Give US Shelter from Callousness
In July 2015, the
country was shocked by the accidental shooting of Kathryn Steinle by a
homeless, undocumented immigrant. Less than a month before, real estate
billionaire Donald J. Trump had announced his bid for the Republican nomination
centered on the idea that undocumented immigrants were an economic and criminal
threat to the United States. Despite the shooting being accidental and the
Mexican immigrant had been deported five times prior, Trump as well as others,
pounced on the story as an illustration as a broken immigration system unable
to enforce its laws. Their
narrative contended that San Francisco’s status as a sanctuary city was the
catalyst for the Steinle shooting given that San Francisco was the first in 1989
to adopt city ordinances that forbade state and local law enforcement from
inquiring about someone’s residency status.
Indeed,
immigration enforcement has been a hot button issue since 1996. Most recently
however, Cal State Long Beach’s police department issued a directive to
withhold cooperation with Immigration and Customs Enforcement by detaining or
even enquiring as to a students immigration status. This is why I choose to
pose the research question, is there a difference in violent crime rates in
sanctuary and non-sanctuary cities? The null hypothesis is that there is no difference,
while the research hypothesis is that there is a difference. My prediction
going in was that there was a negative difference in violent crime rates
between the two categories of cities, that non-sanctuary have less crime. It is
important to note fore the purposes of this paper that the distinction between
sanctuary and non-sanctuary cities is not an empirical data point, rather an
accepted argument by those seeking to have stricter immigration law
enforcement.
The data is a combination of
information pulled from Census.gov Annual American Community Survey Data and
FBI Annual Crime Reports about years 2010 through 2014. The data included
information on total population, naturalized foreign born population, non U.S
citizen population, unemployment rate, median wage, and total number of violent
crimes surveyed from 12 American cities. The cities were broken into two
separate groups, sanctuary cities: San Francisco, Los Angeles, Miami, New York
City, Salt Lake City, Seattle, Portland, and Boston. Non-sanctuary cities:
Phoenix, Atlanta, Birmingham, Indianapolis, Columbia, Dallas, Cleveland, and
Charlotte. It is important to note when considering the variables included in
my data set that it does not include information on the undocumented population
of these cities, for the reason that that group is by definition not documented.
The closest variable representing that information is those in the cities’
population who are not U.S citizens that could include but would not be limited
to the undocumented community.
It is also important to note that the violent crime data from the FBI is
raw data that defines violent crime as murder, non-negligent manslaughter,
rape, robbery, and aggravated assault. As with most social science research I
will be using a .05 significance level.
To answer my
research question I ran three separate statistical tests, an independent T
test, and two different multi variable regressions with different dependant
variables, VPCBENJAMIN and the raw violent crime rate. Ultimately, the model
that proved most accurate was the multi variable regression with the raw
violent crime rate as the dependent variable.
Once that data set
was loaded on to SSPS I recoded violent crime into a new variable VPCBENJAMIN
by dividing the violent crime rate by population/100,000. This allows us to
view violent crime rate per one hundred thousand. This also allows us to compare groups of city populations
based on whether or not they are sanctuary cities despite the differences in
population between the 12 cities. The sanctuary status was defined by the
values 1 being a sanctuary and 0 being a non-sanctuary city. By doing all this
I was able to run an independent T test of the test groups’ sanctuary status
with the recoded variable VPCBENJAMIN. What it found was that there was
statistically significantly lower crime rate in sanctuary cities versus
non-sanctuary cities by about 308.3134. The significance level was .000 meaning that the difference
between the means of these groups is very significant. Since this paper is supposed to be a
more in depth project, it seemed necessary to run other tests including a
multivariable regression to see how foreign populations specifically impact
violent crime rates.
Then, I ran a
multi variable regression with the same recoded dependant variable, with the
predictors population, unemployment, median wage, sanctuary status, naturalized
foreign born citizens, and not US citizens. Non US Citizens and Unemployment
came back as statistically insignificant at the .05 level. Only Naturalized US Citizens had a
positive coefficient. This meant
that as the Naturalized US Citizen population increased, there would be an
increase in violent crimes. For
each Naturalized US citizen added to a city, the violent crime rate per 100,000
would increase by .003. However,
the adjusted R square for this model was only .673 which means the model
overall is not very accurate at explaining the violent crime rate in a city even
with the amount of predictors.
Finally, I ran
another multi variable regression with the same predictors as the previous
regression, but with a different dependent variable, the raw violent crime
rate. I did this because when I
recoded the violent crimes rate to per 100,000, this reduced the amount of
variation in the dependent variable and therefore reduced the statistical
significance of individual independent variables. Running the regression with raw crime rates and population in
the independent variable section should allow for greater significance and a
higher R squared for the model.
The regression showed that now all explanatory variables were
significant at the .05 level. This
revealed that the population and naturalized foreign-born citizens had a
positive coefficient, while all other variables had negative coefficients. Since this model had an adjusted R
square of .986, this seems like the most accurate model to help address the
research question of there is a difference in violent crime rates between
sanctuary cities and non-sanctuary cities. Looking at the coefficients a most
seem to be consistent with common knowledge. For example, as median wage increases, the violent crime
rate decreases. Similarly, as
population increases, so do the violent crimes. Most importantly for this research question though, was that
the coefficient on sanctuary cities was -1393.061 which means that we could
predict a sanctuary city to on average have 1393 less violent crimes per year
compared to a non sanctuary city.
This finding is consistent with the independent variable T test. Given the results of both the T test
and multivariable regression I am forced to reject the null hypothesis that there
is no difference in violent crime rates between sanctuary cities and
non-sanctuary cities. There is
clearly a difference between sanctuary cities and non-sanctuary cities in
violent crime rates. The findings
of this research paper are contrary to popular belief that sanctuary cities are
more dangerous than non-sanctuary cities.
This is relevant to public administrators who consider the political
risks of embracing sanctuary policies. They can take comfort in knowing that a
city can be compassionate without putting its residents in danger of suffering
the effects of violent crime rates.
Admittedly, one
could critic this analysis by saying my sample size was not large enough, or
that I cannot safely make inferences about the relationship between a rates of
undocumented immigrants and violent crime rates seeing as how I have no
variable to directly represent the undocumented population. However, there is
research on that information out there, only it can not be applied to my data
set because that information is statewide data, whereas mine is city data. As I
said before even though the shooting of Kathryn Steinle was accidental, and her
shooter had been deported five times, the cognitive dissidence of those like
Donald Trump is what drives them to make a wrong interpretation of the
direction of violent crimes in relations to a city’s sanctuary policies. They will always maintain that the
undocumented are the cause of the broken immigration system, and not 85 years
of contradictory immigration policies designed to provide the agricultural
industry with cheaper labor that render the undocumented as merely pairs of
arms for labor. One could collect my same set of data from the next five-year
period in 2020 (2015-2019) and compare the coefficients and statistical
significance to measure my accuracy. Finally just out of curiosity, I did a
independent T test of the dependant variable being unemployment, with the test
groups being sanctuary status and found extreme statistical insignificance of
.848 meaning that there isn’t any real statistically significant relationship
between unemployment and a cities’ sanctuary policies.
