“Immigration and Job Creation at the State Level”

Mike Kimel continues his discussion on how Immigration impacts Job Creation.

In this post, I am looking at how the percentage of a state’s population that is made up of immigrants affects job creation.  After all, we hear from some quarters that immigrants create jobs, and from others that immigrants take away jobs that would otherwise go to Americans.  Obviously, the truth is sometimes one and sometimes the other, and mostly somewhere in between.  It depends on many factors, including the nature of the immigrants themselves, who, like anyone else, vary in attitude, capability, sociability, etc.  Even so, understanding whether in aggregate, immigrants help or hinder, and under what conditions, is worth knowing.

As I was pulling data, it occurred to me that I had phrased the question poorly.  The relevant issue is not whether immigrants create jobs.  It is whether immigrants create jobs for the native population.  If you doubt that, I’ve got an experiment for you to perform.  Go to Chetumal.  From the pictures I’ve seen, it seems like a very pretty town in Mexico on the border with Belize.   Meet with the mayor, and tell him (at this time the mayor is a man) you will renovate a factory and put 100 people to work in his town.  I bet he will be ecstatic.  Now explain that your plan involves hiring 110 Belizeans and firing 10 Mexicans who are currently employed.   My hypothesis is that the mayor’s mood will noticeably sour at this point.  You don’t actually have to go with Chetumal to run this experiment.  Pretty much any jurisdiction not run by a US politician will probably do.

Having established the question, here’s how I tackled it.

1.  I found data on the immigrant share of each state from Pew Research.  Data is available in 10 year increments from 1960 to 2010, and then for 2014.  Pew references the American Community Surveys, so I am 99.97% certain that all of the data originates with the Census, but I couldn’t find it there.

2.  I found data on monthly reports containing, among other things, “employees on nonfarm payrolls by state” for every month going back to December 1993 at the BLS

3.  There are three years for which the immigrant data can be matched to employment data:  2000, 2010, and 2014.  For each of those years, I used employment figures for December and the immigrant share of the population to calculate the “native born employment.”  That is, the number of jobs held by native born people.

3a.  Note the implicit assumption that the native born share of employment was equal to the native born share of the population.  That may be an underestimate in places where immigrants are go-getters and bring strong competitive advantages.  Conversely, it will be an over-estimate where the immigrants are less industrious and less competitive than the natives.  I believe this effect will be small.

4.  I computed the growth rate in native employment from 2000 to 2010, and from 2000 to 2014.

5.  I computed the correlation between the immigrant share of the population in 2000, and the growth rate from 2000 to 2010 as well as the growth rate from 2000 to 2014.  They were -0.047 and 0.032, respectively.  In other words, very close to zero.

So, at first glance, immigrants today don’t affect the job market tomorrow.  But the problem is, this analysis lumps Texas with West Virginia.  That’s like comparing oranges and bicycles.  So how do we do this differently?  One way is to use geographic groupings developed by the Census.  For each of those regions, I found the correlation between immigrant share of the population in 2000, and the growth from 2000 to 2014, and also the median immigrant share for the states in the region.  And since a picture is worth a thousand words, I overlaid that on a reasonable looking map showing the Census regions that I found here.

So what do we see here (click on the picture for larger size)?  The red numbers, which are the correlation between the immigrant share of the population in 2000 and the growth rate from 2000 to 2014 are positive in the three Census regions that mostly correspond to the old Confederacy, and negative everywhere else.  That is to say, in the old Confederacy, the more immigrants there were in a given state in the year 2000, the more jobs were created over the next 14 years.  But, in general, outside that region, the more immigrants there were in the year 2000, the fewer jobs were created over the next 14 years.

Why might that be?  Assuming the relationship is not spurious, I can think of a few reasons for the relationship. One scores OK on my personal self-censorship index (i.e., the ratio of how well an explanation fits the facts divided by the expected amount of trouble I will get into for pointing it out) so that’s the one I’m sticking with.  The gray highlighted numbers shows the median percentage of  immigrants for the states in each Census region.  And it does seem there is a rough correspondence between the likelihood that regions with a smaller share of immigrants are those most likely to benefit from more of them.   Put another way  – like just about any other variable, there is an optimal number of immigrants.  If you have too few, added some will generate benefits.  But there is a converse to that statement too – if you have above the optimal percentage of immigrants, adding more immigrants can put locals out of work.

Normally this would be the end of the post, but I feel I should add a few additional comments below.

**************************************************************************************

Comment 1:  Frankly, I don’t think the Census geographic divisions are the right way to do this analysis.    Sure, the Great Lakes states (a.k.a., East North Central) makes sense.  And we aren’t comparing Texas and West Virginia, but we are comparing Florida and West Virginia, which is just as bad.    And worse, we are weighting Florida and West Virginia the same.  Personally, I’d like to see “Large Coastal Economic Powerhouses” – California, Texas, Florida, and New York – which make up a third of the country’s population (correlation = -0.59) and other logical groupings.

Comment 2.   This analysis biases down the negative effect of immigration on jobs, and biases up the positive effect of immigration on jobs.  Assume for simplicity that any state that manages to crack the nut will generate jobs for a 15 year period, no matter what else happens.  If a state pulls that stunt off in 1997, by 2000 its share of the immigrant population will be up, but it will continue generating jobs for at least a decade.  The way I set up this analysis, the newly arrived immigrants “get the credit” for job generation.

Comment 3.  Perhaps this analysis shouldn’t be on the immigrant share of the population as much as on newly arrived immigrants as a share of the population.  After all, most immigrants who have been here at least X years have fully assimilated and are part of “us.”

Comment 4.  As always, I will mail a copy of my spreadsheet to anyone who wants it provided they email within two weeks of the date this post goes up.  You may be in luck beyond that point, but as time goes on, I may switch computers, die, etc.  If you want my spreadsheet, I am at my first name (that is “mike”), dot, my last name (that is “kimel” with only one m) at gmail dot com.

Comment 5.  This one is mostly irrelevant, but provides a bit of the background to this post.  A few weeks ago I had a post using national level data which showed a a negative correlation between the immigrant share of the population and the growth in jobs over the next decade going back to 1950.  In a more recent post I tried to provide a few explanations for why.  The comment section got rather testy and led to what is, as far as I can see, a rather odd post.