RACE TO THE FINISH: Who do you trust!

FRIDAY, NOVEMBER 2, 2012

Epilogue—What the professors wrote: Do you know what the AP meant when its reporters said this:

ROSS AND AGIESTA (10/27/12): Racial attitudes have not improved in the four years since the United States elected its first black president, an Associated Press poll finds, as a slight majority of Americans now express prejudice toward blacks whether they recognize those feelings or not.

[...]

In all, 51 percent of Americans now express explicit anti-black attitudes, compared with 48 percent in a similar 2008 survey. When measured by an implicit racial attitudes test, the number of Americans with anti-black sentiments jumped to 56 percent...

Do you know what the AP meant by those highlighted statements? In all candor, no, you do not.

We clicked to the AP’s report after reading this post, in which Digby said the following: “Read the whole article. The methodology is quite interesting.”

We did as Digby instructed! But in fact, the AP report provides little information about the methodology which led to those highlighted judgments. Even in this companion report, the AP tells us very little about how respondents qualified for the damning judgments expressed in those highlighted claims.

How did the AP reach those judgments? Clicking further, a reader could see the specific questions respondents were asked in the AP’s survey. But how many questions did a respondent have to get “wrong” before reporters Agiesta and Ross felt they could say that he was “expressing explicit anti-black attitudes?”

How many “wrong” answers were required before the AP rendered this judgment? Was it enough to say the wrong thing when asked to react to this statement?

“Irish, Italians, Jewish, and other minorities overcame prejudice and worked their way up. Blacks should do the same without special favors.”

Was only one wrong answer enough? How many wrong answers must people emit before Agiesta and Ross felt free to say that the person in question was “expressing explicit anti-black attitudes?”

Despite Digby’s response, the AP made no real attempt to explain how it reached its Olympian judgments. Readers of that AP report have no idea how the methodology worked, despite the joy with which some progressives rushed to affirm its great work.

We've seen several progressives voice reflexive approval of that AP report. Having said that, we’ll take a guess: Liberal approval was somewhat muted because of one of the AP’s findings—the finding that 55 percent of Democrats “now express anti-black sentiments.” (This finding emerged from the AP’s test of implicit racism.)

Oof! Everybody knows how these gong shows are supposed to work. These show trials are supposed to show that we Democrats are more racially pure than all other groups, by far.

But good lord! By one of the AP’s two "tests" (their term), we Democrats are more racist than independents—and we’re only slightly less racist than the Republicans are! We’ll guess we would have heard louder approval if that unlovely result weren’t there, though some of us plowed ahead with high affirmation anyway.

They seemed to feel this report must be right. Digby praised its methodology, even though it wasn’t explained.

Is it true? Did the AP’s surveys “show [that] a majority of Democrats...hold anti-black feelings?” Did the survey show that 55 percent of Democrats “hold anti-black sentiments?”

Rather gaily, the AP’s reporters tossed the term “racism” around as they made these striking claims. But do you know how they reached these conclusions? Do you even really know what those assertions mean?

Trust us: Almost surely, Agiesta and Ross have no idea how the AP’s car-pool of academics reached their deathless conclusions. On the test of explicit racism, do you think they know how many wrong answers it took to fail the test?

On the test of implicit racism, so they know how many bad reactions to Chinese ideograms were required to qualify for re-education? Do they know how many black Democrats were found to hold those “anti-black feelings” too?

If they knew, they didn’t tell! But this is the way these gong shows get spread all through the land.

(We refer to gong shows on matters of race. Or on imagined events like the Whitewater scandal. Or on imagined statements about who invented the Internet. Or on all that bullshit Colin Powell peddled at the United Nations. General Powell, who Rachel Maddow so adores.)

“In all,” do 51 percent of Americans “now express explicit anti-black attitudes?” To believe that (unexplained) claim, you pretty much have to trust the professors.

We pretty much don’t think you should do that. Here’s why:

We liberals tend to trust the professors, especially when they say the things we like. To some extent, you’re forced to trust them (or not) in a matter like this. In a survey of this type, so much technical claptrap is involved that it is extremely hard for a regular person to evaluate what's being said.

For our money, rational people would stop trusting these profs as soon as they read a few of statements to which they made respondents respond. In our view, you have to be a political illiterate to pose that (decades-old) question about “special favors.”

Happily, the professors posed it. For our money, a properly skeptical person would have stopped trusting them then!

That said, many people will be inclined to assume that these lofty “social scientists” must know what they’re doing. They hail from Stanford and Michigan, institutions of high renown! Almost surely, plebeians like us should defer to their erudition!

It’s easy to succumb to such thinking. Luckily, in the current case, the professors showed us the way to a more sensible posture. They did that when they published their account of how this gong-show got started.

In our view, what follows is really quite striking. It’s hard to know how Stanford and Michigan keep their doors open when scholars from their hallowed halls are publishing clip-clop like this:

PASEK ET AL (2009): The presence of an African-American candidate on the ballot running for President in 2008 raises the possibility that the election outcome might have been influenced by anti-African-American racism among voters. This paper uses data from the Associated Press-Yahoo! News-Stanford University survey to explore this possibility, using measures of both explicit racism (symbolic racism) and implicit racism (the Affect Misattribution Procedure)...

Long before election day 2008, long before the country even knew who the major parties’ nominees for President would be, forecasting models predicted a win by the Democratic Party's candidate. These predictions were based upon a common set of indicators, including the health of the national economy and approval of the incumbent President. The average predicted vote share for Barack Obama across the 9 models shown in table 1 was 53.3 percent, a little smaller than the 53.7 percent that President Obama eventually earned...

But during the summer of 2008, the numerous polls being reported by the news media did not find an Obama lead. For example, as shown in figure 1, the ABC News/Washington Post tracking poll documented essentially no difference between the candidates’ share of the vote among likely voters until late September. Why was this? Even during the summer, the national economy was doing badly and in serious decline, the United States was involved in two wars that were not obviously succeeding at achieving their goals, and approval of President Bush was remarkably low. Furthermore, the proportion of the nation that called itself Republicans had been declining steadily over the prior months. These and other key factors that are thought to influence election outcomes pointed toward a greater Obama lead than was being observed.

Faced with this puzzle, a team of researchers at Stanford University and the Associated Press worked together to generate a series of hypotheses about what might explain Mr. Obama's lagging performance and to test those hypotheses with data from a new survey.

In 2009, that account appeared in Public Opinion Quarterly. It took five professors to write it. It explained the way this project got started—this deathless collaboration between the AP and Stanford.

That account serves one clear public purpose. It helps us see that we should take nothing these professors say on faith.

Good lord! At a mere glance, we can see trouble the professors have avoiding self-contradiction. In the first sentence of that third paragraph, the professors say that
“the numerous polls being reported by the news media” in the summer of 2008 “did not find an Obama lead.” They then suggest, in the very next sentence, that Obama might have had a lead after all. (“The ABC News/Washington Post tracking poll documented essentially no difference between the candidates’ share of the vote.” Our emphasis.)

By the end of that paragraph, Obama did have a lead in the polling; his lead simply wasn’t as big as several “key factors” were said to “point toward.” But the first and last sentences of that one paragraph stand in obvious self-contadiction! Out of this jumble, we get our explanation of this project's origin.

Just from reading that one paragraph, a reader can see how easily the professors self-contradict. That reader may not realize the larger problems with that passage.

One example:

The reader may not realize that there was no ABC News/Washington Post tracking poll in the summer of 2008. This is a somewhat minor point. But the error suggests the professors may not know what a “tracking poll” actually is.

That wouldn’t make the professors bad people. It would suggest that they may not follow politics all that closely.

No, Virginia! There was no ABC/Washington Post tracking poll in the summer of 2008! But that error was tiny compared to the giant misstatement offered at the start of that third paragraph, serving as a potent topic sentence:

“But during the summer of 2008, the numerous polls being reported by the news media did not find an Obama lead.”

How do Stanford and Michigan keep their doors open when their academics publish statements like that? As everyone knows—as anyone can check with a few simple clicks—Obama led McCain in the polls all through the summer of 2008, often by substantial margins. Examples:

The summer of 2008:
Quinnipiac, July 13: Obama 50, McCain 41
Reuters/Zogby, July 13: Obama 47, McCain 40
CNN, July 29: Obama 51, McCain 44
CBS News, August 31: Obama 48, McCain 40
USA Today/Gallup, August 31: Obama 50, McCain 43
Diageo/Hotline, August 31: Obama 48, McCain 39

As everyone knows, polls jump around a certain amount, even when an election is static; we’re presenting some of Obama’s biggest leads tyhat summer. But almost all polls that summer showed Obama ahead, with the margin averaging roughly five points. At one point, one very famous American news org even reported this:

The summer of 2008:
Associated Press/Ipsos, August 4: Obama 47, McCain 41

One year later, the professors said their project began because they wanted to figure out why Obama wasn’t ahead in the polls that summer! (Or whatever. Wasn’t ahead by enough!) And just last week, the Associated Press—whose only poll showed Obama ahead that summer—was still linking to that 2009 publication. Oh, and by the way: Another poll by two major news orgs showed Obama ahead that summer:

The summer of 2008:
ABC/Washington Post, July 13: Obama 49, McCain 46
ABC/Washington Post, August 22: Obama 49, McCain 45

ABC and the Washington Post did not conduct a tracking poll that summer; their tracking poll started in late October. But in the monthly polls they did commission, they showed Obama ahead each time, by three and four points.

Why would these professors have made that first puzzling statement? (“But during the summer of 2008, the numerous polls being reported by the news media did not find an Obama lead.”) We don’t know, but their statement was patently false—and unless you read their shifting claims with great care, the professors seem to explain the origin of this project in terms of that bogus assertion! In our view, no: You shouldn’t put your trust in professors who are willing to publish nonsense like this:

PASEK ET AL (2009): But during the summer of 2008, the numerous polls being reported by the news media did not find an Obama lead. For example, as shown in figure 1, the ABC News/Washington Post tracking poll documented essentially no difference between the candidates’ share of the vote among likely voters until late September. Why was this? Even during the summer, the national economy was doing badly and in serious decline, the United States was involved in two wars that were not obviously succeeding at achieving their goals, and approval of President Bush was remarkably low. Furthermore, the proportion of the nation that called itself Republicans had been declining steadily over the prior months. These and other key factors that are thought to influence election outcomes pointed toward a greater Obama lead than was being observed.

Faced with this puzzle, a team of researchers at Stanford University and the Associated Press worked together to generate a series of hypotheses about what might explain Mr. Obama's lagging performance and to test those hypotheses with data from a new survey.

In that remarkable passage, the professors explain how this project got started. Does their explanation make sense?

Consider:

The professors note that Obama ended up slightly outperforming the forecasts they cite in the November election. In truth, such forecasts are very crude. It probably doesn’t make much sense to shave them down to tenths of a point.

But by the professors’ own method of reckoning, there was no “puzzle,” no “lagging performance,” displayed in the November vote—in “the only poll that counted." Why then did they start this project? They seem to say that a team of researchers...worked together to generate a series of hypotheses,” hoping to “explain Mr. Obama's lagging performance” in the polls which appeared that summer.

But there was no apparent lagging performance in those polls. Obama led in almost every one. On average, he held a lead which seemed to comport rather nicely with those crude election forecasts.

Fish gotta swim and birds gotta fly. In similar fashion, teams of researchers gotta generate a series of hypotheses which might explain lagging performance. That said, nothing the professors say in that passage makes a whole lot of sense. And several key statements don’t make sense because they’re simply inaccurate.

But so what! The professors wrote it; the journal published it. Nobody noticed the self-contradictions or the plainly inaccurate statements. Three years later, the AP is still sending readers to this publication to learn how this project began.

Was there a “puzzle” which needed explaining in the summer of 08? Was there really a “lagging performance” which called for work by a team of researchers?

We’d have to say that there was not—although we’d be inclined to think that Obama may have lost a few points off his percentage of the white vote, due to race, in that November election.

We’d be looking at his apparent share of the white vote, as compared to prior results for Candidates Gore and Kerry. (“Apparent” because exit polls are surveys, not absolute counts. No one conducts an actual count of the white or black vote.) But good God! Attempting to explain why Obama wasn't ahead by enough in July, the professors generated a series of hypotheses according to which 51 percent of Americans “now express explicit anti-black attitudes” (it was 48 percent in 2008), while 55 percent of Democrats “now express anti-black sentiments.”

By the time the professors get done, it's hard to believe that Barack Obama ever got any votes! And no one has the slightest idea how they reached ther judgements.

No, professors: ABC News and the Washington Post were not conducting a tracking poll in the summer of 2008. Much more significantly: Obama was running ahead, by a fairly comfortable margin, in a wide range of major polls.

Citizens, who do you trust?

Are you inclined to trust the people who produced that cock-eyed report in 2009? Are you inclined to trust the news org which is still linking to it? Are you inclined to trust an academic journal which didn’t notice those obvious howlers when this publication was submitted?

Are you inclined to trust the academic departments from which those errors emerged? Reaching our most significant queastion: Are you inclined to trust the people who posed that trap-laden question about those “special favors?”

For our money, people like Digby should stop putting their trust in those people and those orgs—in the types of orgs which have swallowed bullshit from all directions over the past thirty years, but mostly from the right. Those organizations don’t get smarter, or more productive, when they repeat our tribe’s overblown cant.

Question: Do those error-ridden academics understand American politics? Why did they undertake this project, given the state of the polling that summer?

Tomorrow, we’ll offer a bit of a “good news” report from the South. And we’ll offer our novel about the way this deeply unimpressive "journalistic" project got started.

Tomorrow: This just in from the Deep South—and our own origin novel