GMU Ph.D. student Sam Wilson recently mentioned on Facebook that he was using the General Social Survey to test for political externalities of immigration.  He posted a few crosstabs, but nothing more.  I immediately publicly offered him the chance to guest blog his findings for EconLog, results unseen.  Here’s what Wilson found:

It should come as no shock to readers of this blog to discover that a
preponderance of US voters favor rather strict immigration
restrictions. I also expect that many readers are passably familiar
with some of the excuses forwarded by those in the closed-borders camp
and are eminently capable of judging the relative merits of each.
Almost alone among these reasons, political externalities stand as the
one objection that at least partially withstands careful scrutiny
using ordinary economic analysis*.

“Political externalities” is economics jargon. It’s a way of saying
that, in a democratic polity, since we all have to consume what the
majority (perhaps subject to some constraints) wills, any external
shift in the majority opinion towards an unwanted outcome harms
everyone. It’s a form of pollution, in a way. In Arnold Kling’s
terminology, when we resort to using voice, we abdicate our exit
option. By the conventional wisdom, immigrants allegedly corrupt the
national voice.

Bryan Caplan wrote the book on voter irrationality (available here–a
bargain at twice the price) so it is fair to say that you’d be hard
pressed to find another academic as well versed with political
externalities anywhere. Similarly, Bryan makes a habit of
extolling the virtues of free and open immigration. Taken together,
these two policies that almost seem at odds with each other. Using
survey results from the General Social Survey (GSS) and the Survey of
Americans and Economists on the Economy (SAEE), he finds increased
divergence from the informed positions of professional economists
among respondents with low educational attainment. Therefore, if we
know that:

     a) uneducated voters make bad decisions at
the polls and 

     b) low-skilled immigrants tend to be poorly

then it seems to be at least plausible that we, in our capacity
as concerned voting citizens, would want to keep out the riff-raff
simply to preserve the Republic.

This is not a novel argument. Bryan has discussed it at length
online, in print, and in person: 
see herehere,
here here
here for
starters (check
here for a more
through summary). Throughout (and in several lunch conversations I’ve
had with him), he admits that this is the closest thing to a
reasonable objection one might have against open borders**. From that
position, he argues quite convincingly that there are a number of
arguments (not the least of which is the obvious
moral case
 {if you haven’t already read that Cato paper,
please do so now–you won’t regret it}) against accepting the
political externality at face value, but I can’t recall him ever
directly challenging the premise itself using empirical

I intend to remedy this minor oversight. Because this is kind of
long-winded, I will do so below the fold.
Like Bryan, I also turned to the GSS for data, but my approach was
slightly different. The questions I am asking are:

  1. Do immigrants
    substantially different
    opinions on those issues we typically use democratic processes to
  2. If so, how do these opinions vary by subject type and construal
  3. Are there important opinion differences between the median
    native-born American and second or third generation immigrants? That
    is to say, if off-the-boat immigrants have political biases, are they
    passed on to kids and grandkids (alternatively, how successful is
    generational assimilation?)
  4. If differences exist, how strong are they compared to the effects
    of race, gender, education and political affiliation?
The preliminary answers I have gathered point to the following
stylized facts:
  1. For more distant domestic issues like race relations, crime
    prevention, drug treatment and environmental protection, the opinions
    of immigrants and their children and grandchildren are statistically
    identical to other respondents in the sample.
  2. For the near-mode issues of mass transit and the fairly generic
    “problems of big cities”, immigrants and their descendants tend to
    agree that the government spends too little (relative to native-born
  3. On the issue of foreign aid transfers, first generation
    respondents felt the US was spending too little, whereas second and
    third generation were statistically indistinguishable from native-born
  4. On the issue of military spending, first generation respondents
    felt the US was spending too much, and the second and third generation
    respondents were statistically indistinguishable from native-born
  5. For those other questions that showed occasional statistical
    significance, the coefficients tended to be positive for the
    Goldilocks answer: US spending on this issue was “just
Overall, with a few important exceptions, immigrants’ political
opinions just aren’t all that different from anyone
The exceptions tell a Horatio Alger tale of enlightened self- and
family-interest: improved access to mass transit improve employment
opportunities for low-income immigrant families and correcting the
ills of big cities is akin to wishing for better local infrastructure.
The generational stickiness of this opinion is interesting on its own
and may merit closer study, but for now, it’s merely an empirical
The other exceptions conform to my priors. In addition to being a Navy
veteran, my day job is with the military and I am well aware of the
size and the international presence of US armed forces relative to
that of other nations. I am not even slightly surprised that new
arrivals are a bit put off by the scale of the Pentagon’s operations.
Ditto for foreign aid. You don’t have to take a walking tour of Upper
Manhattan to get a sense of just how much wealthier the average
American is than folks abroad. Our streetlights are bright and
ubiquitous, our roads straight and well-paved (I find myself chortling
at accusations of “crumbling” infrastructure, having driven on the E36
in Poland, an exercise in the nail-biting dodging of entirely missing
slabs of roadway for 100km or so), and our water clean and
free-flowing. If you don’t imagine the disparity to be positively
shocking, I urge you to live on your own for a year abroad then judge
again once you return stateside. This disparity by dint of shock value
alone should raise the probability that new arrivals might think the
US should be doing more to help needy foreigners (I might also add
that I’d be interested to see how responses to this question might
change in response to greatly liberalized immigration policy).
Importantly, on the really big-ticket line items like OASDI and
welfare transfer payments, even new immigrants have the same political
opinions as people already here. The myth of the immigrant
welfare-state-expanding freeloader is not evident in the data I’ve
gathered. Furthermore, to the extent that immigrants’ political
preferences diverge from native-born Americans, that divergence is
assimilated away even as quickly as the very next generation. Any
lingering differences can be safely tucked into the “small potatoes”
I cannot reject the null hypothesis that there are no significant
long-term political externalities to open immigration.
As promised, here are some of the technical details (full outreg
tables available upon request, please find my contact information in
the link at the bottom of this post).
>From the comments section of my STATA do-file:
Do-file to investigate the partial effects of immigration on policy
preferences using GSS data drawn 28 Jun 2012
For use with data set “G:\Immigrant Song\4697_F2.dta”
One criticism of open immigration policy is to suggest that recent
immigrants and their descendents will favor a stronger welfare
state and/or more government intervention. To the best of
my knowledge, this assertion hasn’t been adequately challenged using
survey data. 
I use the following measures of preference for larger government
(questions ask if government is spending too little, just right or too
(space exploration)
natenvir (environmental protection)
(health care)
(solving problems of big cities)
natcrime (fighting crime)
(dealing with drug addiction)
(improving national education)
(improving the conditions of blacks)
(military spending)
(foreign aid)
(welfare [transfer] programs)
(highways and bridges)
(OASDI-Social Security)
(mass transit)
(parks and recreation, presumably to include Ron Swanson)
The independent variables of interest are: 
parents born outside the US and 
grandparents born outside the US
These are categorical variables in the survey:
. tab parborn
Freq.     Percent      
    Both born in US |     34,404    
  82.40       82.40
yes, father no |      1,359      
 3.26       85.66
no, father yes |        910      
 2.18       87.84
yes, father dk |         79      
 0.19       88.03
 Mother no, father dk |        
24        0.06      
dk, father yes |          8    
   0.02       88.11
 Mother dk, father no |        
10        0.02      
 Mother dk, father dk |        
45        0.11      
 Neither born in US |      4,911    
  11.76      100.00
              Total |    
41,750      100.00
. tab granborn
     BORN |
     U.S. |      Freq.    
Percent        Cum.
     None |     23,681      
60.24       60.24
        1 |      2,508  
     6.38       66.62
        2 |      4,437  
    11.29       77.90
        3 |        976  
     2.48       80.39
        4 |      7,710  
    19.61      100.00
    Total |     39,312    
I grouped these into categorical variables iparborn and igranborn for
the sake of convenience. When using indicator variable commands, the
omitted category is for native-born respondents, handily
I used these categories to create additional dummy variables for weak
and strong 2nd and 3rd generation immigrants.
1st generation:
= born in US
immigrant = not born in US
2nd generation:
foranypar (either or both parents born outside the US or unknown for
either or both [iparborn categories 2-9])
forbothpar (both mother and father born outside the US [iparborn
category 9 only]
strong: secondgen = native*forbothpar
[blog note: for reporting results here, I used the “extra
strong” definitions for both 2nd and 3rd generation
3rd generation:
thirdgen (any single grandparent born outside the US [igranborn
categories 2-5]
thirdgenstrong (all four grandparents born outside the US [igranborn
category 5 only]
strong: thirdgenonly = thirdgenstrong*(1-foranypar) (all grandparents
born outside US, both parents born in US)
extra strong: tgenonst = thirdgenonly*native (67 changes to
thirdgenonly, used as a robustness check)
Control variables include:
survey year (categorical from 1972-2006)
highest year of school completed
highest degree obtained by respondent
dummy, omit male for consistency
categorical; white, black, other, omit white for consistency
log(real income) in 1984 dollars [I have to double-check if that’s
actually the base year]
partyid: political party affiliation, as follows:
. tab
     AFFILIATION |      Freq.  
  Percent        Cum.
 STRONG DEMOCRAT |      8,023      
15.81       15.81
NOT STR DEMOCRAT |     11,018       21.72
    IND,NEAR DEM |      5,981    
  11.79       49.32
     INDEPENDENT |      7,444  
    14.67       63.99
    IND,NEAR REP |      4,405    
   8.68       72.67
REPUBLICAN |      8,175       16.11
 STRONG REPUBLICAN |      4,970  
     9.80       98.58
     OTHER PARTY |        720
           Total |     50,736
strong democrat or not strong democrat
ind, near democrat; independent; ind, near republican
not str republican, strong republican
I included beta estimates in my outreg tables, and in terms of the
magnitude of effect, political party affiliation, race and gender
tended to outperform immigration status in those cases where
immigration status was statistically significant. If I end up turning
this into a journal-worthy article (and I’m leaning heavily in that
direction), I plan to use better-fit LDV modeling and do some better
data prefiltering, plus maybe track some margins across some of the
categorical variables. 
Additionally, the econometric specifications were not of especially
good fit. Most R-squared values were below 0.05. It appears that there
is a lot of unexplained variation in public opinion. I urge the
cautious reader to remember this when interpreting any survey
results, especially those encountered in the media.

*I count cultural externalities as mostly positive on net: immigrants
bring with them wonderful new food, music, stories and celebrations.
They make America a less boring place.

**He also freely admits that I have a point about screening for
communicable disease and criminal record as part of the entrance
process, but that’s hardly controversial.

I would like to take this space to thank Bryan Caplan for
his extremely generous offer to allow me to post this on
EconLog. As a long-time fan, this is a genuinely overwhelming
opportunity. Thanks also to GMU Professor Garett Jones for providing
the impetus to conduct this investigation and American University’s
Daniel Lin for providing the germ of the idea.

Samuel Wilson is an economics student at George Mason University. He
is a regular blogger along with Michael Munger and Jeff Horn at Euvoluntary