Dishonesty with statistics

Note: In this post, pp is short-hand for percentage points.

Tyler DiPietro mentions a Townhall column, by John Hawkins, called (without any apparent attempt of irony) Shattering 3 Myths About Liberals.
Taylor handles some of the content himself, but refers the more statistical parts to Alon Levy, who briefly gets into the problems in his post Dog Bites Man; Conservative Pundit Abuses Statistics.

While I don't disagree with Alon's post, I prefer to go a little deeper into why the statistics mentioned in the Townhall column cannot in any way be lead to the conclusion that the author reached.

First of all, the offending part:

1) Conservatives are more racist than liberals. Although you will find racists on the left and the right, the left is much more racist on the whole. That's why we still have to create majority black districts in order to get significant numbers of black Americans elected to Congress -- because even though blacks vote overwhelmingly for the Democratic Party, a large percentage of white liberals won't return the favor and vote for black candidates running on the Democratic ticket. According to a study at Yale, "(W)hite Republicans nationally are 25 percentage points more likely on average to vote for the Democratic senatorial candidate when the GOP hopeful is black...In House races, white Democrats are 38 percentage points less likely to vote Democratic if their candidate is black." That 25% of Republicans has nothing to be proud of, but as you can see, the Democrats are far worse.

He also refers to some other statistics to support his case, but I'll only focus on this part.
Hawkins' source is a rightwing news site, that linked on to a Washington Post column.

Unlike the rightwing news site, and John Hawkins, I prefer to look at the primary source, and not a Washington Post write-up of it. The study is called "How Black Candidates Affect Voter Turnout" (pdf), and is available on the internet.

Unsurprisingly there is a little more to the story than meets the eyes.

As Alon wrote:

The most shoddy part of the quote is the ellipsis, which covers several paragraphs in the relevant article. The 25% and 38% figures are not meant to be compared; after all, the 25% figure applies to Senate races while the 38% applies to House races, in which different dynamics might be in play.

That's quite true, and had we cherry-picked other numbers, we could have reached the exact opposite result (Republicans being more racist). Heck, we could just have quoted the study, which says:

For Senatorial elections (column 3), where unfortunately there is much less variation in candidate race, it is Republicans who are less likely to vote for their party's candidate when that candidate is Black. Independents and Republicans are 28 and 25 percentage points more likely, respectively, to vote for the Democratic candidate when the Republican candidate is Black.

Page 24.

For the record, Democrats are 2.4pp more likely to vote Democratic if the Republican candidate is Black and 1pp less likely to vote Democratic if the Democratic candidate is Black.
In other words, the 25pp figure should be compared to 1pp, not 38pp.

When it comes to the House figure of 38pp (which I find rather disgusting to be honest), the corresponding Republican number is actually that the Republicans are 7pp more likely to vote for a Black Republican (or at least 7pp less likely to vote for the Democratic candidate).
However, it should be mentioned that Republicans are 17pp less likely to vote Democratic if the Democratic candidate is Black, while Democrats are 2pp more likely to vote Democrat, if the Republican candidate is Black.

I would hesitate to try to explain the numbers, but the author of the study, Ebonya Washington, puts it down to the perception of Black candidates as more liberal.
If that's the case, a white Democratic candidate would be percepted as similar to a Black Republican candidate. A Black Democratic candidate would on the other hand be percepted as significantly more liberal than a white Republican, and somewhat more liberal than a Black Republican candidate.

Of the 335 contested races with Black candidates, 239 of them have had only one Black candidate. Since only 61 of these have been Republicans, 178 of them must have been Democratic.
It seems significant that 96 of the races were between a Black Republican candidate and a Black Democratic candidate – in other words, in nearly 1/3 of the contested House races involving a Black candidate, the 38pp of the Democratic voters would vote for a Black Republican candidate rather than a Black Democrat candidate.

Now for some different measures of how racist the parties are – let's take the number of Black candidates.
The study doesn't say anything about the exact number of Black candidates that the two parties have put forth, but it's possible to get some general idea of the relative proportion. All numbers are from the study, and relates to the period 1982-2000.

There have been 389 Black candidates for the House in total. That's approx. 9% of all races. Of these, 335 were contested. 89 of the contested races had a Black candidate from both parties, while 61 had only a Black Republican candidate, and 178 had a Black Democratic candidate. That means that out of the 335 contested races, 150 had a Black Republican candidate, while 267 had a Black Democratic candidate. That's 45% and 80% respectively.

There have only been 12 Black candidates for senate (out of 333), of which 7 were Democratic and 5 were Republican. When it comes to Governors, there have been 4 candidates (out of 239), where 3 have been Democratic and 1 Republican.

It seems to me that the Democrats have a better track record of nominating Black candidates (though both the parties' track record is hardly impressive).

And then there is the little matter of actually electing Black candidates.

From 1982 to 2000, in House elections, Black Democratic candidates won 88 percent of their elections while Black Republicans succeeded only 4 percent of the time. In the Senate Black Democrats were victorious 14 percent of the time and Black Republicans never.

P. 11

So in other words, the Republicans nominate and elect significantly fewer Black candidates, but it's the Democrats who are the racists? Give me a break!