Friday, August 4, 2017

The RAISE Act's Skills-Based Visas: A Good Idea Taken to Ivy League Extremes

The Trump Administration signaled its support earlier in the week for the RAISE Act, a piece of legislation reforming the legal immigration system sponsored by Senators Cotton & Purdue. Lyman Stone has a very good run down of the all the details. I disagree strongly with the bill's aim to cut the overall level of legal immigration--in a country like ours that's only going to get older for some time, the conversation should be oriented around the goal of increasing legal immigration--but here I want to talk about one policy mechanism in the bill I do broadly agree with: skills-based visas.

Our current system is complicated, and the RAISE Act does a lot (so again, do read Lyman's summary), but skills-based visas are a way of giving priority to green card applicants with certain markers of achievement. Each achievement earns a certain number of points, and applicants above a certain threshold (in the RAISE Act's case, 30 points) are deemed eligible. Putting aside the RAISE Act's unwise reduction in overall immigration, orienting the legal immigration system towards skills-based visas is a good idea because it facilitates more economic growth and better assimilation.

It should come as no surprise that the devil in these kinds of policies is in the details. The bill chooses to emphasize young people (e.g. 26 - 30 year olds get 10 points, those over 50 get 0), STEM degrees (8 points for a STEM masters, 13 points for a STEM PhD, law degree, or MBA from a US university... I'm not entirely sure why MBAs get weighted the same as PhDs), English-speaking ability (those scoring in the top 10% of the TOEFL test get 12 points, while those in the bottom half get none), and income (having a job lined up that pays more than 3x the state's median household income gets 13 points; less than 1.5x gets no points). There are also a couple of special provisions for extraordinary achievement (25 points for Nobel Laureates, 15 for recent Olympic athletes) and business investment (12 points if you invest $1.8 million in a new US business and maintain it for 3 years). Here is a summary of the criteria and the associated points (source):

A bunch of policy wonks I follow on Twitter had some fun Friday adding up our scores and seeing how we'd fare. For the record, I got a 32 (whew!). But I was surprised how many people I knew--smart Americans who are native English speakers and hold advanced degrees!--wouldn't qualify. Some didn't make enough money, in many cases because they live in DC and are in public service. Others had advanced degrees that nevertheless didn't qualify as STEM, so only got credit for their bachelors degrees.

This made me wonder: how many current US citizens would qualify for a skills-based visa under the RAISE Act?

To answer this question, I used the IPUMS extract of the American Community Survey (ACS) for 2015. I coded all US citizens based on the points assigned in the latest draft of the bill, which you can read here (the points criteria start on page 27). I had to assume that all higher education degrees were from US institutions (biasing points up). I also had to make judgements about what degree fields qualified as STEM and what each person's TOEFL scores would land based on self-described English ability. I also can't account for the extraordinary achievement or business investment criteria (sorry Nobel Laureates). I think my assumptions were reasonable and that the results are not meaningfully sensitive to alternatives, but you can judge for yourself by viewing my Stata code here.

The results surprised me:

  • Among current US adult citizens age 18 or over, only 2.1% (4.8 million in 2015) would qualify for a skills-based visa under the RAISE Act. At the risk of being too pat, I'll note that Harvard's admittance rate last year was 5.4% of applicants. 

  • Of the 750,000 US citizens who naturalized in 2014 in the ACS, only 20,000 (2.7%) would have qualified under the RAISE Act. It's important to note that not all 750,000 would have been admitted under employment-based visas: many for example likely came because of family, and some may have gotten visas through the diversity lottery.
  • Among both 1) all US citizens and 2) citizens who naturalized in 2014, the key factor pushing them above 30 points on average is income. English scores top off relatively quickly, as does age. Greater education gives gradually greater advantages, but the steepness of the income criterion (only those who individually make a salary of 150% of the state's median household income or more get any points from income) means the skills-based visas disproportionately benefit immigrants going into very high-income industries and occupations, as well as those who settle in high-income cities. 

  • Finally, among the 4.8 million US citizens who would pass the threshold, 80% are in management or professional occupations...

  • ...with the largest shares in professional services, finance, and manufacturing.

  • Within the professional services industry, almost half are in medicine, another 13% are in law, and 12% are in higher education. 

Now obviously, these findings may strike many as a feature, not a bug. It depends a great deal on your views of immigration policy. Also, the pool of immigrants is itself influenced by the rules, so the conclusions we draw from these cross-sections shouldn't be strong. 

But it is worth asking: are these where America's greatest needs are? For some of the industries and occupation, arguably yes, but e.g. law and finance are not my ideas of industries hurting for extra workers or where a thoughtful immigration policy would want to augment the (shrinking) US labor force in the future.

I still think skills-based visas are a good way to orient immigration, but the restrictions to the criteria and weights in the RAISE Act right now strike me as too narrow for America's good, to say nothing of the reductions in overall immigration imposed elsewhere the bill and not offset by more employment-based visas. Luckily these can all be changed by Congress. I hope they do.

Sunday, July 30, 2017

Summertime: Is the Living Too Easy?

Senator Ben Sasse (R-NE) thinks young people have lost a sense of work ethic because they don't work as much over the summer as they did when Sasse was a teenager. As a parent of four kids and someone who focuses on labor economics, I found myself ambivalent about the article.

On the one hand, I sympathize with Sasse on the value of work and pushing our kids out of their comfort zones. My first formal summer job at the San Diego Wild Animal Park entailed cooking hot dogs and making cotton candy. It wasn't really "hard" work so much as it was, well, unpleasant work. To be fair, a lot of that I chalk up to being assigned next to the monkey enclosure (ugh). Beyond that though, pay was low, turnover was high, and the opportunity for advancement was limited. Nevertheless I earned some money and learned some humility, both of which I desperately needed at the time. And I was more motivated for college than ever, a privileged opportunity I had that a lot of my co-workers did not.

That said, while times have indeed changed, teenagers aren't really more "idle" over the summer than they used to be. They're just studying rather than working. Below, I've charted the proportion of 16-19 year olds who are in the labor force (that is, working or actively looking for work) and not in the labor force but enrolled in school. After 1989, the data allow us to split school enrollment further into full-time and part-time enrollment.

What this shows is that Sasse is indeed strictly correct that the share of teens in the July labor force has fallen over timer. But teens haven't swapped out work for idleness. They're now far more likely than in the past to be enrolled in school in July, so much so that the total proportion of teens who are either working or enrolled has been relatively stable since 1976, though it fell in the wake of the early-80s, 1990, 2001, and 2008 recessions. Virtually all of this enrollment growth has been in full-time enrollment.
NOTE: Participation pre-1994 adjusted for CPS redesign per Polivka & Miller (1998)
SOURCE: IPUMS monthly CPS extract, @ernietedeschi

Perhaps even more interestingly, this shift towards more summer school enrollment is not limited to just upper-income families, as one might expect. On average, teens belonging to households in every earnings quintile were more likely to be enrolled in July over the 2014-16 period versus the 1994-96 period. And while it's still the case that teens in top-20% households are the most likely to be in the labor force in July, teens in the middle quintile--households in the 40-60th percentile of earnings--are now the most likely to be enrolled in July.

"Q1" = bottom 20% / quintile. "Q5" = top 20% / quintile
Not adjusted for household size
NOTE: Weekly earnings adjusted for top-coding based on methodology of Schmitt (2003)
SOURCE: CPS, @ernietedeschi

Whether you're relieved or alarmed by these findings probably has to do with your prior beliefs about how valuable, or substitutable, school is relative to work. The literature on the benefits of teen employment are mixed and contextual, while the benefits of education are, in broad strokes, fairly clear. Speaking from my own perspective: if my children were faced with a choice between learning to code over the summer versus, per Sasse's experience, detasseling corn and other manual labor that is now done by machine, I'd lean hard on them to take the coding classes.

But that may not be the choice being faced by teens here. There's a lot the data don't tell us about these summer enrollees. We don't know if they're taking high school or college summer courses, or attending SAT test prep sessions. We don't know if they're learning valuable new skills on the margins, or knocking off curriculum requirements so they can have easier schedules during the school year.

And of course, teens may not face a choice at all. Increased college enrollment is rightfully lauded, but it may bring with it greater academic obligations on teens that crowd out their summers, and it would be reasonable to question how necessary this is. Since teens are generally low-skill and low-experience, their employment is particularly sensitive to recessions, broader structural changes, and the minimum wage.  And crucially, we can't even rule out that this is just survey error, and that households are erroneously marking off more teens as enrolled in July simply because more teens writ large are enrolled in school in non-summer months.

So further analysis is merited and we should be cautious about jumping to conclusions. But rather than accepting Sasse's framing out of hand that the decline in teenage summer work is a pure negative, we should ask more questions about why teens are apparently in school more over the summer, and weigh the value of that change.

Personally, I learned far more about actual hard work--versus just unpleasant work--from school rather than my summer jobs. You'd be right to point out that that's not surprising coming from an economist. Then again, maybe it also shouldn't surprise us that it's a successful US senator extolling the value of unpleasant work.

UDPATE #1 (July 31): The possibility that the rise in enrollment is driven by survey error has been nagging me, mainly because of the sheer magnitude of the shift.

Here's the question about enrollment in the Current Population Survey (CPS) that is the basis for the data I'm using:

As you can see,  parents are supposed to respond that their kids are not enrolled if they're on summer vacation. One might argue the wording is confusing because parents are supposed to answer "yes" if teens are on "holiday" or "seasonal" vacation, which refers to e.g. winter or spring breaks. Perhaps summer vacation could be misinterpreted as a "seasonal" vacation. But this at least alleviates some of my nagging skepticism because the guidance is there.

It's also important to keep in mind that households fill out the CPS many times: once a month for four months, then a year hiatus, and then another four months after that. So another possible source of error is inertia in the answers: households just going through and answering the same as they did before, even though the circumstances change for some questions (like their kids' enrollment).

Luckily, there's a way to account for inertial error: we can limit our analysis to just households who are responding to the CPS for the first time (in data parlance, their month-in-sample or MIS equals 1). When we do that, the results don't change meaningfully.

So at least we can take comfort that this specific type of survey error is not bearing on the results, though we can't yet rule out other types of error.

UDPATE #2 (July 31): The rise in enrollment is predominantly in high school classes, though both high school and college have seen a rise.

Thursday, June 15, 2017

Demographics explain all of the recent differences between native-born and immigrant employment in the US

Being an immigrant in the United States is often difficult. Even under the best of circumstances, immigrants can face challenging cultural and linguistic barriers, to say nothing of separated families, traumatic situations back in their countries of origin, and sometimes discrimination in the US.

So it may come as a surprise to see that, taken at face value, immigrants are more likely than native-born residents to be employed in the US. Bloomberg View columnist JustinFox published a chart using OECD data showing that the foreign-born employment premium in the US is unusual among advanced nations: only Israel, Italy, and the US see positive advantages for foreign-born residents. Even in other relatively high-immigrant countries such as Canada and Australia, native-born residents are still more likely to be employed than foreign-born ones.

Moreover, within the US over time, this immigrant advantage is recent but has grown since it emerged. In 1994, 62.9% of adult native-born Americans were employed (solid blue line below) versus 59.7% of immigrants (solid orange line), giving native-born adults a roughly-3-percentage-point premium. But by 2001, the native-born advantage had disappeared entirely and the foreign-born employment advantage kept growing thereafter. It’s come down a bit from its 2014 peak, when foreign-born Americans were 4.1 percentage points more likely to be employed, but it still stands at 3.5 percentage points as of 2016.

However, face value is deceiving here because the native-born and immigrant populations are very different from one another. For example, a larger proportion of the immigrant adult population is in the prime working-age years of 25-54 years old than native-born adults: 61% versus 47% respectively as of 2016. Also, the prime-age immigrant adult population also skews slightly more male than native-born adults (49.7% versus 49.0%). Since 25-54 year olds, as well as men, are more likely to be employed than others,  it would not be surprising that a subpopulation which skews more prime-age and more male would have higher employment rates.

To see how much demographic differences across a broad range of characteristics explain the wedge between immigrant and native-born employment, I reweighted the Census/BLS Current Population Survey (CPS) so that the proportions of the foreign-born population by age, sex, education, region of the country, and metro status matched those of the native-born population each year of the sample. So, for example, since 47% of the native-born population was ages 25-54 in 2016, I reweighted the immigrant population in the survey so that their 25-54 proportion would also be 47% in 2016. This is a quick way to see visually how much demographics explains the foreign-/native-born wedge. The reweighted immigrant employment line is shown in dashed orange.

As you can see, demographics explain a great deal of the difference between native- and foreign-born employment rates. In fact, beginning around 2006, the reweighted line is almost exactly on top of the native-born line, indicating that the demographic characteristics I controlled for explain virtually all of the wedge. This is in itself pretty interesting: once we control for demographic, geographic, and educational differences, immigrants are just as likely to be employed as native-born workers, no less and no more. It doesn’t mean there aren’t other factors working against immigrants, or to their advantage. But it does suggest that once we control for basic characteristics of the population, all these other factors are a wash, at least since 2006.

But note that demographics don’t fully explain the employment wedge before 2006, and in fact, controlling for demography makes immigrant employment rates consistently lower than native-born rates before then. It may be because the types of skills brought to these shores were dramatically different before then even after controlling for education, but it’s hard to adjust for this (the CPS for example usually doesn’t record the last occupation or industry of people out of the labor force, which would be one way of testing if the skill mix of the entire immigrant population had shifted). This may also relate to changes in where immigrants have been coming from, e.g. Latin America versus Asia, but it’s not clear beforehand how this should affect employment, if at all. And of course, discrimination may play a role. Regardless, the lead-up to 2006 remains an open and potentially interesting story, and I may have more to add in a later post.

In the meantime, know that when it comes to the labor market, foreign-born Americans look a lot like native-born Americans.

Tuesday, April 25, 2017

The Arithmetic of Border Adjustment

Just in time for the apparent death of the destination-based cash flow tax, I'm publishing my detailed notes on the step-by-step accounting math of border adjustment and the dollar. Now to be honest, this paper was partially just a self-exercise in practicing LaTeX and R Markdown. But more importantly, I've been frustrated that there isn't a single document with a detailed walk-through of why, for example, economic theory predicts a 20% DBCFT leads to a 25% dollar appreciation, why this predicted response is by the real and not nominal exchange rate, and why theory concludes a DBCFT is trade neutral. I hope that this paper serves that purpose, and that the next time Congress has a crazy fling with the idea of border-adjusted taxation it can help clarify the ensuing debate.

The RAISE Act's Skills-Based Visas: A Good Idea Taken to Ivy League Extremes

The Trump Administration signaled its support earlier in the week for the RAISE Act, a piece of legislation reforming the legal immigration ...