After all, do guns increase or decrease crime? Let’s see the data
Author: Carlos Goés
Originally Published at Peace and Conflict Monitor on 07/06/2016
Few subjects are as politicized as gun ownership and its relation with violence. Both sides of this issue – those who want to restrict access to guns and those who want to increase it – are endowed with reasonable arguments and specific data to sustain their arguments. Interestingly, this is also a contention that transcends the typical ideological barriers. In Brazil, for instance, far-right militarist Representative Jair Bolsonaro defends civilian firearms ownership, while the far-left Party of the Workers’ Cause argue that an armed population can help to combat crime. Similarly, we can find pro-gun control arguments on both the right and the left.
But who is the right one in this story? The most correct answer is that both are partially correct, but the conclusions of the two sides are wrong.
“More guns, more crime” or “more guns, less crime”?
You have probably heard both contradictory and reasonable arguments. Those in favor of disarmament say that, with less available guns, it will be more difficult for criminals to commit violent crimes. Those that support civilian gun ownership say that, with the low probability that civilians have guns; criminals will be more willing to attack them since the retaliation risk is low.
What the two sides do not notice is that, actually, these effects are not mutually exclusive. When rationalizing the incentives and disincentives for crime of a particular policy (something jurists and economists name “economic analysis of law”), a more neutral posture leads to the observation that gun availability actually have two simultaneous effects on violence:
- Promoting effect: Just like what happens with drugs, prohibiting gun market reduce the total supply of guns and, as a result, increase the price of guns in the illegal market. This does not mean that it will be impossible for criminals to buy guns – but it means that it will be more expensive – and the expected effect is that, because of the higher cost, criminals have less access to guns. In short, the violence promoting effect of guns is: if guns were legalized, the price of weapons would tend to decrease, their access to criminal would increase, and more violence would occur.
- Dissuasive effect: Criminals, like every other human being, are averse to risk. This does not mean that their risk aversion is the same as non-criminals, but simply that, everything else being constant, they prefer to take less risks – as an example, they would prefer that no one else but themselves had guns – and greater risks must be compensated with greater rewards. In short, the violence dissuasive effect of guns is: with legalized guns, the civilian population access to guns will be greater and the uncertainty on the civilian crime victims being armed will increase. Thereby, the risk to the criminal also increases and, being him averse to risks, he will also tend to commit fewer crimes.
Guns actually provoke these two contradictory effects. On the one hand, they increase crime, on the other, they decrease it. Once the existence of these two effects is admitted, what really matters, for the effects of public policy, is the liquid effect of guns:
Liquid effect of guns on violence = (Promoting effect) – (Dissuasive Effect) And what is the liquid empirical effect?
Phenomena like violence are complex. Just think on how there is a myriad of variables that could potentially interfere on violence: poverty, inequality, education, culture, racism, sexism, etc. Hence, isolating which is the specific effect of gun policies on this complex phenomenon is a difficult task – which does not mean that it cannot be done with the appropriate statistical tools.
Given the difficulty, even after we control all those factors, it is not surprising that there studies that found a positive correlation between gun access and gun-related deaths and others that say that more guns leads to less homicides. Many times, the conclusion is guided to which is the indicator that it is chosen to the analysis.
Look at the first two graphics, for example. They compare legislation restriction on guns with violent death rates in different American states. If the rate of gun-related deaths is used, it seems to be a relation between violence and a more restrictive legislation. But if the total homicide rate is used, the relation disappears.
How is this possible? The first element is that the first graphic includes suicides. Even though suicides are important, in qualitative and moral terms, they are different from the use of violence against other people. Therefrom, disarmament advocates usually talk about violence against third parties. Another possibility is the existence of a substitution: if potential murders do not have access to guns, they would simply look for other methods to commit crimes. With that, the total homicide rate would not be affected, but the gun homicide rate would.
When we observe contradictory results in serious studies on complex problems like this one, the best answers that we could find are on the “meta-analysis”. Those “meta-analysis” add several statistical studies and reach one single result. The idea is that, even though several studies could have problems, when aggregated, biases tend to disappear. The most important meta-analysis in this subject is a study by the U.S. National Academy of Sciences (NAS), of 2002. After reviewing hundreds of studies, the NAS concluded:
Despite a large body of research [on the subject], the committee found no credible evidence that the passage of right-to-carry laws decreases or increases violent crime (NAS, p. 2).
In other words, the liquid effect of easier access to guns is undistinguishable from zero (what, in statistical jargon, is called “statistically insignificant”).
Comparing homicide rates and gun ownership per capita in rich countries – those Organization for Economic Co-operation and Development member-states, that include Western Europe, United States, Canada, Japan, and some other countries such as Chile, Mexico, and Turkey –, the relation also seem to be close to zero. This is maintained even when statistical exceptions, which could distort the general result (as in Mexico, for example, that have extremely high homicide rates, but because of drug cartel, not because of guns), are excluded. See below.
The result contradictions and the absence of relation captured by the meta-analysis show one thing: you should be very skeptical in face of anyone that makes peremptory statements indicating that weapons reduce or increase violence. In this case, both gun rights and disarmament advocates are wrong. When hundreds of studies are aggregated and several factors are controlled, the liquid effect seems to be close to zero.
OK, what about Brazil?
Even though the best available studies on the subject focus in the United States, it is important to see what the data available in Brazil is. A good part of the researchers try to use the Disarmament Statute as a point of discontinuity in Brazil’s homicide rates. The idea is that, if all the other variables that determine homicides kept their previous tendency and only one thing changes (in this case, gun laws), this variable will explain any changes in the homicide rate. This is a reasonable strategy and, at first sight, the Disarmament Statute seems to have had an effect on the reduction of the gun homicides. See below.
Daniel Cerqueira, an economist who wrote a great doctoral thesis on the subject using data of municipalities from the State of São Paulo, found a similar relation. Is this enough to conclude that fewer guns mean fewer crimes? Maybe. The problem is that Cerqueira, when statistically calculating the relation between homicide rates and gun ownership rates (named “elasticity” in Economics jargon), ended up ignoring other important variables for homicide rates that passed through historical changes more or less at the same time that the homicide rates decreasing in Brazil: income inequality and poverty rates.
The graphic below show the relation between the deviations of gun homicides and income inequality (measured through the GINI index). Even though it is not possible to say there is causality between the two, there is a meaningful correlation between the two variables.
It is possible to do a similar analysis using data that compare poverty rates with the homicide rates. Poverty, which could certainly influence homicide rates, dropped to a historical low from 2003 on – that coincides with the promulgation of the Disarmament Statute. This further complicates the presumption that everything else was kept constant and that the cause of the drop in homicides must be explained by the Statute and derail the point of discontinuity strategy.
A good analyst would recognize that, in the same way that the hypothesis that the Disarmament Statute reduced criminality is reasonable, it is also reasonable to think that reduced income inequality or poverty lead to reduction of gun homicide rates. Exactly because this subject has not yet been studied as in depth in Brazil as it has been in the United States, this is still an open question. The skeptical position, which holds that there is not enough evidence for either side of this debate, still seems to be the most honest one.
In theory, guns could promote as well as dissuade crime. Thus, what matters for the analysis of gun access public policies is the liquid effect of this simultaneous relation of promotion and dissuasion. When aggregated in meta-analysis, the best studies on the subject – conducted in the U.S.A. – indicate that the liquid effect seems to be close to zero. In rich countries (OECD members), there is not even a correlation between homicides and gun ownership rates. In Brazil, the issue is still open to debate: it is possible that the Disarmament Statute has reduced gun homicides, but without considering other important variables – like the decrease in inequality and poverty – it is impossible to trust the results. In the end, both disarmament and gun rights advocates seem to be wrong: guns neither increase nor reduce crime. The available evidence indicates that, probably, they simply do not effect crime rates one way or the other.
Bio: Carlos Goés is an economics analyst with interests in econometrics, the economics of development, political philosophy, and anthropology. He completed his Masters in International Economy at Johns Hopkins University and his undergraduate in International Relations at the University of Brasília. Nevertheless, he guarantees that he learned a lot more paying library fines and having beers with friends than in the classrooms.