$20K Waiter to $100K Data Scientist in 3 Months?

A number of readers have called my attention to the odd story of Paul Minton. After college graduation in 2011, he worked as a waiter with income about $20,000 per year in San Francisco for a couple of years. But after a 3-month “boot camp” course in data science, he somehow landed a $100K job in the field.

So, is this a misleading PR stunt by industry lobbyists, as some of my readers suspect? Could be, but here I’ll tell you why it’s plausible, and then address the question of whether graduates of boot camps are generally skilled enough to be paid that much right out of boot camp. (Short answer: No.)

First, note that we’re not talking about software developers here. Data scientists (a term I dislike but find convenient to use anyway) do write some code, but only as a means to an end, which is to develop predictive models. It’s the Big Data field you’ve been hearing about (though often the data sets are not that large).

In one sense, the barriers to entry are low. Someone with good quantitative intuition and drive can do well, the most celebrated case being that of Nate Silver, well known for his accurate predictions of the last two presidential elections, in spite of not having a broad knowledge of the field. (He didn’t do well when he tried predicting the UK election, showing that his intuitive understanding of US politics was his key to success, rather than knowledge of statistical methods.) Actually I’ve known guys wearing backwards baseball caps — nothing wrong with caps, just making the point that they were not PhDs — who’ve developed successful data science consulting businesses with almost no background.

According to Minton’s LinkedIn entry, he actually had some prior background in data science, including three MOOCs (large online courses). While these courses tend to be shallow, one can at least learn the terms and get some experience with real data. In fact, one person won a Kaggle data competition armed with nothing more than a MOOC from Coursera, where Minton had taken courses prior to the boot camp.

The boot camp included several useful projects, again according to Minton’s LinkedIn page. The company that hired him probably gave him a data set to analyze as part of the interview process, and he would have felt comfortable with it.

Most importantly, it’s clear that Minton talks a good game. I don’t mean that pejoratively; on the contrary, he appears to have real enthusiasm for this field, which is a huge point. He follows Kaggle, for example, in the same way that some follow the NBA. This trait is sadly missing in most students I interact with (grad and undergrad, domestic and foreign, UCD and elsewhere). Employers love this, for good reason.

So, yes, it does sound plausible that Minton was hired into a $100K data scientist job in spite of limited background. But that doesn’t mean employers who hire such people are making good decisions. There really is a major value to employers in hiring people who have extensive background in the subject, rather than the Cliff Notes version. And though I’m speaking generally here, some of Minton’s postings do suggest to me that he would benefit from a deeper pursuit into the field.

What is most worrisome about all this is that I know two older people, who live in the Bay Area like Minton, who are much more qualified, yet cannot get a job. Both have better formal training, both have more work experience, both have good “soft skills,” yet employers pass them by. Good for Minton, whom I’m sure will be a big success, but something is very wrong here. The real question is not the one my readers brought up — is Minton’s story for real? — but rather, why these two older workers are being ignored.

The Biggest Threat to the H-1B Cap

In the last couple of years, a number of proposals have surfaced in Congress to greatly expand the H-1B program, as much as tripling the yearly cap. Many people are concerned about the major impact this would have on U.S. tech workers. (Clearly, I share this concern, but would point out that my other major concern is the lower average quality of the H-1Bs.)

Yet it seems unlikely that those on GridLock Hill will enact such proposals, as they are tied to amnesty for unauthorized immigrants. Next year is a presidential election year, and what with the recent alleged murder of a young woman in San Francisco by an unauthorized immigrant, the murderous rampage in Tennessee by a legal immigrant, Donald Trump’s ill-considered yet apparently quite popular remarks, the SCE and Disney H-1B incidents, and so on, a direct increase in the H-1B cap looks unlikely for the next year or so.

But what IS likely is an indirect way to attain much the same effect as increasing the H-1B cap, in the form of an executive action extending the Optional Practical Training (OPT) portion of the F-1 student visa. OPT allows foreign students to work in the U.S. after graduation for 12 months in general, 29 months for STEM grads. For the latter, then, OPT would run up to 6 years. Unless you believe in wild coincidences, it seems pretty clear that the Dept. of Homeland Security (DHS) chose this 6 figure to coincide with the 6-year duration of the H-1B visa. Moreover, DHS has explicitly stated that that their proposed extension of OPT is motivated as a solution to the “problem” of there not being enough H-1B visas available. In other words, DHS is overtly stating that they intend the OPT extension as an end run around the statutory H-1B cap.

I would guess that DHS’ proposal will be adopted. The industry lobbyists will push hard for it, and sadly, there will not be much opposition. Senator Grassley has expressed concern, good, and of course there is still the lawsuit, which seems to be holding up fairly well so far. But OPT is not widely understood, nor is there sufficient understanding among critics of H-1B of the role of AGE in the whole mess. Review: Most H-1Bs are young, thus cheaper than older (35+) Americans. The legal wage floors for H-1B depend on experience, thus pretty much on age, so hiring young H-1Bs in lieu of older Americans is legal. As I’ve said, in the SCE/Disney cases, AGE was the key factor underlying the wage savings accrued by hiring H-1Bs. OPT workers, by virtue of being new/recent graduates, are almost all young. So, if the DHS proposal goes through, it will become a de facto expansion of H-1B.

In addition, there is my hobby horse, the incorrect assumption — even by many critics of the H-1B program — that the H-1Bs who are hired from U.S. university campuses, rather than imported from India by the Infosyses, are the “good” H-1Bs. Once again, I cannot emphasize enough that this is not just a misunderstanding, but one that is likely to have major consequences, the current OPT proposal being a case in point.

A number of critics of the H-1B program emphasize that the OPT workers enjoy a hiring advantage over Americans in that the OPTs are exempt from employer Social Security tax contributions. Yes, that makes OPTs cheaper to hire than American new grads, but the big savings come from hiring the young OPTs instead of the older Americans.

So, both the Googles and the SCEs will benefit if the DHS proposal is adopted, in almost exactly the same way as from an H-1B increase.

Startups Go Belly Up: Some Implications for H-1B

One of the industry lobbyists’ favorite arguments for H-1B is that immigrants fuel the creation of tech startups. You can imagine how much this idea appeals to members of Congress and the President, quite apart from the influence of the industry’s largesse in campaign contributions.

What is almost never mentioned, though, is that the vast majority of tech startups fail. Accrding to this news report on the demise of Homejoy Inc. the failure rate is 90%. Homejoy, “the Uber of the home cleaning sector,” went down in flames in spite of an excellent pedigree, the report points out:

The company, which emerged from the legendary Y Combinator, had secured $38 million in venture funding in late 2013 from prominent backers including Google Ventures, Redpoint Ventures and Max Levchin, PayPal Inc.’s former chief technology officer.

The government data show that Homejoy did employ H-1Bs. The foreign workers were not necessarily underpaid (remember, this must be gauged relative to wages for Americans with the same qualifications, impossible to tell from the data), but there certainly is a suggestion that Homejoy seemed to know how to exploit its workers. Like Uber, Homejoy has been embroiled in a controversy over whether their drivers/cleaners are their employees. One Web report, titled “Silicon Valley startups too often exploit the defenseless”, has this comment:

Whether it’s Homejoy or Alfred Club, well-funded Silicon Valley startups are seriously exploiting the impoverished and using dubious ploys not to pay employment taxes and insurance.

Homejoy cleaners gets $19 per house and at least some of them are homeless. Alfreds are servants that run errands for you and keep the house clean for $99 a month. Even though they are at the beck and call of the startup and the client, they are independent workers and thus the startup pays no employment tax and provides insurance. And the pay is a pittance.

Granted, I’m using circumstantial evidence here to paint Homejoy with the H-1B-abuse brush, and  again, they do seem to have paid their H-1Bs above the legal prevailing wage, though as I have explained before, the latter is a lowball figure, typically well below what the worker would command in the open market. I will simply say that this is the “Let them eat cake” attitude I find common in Silicon Valley, both in public (see the Dalgaard remark) and in private.

But in any event, the salient issue is that Homejoy failed. Again, the high-pitched claims by the industry lobbyists about the U.S. needing H-1Bs to fuel a startup boom must be viewed in this context.

Moreover, what happens to the H-1Bs who had been employed by Homejoy? Well, some of Homejoy’s workers were picked up by Google, and if any were H-1Bs, then they were “freebies” for Google. Under H-1B policy, workers who transfer from one firm to another, or who renew their visa, are exempt from the H-1B cap. So it’s in the interests of Google and the other big guys to have a large number of startups out there; when the startups fail, as the vast majority do, Google and the others get to circumvent the H-1B cap.

Often things are not as they seem on the surface, especially if that surface was created by PR specialiists.

Second Plaintiff in Age Discrimination Case Against Google

There is such a youth image of Silicon Valley nowadays that many people take it for granted that only the young are capable of good programming. Indeed, Facebook CEO Mark Zuckerberg publicly said so, and though Zuck later apologized, the image remains.

It is, of course, just that — an image. Experience does count in software development. I’ve been programming since my freshman year in college, eons ago, and would contend that I’m better than ever; I learn new languages more quickly, anticipate bugs better, have a better understanding of the future extensibility of my code, and so on.

The industry has generally conceded that its programmers tend to be young, but justifies it with the claim that only the young people, being recent graduates, know the latest technologies. I’ve refuted this in my University of Michigan law journal article, but my quick retort is, “Hey, who taught those new grads those new technologies — old guys like me!” So, yes, many older people know the new stuff too. (While I’m on the subject, I should plug my latest book, Parallel Computation for Data Science, chock full of the latest technologies!)

Moreover, the pair of plaintiffs currently suing Google for age discrimination claim to know current technologies, and assert that Google’s own review process found them to be highly-qualified. If so, the firm can hardly counter that the plaintiffs were rejected because these two workers weren’t up-to-date in their skill sets — as opposed to rejecting them due to a dislike of older people.

I’ve found over the years that the age problem, rampant in Silicon Valley, is largely fueled by the easy availability of foreign workers — H-1B, L-1, F-1 (foreign student visa, with Optional Practical Training work rights after graduation) and so on. The vast majority of the foreign workers are young, so the effect is that these programs greatly swell the young engineer population from which the employers can choose.

In addition, there are Vicious Cycle effects. The more young people are hired, the more there is an expectation of hiring young (which saves a lot in wage costs). This goes double for the foreign hires, who typically don’t come from cultures that are sensitive to discrimination issues. Indeed, I often hear immigrants scoff at the U.S. for its sensibilities in this regard. I believe that many immigrant engineers are only vaguely aware that age discrimination is considered unethical and is illegal. I suspect that this played a role in the bad phone interview that Plaintiff Heath says he had at Google.

As some of you know, I was an expert witness in an age case against Google that became quite prominent in the employment law field for certain legal-principles rulings. The case was eventually settled. The plaintiff claimed, and it was not disputed by Google, that he was the subject of overt hostility due to his age. In addition, there was statistical evidence showing disparities in sizes of bonuses, given similar performance ratings.

It’s been my impression that since that time, Google has tried to show a better attitude toward older workers. Of course, the word show may or may not be accompanied by actual improvement. I’ll mention a couple of possibly relevant recent incidents.

About a year ago Google came to my campus, UC Davis, to recruit, giving a formal presentation in a large lecture hall. There were four Googler presenters, including one of my students from a few years ago, who is one of the sharpest I’ve ever taught, thus speaking to Google’s high standards. Among the other three on the panel, two were also recent grads (one of whom I also had taught), and the fourth was older, I’d say early 50s.

The older guy made a point of saying — out of the blue, not in response to a question — that in Google interviews, older applicants are asked exactly the same questions as are the young’ns. But that ignores the likely fact that Google expects a higher bar for the older applicants in their answers to those same questions.

In that light, it’s interesting that a reader who attended the 2014 Grace Hopper conference told me that Alan Eustace, Google’s senior vice president of search, had stated, “It’s very difficult to justify hiring an older applicant.” Upon hearing this, I immediately went to the Hopper web page to view the video (the session was titled the Male Allies Panel), only to find that a message saying that the video had been removed! Deleted from YouTube too. I wrote a couple of e-mail messages to Hopper organizers, including one academic; no one even replied.

But isn’t it proper to ask more of the older applicants? Maybe the answer is Yes in general, but Google is a firm that hires a large number of foreign workers, sponsoring many of them for green cards. And in the latter case, Google is certifying that no interested and qualified Americans could be found to fill the position. So Eustace’s alleged remark would be quite relevant.

Once again, the connection between the age issue and H-1B is one of the two main source of abuse of the program. (The other is use of green cards to render foreign workers as de facto indentured servants.)

Barking Up the Wrong Tree (continued)

Some of you readers may have heard of Virgil Bierschwale, a veteran programmer (twice over, as you’ll see) who has long been out of work. He believes, with considerable reason, that Congress has stacked the deck against people like him with its H-1B work visa policies, and he doesn’t mince words on his blog.

Virgil is probably the type the industry lobbyists love to conjure up in the minds of the press and Congress, an older white male, navy vet, likely conservative in his politics. The lobbyists probably want people to add to that image with traits such as xenophobia, complacency (leading to technological obsolescence) and so on. I doubt that Virgil is like that, but it is certain that the lobbyists want people to think, “Well, poor guy, but no wonder he got replaced by an H-1B, and maybe he deserved it.”

Except that he didn’t get replaced by an H-1B. Yes, he is unemployed, and yes he believes that the H-1B program is a major factor. But Virgil has never been REPLACED by an H-1B. He does believe that H-1Bs are being hired INSTEAD OF himself. And not only that, he knows a number of other software developers, including a former H-1B, who are in the same situation — years of valuable experience, yet can’t get a job in spite of seeing lots of foreign workers being hired.

I don’t know Virgil well enough to speak to his technical acumen, and it’s certainly possible that there are multiple causes for his difficulties in finding work. But I do know a number of techies whose technical skills I know to be first-rate and whom I know to have all the needed “soft skills,” but who have really struggled in the job market. And in all these cases, it is clear to me that employers are hiring H-1Bs INSTEAD OF these Americans — and NOT using the H-1B program to REPLACE Americans.

Those readers who know this blog especially well have probably guessed by now that my theme in this post will be a frequent one of mine, the counterproductive focus in the H-1B debate on the IT services firms such as Infosys, and the concommitant, incorrect view that firms like Intel use the visa responsibly. I’d been meaning to write about Virgil for a while, but today I saw a Computerworld article that caused me to realize that he is a perfect example of my point, which is that most American victims of the H-1B program are more like Virgil than they are like those who were laid off by SCE and Disney and replaced by H-1Bs.

I don’t mean that most of the victims are white, male navy vets like Virgil. No, what they share with Virgil is that employers hire H-1Bs INSTEAD OF them rather than REPLACING them by the foreign workers. Yet the H-1B debate has devolved almost solely to concern over the SCE/Disney replacement scenario.

As I have warned repeatedly, this focus is not only inaccurate (again, most victims are like Virgil, who has never been replaced by an H-1B) but also destructive, as it will likely lead in the end to legislation which “solves” the problem by clamping down on the Infosyses while rewarding the Intels with an increased H-1B cap.

We’re seeing this more and more, with politicians promising to prevent replacement-by-H-1Bs. For  example, I reported recently that the California state legislature is upset about the SCE incident, but their solution is simply to ban the utilities from using “indirectly hired” workers, meaning supplied by the Infosyses. This exhibits complete ignorance of the problem; as I’ve explained, if Congress increases the H-1B cap and otherwise liberalizes foreign tech worker programs, SCE will have no trouble getting as much cheap labor as they want, without going through the Infosyses.

The Computerworld piece is quite interesting in illustrating the fact that the Obama administration’s actions on H-1B, if there will be any at all, will deal only with the replacement-by-H-1Bs problem. The administration will continue to push for an INCREASE in H-1B and other foreign tech worker programs, falsely extolling them as vital to keeping the U.S. in its world-dominant position on the technology stage.

In the article, Professor Ron Hira contends that the administration could solve the replacement problem by executive action, in contrast to the administration claim that only Congress can do this. Ron cites the wording in the H-1B statute that bans the hiring of H-1Bs if it would “harm the wages and working conditions” of the American workers.

Though I think this narrow focus is the wrong way to go, it is worth discussing. If the administration is not even willing to attack the narrow problem, prospects for solving the broader one are grim indeed.

Clearly Obama has been happy to bypass Congress on immigration issues, by issuing his own executive orders of questionable legality. But if he wants to take refuge in legalities, I believe that legally he is off the hook in terms of Ron’s idea.

John Miano has pointed out that by delineating special circumstances in which employers are banned from replacing Americans by H-1Bs, Congress has, legally speaking, expressed its intention that in all other settings such replacement is just fine. I would add that similarly, by setting a four-tier rule for wage floors for H-1Bs, broken down by four levels of experience, Congress has given its blessing to employers who hire younger H-1Bs instead of “Virgils.” So the clause in the statute regarding harm to American wages arguably applies just to Americans of a given experience level. Hiring young H-1Bs in lieu of older Americans is thus no violation, and as I’ve noted in the past, this is the main factor behind the legally-compliant actions by SCE and Disney too, as well as the Intels.

Innumeracy in Politics, Alleged Murder in SF by an Unauthorized Immigrant, Etc.

Note to the reader: As my posting title indicates, I’m going to address two very different topics in this posting. They in fact are connected, but one of them — the senseless killing of a young woman — is far more important, not to mention attention-getting, than the other. Nevertheless, the killing sparked a bitter debate on immigration policy, and that debate in part revolves around the misunderstanding of quantitative information. Thus I feel I ought to write about both.

Following up on my last post, “Prominent Political Scientist Decries STEM “Frenzy”, I feel compelled to comment on an irony: In my experience, many of those who frantically call for more students to study STEM are the same types who loathed STEM when they were in school, and disdain it to this day.

As someone coming from a math/statistics background, what strikes me most is the widespread innumeracy (quantitative analog of illiteracy) that I observe among people in the public policy arena. As one otherwise insightful NPR reporter once said to me, only half jokingly, “We’re innumerate and proud.”

I’m not talking about, say, differentiable manifolds, but simple primary school stuff like rates and proportions. For some reason, it’s really difficult for a lot of people in the policy area to understand the difference between rates/proportions on the one hand, and absolute numbers on the other.

A common example concerns the studies that cite the number of patent applications filed by immigrants in tech. Well, of COURSE there are a lot of immigrant inventors in tech, because there are a lot of immigrants. What matters is the per capita rate, and it turns out that the native rate is higher. Given the lack of a STEM shortage and thus the direct and indirect displacement of Americans from STEM, that means that STEM-based immigration is causing a net loss in patent production.

Which brings me to the alleged murder of a young woman in San Francisco by a Mexican national named Sanchez with seven felony convictions and five deportations. Again, it’s hard to speak of something like innumeracy in the context of such a sad event, but the subject matter — immigration and crime — does come up in the debate over immigration, and it’s vital to look at the issue carefully. In this case, by the way, it will turn out that the absolute number, not a rate, may be more relevant.

This was the topic this morning in the first hour of a highbrow talk show hosted by SFSU Professor Michael Krasny, with the focus on the fact that San Francisco is officially a “sanctuary city,” meaning that city agencies are directed to avoid cooperating with federal immigration authorities, to the extent allowed by law. Before I get to the innumeracy issue, I need to build the background.

The guest on the side of criticizing SF city policy was former U.S. Attorney Joseph Russonello, who was present for the full hour. On the “It’s not SF’s fault” side were SF Sheriff Ross Mirkarimi for the first half hour, and civil rights attorney Angela Chan in the second half.

Russonello emphasized various arcane points in the law. While I’d say he made the best arguments, he did not present himself well, often rudely interrupting the other two guests and getting into more legal detail than most listeners wanted to hear. Russonello also stated that, contrary to the claims of immigrant advocates, Sanchez’ prior crimes should be considered violent, as the circumstances indicate that he is with the Mexican drug cartels.

Mirkarimi was very polite, and claimed that he, who heads the agency that released the suspect to the streets, did so because the law required it.

Chan is with the Asian Law Caucus, by far the most radical of the various Chinese/Asian political organizations in the Bay Area. She has been active in preserving and extending SF’s (and the state’s) sanctuary status. She was extremely combative, invoking the “R word” (racism) a number of times, and most importantly for the innumeracy point, insisted repeatedly that this tragic case is not about immigration, and shouldn’t be used to tighten up policy, as many are urging.

Chan, in making her point that the question of this murder should be decoupled from the immigration issue, cited research finding that immigrants are five times less likely to be incarcerated than are natives. Russonello said that there are 175,000 foreign nationals in U.S. prisons. In other words, Chan was claiming a rate,  while Russonello was discussing an absolute number. Chan simply could not understand the difference, and indignantly but wrongly treated his number is inherently contradictory to hers, and thus claimed his number must be wrong. Sadly, Chan showed herself to be among the “innumerate and proud.”

It matters. Russonello’s point, I believe, was that the proportion of immigrants in prison is not what a lot of people care about; instead, those people would view it as indicating that without immigration, we’d have 175,000 fewer dangerous people in our society.

In other words, it becomes a tradeoff. In many people’s minds, immigration brings certain benefits, such as a better economy or a more diverse populace. (Personally, I doubt the first, but support the second.). Immigration policy is really a cost/benefit tradeoff. In this light, it’s important to understand exactly what the costs and benefits are, and concerning crime, the Russonello numbers, if correct, must be taken into account.

Now, what about the paper cited by Chan? It was written by immigration advocates, and much more importantly, published by the research arm of the American Immigration Lawyers Association. As such, it is inherently biased, as is so much “research” on H-1B that is actually funded by advocacy groups. The report cited by Chan has NO formal statistical analysis such as computing margins of error, etc.; that 5X figure is especially suspect in that light.

I haven’t gone into this area of immigration research, and really don’t know whether the immigrant crime rate is higher or lower than the native one. If you are interested, CIS, an advocacy group on the other side of the immigration issue, has an interesting analysis, but again, I haven’t delved into the matter enough to say who is right.

What I can say, though, is that for this particular issue, the absolute numbers seem to be much more relevant than the rates. Russonello’s figure is pretty scary.