Opinion, Berkeley Blogs

How to lie with statistics: Job losses, women, and presidential candidates

By Rosemary Joyce

In case anyone hasn't heard yet, yesterday apparent Republican candidate for president Mitt Romney rolled out his argument for the women's vote in November. And it was a doozy: 92.3% of jobs losses during the Obama presidency belonged to women. 92.3%! Can you believe it???!!

Well, no. You can't.

Not that the numbers are made up. The calculation is technically precise.

But here's where this gets interesting: while precise, the number gives an impression that is fundamentally inaccurate.

Explaining why, though, is challenging because to understand it, you need to be able to follow not just the raw math, but the context that makes this "statistic" misleading.

NPR provides the clearest explanation. The Romney campaign uses numbers comparing jobs today and jobs in January 2009, when Barack Obama was inaugurated. NPR's Scott Horsley explained:

the U.S. economy has actually been adding jobs for the last year and a half, but we're still not back to the point where we were in January of 2009, when President Obama took office. We're still about 740,000 jobs short.

Now, if you just look at the number of women who are working today, that's down about 683,000 from where it was in January of '09. If you divide 683 into 740 that is 92 percent...

So that means the Romney campaign is telling the truth, right?

Sorry: no.

The Romney campaign is trying to use statistics– a startling, even unlikely statistic– to persuade women that Barack Obama has been bad for us. As the redoubtable Ezra Klein notes, the statistic he is fighting here is a different one: a 19 point gap in women's preferences if the presidential race were held today, in favor of Obama over Romney.

Basically, the argument is, we'll see your 19 and raise you 92.

So how might we actually assess differences in men's and women's employment experiences during this most recent downturn?

First, let's acknowledge that politicians have less influence then they might want you to believe during a campaign– especially presidents, who must enlist congress in support of any policy initiative. Giving them credit or blame for the way an economy sheds jobs may help a campaign find a startling number, but it is unlikely to reflect real policy directions, let alone gender politics.

Quoting Horsley again, by January 2009

the economy had already lost some four and a half million jobs. And those job losses were disproportionately among men.

Does that mean Barack Obama's predecessor, George Bush, was anti-male? Of course not. What it reflects is the way that men and women are distributed in the work force. The first sectors affected by this recession, construction and manufacturing, were disproportionately male; recent job losses have been coming from parts of the economy where women make up a higher proportion of the work force.

Think teachers here.

So, what has happened to employment for men and women? let's start measuring from the beginning of the recession, not from some arbitrary point.

According to the Washington Post's Glenn Kessler, counting from December 2007– which is when the recession began– the total job loss was 5 million jobs, 1.8 million of them previously held by women. That means 36% of job losses in this "great recession" were sustained by women.

Really, there should be no reason to even write about this claim. It is clearly political, and the number the Romney campaign came up with simply doesn't match anyone's lived experience– because what we all have seen is about two men losing jobs for every woman laid off.

But that number is getting repeated and repeated and repeated. And weirdly, the various news media cannot seem to find a way to say clearly that it is a lie.

So, for example, in its "Fact Checker" feature, the Washington Post writes

We cannot fault the RNC’s math, as the numbers add up... at this point we will give this statistic our rarely used label: TRUE BUT FALSE.

Ah, no. The numbers may add up; but the argument they are used to advance isn't even the slightest bit true– it is entirely false.

I am not any happier with the fact-checking website Politifact, which rates the claim "Mostly False" and says

There is a small amount of truth to the claim, but it ignores critical facts that would give a different impression.

"Ignoring critical facts that would give a different impression" is a very good paraphrase of a much simpler, old-fashioned word that applies here: "lying".

How to Lie With Statistics is the title of a book that was required reading when I was an undergraduate taking statistics in the Sociology Department at Cornell University. Originally published in 1954, I find, it is still in print. Google Books offers access to the 1993 reprinting.

The good folks in the media might do well to read it as we launch into what promises to be a season of what the author, Darrell Huff, called "stasticulation": "misinforming people by the use of statistical material".

Huff titled his book harshly for a reason. As he explained,

whoever the guilty party may be in any instance, it is hard to give him the status of blundering innocent...As long as the errors remain one-sided, it is not easy to attribute them to bungling or accident.

Stasticulation season is upon us. And the media, unfortunately, are apparently unable to call it out when they see it. As Huff reminded us more than half a century ago:

The secret language of statistics, so appealing in a fact-minded culture, is employed to sensationalize, inflate, confuse, and oversimplify.

So when the Romney campaign trots out their claim that 92% of job losses were incurred by women, try this other statistic out, from the Bureau of Labor Statistics:

women actually make up a larger share of the workforce now than they did in Jan. 2008 before the financial meltdown, and since January 2009, it is a statistically insignificant change. In January 2008, women made up 48.8 percent of the workforce; in January 2009, 49.5 percent; now 49.3 percent.

If you can square these two "statistics": women lost 92% of jobs but somehow stayed at close to 50% of those employed — then you don't need to read How to Lie with Statistics. But judging from the coverage I have seen, we might all want to send copies to our favorite news media.