Charles Blow in the New York Times last year made the very common argument: “America has the highest gun homicide rate, the highest number of guns per capita . . . .” In another story, the New York Times quotes researcher David Hemenway as claiming: “Generally, if you live in a civilized society, more guns mean more death.” CNN’s Piers Morgan believes: “America has the worst incidents of gun murders of any of what they call the civilized world.” Bloomberg’s Businessweek also made similar claims this spring. The one common feature for these claims is that they rely on the Small Arms Survey.
So how do homicide rates compare across countries? (Click on figures to make them larger. It is necessary to greatly enlarge the picture to read the names of individual countries. UNODC data. Here it is as an Excel file.)
Much of the debate is focused on gun ownership rate data for 109 countries from the Small Arms Survey. There are real problems with this survey. For example, the rates of gun ownership for Israel (7 per 100 people) and Switzerland (supposedly 47 guns per 100 people) . Anyone who has ever been to Israel knowns that this estimate is ridiculously low. Indeed, over time about 12 to 15 percent of the adult Jewish population in Israel is allowed to carry handguns in public.
The problem with this survey excludes weapons that are technically owned by the government. The vast majority of guns in Israel are technically owned by the government, but if people have possession of guns in their homes for decades, the issue should be that public possession, not who technically owned the guns. Similarly, at that time of the Small Arms Survey, all able bodied Swiss males between the age of 18 and 42 kept their military weapons in their homes. After age 42, they could apply for permission to continue to keep their military weapons. Israeli guns are also excluded for the same reason.
The Small Arms Survey claims that the United States has by far the highest level of gun ownership, with 88.8 guns per 100 people. Both Israel and Switzerland probably have much higher gun ownership rates, but including them the way the Small Arms Survey does biases the results to The US gun ownership is so high compared to other countries that it drives any regression results.
There are also other problems with the survey. For example, a much better measure of gun ownership would be the percentage of the population owning guns, and not the number of guns per 100 people as used by the Small Arms Survey. Presumably, the issue is whether people have access to guns, not the number of guns greater than one that an individual has access to.
In addition, what most people don’t understand is that homicides are not the same as murders. Homicides add together murders and justifiable homicides (when a police officer or a civilian kills someone in self-defense). The homicide rates are likely underestimated in a number of countries, particularly the worse ones. In Venezuela, in 2015, the government claims that there were 18,000 murders, while the independent Venezuelan Violence Observatory says the number is almost 28,000.
Finally, murder isn’t a nationwide problem in the United States; it’s a problem in a very small set of urban areas. In 2014, the worst 2 percent of counties accounted for 51 percent of the murders. 5 percent of counties account for 68 percent of the murders. Yet, even within these counties with all these murders, there are large areas without any murders.
Looking at all countries for which the Small Arms Survey measured gun ownership, and using the Small Arms Survey data the way that it measures gun ownership, implies that more guns equals fewer homicides.
Usually only a small set of countries are used in any comparison, typically limited to so-called “civilized,” as Hemenway or Morgan calls them, or “developed” countries. It isn’t clear what is meant by “civilized” countries, so what can Americans learn from these other “developed” nations? Using the developed nations as defined by the Organisation for Economic Co-operation and Development (OECD), developed countries in fact show that more gun ownership as measured by the Small Arms Survey is associated with fewer homicides. First, this is how homicide rates vary across developed countries.
The relationship between homicide rates and the supposed measure of gun ownership provided the Small Arms Survey shows that even with their obviously biased measure of gun ownership, more guns ownership is associated with fewer homicides, though the relationship is not statistically significant.
Because the US is claimed to be such an outlier, it makes the relationship between gun ownership and homicides less negative than it actually is. (Regressions fit the regression line to “minimize the sum of the squared errors” so you can see how much extra weight one extreme value is given.) But so what can Americans learn from these other “developed” nations?
Yet, even though the cross-country data implies that more guns equals fewer homicides, this type of comparison isn’t very convincing. There is a real problem in using cross-sectional data. Suppose for the sake of argument that high-crime countries are the ones that most frequently adopt the most stringent gun control laws. What if gun control actually lowered crime, but not by enough to reduce rates to the same low levels prevailing in the majority of countries that did not adopt the laws. Looking across countries, it would then falsely appear that stricter gun control resulted in higher crime. Economists refer to this as an “endogeniety” problem. The adoption of the policy is a reaction to other events (that is, “endogenous”), in this case crime. To resolve this, one must examine how the high-crime areas that chose to adopt the controls changed over time —not only relative to their own past levels but also relative to areas that did not institute such controls. Below is part of a long discussion in The Bias Against Guns, Chp. 5 (More Guns, Less Crime also has a long discussion in Chp. 2).
Unfortunately, many contemporary discussions rely on misinterpretations of cross-sectional data. The New York Times recently conducted a cross-sectional study of murder rates in states with and without the death penalty, and found that “Indeed, 10 of the 12 states without capital punishment have homicide rates below the national average, Federal Bureau of Investigation data shows, while half the states with the death penalty have homicide rates above the national average” (Raymond Bonner and Ford Fessenden, “States With No Death Penalty Share Lower Homicide Rates,” New York Times, September 22, 2000, p. A1.). However, they erroneously concluded that the death penalty did not deter murder. The problem is that the states without the death penalty (Alaska, Hawaii, Iowa, Maine, Massachusetts, Michigan, Minnesota, North Dakota, Rhode Island, West Virginia, Wisconsin, and Vermont) have long enjoyed relatively low murder rates, something that might well have more to do with other factors than the death penalty. Instead one must compare, over time, how murder rates change in the two groups – those adopting the death penalty and those that did not.
It is because of this concern that we also provide another post that looks at crime rates before and after regulations such as bans.
Finally, as an aside, one has to be very careful in making comparisons across countries because numbers are not always comparable. For example, homicides in England and Wales are not counted the same as in other countries. Their homicide numbers typically “exclude any cases which do not result in conviction, or where the person is not prosecuted on grounds of self defence or otherwise” (Report to Parliament). A more detailed discussion of the difference between “offenses initially recorded as homicide” and “offenses currently recorded as homicide” in England and Wales based on the outcomes of trials is available starting on page 9 here. While this adjustment reduces overall homicides by about 15 percent, it has a larger impact on firearm homicides because those tend to be the ones most likely to involve gang fights that are much more difficult to solve. The problem isn’t just that it reduces the recorded homicide rate in England and Wales, but there would be a sizable reduction in the reported US homicide rate if this approach were used here. For example, from 2000 to 2008 only about 62 percent of US homicides are even cleared by arrest. The numbers in the UK appear to be only adjusted based on cases where charges are brought. In that case, it is useful to note that in the US only about half of those arrested are eventually convicted (also here).
A time-series discussion of how crime rates change before and after gun bans is available here.
Other Countries bias down their homicide numbers
You also need to be very careful before relying too heavily on homicide rates in other countries. If the Unites States is relatively more accurate in measuring its homicide rate and other countries try to hide their rates, it will look make it look like the US has a relatively higher rate than it actually does. Take two examples.
Argentina — Some countries might deliberately mischaracterize homicides into another category.
The breakdown of official statistics also raises the critical question of whether the extent of youth homicides is being obscured. With gun suicides and accidental deaths separately categorized, many of the violent deaths involving a firearm that La Nacion reports are currently listed as “unknown intent” could be homicides. . . .
UK — Homicides in England and Wales are not counted the same as in other countries. Their homicide numbers “exclude any cases which do not result in conviction, or where the person is not prosecuted on grounds of self defence or otherwise” (Report to Parliament). The problem isn’t just that it reduces the recorded homicide rate in England and Wales, but what would a similar reduction mean for the US.
If taken literally, and there is significant evidence that in practice the actual adjustment is no where near this large, a simple comparison can be made. In 2012, the US murder rate was 4.7 per 100,000, a total of 14,827. Arrests amounted to only 7,133. Using only people who were arrested (not just convicted) would lower the US murder rate to 2.26 per 100,000. More information on the adjustment for England and Wales is available here and it suggests that while many homicides are excluded it isn’t as large as it would appear (in 1997, the downward adjustment would be about 12 percent).
Many gun control advocates prefer to look at only firearm homicides, not total murders. The United States has neither the highest firearm homicide rates for all countries or for developed countries. Among OECD countries, Mexico has the highest firearms homicide rate, with a rate about 3 times higher than the US rate. Brazil’s and Russia’s are much higher, though Russia does not report firearm homicides so it is only a guess for that country.
By the way, despite Israel and Switzerland having very high gun possession rates, their firearm homicide rates are extremely low. In the data shown below, Switzerland had a firearms homicide rate of 0.77 per 100,000 people and Israel has a rate of just 0.09 per 100,000.
Note that there are many countries that clearly have higher gun homicide rates than the United States that don’t have data available. Indeed, while 192 countries report total homicides, only 116 countries report firearm homicides. The average homicide rate for the countries that don’t have firearm homicides is 11.1 per 100,000. The median homicide rate for those that are missing is 8.7 per 100,000. Among the countries with higher homicide rates is Russia with a homicide rate of 11.6. The bottom line is that the countries that are missing the data are among the worst homicide countries.
Again, all the concerns provided over relying on cross-sectional data still apply here. In addition, the firearm homicide data is not available for many of the countries with the highest homicide rates, suggesting that this cross-sectional comparison is even much more misleading than the discussion on homicides. Click on figures to enlarge.
The UNODC homicide report from 2013 is available here.
Regression estimates for developed countries.