Infographic: do more guns = more freedom?
Let me preface this very controversial topic by saying that this is merely an exploratory project that I took on as part of a project for a data science course. The question I posed in the title is something that has personally been on my mind ever since I moved to the US from Europe and I have struggled to understand America’s obsession with guns.
As a somewhat sheltered scientist who came to discover American history prohibitively late in their life, I have resorted to what I know best (numbers and graphs!) to provide some data-driven insights for myself and potentially other confused Europeans like me.
The questions and exploratory data analysis in this post are designed to place the US on a global map in terms of firearm ownership, firearm-related deaths and police-caused fatalities, and look for potential trends in how these are related to personal freedom across countries ranking high, medium and low on a Personal Freedom score scale.
The data
The data used here comes from several sources:
- IHME-GBD data set — generated using the GBD Results tool. Contains firearm-related, police-related and other death rates per country, per year for years 2013–2019.
- global gun ownership —data from the Small Arms Survey (2017)
- freedom index — data from the 2020 report of the Cato Institute, downloaded from Kaggle https://www.kaggle.com/gsutters/the-human-freedom-index?select=hfi_cc_2020.csv (most recent year available in the data set is 2017)
The data has been transformed for the purposes of this project and provides per-country statistics for 2017 on the following features:
- firearm-ownership_rate — firearm ownership per 100,000 citizens
- violence-firearm_rate — firearm fatalities per 100,000 citizens
- police_rate — police fatalitices per 100,000 citezens
- pf_score — Personal Freedom (PF) score (0–10)
- ef_score — Economic Freedom (EF) score (0–10)
- hf_score — Human Freedom (EF) score (0–10)
Do more guns = more gun/police fatalities?
Before we go into the somewhat abstract definition of freedom, let’s try to answer a simpler question: “Do countries with high rates of firearm ownership show higher rates of firearm and police fatalities?”.
To answer this question, let’s plot the per-country data on firearm ownership, firearm-related deaths and police fatalities, ordered by firearm ownership rates. If any trends exist between the rate of firearm ownership and firearm/police fatalities, we should see all bars corresponding to each country generally decrease from top to bottom.
At this point, you may be wondering why I have decided to include police fatalities in this analysis. This is again motivated by the somewhat peculiar situation in the US, where often you will hear that police officers fatally shot someone in self-defense because they mistakenly or correctly assumed the victim had a gun. So, temporarily ignoring all implicit biases that go into this assumption, I couldn’t help but wonder if the high rate of firearm ownership in other countries contributes to police officers making these assumptions which would show as higher rates of police-caused fatalities.
Here’s the plot:
There are several things to note here:
- There are no distinct trends in the firearm and police deaths depicted above. Many of the high-ownership rate countries report significantly lower firearm and police deaths compared to countries with lower firearm ownership rates.
- The countries ranking high on firearm ownership are very diverse: some are peaceful countries that are rarely involved in international conflicts (Iceland, Switzerland), some have had conflicts on their territory in the last 30 years (Serbia, Montenegro, Bosnia & Herzegovina, Lebanon) and some have ongoing wars on their territory (Saudi Arabia, Yemen).
- A significant portion of the low-ownership ranking countries show rates of firearm and police deaths comparable to or higher than countries with much higher firearm ownership.
What does this tell us? Nothing unexpected, honestly. The world is an incredibly diverse place and each country has its own laws, politics, economy, culture, social dynamics and stability that drive these numbers much more than the simple rate of firearm ownership.
So let’s try to take some of that information into account through the freedom score!
Does personal freedom make a difference?
The Personal Freedom (PF) score used here was taken from the Cato Institute Freedom Index report, which is calculated using data on rule of law, security and safety, movement, religion, association and assembly, expression and information, identity and relationships.
It is separate from the Economic Freedom (EF) score, and both of them make up equal parts of the overall Human Freedom (HF) score. I have chosen to use PF over EF or HF, mostly because the justification for the high rate of firearm ownership in the US is closely related to individual liberty and prevention of “an over-reaching federal government”.
The graph of PF score as a function of firearm fatalities shows the majority of the countries show a rate of < 50/100,000 fatalities regardless of PF score. A notable exceptions are countries with PF scores between 6 and 8, Venezuela and El Salvador as the ones ranking highest in firearm fatalities, and the United States, which displays high rate of firearm fatalities compared to other countries with similar PF score, but also a high rate of police fatalities.
Let’s also look at the PF score as a function of police fatalities. Here we can see that lower PF score countries record significantly higher police fatality rates, with the exception of the United States, which records the highest rate of police fatalities for all countries with PF score > 8.
Note also that in both of the graphs above, the firearm ownership rate (depicted in the point sizes) does not play a significant role in the distribution of countries across the plots.
Overall, firearm and police fatalities do not show a 1:1 relationship with PF score, but the above graphs indicate three categories that we can group countries in:
- high PF (PF > 8) countries — distinguished by (on average) high firearm ownership, low rate of police fatalities and low rate of firearm fatalities.
- medium PF (6 < PF < 8) countries — distinguished by high rate of firearm fatalities, intermediate rate of police fatalities and a range of firearm ownership rates.
- low PF (PF < 6) countries — distinguished by a wide range of very low and very high firearm ownership rates, high rate of police fatalities and (on average) low rate of firearm fatalities.
Let’s use this categorization for a more in-depth look into the ultimate question.
Do more guns = more freedom?
To attempt to answer this question, let’s compute the correlation coefficients (CC) between the variables in our data set.
Statistics refresher: CC of 0 means there is no correlation, a positive CC indicates correlation (with CC=+1 being perfect correlation) and a negative CC indicates anti-correlation (with CC=-1 being perfect anti-correlation).
There are different correlation coefficients one can use. For this analysis we’ll use the most common Pearson CC (or r) which measures if the two values follow a linear relationship; and the Spearman rank coefficient, which measures if two values follow a monotonic (not necessarily linear!) relationship.
The Pearson method is more sensitive to outliers than Spearman, which is why for the global and group analysis we’ll use the Spearman CC (we care about overall trends and would like to decrease the effect of outliers!).
The global map of correlation for our variables has several key features:
- all freedom index scores show high positive correlation, which is no surprise. PF and EF both go into the value of HF, while it is reasonable to expect that a country with high economic freedom will have high personal freedom and vice versa (with some exceptions);
- there’s a strong negative correlation between the rate of police fatalities and all freedom indices;
- to a lesser extent, the freedom indices are positively correlated with firearm ownership rates and negatively correlated with firearm fatalities.
Let’s break down the correlation of firearm ownership rate with the other variables by the three PF-based groups we introduced earlier:
One interesting thing that we can note in the above plot is that there seems to be an almost inverse trend between the high and low PF countries, in particular in terms of the PF score. Firearm ownership is somewhat correlated with PF for high PF countries (correlation coefficient ~ 0.3), while it is anti-correlated to a similar extent for low PF countries (correlation coefficient ~ -0.3). This indicates that gun ownership may be related to personal freedom only in more developed countries.
The other interesting observation comes in the rates of firearm and police fatalities between different groups. For both high and medium PF countries, firearm ownership and violence are somewhat correlated, while police fatalities less so. The situation in low PF countries is the opposite: there is no correlation between firearm ownership and violence, but police fatalities show a higher correlation coefficient compared to the high and medium PF group.
Ok, but what about the United States?
Let’s look at the high PF group in more detail now to determine its overall trends. Instead of just using a correlation coefficient, we’ll look at a correlation plot.
Each panel of the correlation plot depicts a scatter plot cross section of two variables, with histograms of each variable across the diagonal panels. A linear fit of each cross section is represented by a straight line, while the uncertainty of the fit is given as a shaded area around the line.
We immediately notice that there is one strong outlier in all panels of the correlation plot above. And we already know (from the earlier scatter plots) that the point corresponds to the US. If we look at the resulting linear relationships it’s immediately evident that the outlier is driving the overall positive slope, which would result in a high positive Pearson coefficient. So what happens to the Pearson coefficient if we remove the US from the data set? Here’s the result:
That’s a significant change! Both the firearm and police fatalities show significant decrease in the correlation coefficient after removing the US! But the correlation coefficient between firearm ownership and personal freedom more than doubles.
So what can we conclude from this? — Compared to similarly developed countries, the high firearm ownership rate in the US seems to be related significantly to the high firearm violence and high police fatalities rates, but not as much to the personal freedom score!
Key takeaways
- There are no clear correlations between firearm ownership, firearm-caused deaths and police-caused deaths on a global level. This is understandable because a lot of these incidents are driven by the laws, politics, economy and stability within a country and these differ greatly on a global scale.
- Firearm-caused deaths show the widest range in medium-PF countries (6 < PF score < 8), while police-caused deaths are highest for low-PF (PF score < 6) countries. This is logical if we take into account access to firearms and political instabilities in both of these groups.
- USA is a clear outlier in all of these categories within the high PF group. Its firarm-caused death rates are closer to the mid-PF countries (with much lower firearm ownership rates), while police-caused death rates fall on the boundary of the values spanned by mid-PF and low-PF countries.
- The firearm ownership rate in the US does not follow the trend of the rest of the countries with high PF score. Its removal from the dataset significantly decreases the correlation between firearm ownership and firearm/police fatalities, but slightly increases the correlation between firearm ownership and personal freedom.
_______________________
A blog post is not nearly sufficient to cover all the bases and insights on a topic, so what I do regularly (sometimes regrettably) is read the comments for additional information or alternative interpretations. That said, if you have any thoughts or insights on this topic, additional data sources and features to look at — I implore you — comment on this post and help me out on this journey of self-education through data!
