By the Numbers: Is the US Under-Policed?

Filed under: “more than you wanted to know.”

Alex Tabarrok had an interesting post about police vs. prisons at Marginal Revolution that I thought deserved some attention, given my long and interminable series on policing in the United States.

First, the basic economics of crime. There are essentially two ways for government to change the cost-benefit calculation for criminals: increase the probability of being caught (more police), or increase the severity of the punishment of when you are caught (more prison).

For several decades, the United States has chosen the more severe punishment path. Prison sentences are long and compounding, and today, the US has the largest prison population in the world and the most prisoners per capita. The Council of Economic Advisers (CEA) concludes:

Growth in U.S. incarceration has been fueled by criminal justice policies. If prison admission rates and average time served in prison remained the same as they were in 1984, research suggests that State imprisonment rates would have actually declined by 7 percent by 2004, given falling crime rates. Instead, State prison rates increased by over 125 percent.

prisoners per capita

Most economists and criminologists now agree that more severe punishment is a bad way to discourage crime. The same CEA report concludes,

Economic research suggests that longer sentence lengths have little deterrent impact on offenders. A recent paper estimates that a 10 percent increase in average sentence length corresponds to a zero to 0.5 percent decrease in arrest rates. …

Expanding resources for police has consistently been shown to reduce crime; estimates from economic research suggest that a 10 percent increase in police force size decreases crime by 3 to 10 percent.

The research on this is sound and compelling, and I don’t dispute it.

But there’s also an apparently shocking empirical claim in the CEA report about how the US compares internationally on investment in police vs. prisons. The report finds that the United States has way more prison guards per capita than the world average — not surprising since the US has way more prisoners per capita.

But the strangest finding, Tabarrok points out, “is that on a per-capita basis we employ 35% fewer police than the world average. That’s crazy.”

police etc

I admit, that is crazy. Tabarrok and the CEA both emphasize that simply having more police on payroll isn’t necessarily going to be an improvement, because cops also need good strategy, good training, and good relationships with the community to be effective. But the research shows, all else equal, more cops reduce crime significantly (at least violent crime).

So it is incredible that the US would be so under-investing in police. The United States is the richest big country in the world — comparatively well governed and also fairly violent. Could we really have a third fewer police per capita than the rest of the world? And what would it mean if we did?

I dug into this number a bit. The source for the CEA chart is from this UN report by Harrendorf, Heiskanen, and Malby (2010), which in turn uses data reported to the UN by each country, up through 2006.

#1. Don’t Do That

The first thing to note is that the UN report cautions about the problems of doing what the CEA does:

Attempting to measure the total police personnel with only one value, one has to keep in mind the shortcomings of such an approach…. Figures might include (or not include) data on criminal police, traffic police, border police, gendarmerie, uniformed police, city guard or municipal police, but also customs officers, tax police, military police, secret service police, police reserves, cadet police officers or court police.

Apart from this, the way of counting personnel might differ (e.g. heads vs. budget posts, which will make a difference when counting part-time personnel). Therefore, comparability could be considered fairly weak.

One cannot be sure that each and every country was able to exclude support staff from their data, because this would depend on the statistical possibility to do so. Also, it is not fully clear whether, apart from support staff, other civilians in the police force are included or only uniformed police are counted.

But the CEA report sweeps aside these concerns and simply makes a generic comparison between the world average of police per capita and the US figure. As I dug into the numbers, it became clear how problematic this is.

#2. The US Figure Is Off

The CEA figure depends on numbers from the UN report, which in turn depends on numbers reported to the UN from the FBI, which in turn depends on numbers reported to the FBI by various law enforcement agencies.

The US figure in the UN report comes from the FBI’s 2006 Uniform Crime Reports. In 2006, the FBI recorded 683,396 full-time police officers across the US. The UN took this figure and arrived at 224 police per 100,000 Americans, which is indeed 35% less than the world average of 342 per 100,000.

But crucially this number is not complete, as the FBI makes clear. Participation in the survey is voluntary, and typically only 75-80% of police agencies report data to the FBI each year.

Fortunately, the Bureau of Justice Statistics (BJS) conducted a complete census of police every four years from 1992 to 2008. In 2004, BJS found about 732,000 full-time state and local police; in 2008, 765,000 — about 251 officers per 100,000 in both years.

That’s 12.3% more than the UN figure, or only about 27% less than the world average.

#3. National Law Enforcement

Another thing to note is that the federal government has law enforcement agents that carry guns and make arrests. Federal agents are a significant share of law enforcement operations and resources in the United States, and federal prisoners were counted in the prison statistics. However, federal officers were not included in the FBI figure cited by the CEA.

Fortunately, the BJS also conducts a census of federal law enforcement officers. In 2004, there were 105,000 federal agents; in 2008, 120,000 federal agents. Let’s strike an average to get 2006 — say 112,500 federal agents. That’s an additional 37 federal officers per 100,000, bringing the US total to roughly 288 officers per 100,000. The true US number cannot be less than this.

That’s 28% more than the UN figure, or only about 15% less than the world average.

#4. Civilian Staff

In the US, as of 2008, one third (32%) of state and local law enforcement personnel were civilian staff, not “sworn officers” (police with arrest powers). The UN report notes that it’s not clear if other countries count or exclude civilian staff, but either way, civilians make a difference to the effective size of the police force. It’s not obvious whether it’s better to have cops doing their own paperwork, but in theory, more civilians increases the capacity of sworn officers to do actual police work.

While the number of sworn officers per capita has remained roughly constant since 1996, the number of civilian staff per capita grew steadily by 25% from 1996 to 2008.


By 2008, there were 373 officers and staff per 100,000. Throw in the feds, and we’re over 410 law enforcement per 100,000 — 20% greater than the world average.

To be clear, it is not clear if we should count these civilians or if other countries did — and that’s the point. This just establishes how large the margin of error could be here.

#5. Part-time Staff

As the UN report noted, it’s not clear if other countries included part-time employees. In the US, in 2008, BJS found roughly 100,000 part-time state and local law enforcement employees, including 44,000 sworn officers. Counting the part-time officers adds another 15 per 100k residents to our figure.

Putting it all together shows the range of possible figures that could have been reported for the US.

Again, it’s not clear if other countries report only full-time staff, full-time equivalent staff (two part-timers = one-full timer), or total staff regardless of hours worked.

But given how distorted and incomplete the US data was, we have no reason to think other countries are delivering immaculate figures.

#6. Outliers and the “World Average”

In Venezuela, according to UN data, there were only 16 police per 100,000 people; in Syria, only 10 per 100,000. The UN report calls attention to these two outliers as examples of data that is almost certainly wrong and incomplete. Even in 2006, these countries were not exemplars of law and order, but these figures are clearly not credible. Even Bangladesh and Kenya (the next lowest countries) reported 80 and 98 officers per 100,000, respectively.

On the high end, Bahrain (1,867 police per 100,000 residents), Brunei (1,087), Kuwait (1,065), Montenegro (891), and Mauritius (777) swing the world average massively upwards. If “world average” simply meant total cops divided by total populations, this wouldn’t matter. However, the world average is the average of all the countries‘ averages. So even though these are all tiny countries, their per capita rates are weighted equally in the world average.

It’s clear that 342 per 100,000 is not an accurate reflection of the global norm. The OECD average is 294. The world median is 303.

If you control for population — divide total cops by total population — the global mean is just 211 police per 100,000 people. This is the comparable figure for each individual nation’s average. By this measure, the United States has 36% more police than the world average.

#7. What Does It All Mean?

I dunno. Look at these averages by region, from the UN report:

police 100000

What does it mean that the US has (at least) 288 police per 100,000 people, compared to the Middle East average of 435? Or the Central Asian average of 326? What does it mean that Italy has 485, Germany 303, Canada 191? Is it bad? Good?

Some countries have a lot of police per capita because they are tiny, and you just need a certain absolute number of cops to keep order anywhere. Some combine their army, militia, and police as “security services.” Some have few police because they are very peaceful, some because they are too poor to afford them. Some countries have a lot of police because they are very violent; some are (arguably) peaceful because they have a lot of cops.

Some are corrupt and dangerous, some are safe and efficient. You can’t tell anything about how one country’s police force compares to any other — or all the others — just by looking at this figure. Which brings up the final point:

#8. The US Is Not a Monolith

Check out this chart from the 2008 police census, showing variations in police per capita:


Drilling down to cities: In Washington, DC, there are 612 police per 100,000 residents. In New York, 417. In Baltimore, 474. You can’t even tell very much about the US from its own national average, or about a state from its state average.

In conclusion, there’s good evidence that, all else equal, more cops means less crime, on the margin. Some places in America desperately need security, and more cops might make sense. (It certainly makes more sense than more incarceration, longer prison terms, and mandatory minimums, which increase recidivism and help shore up gang membership.)

But there’s no good reason to assume (based on this raw comparison) that the US is in general under-policed, compared to the rest of the world.

As always, if you spot an error in my data, please let me know, and I will publish a correction.