The Sweep: The Folly of Issue Polling

Afghanistan edition.

We have tried to explain the problems with “issue polling” in this newsletter and on our podcasts before. But I’ll summarize again here. Asking people to tell you who they will vote for—and whether they will vote—is hard, but it is a concrete question about a future event that the person has the information to answer. Maybe because we can compare the responses to the actual results in a matter of weeks or months, we tend to take them with a grain of salt. And we expect pollsters to refine their methodology to improve their accuracy every cycle. 

In comparison, asking people about their feelings on an issue of public policy is generally silly. First, very few issues are actually simple enough for a statistically significant number of people to know much about. Second, it’s hard—in some cases, arguably impossible—to phrase questions about policy in a non-leading fashion. Third, and most important to my mind, the sentiments supposedly being measured can resist easy categorization. In election polling, the respondent has a finite number of options, and the question is about a single concrete action: For which of these candidates, if any, will you pull the lever? But issue polling is about sentiment—and people can have contradictory and complicated feelings about an issue that are hard to capture in “agree” or “strongly agree” responses. 

And yet, politicians and government leaders love issue polls. So do corporate boards, for that matter. They provide data—even if the data is crap—to back up someone's argument, and they give a sense of direction to someone whose chief goal is to win reelection. As the French revolutionary Alexandre Auguste Ledru-Rollin may or may not have said, “There go the people. I must follow them, for I am their leader.”

I’m sure you can guess where I’m going with all of this. For years, news stories on Afghanistan have provided data to back up the idea that Americans don’t want to be there. Here are some of the questions and results from polls In April and May of this year:

The Hill asked: “Do you approve or disapprove of Biden’s plan to remove the troops from Afghanistan?” 73 percent of respondents approved; 27 percent disapproved.

Fox News asked: “Do you think the U.S. should remove all military troops from Afghanistan, or should some U.S. troops remain for counterterrorism operations?” 50 percent answered that some troops should remain; 37 percent responded that we should remove all troops.

The Economist and YouGov asked: “Do you approve or disapprove of the plan to withdraw all U.S. forces from Afghanistan by September 11, 2021?” 58 percent approved; 25 percent disapproved.

Quinnipiac asked: “As you may know, President Biden has decided to withdraw all U.S. troops from Afghanistan by September 11, 2021. Do you approve or disapprove of President Biden's decision?” 62 percent approved; 29 percent disapproved. 

The predictable headlines blared things like “Overwhelming majority backs U.S. withdrawal from Afghanistan.” But there’s something odd just in the data above. The response to the second question is totally at odds with the third and fourth. Six in 10 want all troops gone in the last two, but only four in 10—the inverse—want all troops gone in the second one. What’s going on? 

For starters, a lot of people just don’t know or care. Back in March, an interesting nugget from Brookings noted that “in a recent poll conducted in the fall of 2020 by the National Opinion Research Center (NORC) for researchers Peter Feaver and Jim Golby, only 59 percent of survey respondents answered the question about withdrawing troops from Afghanistan.” Relatedly, the New York Times noted that “just 12 percent of Americans said they were closely following news related to the U.S. presence in Afghanistan.” So a lot of people didn’t have an opinion, and those that did were more than willing to tell you that their opinion wasn’t a carefully considered one after a dark night of the soul. 

Another thing you may have noticed is that three of the four versions of the Afghanistan question reference President Biden. (The third question doesn’t use his name, but in the full script, the previous three questions each referenced Biden’s “clear plan for Afghanistan,” “Biden’s handling of Afghanistan,” or some such.) So even if you don’t have much of an opinion about what we should do in Afghanistan, the question is also a referendum on a president that the respondent supports or opposes. 

And then there’s the big problem: you’re asking people to give a scripted answer to a simple question about an issue that is deeply complex. 

Let’s walk through some possible answers with a hypothetical American. Do you want every American serving in Afghanistan to come home? Yes. What if it means the Taliban takes back control of the country? Still okay. What if the withdrawal embarrasses the United States globally and it empowers Russia and China? Well, still good. What if it empowers al-Qaeda to launch another 9/11-style terrorist strike against the US? Oh, in that case, no! 

Would you say this person supported Biden’s plan to withdraw troops or not?

If only 12 percent of Americans were following the news in Afghanistan closely until last week, that number is significantly higher now. And as they tune in, the numbers are shifting rapidly just as we would expect. According to Politico, “49 percent of voters continued to support the withdrawal, down from 69 percent in April.” Mention al-Qaeda and the number drops even further. 

This is why you never make decisions based on issue polls. People are smart and their opinions are nuanced. So are questions of public policy. For all the reasons that the Afghanistan polling highlighted above was stupid, so is the vast majority of polling on gun control, abortion, climate change, affirmative action, etc. Trends are meaningful when the same polling firm asks the same question over time—though even then it’s easy to mistake causation for correlation—but issue polling as a general matter should be ignored. The collapse of Afghanistan shows us exactly why.


Let’s turn to Chris, who has some thoughts on that sweet, sweet census data.

Whitewashing the Census Numbers 

The coverage of the shrinking share of the American population that calls itself white revealed in the 2020 Census was sometimes gleeful in its schadenfreude. Some of it was intended to provoke alarm among the still-majority white population. But almost all of it was rooted in the idea that more diversity and less whiteness was significant as a political proposition—that these demographic changes will drive a political shift to the left and away from traditional values.

Not only is that political forecast dubious on its own, even the data on which it rests are more than a little fishy. 

Here’s the gist of what the census numbers say: The white share of the population fell from 64 percent in 2010 to 58 percent in 2020. The raw number of whites shrank from 196 million to 191 million. Unless you count whites who also consider themselves Hispanic. Or is it Hispanics who also consider themselves white? There were 12.6 million of those folks in the Census. Should we consider them part of the nationalist right’s feared “great replacement” of good old white American culture? Or are they tainted by their white identity, and do they thus threaten to slow the end of American white supremacy? 

Pretty confusing, right? You ain’t seen nothing. In the box one could check to declare his or her whiteness, respondents were asked to say what kind of white they were. The instructions were to check the box and then write in their specific heritage: “Print, for example, German, Irish, English, Italian, Lebanese, Egyptian, etc.” Arabs are whites? Or only some Arabs? Does it depend on pigmentation? What if you are a very dark-complected Syrian-American but really love NPR and purebred dog shows? The Census doesn’t say. It was up to each respondent to declare themselves white, black, American Indian, or one of 11 different categories of Asian or Pacific Islander. 

That preposterous dogpile of pseudo-scientific racial and ethnic labels may sound laughable to the ears of Americans who have grown quite comfortable with racial mixing, but it’s nothing new.

The Ku Klux Klan as we understand it in popular culture today—the hooded, white-uniformed, organized hierarchy of kleagles and exalted cyclopses—was not so much a product of the Reconstruction South as it was of the industrial North of the 20th century. The first Klan, the one led by former Confederate officers, was a true terrorist organization. They conducted deadly nighttime raids aimed at driving out carpetbaggers and subjugating newly freed slaves after the Civil War. And they were successful, too. Even by the time Gen. Nathan Bedford Forrest, the first grand wizard of the Klan, resigned his post in 1869, much of the work of oppression could be done openly by governments.

But that was not the Klan that marched 30,000 strong down Pennsylvania Avenue in Washington in August 1925 or the one that nearly one in three white men in Indiana joined. The Klan of that era was certainly inspired by the terror rides of Forrest and his fellows, or a romanticized idea of them. The depiction of the Reconstruction-era Klan in the first real Hollywood blockbuster, 1915’s The Birth of a Nation, was as heroic defenders of civilization and the social order. The movement that resulted was something like the Oath Keepers and similar kooky groups we know today, but also the very popular fraternal organizations of their time, like the Masons, Odd Fellows, and Moose. Part political action committee, part social club, part pyramid scheme, the new Klan Inc. grew quickly in places far from the South and without any significant black populations.

The reason for the organization’s rapid expansion wasn’t just good marketing and friends in high places. It was that the new Klan zeroed in on the popular anxiety of the era: immigration. Great panic greeted the 15 million new immigrants, mostly from Southern and Eastern Europe, who arrived between 1900 and 1915. That would be like adding 65 million new immigrants to America over the next 15 years—more than quadruple the number we expect to receive. By 1910, three-quarters of New York City’s population was made up of immigrants. 

Nor were these the fair-skinned Northern Europeans of the previous immigrant waves. And not only were the new immigrants mostly Catholics, but there were lots of Jews and Orthodox Christians, too. The millions of members of the new Klan were perhaps more radical on the subject of immigration than most of the middle-class, white Protestants from which they drew their numbers. But the Klan’s message of “true Americanism” juxtaposed against the alleged immorality, sloth, and political radicalism of immigrants from places like Russia, Poland, and Italy resonated with many in the mainstream. Part of what kept pastors and civic leaders from publicly denouncing the Klan was a general agreement among many in the broad elite that something needed to be done about the erosion of white American values and the rise of Bolshevism and anarchism from these olive-complected foreigners. 

But what happened? As it turned out, it didn’t take long for America to change the immigrants and for the immigrants to change America. America got a taste for pizza and pierogis, and the immigrants developed an understanding of representative democracy. Within a generation or two, Americans ceased to place any meaningful difference between the Northern European ethnic stock from before the 20th century and the folks whose families came later. Indeed, the new census counts as white many Americans whose grandparents would have been targets of the Klan of the 1920s. The Hispanic, Asian, and African families that are either being cheered or feared for ending white rule today will be no different than the Italians, Poles, and Lebanese of a century ago. America will change their families and their families will change America. 

The single most important change in the new census numbers is probably in its definitions. Until 2000, respondents had to choose one race only. In 2010 and again in 2020 the Census Bureau has tried to account for multiracial families. This time it paid off with a 276 percent increase in the share of mixed-race individuals—now 34 million Americans. But that’s been going on all along, it’s just now the Census Bureau is doing better at capturing it. The trend will continue until, yet again, the terms and concepts we use now for race and ethnicity will sound as hopelessly backwards as the ones from the time of the second Klan. We will make America new again. And that’s not good for one party or ideology or another. That’s good for all of us.

Correction, August 18, 2021:
An earlier version of this newsletter confused “pierogis” with “pirogues.” The former, of course, is a European dumpling; the latter is a type of canoe.