Biden’s job approval has entered dangerous territory
In 1880, the Americans did something momentous: They all elected their delegations to Congress the same year. Before that, the elections had been a hodgepodge. For example, the first elections of 28th Congress, which apparently met between 1843 and 1845, was held in Missouri on August 1, 1842. But only five other states held elections that year; almost every other state held them during the odd year 1843. Maryland was finally able to hold its elections on February 14, 1844, half a year before the presidential election and less than a year before the presidential election. Next Congress called.
Even after 1880, a truly uniform election day did not come until 1960, when Maine abandoned its stubborn insistence on holding Congressional elections in September. But for all intents and purposes, 1882 marks the birth of the phenomenon we know today as the year of the midterm elections.
It also gave rise to one of the rare political observations on which political scientists and “lay” election observers almost unanimously agree: midterm elections tend to go badly for the party occupying the White House. James Garfield barely won the presidency in the 1880 elections: he won the Electoral College by 60 votes and the popular vote by less than 2,000 votes. Two years later, Republicans lost 34 House seats (there were only 325 representatives at the time, so this equates to more than 40 seats in the House today).
The president’s party has lost seats in almost every subsequent presidential election. There have been four exceptions: 1902, 1934, 1998, and 2002. However, the 1902 midterm probably does not deserve to be counted in this category, as the size of the House had expanded after the decennial redistricting. On average, the ruling party has lost about 30 seats in the midterm elections, but this statistic includes years like 1894, when the Democrats lost 125 seats, almost half of their caucus. Postwar elections have been quieter, with the president’s party losing around 25 seats.
In the Senate, it’s a somewhat different story. Because chamber seats are staggered, with only a third of them in any given year, a party can enter a predictably difficult election with very few vulnerable seats for elections. On average, the president’s party still loses seats, about four, but since the direct election of senators began at the national level in 1914, the president’s party has won seats or even broken in eight elections, about one of every three.
The most important predictor of a party’s midterm performance is the approval rating for the president’s office. So, for example, the years in which the president’s party has gained ground in the House during a midterm election have been years in which the president has been exceptionally popular. The 1998 midterm elections occurred in the context of a massive economic boom and an unpopular attempt at impeachment by Republicans, while the 2002 midterm elections occurred while George W. Bush was still enjoying his post-9/11 recovery. . Both presidents enjoyed approval ratings in the mid-1960s on Election Day.
While we do not have poll data for FDR in 1934, one can imagine that approval of his work was probably well into the 1970s. On the other hand, abysmal elections have occurred for parties in power when the president was in power. Unpopular General: Witness 1994, 2010, and 2018. Therefore, the trend in approval of President Biden’s job should be concerning to Democrats:
Let’s look at the implications of this, first through the lens of the Senate elections, then through the lens of the House elections. I have explored the Senate elections in this article. As the graph near the bottom indicates, when Biden hovered around 52% approval for the job, Democrats were largely favored to maintain control of the Senate. The most likely outcome for the party was that it would actually win a seat.
But Democrats’ chances of occupying the Senate were set to decline rapidly if approval of the president’s job fell further. At 50%, they would be expected to break even, but would lose control about one in four times. With 49% approval of the job, the Senate is considered 50-50.
46%? Democrats would lose between four and zero seats 95% of the time, with an expected result of two seats. They only retain control about 4% of the time. Obviously, this is an outcome Democrats would like to avoid.
And the House? Modeling aggregate House elections is a treacherous task, as there is too much foam at the individual House level. Nonetheless, we can extract some broad generalizations from past choices, using factors that I have referenced earlier.
The first relevant piece of information for the data is the degree of exposure of the president’s party. In other words, midterm elections tend to mean regressions. When a president’s party performs well above the average of previous elections in the presidential election, the tendency is for its midterm losses to be more severe. This is because large House gains often require overextension into opposition party areas.
When Barack Obama won the presidency in 2008, more than 40 Democratic House candidates prevailed in the districts John McCain won, even as McCain was losing the national popular vote by seven percentage points. Not surprisingly, then, the Democrats took a beating two years later. Along the same lines, in 1998 Republicans were still defending many fringe districts from their 1994 wave. Democrats entered that election with minimal exposure, which helped secure their victories.
Democrats begin this cycle with a relatively low degree of exposure, consistent with a loss of about 18 seats. But if anything, this exaggerates your problem. Very few Democrats represent Trump districts, and only 20 represent districts that were more Republican than the country as a whole. At the same time, we don’t know what the effect of redistricting will be. (Click on the graphic to enlarge the image).
A Republican Party as overwhelmed as the Democrats today suffered massive losses in 1974 and made modest gains in 2002. Another factor we consider is the growth of the economy. Traditionally I have used real disposable income as a proxy. As we can see, personal income growth correlates with better party performance in midterm elections. Years with soft or negative income growth, such as 1974, 1982, 1994, and 2010, tend to go badly for the party in power. Years with strong growth, such as 1998 and 1978, tend not to be events, or even net positives for the president’s party.
We don’t know what actual disposable income will look like in 2022, especially given the various coronavirus relief laws from 2020 and earlier this year. But for now, let’s assume growth is strong. This would also suggest modest Democratic losses, perhaps on the order of 6% of the caucus (or roughly 13 seats).
However, another factor is very important: the approval of the president’s job. Even in years of strong economic growth, like 1966, if the public is unhappy with handling a non-economic issue, your party will suffer. This is probably the strongest predictor we’ve seen:
Cases that stray out of line can be largely explained in terms of other issues we’ve witnessed: Democrats outperformed in 1978 because of the economy and in 2014 because they had so little exposure. In 2010 and 1974 they underperformed due to the economy and, in the case of 2010, overexposure.
What we see from this is that the decline in approval of President Bidens’ job from about 52% to 46% is consistent with a loss of an additional 5% from the Democratic caucus, or an additional 11 seats. Putting them together in a single regression analysis, we account for about two-thirds of the overall variation. This is not so bad, given the amount of randomness we have to explain here. The model generally suggests that declining approval of Biden’s job will likely cost Democrats about eight seats.
Part of this is likely to be a throwback to Afghanistan’s failed pullout, which may roll back in the coming months. It may also be the case that there is some reluctance on the part of Democrats to respond to polls at this time, artificially deflating approval of their work.
Regardless, the decline in approval of Biden’s work is consequential. Right now, he’s probably sitting right under the “Mendoza lineThat would represent a “normal” midterm, perhaps with no losses in the Senate and low double-digit losses in the House. If the president recovers, Democrats will have a successful 2022, even if they narrowly lose the House. However, if it goes much lower, it could turn into a nasty defeat.