**Update through 2013 available here.**

How many cops are there in the United States? In 2008, according to the Bureau of Justice Statistics, there were 765,246 full-time police officers in the United States — roughly 251 police per 100,000 residents.

Since then, we haven’t had a good official count, but extrapolating from the most recent Census Bureau data, I estimate that, as of 2014, there were 765,980 police in the United States, roughly 241 police per 100,000 residents. This estimate is probably a bit on the low side. Here’s how I got there.

**Many Data Sources, All of Them Flawed**

This problem is so vexing because the question is so simple, but the data are so difficult to find and interpret.

First, here’s a quick lay of the land on the data:

**FBI: **The FBI’s Uniform Crime Reports contain a figure for sworn officers reported by police to the FBI each year. That figure comes from state and local police departments that participate in the survey. The trouble with the FBI figure is that the share of police agencies that participate varies widely each year, from 60-80 percent, so it’s not terribly informative.

**BLS: **The Bureau of Labor Statistics has an estimate for the number of “police and sheriff’s patrol officers,” which is a strange and arbitrary category. Even worse, the BLS’s definition for police abruptly shifted in 2003 from “police and detectives, including supervisors” to “police and sheriff’s patrol officers,” a definitional change that instantly halved the number of recorded workers. This makes it useless to compare trends over time.

**NLEOMF: **The National Law Enforcement Officers Memorial Fund claims that there are “more than 900,000 sworn law enforcement officers now serving in the United States,” a number they’ve been throwing around for several years. Where did this figure come from? When I asked, NLEOMF said it was “a general estimate derived from discussions with various law enforcement agencies.” Unfortunately, “general estimates derived from discussions” sounds like we’re into “wild guess” territory.

**Census Bureau: **The US Census Bureau conducts an annual survey of state and local government employees (last released in 2014), which includes an estimate for the number of sworn police officers. Unfortunately, once again, participation varies, so it’s not immediately clear how closely the Census estimate tracks the true figure. (More on this shortly.)

**BJS: **The Bureau of Justice Statistics has the only accurate figures for the number of police officers in the United States. Every four years, from 1992 to 2008, BJS conducted a complete Census of State and Local Law Enforcement Agencies, recording the total number of sworn police officers. A reasonable bit of interpolation for the three years between censuses gives us a pretty good estimate for the true number of cops each year from 1992-2008. Unfortunately, the 2012 BJS census was chronically delayed until finally being cancelled. BJS then commissioned a census for 2014, but it too has been delayed several times and still hasn’t been released. So the last true figure we have for the number of cops in the United States is eight years old.

**Filling in the Gaps**

However, the Census Bureau has estimates for the number of police every year through 2014. The trouble with those figures is that we don’t know how closely they reflect the true figure. But because the BJS gives us the true figure (or something very close to it) for 1992-2008, we can compare the Census estimate to the BJS number for those years and see how closely they track each other.

The good news is that they track each very well, on average.

From 1992-2008, the Census Bureau estimate averaged 88.6 percent of the true figure, with a standard deviation of 1.6 percentage points. Even better, the Census estimate got closer to the true figure over time, and the variation decreased: the average since 2000 was 89.6 percent, with a standard deviation of only 0.7 percentage points.

Using this relationship, we can project, based on the Census numbers, what the true figures would be for 2009-2014. We also want to see how well this average projection matches up to known figures, so here it is extended into the past, as well as the future. Using the Census numbers and an average relationship of 88.6 percent, we get a projection that looks like this:

This chart strongly suggests that the number of full-time police officers has been flat or falling. This projection gives us an estimate of 765,980 police officers in 2014.

**It’s Always More Complicated…**

However, there’s a bit more to the story. Since 2007, the Census Bureau has reported a “response rate” for data on full-time employment from state and local governments; that is, how many state and local government entities they surveyed reported the full data. Basically, the response rate tells you how complete their data is for a given year.

From 2007-2011, the response rate averaged about 88 percent, but from 2012-2014, the rate was only about 81 percent.

Year | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

Census response rate | 86.9% | 90.6% | 86.9% | 89.5% | 87.1% | 81.5% | 79.6% | 81.9% |

Share of officers found by Census | 89.5% | 90.5% | – | – | – | – | – | – |

This suggests that part of the steep decline in the number of police officers was due to lower reporting in 2012-2014. Unfortunately, we don’t know what the response rate was in prior years, so we can’t see if this kind of drop in reporting is a common variation or how it correlates with changes in the share of police found by the Census.

Because the BJS numbers end in 2008, we have only have two years of data to compare, but between 2007-2008, the Census response rate increased by 3.7 percentage points, which translated into only a 1 percentage point increase in the share of police officers found by the Census (from 89.5 to 90.5 percent). But we don’t know if this is typical or representative.

Another point to consider is that the *projected* number of police officers for 2007-2008 overestimated the true figures. So it’s possible that when the response rate is that high (87-90 percent), the true figure is lower than our formula projects. This would imply that the projected highs of 2009-2011 (when the response rate was high) are somewhat overestimating the true figure, while the projected lows of 2012-2014 (with a lower response rate) are somewhat underestimating the true figure.

The truth is probably in the middle. If we fit a polynomial trendline onto the true BJS data, we get a very good fit (R² = 0.9967), and when we project it forward, it just so happens to fall right between the highs of 2009-2011 and the lows of 2012-2014 — and close to a standard deviation of both. Just like Goldilocks.

All of this strongly points towards a flat or possibly even declining trend for the number of police officers in the United States.

When you look at the number of officers per capita, using these data, you can see that police per 100,000 residents is also likely in decline.

*R² = 0.9618*

If this is true and the trend continues, it might well be cause for concern, because, all else equal, more cops means less crime.