumping to 40 percent, well above any previous results. At the same time, independents fell to their low over this period, while Democrats dipped from their highs to a middle range of 38 percent.

Republicans were unable to solidify their new majority, however, and the next three elections saw the GOP slip 7 points to a new low, while independents increased by 6 percent, as the margin between the two went from 40 percent-22 percent in 2002 to 33 percent-28 percent in 2008. 

Democrats were remarkably consistent during this same time, making up 38 percent of the electorate in the 2002, 2004 and 2006 elections. Then, in 2008, they returned to their previous high of 40 percent, last seen in 1996 when Bill Clinton won re-election and in 1986 when Democrats retook the Senate.

ICYMI: What You Need to Know About Voter Registration and Turnout This Midterm Season

But from 2008 forward, voters seem to have boarded a bit of a political roller coaster, disrupting the patterns of past elections. Republican makeup of the electorate has ranged from 33 to 36 percent over this time frame, while independents came in between 28 and 30 percent.

Democrats’ ranged from 35 to 40 percent, but they were plagued by significant swings between presidential and midterm elections, dropping 5 percent from 2008 to 2010 and 4 percent from 2012 to 2014. In 2016, the party ID margin favored Democrats by 3 percent, which was similar to the presidential turnouts in 1984 through 2000. 

In terms of party ID, however, 2016 was a mixed bag for both parties. Democrats’ 36 percent party ID was the lowest to date in a presidential election. The Republicans’ 33 percent tied an overall low, matching the previous two presidential turnouts. Independents, on the other hand, reached their previous 1990 high of 30 percent, but notably, it was the first time they hit this mark in a presidential year.

Crashing the party

So what can we take away from this data that might affect 2018? Since 2008, we have seen a return to some older patterns: Democrats doing better in presidential years, with Republicans improving in midterm years. Still, both parties’ IDs are decreasing as more voters now identify as independents — not a positive trend for either party. It means independents will tip the election in November again, as neither party is big enough to win outright.

The good news for Republicans is that they have held the House in five of the last six midterm elections, including one with a sitting Republican president, although that was in 2002. But the challenge is that, in all but one of those victories, they were even or held the party advantage, which they don’t now.

For the two elections where the party advantage was even, Republicans were able to regain the majority, once after 40 years in the minority (1994) and again in 2010 after pundits declared a 40-year Democratic majority was possible. Additionally, Republicans have won in general election environments where they trailed in party identification by 6 points (2012), 4 points (1996) and 3 points (2000, 2016).

For Democrats, the good news is that the party advantage historically increases slightly for the party out of power in midterm elections. When they won the House in 2006, the party ID margin favored them by 2. In 2016, they held a 3-point party advantage. So in theory, that advantage should be increasing.

The challenge for Democrats is that overall their party advantage has been declining. In presidential elections, their party lead has gone from a 7-point advantage in 2008 to 6 points in 2012 to only 3 points in 2016, which represents significant slippage.

In fact, in the 2016 election, Democrats made up only 36 percent of the electorate, their lowest point in a presidential year since 1984. In the last two midterm elections, Democrats made up 35 percent of the electorate — again, the lowest they have been since 1984. Finally, the Democrats have never had greater than a 3-point party ID advantage in a midterm election.

Party ID in a survey will significantly affect the outcome of the generic ballot, along with other questions, and it always has a bearing on the quality of a poll’s sample and thus the validity of the results. Historical data on party ID can help guide us through what is likely to be a mountain of data over the next few months.

For the moment, the fall election is wide open, and analysts need to look closely at a survey’s methodology before leaping to conclusions. 

What to do with that cup of coffee?  “Calm down and carry on.”