Alan Brinkley’s piece in Saturday’s Wall Street Journal is subtitled: “Millions of voters have moved out of the political party system. The decline of loyalty has made politics less stable and predictable—and has resulted in close elections.”
This statement, and Brinkley’s thesis, is dubious. Elections are not getting closer, less predictable, or more volatile. And voters’ loyalty to political parties is quite strong and has been getting stronger. I elaborate below. This is a lengthy post, but I think it is important to dispel Brinkley’s misconceptions, which seem to me typical of much conventional wisdom about elections.
But for the past 40 years, close and unpredictable elections have increasingly become the norm.
Have presidential elections become closer in the past 40 years? Below is the absolute value of the difference between the Republican and Democratic candidates’ vote share, from 1964-2004, with a linear trendline to summarize any general trend.
From this it appears that elections are closer, mainly because of 2000 and 2004. But what if we go back to 1960 instead of using Brinkely’s arbitrary 40-year period?
Now the apparent trend toward closer elections is more muted.
What if we exclude the election of 1964?
Now there is no substantively important trend.
In short, Brinkley’s characterization of increasingly “close” elections is contingent on starting in 1964 and depends heavily on the 1964 result.
What about his claim that elections are more “unpredictable”? This is theoretically distinct from “closeness.” A close election may be entirely predictable, after all. One way to do this is to examine the performance of an election forecasting model to see whether its error—the gap between what it predicted and what actually happened—has increased over time. If so, then elections are increasingly unpredictable.
Here is a graph with the absolute value of the forecasting error in Douglas Hibbs’s bread-and-peace model, 1952-2004. (See Table A-1 of this pdf for the data.)
Again, the data are equivocal. Yes, the model didn’t predict 1996 and 2000 particularly well, but it was extremely close in 2004. There is no consistent increase in “unpredictability.”
Larry Bartels has conducted a more rigorous analysis of electoral predictability, extending across 1868-1996 and using different kinds of data than I did above. The paper is here (gated). The upshot of his analysis is “Whereas the presidential elections of the 1960s and ’70s were highly volatile by historical standards, the five most recent presidential elections have evidenced levels of volatility below the long-run historical average.”
In short, elections have not become more unpredictable or volatile.
Brinkley goes on to argue that the apparent unpredictability has arisen because parties no longer provide a stabilizing presence:
It  was, rather, the beginning of a massive de-alignment—the movement of millions of disillusioned voters out of the party system altogether. By the early 1970s, nearly a third of the public identified themselves as “independents,” affiliated with no party. Many of them ceased voting entirely. Others began to pick and choose candidates on whatever criteria mattered to them at a given moment. Serious third party or independent candidates, a rarity through most of the 20th century, have siphoned off significant numbers of voters from the major parties in five of the last 10 elections. Turnout in presidential elections, rarely below 60% in the first seven decades of the 20th century, dropped dramatically after 1968, and dipped below 50% in 1996. Party loyalty, in short, is no longer a strong factor in the decisions of voters, and with its decline has come a less stable and less predictable political landscape. By the late 1970s, political scientists and historians were no longer paying much attention to parties and were focusing on social movements instead.
This is mostly wrong, or at least misleading. Brinkley conveniently stops his turnout story in 1996, ignoring the increases in 2000 and 2004. As a function of the voting-eligible population, turnout in 2004 was 61% (see here), which is only a couple percentage points lower than in the 1960s.
Moreover, Brinkley fails to note that most independents have a partisan leaning; they are “independent leaners.” The number of true independents is small and has been shrinking in recent years. See here.
Moreover, most of these independent leaners act just like loyal partisans. Here are graphs of the percentage of Democrats and Republicans voting for their party’s presidential candidate. Each partisan group is broken down into self-identified strong partisans, weak partisans, and independent leaners.
On average, independent leaners are no different than weak partisans.
These graphs also show another telling feature of American politics: partisan loyalties have grown stronger over time. Democrats and Republicans have become more likely to vote for their party’s nominee for president. Brinkley is pretty much wrong when he writes:
But in presidential elections…parties have increasingly little meaning to voters who, whatever their formal affiliations, cross party lines often and without hesitation.
He is also wrong when he writes:
One result has been the end of the once-strong connection between the election of presidents and the election of members of Congress.
He notes the presence of divided government, and conjures up a weird statistic—whether states simultaneously elected a president candidate and a majority of their congressional delegation of the same party. Here’s a more relevant statistic: what percent of voters vote for one party’s candidate for President but another for Senate or House. In other words, what percent of voters split their tickets?
The percent of voters who split their tickets has declined sharply in recent elections—back to 1960s levels in fact.
In sum, Brinkley’s obituary for political parties is unnecessary, and his account of electoral behavior basically ignores every recent trend and a lot of political science scholarship published since, well, the 1970s and early 1980s. There are actually fewer true independents now than in earlier years. Most people who claim to be independent are in fact closet partisans. And the 90% of the American public that is partisan is actually more loyal now than before. Thus, it is little surprise that elections do not appear systematically less predictable or volatile.