Documente Academic
Documente Profesional
Documente Cultură
HOW DOES LEGAL IMMIGRATION EFFECT VIOLENT CRIME RATE IN THE UNITED
STATES?
Jordyn Horvath
Abstract
This study explored the impact of legal immigration populations in each state and violent crime
rate’s in each state. Using data from census.gov and AmericanFactFinder.com, I composed full
population, legal immigration, percent of people below the poverty line, and violent crime rate
percentage for each state from 2010 to 2017. Monitoring for regional and population differences,
immigration population and Y being crime rate per 100,000 people. I hypothesize that there will
be no effect on violent crime rate because of immigration population. My results were there is a
Keywords: Immigration, population, crime rate, violent, legal, region, state, year
The Effect of Legal Immigration on Violent Crime Rates 3
1. Introduction
In the United States, population has grown every year, and keeps growing. With with legal
immigration, it has fluctuated greatly from decade to decade. According to the U.S. census,
during 1880 and the 1920’s there was an annual rate of 600,000 people immigrating to the
United States. (Immigrationeis.org) The United States called this immigration fluctuation period
the “The Great Wave”, but at the time it was not as directly frowned upon by many people, or it
wasn’t criticized as often by the news or politicians. There was a huge need for factory workers
at this time and this position was quickly filled by European immigrants.
As time has gone on, there has been an overall increase in immigration to the United States.
Not for work but this time for the idea of the “American Dream”. I have always been very
fascinated with why people think America is so much better than their own countries, and this
I have also been intrigued most of my life by crime, whether in books, movies, or television
shows, it is something I enjoy learning about. This is how I got the idea to see if there is any
correlation between crime and immigration in the USA. Right away, when looking over previous
studies on the topic, I was able to see a trend in no correlation or a negative correlation between
My research question is “How has legal immigration effected violent crime rate in the United
States?” To conduct this study, I examined the population, legal immigrant population, violent
crime rates per 100,000 people, and people below poverty in each state and how they have
changed over the past eight years. Then, I examined the differences in regions and states to see if
2. Literature Review
In the summer of 2007 immigration rates completely spiked and the study also found a
spike in crime as well. It was found that third generation immigrants had a larger likelihood of
committing a crime than a first generation immigrant, the researcher found this because he
believed it was likely first generation immigrants were more afraid of being caught by
authorities. (Sampson. Robert J. 2008). Many studies from this time looked over crime and
immigration to find relationships between the two. “In today’s society, then, I would hypothesize
that immigration and the increasing cultural diversity that accompanies it generate the sort of
conflicts of culture that lead not to increased crime but nearly the opposite.” (Sampson).
Often people stake claim and assumptions on the idea of crime and immigration being
related, looking into large metro areas and crime rates it was found that the immigration vs.
crime question is difficult to answer and there is so many for questions to be asked. (Williams.
2002) In most of the studies, there is no definite answer to be had, because of the majority of the
“Exploring the Connection between Immigration and Violent Crime Rates in U.S. Cities
1980-2000” by Graham C. Ousey and Charis E. Kubrin. Like the previous study there was many
inputs on public opinion on whether or not the immigration increase has anything to do with the
increase or decrease in crime rates. That being said this basically had the same question as the
studies prior to this one. The data this study collected was focused around the Cities in the US
The cities studied had to have a minimum population of 100,000 people and in the united
states during this time period 173 cities met this-criteria. The study identifies the dependent
variable as being the violent crime rate and the independent variable being the percentage of the
The Effect of Legal Immigration on Violent Crime Rates 5
population made up of foreign-born humans who have immigrated in the past decade. They also
identified the second independent variable as being “Measure of linguistic isolation” they talked
about including people who speak English “Not Well” and “Not at All”.”. The study also
discusses the analytic strategy and how there may be some fixed effect linear regression models
in the data. All of the studies or literature I found, had the same or similar results, finding no
negative correlation between immigration and crime rates in the United States.
The Effect of Legal Immigration on Violent Crime Rates 6
3. Data
My research tests whether legal immigration rates have a positive or negative effect on
violent crime rates. I hypothesize that there is a negative or no correlation between the two. The
reasoning for that hypothesis comes from previous studies and existing literature and the
relationship between immigration and crime rates in the United States. States with a higher
immigration population have been speculated to have a higher crime rate. This is something
almost assumed by large numbers of people. The data looks at eight years in all 50 states,
spreading from the years 2010 to 2017. I chose these years because they were the most recent
and after the year 2007 there was a large increase in immigration again and I thought it would
show clearly within the data I used. The data also covers the population, immigration population,
amount of the states below poverty level and violent crime rate for each of those years as well as
what regions each of the states are in, in the United States. My simple equation will look like
“catch all” for all other significant actors that you haven’t included in the equation. I originally
wanted to look at nonviolent crimes but figured with violent I would be able to have a more
robust finding within my data. I got almost all of my data for population, poverty levels and
immigration population from census.gov, I found and analyzed all of my crime rate reports from
the website AmericasHealthRankings.org, this website proved super helpful in covering crime
Variable Descriptions
ImmPOPper: Foreign born citizens in each state for each year covered
W: West region
S: South region
Summary Statistics
The standard deviations for population, immigration population, below poverty level, and
lnpopulation, and crime rate per 100,00 people are very high. This means that those three
variables are more widely spread and more reliable because of the larger range. The other four
variables have significantly lowers standard deviations from the first three, meaning that they
4. Analysis
Equation 1.
Yi(CrimeRate)=B0+B1X1i(ImmPOPper)+Ui
The null hypothesis for this model was that immigration population doesn’t have a direct
effect on the violent crime rate. I am rejecting the null hypothesis because the p-value is low and
the t-stat is high. The immigration population had a t-stat of 3.83 which is over the 1.96 at 95%
level, making this statistically significant. My results say that an increase in immigration causes
an increase in crime. This is clearly a result of OVB (omitted variable bias), because many things
can also effect the violent crime rate. For example, both crime rate and immigration are
increasing in the United States over time which is very obvious throughout my dataset.
Interpretation:
A 1 unit change in immigration population is associated with a 4.11 change in crime rate
Equation 2.
Yi(CrimeRate)=B0+B1X1i(ImmPOPPer)+B2x2i(povLevel)+Ui
The null hypothesis for this model was that immigration population doesn’t have a direct
effect on the violent crime rate. I am rejecting the null hypothesis because the p-value is low and
the t-stat is high. The results show that 2 of the variables are above the T-stat of 1.96 at the 95%
except for violent crime rate. The Coefficient also went down. I have included an extra X
variable(povLevel) to see if that would have an effect on my Y, which helps because now my
entire Coefficient isn’t being absorbed for immigration rate. My results say that an increase in
immigration causes an increase in crime. I am also rejecting the null hypothesis that poverty
level has no effect on violent crime; but have found that higher poverty means higher crime.
Interpretation:
A 1 unit change in immigration population is associated with a 5.92 change in crime rate
per 100,000 people. A 1 unit change in percent of people below the poverty level is associated
Equation 3.
Yi(CrimRate)=B0+B1X1i(ImmigrationPop)+B2X2i(BelowPovertyLevel)+
The null hypothesis for this model was that immigration population doesn’t have a direct
effect on the violent crime rate. I am rejecting the null hypothesis because the p-value is low and
the t-stat is high. The results show that 5 of the variables are above the T-stat of 1.96 at the 95%
except for violent crime rate. I have included an extra X variable(povLevel) to see if that would
have an effect on my Y, which helps because now my entire Coefficient isn’t being absorbed for
immigration rate. I have also included three control proxies () in my regression equation to test
if regions would have an effect on my regression equation. My results show that I am rejecting
the null hypothesis that poverty level and immigration population have no effect on violent
crime. They actually have a positive effect on violent crime, and I can confidently say my data
are robust. The added regions to this regression equation show that the Midwest, West, and
The Effect of Legal Immigration on Violent Crime Rates 12
South are higher in violent crime in comparison to the Northeast. Even though all of my
regression equations, show that there is a positive correlation with the X variables and the
Violent Crime Rate, I still have a lot of OVB out there to be discovered.
Intrepretations:
A 1 unit change in immigration population is associated with a 6.96 change in crime rate
A 1 unit change in percent of people below the poverty level is associated with a 9.82
Robust Test
Coefficient
P-Value
The table above is a test for robustness among my data. Robustness is when your
empirical results are strong and consistent amongst a number of different models or sub samples.
To have a robust data set, you want to have the same sign (+/-) and same significance among
your P-Values and Coefficients. It is a model of my three regression equation models, and the
top line of each variable is the Coefficient and the bottom are the P-values. My results are telling
me that for immigration and percent of people below the poverty line they both are extremely
robust because of how low the P-Values are. In the model above, for example, immigration from
all three models are way smaller than a P-value of .05 at the 95% level, so we reject the null
hypothesis. And the same goes for povLevel, both of the P-Values in Models 2 and 3 are
Multicollinearity
X1i(Immpop)=B0+B2X2i(PovLevel)+B3X3i(lnpop)+Ui
Next, I wanted to test for perfect and imperfect multicollinearity(MC) because I wanted
to see if even though my data are robust and statistically significant if they had any
multicollinearity. MC is when two or more of your X’s are highly or perfectly correlated. This is
bad! There are two types of MC, perfect and imperfect. Perfect MC is when regressions fail
completely, when 1 regressor (X) is a perfect linear function of another regressor (X). This
makes it impossible to compute OLS (ordinary least squares). So, on a data set, perfect MC
would be where X1i and X2i are exactly the same variable. Imperfect MC is when two or more
X’s are highly but not perfectly correlated. Which means they may not produce the exact same
effect on Y but they can be very similar. For running a test for MC in my data, I pushed my X1i
variable which was immigration population to be my new Y independent variable and then my
The Effect of Legal Immigration on Violent Crime Rates 15
new primary X2i dependent variable is people below poverty level and my X3i dependent is the
population total percentage for each state. My data proved to not be Multicollinearity because
Next, I tested my data for VIF, which is a statistic that quantifies the level of inflated
variance of our estimates, as a result of multicollinearity. The way to test for VIF, is the equation
good! If this VIF is less than 10 there is no issues with imperfect or perfect MC. I ran this same
equation as having poverty level as my Y and then moving immigration population back into the
X section to test for the same results and I got the same results, under 10 so no issues with
Yi(crimerate)=B0X1i(immPOPper)+B2X2i(PovLevel)+B3ln(X3i)(lnPOP)+
The table above shows that I used the Linear Log model, which is when one’s Y
variable is not log transformed, but at least one of the X Variables are. I used the linerar log
model because in population there is such a large range, I thought talking in percentages would
make this easier. For my data, I added in another x variable, so I chose to add in total population
of the states, used the logging equaition (=ln(pop)) to do this, and then added it as my B3X3i
variable. Looking at the table above, I am able to reject the null hypothesis of immigration
having effects on violent crime rate, because all of the data except lnpop are above a T-Stat of
1.96 at the 95% level. A 1 % change in immigration population is associated with a .01(240.13)
unit change in crime rate per 100,000 people, holding all else constant.
The Effect of Legal Immigration on Violent Crime Rates 17
Yi(CrimeRate)=B0+B1X1i+B2X1i^2+B3X1i^3+B4X1i^4+B5X1i^5+Ui
I identified that in my scatter plot there was no linear relationship, and found at least five
inflection points, and decided to run a general non linear equation to see if making changes to my
X would effect Y. My primary goal of this was to find how the changes in my X’s(Population)
are affecting Y(CrimeRate). I chose total population with its large range for the X in my
equation, also because it is a continuous variable, and I ran it to the 2nd, 3rd, 4th and 5th power, to
be able to run the regression. I found that the increase in population has a nonlinear effect on
crime rates. So I performed sequential hypothesis testing and found the model that produced the
most precise estimates. Final thoughts on this are that my data shows that when population
grows, the effects just keep growing larger when population gets high!
The Effect of Legal Immigration on Violent Crime Rates 18
5. Results/Analysis
The data and statistical tests concluded that immigration population might have
some effect on violent crime rate per 100,000 people. However, overall population,
percent of people below poverty level, and which region the state was located in also
had some effects on violent crime rate as well. All of the variables I used had a
positive relationship with violent crime rate. So if there was a higher immigration
population, there was a strong chance there will be a higher percentage of violent
This was not orginially the forseen results because of the recent and modern
research and studies done on the topic. My orgininal hypothesis was that immigration
population had no effect on violent crime rate. When I added the other variables I was
able to see that overall population as well as percent of people below the poverty line
and these extra X variables also have a positive effect on violent crime rate, meaning
Now I understand that the with the population growing constantly this creates
growth in crime rate, I think that I still have some omitted variable bias in my study,
there is a lot of other factors I could have looked at for this study. I believe other
factors like violent crime rates committed by legal immigrants and not total
population would have a different effect then what my data are showing.
I believe this is such an important topic and there are so many new ways to
research and study it. If I were to further my research I would want to look at more
The Effect of Legal Immigration on Violent Crime Rates 19
specific areas, genders, ages and even education levels to see how those had an effect
on violent crime rates. In other literature there was a study where they examined
social structure and were able to find a negative correlation with that and crime rates, I
References
“Violent Crime Rates for Every State in America (Hover over States to Display
Rankings).” Violent Crime Statistics by State | Interactive Crime Map 2017, www.santarosa-
lawyer.com/united- states-crime-map-2017/.
Ousey, et al. “Exploring the Connection between Immigration and Violent Crime Rates in U.S. Cities, 1980–
2000 | Social Problems | Oxford Academic.” OUP Academic, Oxford University Press, 30 July 2014,
academic.oup.com/socpro/article/56/3/447/1707591/Exploring-the-Connection-between-
Immigration-and.
Sampson, Robert j. “Rethinking Crime and Immigration.” Sage Journals, 1 Feb. 2008,
www.bing.com/cr?IG=55F1644B639F4503BF025E9C241F9F03&CID=340780A3D9D366102C1C8B40
D87C6742&rd=1&h=U2HWQgIotIVQt4MsqrEwaiS0VtKUI6vgZWKeNGqtihs&v=1&r=http%3a%2f%
2fjournals.sagepub.com%2fcontent%2f7%2f1%2f28.full.pdf%2bhtml%3fijkey%3dNY%252FHB1XH
dlJDA%26keytype%3dref%26siteid%3dspctx&p=DevEx.LB.1,5514.1.
Data Access and Dissemination Systems (DADS). “Your Geography Selections.” American FactFinder - Results,
5 Oct. 2010,
factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid=PEP_2017_PEPANNRE
S&src=pt.