A few months ago, a journalist from a left-wing (not merely liberal) publication was interviewing me about the H-1B work visa. He had quoted me before, and was generally sympathetic to my views. At one point, the conversation turned to Pres. Trump’s views on immigration. I mentioned that though I hadn’t voted for Trump and had been critical of him in this blog and elsewhere, I feel he is often treated unfairly by the press, e.g. on the issue of immigration and crime — contrary to press claims, Trump never said that most immigrants, or even most Mexican immigrants, are criminals. His choice of words was, as always, completely blunt and unrefined, and without the obligatory disclaimers, but all he really said was that immigration brings some crime.
The reporter was mystified by this. He replied, in a “this is the fundamental truth and it’s a settled issue” kind of tone, “It’s well established that the crime rate among immigrants is lower than in the society at large.” This statement has been questioned by some, but my point is that it is not relevant; if safety is the concern, then what matters is the absolute number, not the rate.
According to the U.S. Dept. of Justice, as of 2013 there were over 70,000 noncitizens in state and federal prisons. In addition, there is a certain number who haven’t been caught, or are awaiting trial, some naturalized citizens etc. Some are white collar criminals, of course, but even if the number of immigrants in prison for violence, drugs, prostitution/trafficking, burglary and so on is “only,” say, 50,000, that should worry almost anyone. This is 50,000 criminals who would not be in the U.S. if we had zero immigration. In terms of concern for our person and our property, this 50,000 figure is what counts, not the rate, i.e. not the percentage of immigrants who commit crimes. Even the left-wing reporter seemed taken aback when I pointed this out.
Clearly, I am not advocating a zero-immigration policy. No government policy is without drawbacks, and those of us who value immigration (who, I claim, comprise the vast majority of Americans) know that there will be some downsides. The questions, though, are: How many? Who? With what responsibilities? And so on. More simply: Where do we draw the line?
Whether deliberately or unconsciously, by simplistically dismissing the crime issue by saying, “The immigrant rate is lower,” the politicians and the press are not allowing the American people to develop informed opinions on this vital issue. This is criminal. (Pun intended; originally I planned to title this post, “Criminal Statistics.”)
Keep this in mind in the coming weeks especially, as the trial of the accused killer of Kate Steinle in San Francisco is now starting. You will hear repeatedly from the pundits and immigration advocates that “The immigrant crime rate is lower,” when in fact that really is not the issue for public safety.
The issue of Pres. Trump’s temporary immigration ban from certain majority-Muslim countries is quite similar. Neither Trump, AG Sessions, White House Adviser Steve Bannon nor any other policymaker has claimed that most Muslims are possible terrorists. Obviously the rate is quite minuscule. And I personally opposed the ban even when it was just in the talking stage.
Nevertheless, it is a legitimate issue, not something for the Trump bashers. Remember, the Trump policy, still in litigation, took an Obama policy as its foundation, even listing the same seven Middle Eastern countries. And as with the crime case, the absolute number of terrorist incidents is what matters, not the rate.
Yet it is the rate, not the absolute number, that is the focus of an article in Scientific American, reprinted in today’s San Francisco Chronicle, titled, “Why Data Science Argues against a Muslim Ban.” This one is especially insidious, as it is cloaked in one of the hottest areas in today’s tech industry, Data Science. (Also known as Artificial Intelligence, Big Data, Predictive Analytics and so on.) The author is Eric Siegel, founder of the Predictive Analytics World conference series.
As you read the article, watch for the tell-tale signs, e.g. phrases like “more likely” and “less likely” — in other words, rates. That, after all, is what Data Science (DS) is all about. For instance, one of the big applications of DS is marketing, i.e. identifying customers who are more likely to purchase a certain product if shown an ad for it. So, let’s take a closer look at that, putting aside the absolute numbers vs. rate question, and focus on rates. They will be quite relevant to the Trump policy, in ways that Siegel isn’t telling you.
A good data scientist is really interested not just in direct rates, but also expected utility. If say in real estate even the most likely customers have low individual rates of responding to an ad for a $20 million estate, it still makes economic sense to target those most likely customers, since the payoff is so great.
In the case of potential terrorism, the expected utility is negative. But for the very same reasons, it pays to target the most likely candidates, even if their individual probabilities of terrorist activity are quite low. As Siegel points out, it would be ideal to use personal behavior to try to compute such probabilities. Yet look at the outrage from the immigrant advocates when the U.S. government has recently been attempting to do exactly that, by requiring that travelers from the specified countries provide their social media passwords.
Highly intrusive? Of course. But when weighed against the possibility of terrorism on U.S. soil, we must strike a balance, and it again becomes a question of, Where do we draw the line? Siegel is hypocritically obfuscating that issue.
And there is more: As we all know, the Obama people constructed the list of seven countries for travel restrictions (though not outright bans) specifically because those countries were identified as harboring terrorists. Any data scientist worth her logistic regression coefficients would use this information in her prediction model. Yet this is precisely what the article here is objecting to, Pres. Trump’s singling out those seven countries, because Siegel doesn’t like the fact that they are majority-Muslim.
Returning to the issue of absolute numbers, so far we have had just a few per year in the U.S., say 5-6 including foiled plots. Is that tolerable? What if it were to rise to 25-26?
So, one more time: The question is, Where do we draw the line? Do we want Trump’s line, Obama’s line, or Siegel’s line? These are ordered from safest to most dangerous, while also being ordered from least- to most-protective of civil liberties. You be the judge. Just don’t let people like Siegel fool you in the name of Data Science.