Donald Trump's recent failed attempt to surprise the political world with a sizable group endorsement by black ministers occasioned a very sharp observation from Joy Reid on The Last Word. After Jonathan Allen noted that Trump was desperately looking for "a racial or ethnic or any other type of minority that he can go to and not already have basically poisoned the well,” Reid helpfully clarified the why of it all: “Republican primary, that's not about black and Latin voters, because there really aren't any in the Republican primary,” Reid said. “That's about white suburban voters who want permission to go with Donald Trump.”
Trump's situation is anything but unique—it's just a bit more raw than it is with other Republicans. Ever since the 1960s, as Richard Nixon's Southern Strategy was being born, there's been a ongoing dilemma (if not huge contradiction) for the erstwhile “Party of Lincoln” to manage: how to pander just enough to get the racist votes they need, without making it too difficult to deny that's precisely what they're doing.
There are a multitude of cover stories involved in facilitating this two-faced strategy, but one of the big-picture ways it gets covered is with a blanket denial: It wasn't Nixon's race-based Southern Strategy that got the GOP its current hammerlock on the South, it was something else entirely. Say, the South's growing affluence, perhaps, or its “principled small-government conservatism,” or the increased “leftism” of the Democratic Party on “social issues”—anything, really, except racial animus. Anything but that. (It's akin to the widespread beliefs that the Civil War wasn't fought over slavery, or that the Confederate flag is just a symbol of “Southern pride.")
Most who make such arguments are simply mired in denial, or worse, but there are several lines of argument seemingly based on objective data in the academic literature. But a new National Bureau of Economic Research working paper that Sean McElwee recently referred to should put an end to all that.
“Why did the Democrats Lose the South? Bringing New Data to an Old Debate,” by Ilyana Kuziemko and Ebonya Washington, does three key things: First, it uses previously overlooked data—matching presidential approval against media coverage linking President Kennedy to civil rights—to shed light on a key transition period—broadly, from 1961-1963, narrowly, the spring of 1963—when the Democratic Party clearly emerged as the party of civil rights. Second, it uses another new source of data—responses to the “black president question” (first asked by Gallup in 1958), whether someone would support a black (originally “negro”) candidate for president, if nominated by their party—as a measure of “racial conservatism” to analyze the contrast between the pre- and post-transition periods.
As McElwee reported, the paper “find[s] that racism can explain almost all of the decline of Southern white support for Democrats between 1958 and 2000.” Indeed, it explains all of the decline from 1958 to 1980, and 77% of the decline through 2000. (The authors prefer the 1958-1980 time-frame, since Jesse Jackson's candidacy in 1984 and 1988 “may have transformed the black president item from a hypothetical question to a referendum on a particular individual.”) Third, the paper looks at the other explanations—the cover stories—and finds they have only a marginal impact, at best. (Although its focus is Southern realignment away from the Democratic Party, the GOP has obviously been gaining strength at the same time as a direct result.) It also sheds light on an early phase of dealignment, starting when Truman first came out for civil rights in 1948, leading to the Dixiecrat revolt.
Before turning to the paper itself, I want to recall a point I made last year: so-called “principled conservatism” is itself heavily determined by anti-black attitudes. Southern racial conservatives had been closely tied to the Democratic Party for generations before Truman came out for civil rights in 1948, but the 1960s stand out as a decisive turning point. Among other things, I pointed out (a) that George Wallace himself had disavowed explicit racism by the end of 1963, turning to a classic articulation of anti-government/anti-"elite" conservative themes, (b) that there are both international and U.S. data showing that welfare state support declines as minority populations increase, and (c) that even attitudes related to spending to fight global warming are strongly influenced by anti-black stereotypes.
With all that in mind, there's no reason at all to assume that any form of conservatism in America can be separated from white supremacism. We can pretend otherwise for the sake of running thought experiments, data-analysis, etc. and there can be some value is doing this—or I wouldn't find this paper so important. But we should never forget the larger reality: we are not operating in blank-slate situation, where all hypothesis may be considered equally, in abstract purity. White supremacy is the default condition for everything in America, only the strength and salience of its impact varies from situation to situation.
Keeping all that in mind, let's now turn to the important lessons this new paper has to tell us. As I said, it does three key things—sheds light on the 1961-1963 transition period, contrasts the pre- and post-transition periods to show the overwhelming impact of race, and examines other explanations, finding their impacts to be marginal, at best. The second of these is key, but is only possible as a result of identifying the transition point, which is crucial to making sense of everything else—both the central role of race, as well as the relative insignificance of other factors.
As the authors note, there are plausible reasons to consider alternative explanations—dealignment took a long time, but civil rights only briefly registered as the top issue, in contrast, the South's economic gains were more gradual, better matching the gradual shift away from the Democratic Party. But what's missing from those arguments is a full range of data on racial attitudes, particularly straddling the transition period when the Democrats emerged as the party of civil rights. The authors note that “those authors using the cumulative ANES [American National Election Survey] to address the role of racial views on party alignment typically begin their analysis in the 1970s, well after the Civil Rights and Voting Rights Acts.” Thus, they're entering the picture very late in the game.
Building a strong case for factors other than race based on data that excludes the 1960s is like the old joke about the drunk looking for his car keys under a streetlamp, where “the light is better” rather than in the darkness up the street, where he dropped them. What sets this study apart is the uncovering of a new light source to shine into that darkness, and the excellent use that is made of that new light.
Informally, qualitatively, historians and others clearly understand when the Democrats emerged as the party of civil rights—it happened in the early 1960s, when first Kennedy, then Johnson, allied with the Civil Rights movement, introducing and passing both the Civil Rights Act of 1964 and the Voting Rights Act of 1965. The significance of this alliance was dramatically visible in the presidential election map of 1964, when Barry Goldwater—who voted against the Civil Rights Act—lost every state in the nation outside of his native Arizona, except for five Deep South states, all but one of which had gone Democratic just eight years earlier, when Dwight Eisenhower had swept almost all the rest of the country. The electoral maps of 1956 and 1964 are almost mirror images of one another, a dramatic reversal unique in American history, and a stark indication that the transition the authors are looking for took place before the 1964 election.
The authors begin their effort to nail down this transition time by turning to questions a about support for school integration in the ANES. Although different wording is used in 1960 compared to 1964 and 1968, the sense remains constant. They find that “in 1960, only 13% of Southern whites see the Democrats as the party pushing for school integration, 22% say Republicans, and the rest see no difference. Non-Southern whites see essentially no difference between the parties.” But four year later, a dramatic shift has taken place. “By 1964, 45% of Southern whites now see the Democrats as more aggressive on promoting school integration, whereas the share seeing Republicans as more aggressive has fallen to 16%. Non-Southerners' assessment shifts similarly. The large gap in voters' perception of the parties on school integration that emerges in 1964 holds steady in 1968.”
With the 1960/1964 window to start with, the authors then look for when that change might have occurred, using two data sources—Gallup poll measures of presidential approval, and media mentions associating the president with civil rights, both based on the fact that presidents do so much to influence the perception of their parties. They focus primarily on the New York Times, supplemented by looking at two Southern papers as well—the Dallas Morning News and the New Orleans Times-Picayune. They note “two short-lived spikes” of media mentions when Kennedy's administration intervened in support of activists in 1961 and 1962, but a steep increase begins in May 1963, as Birmingham becomes the focus of national attention, eventually leading to mass arrests—including schoolchildren as young as grade school, along with mass beatings and the use of firehoses and dogs against protesters...all broadcast live on national TV—Martin Luther King's “Letter from Birmingham Jail,” the eventual involvement of Attorney General Robert Kennedy, and finally President Kennedy's televised address proposing a legislative end segregation, what would eventually become the Civil Rights Act of 1964. The NYT article count peaks in mid-June, but remains elevated above its pre-April levels through the rest of Kennedy's presidency. Similar patterns are seen in the other papers as well.
The authors then compare this media coverage with the presidential approval levels, and find that Kennedy's approval plummeted dramatically as he took up civil rights, but only in the white South:
The most striking result is the 35 percentage point drop in his support among whites in the South (compared to no change among other whites and a rise among all blacks) between the April 6th and June 23rd 1963 Gallup polls (which correspond to a surge of articles covering Kennedy's support of protesters during Martin Luther King's Birmingham campaign in May and the president's televised proposal of the Civil Rights Bill on June 11th). Smaller Civil Rights moments (e.g., the integration of Ole Miss in September 1962) also match up to significant dips in Kennedy's relative approval among Southern whites.
Testing the significance of other events and issues “news regarding Cuba, the Soviet Union, Social Security, etc.” does not diminish the “overwhelming explanatory power” that civil rights has in predicting his white Southern drop in popularity.
Later in the paper, the authors also look at polling match-ups of Kennedy vs. Goldwater, which began in February 1963. They note strikingly similar results:
During our key period of the spring of 1963, Kennedy goes from having a healthy, thirty percentage point lead over Goldwater to being thirty points behind him. White non-Southerners remain rather aloof toward Goldwater....
The result from the presidential match-ups suggests that Kennedy's decline in approval documented in the previous subsection did not reflect mere short term annoyance. Within months of Kennedy's association with civil rights, half of his Southern white supporters shifted their backing to a candidate who was from a party they had shunned for a century but who was not believed to support civil rights.
Thus, they not only identify a very narrow window in time when the Southern perception of the Democratic Party changed dramatically, they also show the specific actions involved. There is no mystery about what precipitated Southern antipathy. No guesswork required. It's noteworthy that this shift came before the Civil Rights Act was actually passed. With these dates firmly established via the Gallup presidential approval poll, the dates of the pre-transition and post-transition period are set, and the rest of the analysis can proceed.
The heart of the paper is a “triple-difference analysis,” a regression model analysis designed to show “how much of the  pre- versus post-period decrease in Democratic party identification  among Southern versus other whites is explained by  the differential decline among those Southerners with conservative racial attitudes?” The design of this analysis—including the identification of a clear-cut transition point—brings the question of race's role sharply into focus, in a way that other work, using only more recent data, simply cannot hope to match.
While the transition in the perception of the national Democratic Party took place over just a few months, there was a much longer gap between polls asking the black president question, but it's crystal clear how significant this transition period was:
In the South, conservative racial views strongly predict Democratic identification in the pre-period, but this correlation is wiped out between August 1961 and August 1963 (the last poll of the pre- and the first poll of the post-period, respectively).
As for whites outside the South:
We find that racial attitudes have little if any explanatory power for non-Southern whites' party identification in either period.
This is not to say that whites outside the South are free of racism, it's just that racism didn't play a significant role in party identification, either before or after Kennedy's move to embrace civil rights.
Focusing in on the key question the paper seeks to resolve, the authors write:
Most important to the question at hand, the entire 17 percentage-point decline in Democratic party identification between 1958 and 1980 is explained by the 19 percentage point decline among Southern whites with conservative racial views. Extending the post-period through 2000, 77% of the 20 percentage-point drop is explained by the differential drop among Southern whites with conservative racial views.
As for the third main point I described above—the insignificance of other factors—they add:
This pattern of results is robust to controlling flexibly for socioeconomic status measures included in the Gallup data and is highly evident in event-time graphical analysis as well.
There are other possible factors of course, other issues to consider. Indeed, considering the sharply decisive nature of the key results just described, these other factors' role as potential cover stories is only intensified, which is why it's important that the authors devote significant attention to testing for the possible influence, primarily within the same sort of analytic framework. These other explanations also turn out to have marginal impacts, at most.
Because excuses and cover stories for racism are so popular, it's worth taking a hard look at the alternatives which are presented and refuted in section six of the paper, which considers three main forms of alternative arguments: first, that the decline of the Democratic South was due to rising partisan polarization based on issues others than race; second, that it can be explained in terms of economic development or changing demographics; and third, that the large-scale timing of the decline is wrong for a racial explanation.
With regard to the polarization argument, the authors note, “If Southern whites have always been more conservative—especially economically—than other whites, then rising polarization could lead to differential exodus of Southern whites from the increasingly more liberal party.” In addition, the “black president” question could simply be a proxy conservatism in general, so that its significance is misconstrued by the analysis.
To counter the first argument, they turn the the 1956 ANES, and find very few regional differences in white attitudes between the South and other regions, except for race:
We find no significant differences by region on job guarantees, tax cuts, the appropriate in legislators' support for the 1947 Taft-Hartley Act) labor unions, and the regulation of housing and utilities. Southern whites actually supported government provision of affordable medical care at significantly higher rates, but were significantly less likely to support federal financing of local school construction.
Although the authors don't mention it, local school construction was obviously a subject closely associated with race just two years after Brown v. Board of Education. They go on to say:
In comparison, there are similarly few—in fact only one—significant regional difference on foreign policy preferences during this era, but large and significant differences, as expected, on Civil Rights.
In short, they summarize, “This analysis paints a picture of broad pre-period regional consensus (among whites) on policy issues outside of civil rights.”
Next is the question of whether the “black president” is merely a proxy for conservatism generally, or for specifically Southern values and concerns (a black candidate assumed to come from/represent the urban North). This hardly seems possible given the data just cited above, but there's a way to directly test the proposition, anyway—a similar set of questions about “whether respondents would refuse to vote for a female, Catholic or Jewish nominee from their party.” These all would also have violated Southern social conservative norms, and it's fairly straightforward to test them using a variant of the regression model used to establish the main result. Although there's a clear correlation between rejecting all these possible candidates, precisely because broad-based social conservatism is real, the authors report, “Southern dealignment during the post-period is driven by those with conservative views on racial equality, even after we control for (highly correlated) views toward women and religious minorities.”
Next up is the question of whether economic development or changing demographics can explain the dealignment. It's superficially plausible, as the South grew significantly less poor during this time period. Because only six Gallup surveys from 1958 through 1980 included an income control, the authors turn to the same ANES dataset that's been used by others in support of the income argument. They note that “most of the work that finds evidence of income as a driver actually uses cross tabulations and does not, in a regression sense, partial out what share of the total dealignment is explained by differential income growth in the South,” which is precisely the purpose of regression model.
So, using the same transition period division as before, they test for impacts of income, as well as urban/suburban/rural status. Once again the results are consistent, “We find no role for economic development—even broadly and flexibly defined—in explaining the differential decline in Democratic allegiance among Southern whites after 1963.”
Next, they test for two other possible influences—that of population differences due to in-migration or the coming of age of younger voters. This is easily done by restricting their sample to Southerners born before 1941, which "shows a post-period drop in Southern Democratic attachment that is 92% of the size of the drop in the full sample." Thus the impact of Northern in-migration and different attitudes in succeeding generations is relatively minor. Once again, non-racist explanations fail.
Finally, they seek to clear up questions about the timing of Southern dealignment with Democratic Party. While it's post-1960 timing seems to make sense in terms of the claims advanced, there's an earlier history of dealignment to account for, right up to the 1960 election itself. For the post-1960 era, they report:
Between 1960 and 1970, Democrats lost on average over two percentage points per year among white Southerners relative to other whites, whereas there was no additional loss between 1970 and 1980 and the aggregate 1970-2004 rate was below 0.4 percentage points per year.
But what's “less consistent” is the fact that there was earlier, albeit slower process of dealignment among white Southerners, and Kennedy in particular was already weaker than usual for a Democrat in the South in the 1960 election. Kennedy's weakness, however, is not that hard to explain—anti-Catholic bias is the culprit. As the authors note, the only previous Catholic candidate, Al Smith in 1928, was even more notably weak: “Democrats lost six Southern states that election, five of which had not voted Republican since Reconstruction.” [Compare map from 1928 to 1924.] There was no survey data at that time, but there was plenty of qualitative evidence. After all, the KKK of that era was primarily anti-Catholic, as opposed to its earlier anti-black incarnation.
For Kennedy's election, however, there was supporting survey data:
In a 1958 Gallup poll, 48 percent of Southern whites state unwillingness to vote for a Catholic president, compared to only 22 percent of whites elsewhere. In the 1960 post-election portion of the ANES, 29 percent of whites in the South said the most important reason they did not vote for Kennedy was his Catholicism, compared to 15 percent elsewhere.
Not only did Kennedy's Catholicism depress his vote in the South, it boosted it elsewhere:
Catholic voters (94% of whom lived outside the South) mobilized in support of Kennedy, further shrinking the South-versus-non-South advantage Kennedy received.
Hence, there's not really anything there in the 1960 election result that needs explaining in terms of dealignment. There is some dealignment after the election, in 1961, however, before the period highlighted in the main analysis, and this turns out to be explained by the “no Catholic president” response, which unlike the black president response is “a one-off effect,” the authors note, “1961 (the first poll following Kennedy's election) is a huge outlier: those with anti-Catholic views are roughly 27 percentage points less likely to identify as Democrats,” a never-repeated level of impact.
With that set aside, what needs to be explained is the longer trend, beginning in the late 1940s. Here, there are three significant findings, two dealing with Truman and the Democrats, one with Eisenhower. First, the authors use a combination of polling and civil rights media mentions similar to that used with Kennedy to show how Truman's increased involvement with civil rights contributed to Southern dealignment. The authors note:
Truman has little connection to Civil Rights until early in 1948, when we see the number of articles linking him to the issue rise and remain high throughout the year (a year which saw him introduce Civil Rights legislation to Congress in February and, via executive order, desegregate the military and the federal workforce in July).
Presidential approval polls were not yet commonplace, so intention to support Truman's re-election was used instead. They found that “essentially all of the decline in Truman's support in the South occurs between the November 1947 and the March 1948 surveys, the last survey before and the first survey after his February 1948 introduction of Civil Rights legislation, respectively.” The results are regionally concentrated, as expected: “while Civil Rights activity costs him (hypothetical) votes outside the South, the effect is two to three times as large in the South.”
This is hardly news in one sense—the Dixiecrat revolt made it obvious that Truman and the Democrats paid a heavy price in the South for their civil rights activism in 1948 [election map]. But because the findings align with what's already known, they validate the soundness of the approach used in the case of Kennedy.
The second finding, however, breaks new ground. Using the 1952 ANES, they found the following signs of race-based party defection:
Being against ensuring fair employment opportunities for Negroes predicts both intragenerational defection (i.e., having once identied as a Democrat but now identifying as a Republican or Independent) and intergenerational defection (having grown up with parents who were both Democrats but identifying as a Republican or Independent), though has little predictive power for defection in terms of current Democrats voting for Eisenhower.
Thus, the ongoing impact of Truman's civil rights activism in 1948 did play a role in the ongoing trend in the 1950s. It was not a one-time impact, like Kennedy's Catholicism.
The third finding was Eisenhower's civil rights involvements cost him as well—contradicting the claim that increased Southern support for him and the GOP in the 1950s indicated that race was becoming irrelevant. The same approach was used with Eisenhower as with Kennedy and Truman—matching media mentions of him and civil rights with changes in polling—approval polls this time. The results are much the same as for Kennedy and Truman, though for Eisenhower, they came after his last election campaign was safely behind him:
There is a clear increase in Eisenhower's connection to Civil Rights in the fall of 1957; he sent federal troops to enforce the court-ordered desegregation of Little Rock Central High School on September 24th of that year. In fact, his relative approval in the South declines by 25 percentage points between the polls of September 21 and October 12.... And just like for Kennedy, we see that Eisenhower paid an approval penalty in the South when the news made mention of him alongside Civil Rights (regardless of the search terms we use to identify articles), contradicting the claim that Southerners were not upset by his Civil Rights gestures.
Summing up, this paper clearly demonstrates that racial hostility was the dominant factor in the decline of Democratic power in the white South, while showing that a range of alternative explanations have only a marginal impact at best. It not only shows this for the 1958-1980 and 1958-2000 periods, which are its primary focus, but for the period starting with Truman in early 1948 as well, although not with a unified quantitative model across the entire period.
This should effectively put an end to the contrary arguments, but given how deeply white supremacy permeates our culture, I wouldn't hold my breath. After all, clowns like Trump will always be highly motivated to deceive, there's no shortage of accomplices to help them (like Harlem-based, Trump-endorsing pastor James Manning, who “once likened Obama, America's first black president, to Adolf Hitler and has frequently said the president is secretly gay”), and millions of respectable white voters will be all too eager to be deceived.
So don't expect the nonsense to magically stop, just because it's been exposed. But it is a tool you can use to shine some light, when you have an opening. I would keep this paper's findings close at hand, for whenever someone attempts to pretend that racism has nothing to do with the GOP's stranglehold on Southern politics—or America's.