More “fun with statistics”

I know, statistics might not sound that interesting at first, but wait!  Last time I wrote about statistics here, I showed you how news anchors get it wrong when they say that a political race with Candidate A polling 54%, Candidate B polling 46%, and a 4% margin of error is a “statistical dead heat” (in that example, there’s actually a 95% chance that Candidate A is in fact leading the race).

First up in this installment of “fun with statistics,” you may have heard that the Department of Justice is challenging a new Texas law that requires voters to present i.d.’s at polling places in that state.  Now you may be wondering why anyone would object to voters having to prove that they are who they say they are before they can vote, other than to facilitate election fraud of course.  Well, the D.O.J. is using statistics to argue that the Texas voter-i.d. law is unconstitutional because it will have a “disparate impact” on Hispanic people, i.e. it will prevent more Hispanic people than Caucasian people from voting.  Is that true?  Maybe so, but wait!

The D.O.J.’s analysis and argument are flawed, perhaps accidentally, perhaps knowingly.  Why?  Because in determining whether there’s an unconstitutional disparate impact, the relevant comparison isn’t just the raw number of Hispanics who are likely to be turned away from the polls for lack of a credible i.d. relative to the raw number of Caucasians who are likely to be turned away for lack of a credible i.d. — it’s not even just the percentage of Hispanics who are likely to be turned away for lack of a credible i.d. relative to the percentage of Caucasians who are likely to be turned away for lack of a credible i.d.  The relevant comparison is actually twofold:

1) The percentage of Hispanics who are likely to be turned away for lack of a credible i.d. while attempting to vote illegally (“true rejections”) relative to the percentage of Caucasians who are likely to be turned away for lack of a credible i.d. while attempting to vote illegally (statisticians might call this the “sensitivity” of the i.d. requirement as a detector of attempted election fraud, and obviously, we want as close as possible to 100% of attempted illegal voters of both races to be turned away), and

2) The percentage of Hispanics who are likely to be turned away for lack of a valid i.d. while attempting to vote legally (“false rejections”) relative to the percentage of Caucasians who are likely to be turned away for lack of a valid i.d. while attempting to vote legally (statisticians might call this the “specificity” of the i.d. requirement as a detector of attempted election fraud, and obviously, we want as close as possible to 0% of attempted legal voters of both races to be turned away).

In other words, if the i.d. requirement accurately disqualifies an equivalently-high percentage of Hispanics and Caucasians who are trying to vote illegally, and if the i.d. requirement also inaccurately disqualifies an equivalently-low percentage of Hispanics and Caucasians who are trying to vote legally (i.e. if the i.d. requirement has equivalent sensitivity and specificity with respect to both Hispanics and Caucasians), then its impact will not be disparate, even if greater raw numbers or even greater percentages of Hispanic voters than Caucasian voters are turned away.

Here’s an analogy that might help:  The D.O.J.’s argument would be like arguing that our laws against insider stock trading are unconstitutional because significantly more Caucasians than members of any other racial group get convicted of insider trading.  That’d be a flawed argument (and bad public policy) if significantly more Caucasians than members of any other racial group actually commit the crime of insider trading (which, statistically speaking, they do).  Thus, it would also be a flawed argument (and also bad public policy) to invalidate the Texas voter i.d. law simply because it’s more likely to disqualify Hispanics than Caucasians if significantly more Hispanics than Caucasians attempt to vote illegally in that state (which would have to be extrapolated from election-fraud statistics from past elections).  Bottom line:  With respect to the new Texas voter-i.d. law, the D.O.J. is jumping to a conclusion that is not necessarily supported by the statistics that it’s using.

Switching topics now, here’s a statistic worth noting as we try to reduce unemployment in the U.S.A.:  In virtually every society that offers its citizens unemployment benefits, dramatically-elevated percentages of benefit recipients tend to find jobs within a few weeks of the benefits’ expiration.  It’s true in Europe, where the benefits can last as long as five years, and it’s true here in the U.S.A., where the benefits currently can last as long as 99 weeks.  In other words, many recipients appear to wait until a few weeks before the benefits run out before they get serious about looking for work and/or get their expectations about the kind of work they’d prefer into line with the kind of work that’s actually available to them.  And at that point, large percentages of them tend to find work.

From a behaviorist perspective, then, it appears that many unemployed workers are significantly less motivated to find and/or accept employment when they’re expecting a steady flow of benefits.  So, what does this mean?  Well, if we believe that the best thing for both unemployed individuals and for our economy as a whole would be for those individuals to find ways to once again do something of value (create a good or perform a service) with their time and talents, then it may be neither the most compassionate nor the most practical public policy to offer unemployment benefits that can last almost two full years.

See how much fun statistics can be?

Advertisements

Comments are closed.

%d bloggers like this: