Media Malfeascance

I know I had a rant some time ago regarding the apparently deliberate misrepresentation of results from studies as reported in major newspapers. Here is another example from the New York Times, Discrimination in Housing Against Nonwhites Persists Quietly, U.S. Study Finds by Shaila Dewan.

Stipulated that discrimination is bad and should be discouraged. This is about the reporting and not the importance of the subject. The money paragraph.

“Although we’ve come a long way from the days of blatant, in-your-face housing injustice, discrimination still persists,” Shaun Donovan, the department’s secretary, said in a telephone conference on Tuesday unveiling the findings. “And just because it has taken on a hidden form doesn’t make it any less harmful.”

In each of the study’s 8,000 tests, one white and one minority tester of the same gender and age, posing as equally well-qualified renters or buyers, visited the same housing provider or agent. In more than half the test cases, both testers were shown the same number of apartments or homes. But in cases where one tester was shown more homes or apartments, the white tester was usually favored, leading to a higher number of units shown to whites overall.

Those last two sentences caught my eye. That's an awfully stilted way of putting it. What are they hiding? Why the awkward wording? Well, as it turns out, there's a reason. When we think of discrimination occurring, we want a very simple morality tale that's quite clear, but its not quite that, should I say it?, black and white.

Turns out that there is a lot of noise in the system. Real estate agents show different viewers more options sometimes to whites, sometimes to blacks and sometimes to Hispanics. And by big numbers. Let's walk through it. With my skepticism aroused, I went back to the original report, Housing Discrimination Against Racial and Ethnic Minorities 2012. The critical information is forty pages in, Exhibit IV-1: Summary Measures of Discrimination Against Minority Renters.

When the HUD field testers (of each racial group) met with real estate rental agents, what do you think happened? From the news article you would expect that "But in cases where one tester was shown more homes or apartments, the white tester was usually favored, leading to a higher number of units shown to whites overall." Sounds like a lot right? But what does it really mean? We can figure it out from the table.

To me, there are two quite critical pieces of information buried in those sentences that should receive a lot more attention than they do.

In the first case, it is apparently common in these tests for there to be random variation in the number of units shown. Comparing how Blacks and Whites are treated (as paired comparisons), in 54% of the cases, they were both told about additional units (i.e were treated identically). In 28% of the cases Whites were told about extra units but blacks were not, and in 19% of the cases Blacks were told about extra units and Whites were not. So, of a hundred times visited, in only a net of 9% of the cases were Whites treated better than Blacks. I would put this as Agents treated applicants equally (within random variation) 91% of the time. Wow! That's pretty good. Can always do better, but that is still pretty good. And 19% & 28% tells me there is a lot of random variation in the system. So the big improvement for all renters is not in reducing the 9% structural discrimination by agents but in reducing the 19-28% random variation.

Doesn't that picture sound a lot better than "the white tester was usually favored, leading to a higher number of units shown to whites overall" which sounds far more pervasive?

Run your eye down the numbers in the table. There's only a couple of other instances where the discrimination is as high as 9%. What is striking to me is how close the numbers all are. When meeting with agents, Whites are shown only 3% more units than Blacks (given the random variation). Again - Wow! There's less than 1% variation in rents quoted across all the ethnic groups.

Now there is still discrimination to be squeezed out of the system, and the remaining amounts will be harder given the random variation and most of the cheap and easy remedial actions having already been taken, but let's talk about the situation as it is rather than paint false pictures.

It's a scandal that there is any discrimination but there's an even bigger scandal when the degree of discrimination is misrepresented. Since it has to do with numbers and statistics this is probably simply a matter of journalistic ignorance and laziness. A harsh judgment supported by the fact that much of the news article is simply a parroting of the report's executive summary. And it is no surprise that an agency would exaggerate the circumstances - their funding depends on there being a continuing crisis for them to solve. But we should expect more from our news media than this.