Thursday, December 10, 2009

How Safe Are Police Jobs?

How safe is it to be a police officer? [Update: the shorter version is here.]  I tried to answer this question based on the following charts--BLS Chart 1 and BLS Charts 2. I focused on injuries and fatalities.

Injuries: nationwide, there were 14,500 total recordable police injuries in 2008. Assuming the U.S. has between 435K and 800K police officers, officers have about a 3.3% chance of being injured each year. [Note: see end of post for updated stats on the number of local and state law enforcement.] The BLS's injury stats are similar to the FBI's stats (about 15,000 injuries in 2008); however, I am skeptical about relying on injury stats because it is unclear exactly what qualifies as a recordable injury. Different police departments probably report different kinds of injuries. While I don't think police officers are filing reports on paper cuts, I also don't think every recordable injury is necessarily serious. For me to rely on injury rates, I would want to see only serious injuries recorded, such as those that drew blood, led to an ER visit or sprain/fracture/break, or required physical rehabilitation or cortisone shots. I would also prefer to see injuries that occurred while an officer was patrolling a beat listed separately from other stats (I personally feel that patrol officers have dangerous jobs, and the data ought to differentiate between a desk job and a patrol job). Due to privacy rights, I don't think such detailed medical stats are available to the public.

Fatalities: as morbid as it sounds, dead bodies might be the best statistical evidence to determine a job's safety. In death, there is no middle ground, and no room for varied interpretation--you're either dead or you're not. The BLS has compiled fatality rates based on occupation, but my knowledge of statistics isn't good enough to tell you whether the method it uses to calculate fatality rates is reliable. As I explain below, I have concerns about some of the charts on the BLS's website.

According to one BLS chart, 111 police and sheriff's patrol officers died on the job in 2008. When total hours worked are factored in, the BLS comes up with a 15.6 fatality rate. There are 133 categories of different occupations listed in this particular BLS chart. If 66 of the occupational categories have fatality rates below 15.6, then you could contend that being a police officer would be one of the safer jobs. However, only 28 occupational categories had fatality rates higher than 15.6. When I accounted for a possible error rate of 1.7 (meaning I would include any fatality rate 14 or above), just 33 categories had higher fatality rates. See HERE for stats. Based on this single chart, it appears that being a police officer is not one of the safest jobs; in fact, it is only in the 25th percentile for non-fatal occupations.

At the same time, the BLS chart used above is hard to follow, because its categorization of various occupations seems subjective. For example, 3 of the 130 occupation categories are listed only as "Government" "Federal," and "State" occupations. It's impossible to tell what jobs/fatalities were included within those categories, or why the statistician didn't lump those occupational deaths into another more specific category. In short, there doesn't seem to be much rhyme or reason in the categories chosen, and it's impossible in some cases to figure out why a particular job/fatality was placed in a particular category. Consequently, the cited fatality rates may be unreliable because they depend on a statistician's subjective assessment about where to include certain deaths. For instance, would a farmer's death go into "farmers and ranchers" or the "farming, fishing, and forestry" category? If we lump farmers/ranchers with "farming, fishing, and forestry," then the total hours might be reduced, and the number of deaths might increase. As a result, the fatality rate listed for "farming, fishing, and forestry" would be higher because of a statistician's decision, not the actual safety of jobs within the occupational category.

Furthermore, in that particular chart, only one category exists relating to police ("police and sheriff's patrol officers"). Meanwhile, the BLS chart splits professional service positions--which are generally safer--into several different categories. It's like having five separate categories for food services (tacos, hamburgers, etc.) and just one for police services--the fatality rate for the single occupation category is likely to be higher than comparative categories. In short, the previous BLS categories seem too broad and subjective to be particularly useful.

Not satisfied, I looked at another BLS chart. This particular chart listed 133 fatalities under "police protection," which is significantly different from the 111 figure cited in the more generalized chart. See HERE. Even with this discrepancy, the numbers in the chart seem more reliable because they are more specific, and the occupation categories are more detailed. However, these more detailed numbers are useless out of context, because a fair analysis must include the number of people employed in each particular job category. (Having one death out of a million employees is much less statistically significant than ten deaths out of fifteen total employees.) Unfortunately, the BLS does not include the total number of employees in the same chart as the number of workers annually injured within each occupation.

Below are the results of the more specific numbers, in order of most dangerous jobs to least dangerous jobs:

Construction (1,005)
Natural Resources and Mining (MSHA-regulated?) (836)
Agriculture, Forestry, Fishing and Hunting (661)
Transportation and Warehousing (797)
Professional and Business Services (410)
Subcategory: Lawyers (11)
Manufacturing (406)
Admin and Support and Waste Management and Remediation Services (340)
Public Administration (includes police, firefighters, environment quality, economic programs, etc.) (315) Subcategory: Police Protection (133)
Subcategory: National Security and International Affairs (53)
Retail Trade (290)
Mining (non-MSHA regulated?) (175)
Wholesale Trade, i.e. Wholesalers (175)
Other Services (everything from nail salons to dry-cleaning) (173)
Accommodation and Food Services (146)
Health Care and Social Assistance (131)
Arts and Entertainment, and Rec, i.e., Sports, Resorts, and Gambling (99)
Real Estate and Rental and Leasing (81)
Educational Services (76)
Utilities (51)
Information, i.e. Media Services (48)
Finance and Insurance (24)

To conclude, police officers appear to have a 2% (FBI stats) to 3.3% (BLS stats) annual injury rate. I wonder: if different people get injured each year, then wouldn't a police officer's risk of injury after ten years of service (assuming the officer has never been injured) rise to at least 20%, and then after 20 years, to at least 40%? Or is it like flipping a quarter, where the pattern of the previous flips doesn't change the 50/50 likelihood of either heads or tails on the next coin flip?

[My sister explains that from a statistical perspective, the percentage would not go up. "It would be 2%/year no matter how many years you work. Think of it this way: if 2/100 people get injured in 1 year, in the next year, the other 98 people won't have a 4% chance of getting injured since they'll have the same 2% chance (this will include the 2 people already injured). The 2% is the injury rate/year. So if we worked for 50 years, the percentage wouldn't go up to 100% because there's no way of knowing if you'll get injured or if you won't get injured again."]

In any case, as I pointed out earlier, due to privacy rights and different police department procedures for reporting injuries, it is difficult to rely on the recordable injury stats. Days off post-injury isn't necessarily an accurate benchmark, either--I remember one parking meter officer took several weeks off after being slapped by someone who didn't appreciate being given a ticket. I wish we had more detailed injury stats for police officers, but I can't seem to find them anywhere. All we know is that officers have a 2 to 3% chance of being injured on the job. (For a reliable analysis of job safety, I agree I would need to analyze job injuries among all occupations, but I just don't have the time to do so, especially when I question the reliability of the injury data itself.)

Also, if we use the lower 435,000 total police employees number (I estimate that there are between 435,000 and 800,000 police protection employees), and assume 131 fatalities per year, then the annual fatality rate for police officers would be 0.3%. If we use the higher 800,000 number, then the rate goes down to just 0.016%. It's important to note that the majority of annual officer fatalities are caused by car accidents. (Jobs involving long periods of driving are more dangerous than jobs that don't require driving.) Thus, to the extent that we worry about violent felons or general people attacking and killing police officers, the stats seem to show that an officer is most in danger when driving his/her own car, not when interacting with the public. (Of course, I am assuming that the annual numbers for injuries and deaths remain somewhat constant, and a quick glance at previous year numbers indicates my assumption is not unreasonable.)

Until the BLS lists the number of people employed within each of its listed occupational fatality/injury categories and also categorizes "recordable" injuries in more detail, there is no easy way to fairly judge the relative safety of any particular occupation. Right now, though, I'd say a police officer working in a random location has a low chance of being killed on the job, especially if s/he isn't driving.

Overall, my impression is that people concerned about safety definitely want to avoid jobs in construction, mining, truck driving, logging, and animal production/husbandry jobs (ever get kicked by a horse?). Large furniture-making jobs (which probably require steel and hot liquids) don't seem safe, either.

Update: my sister points out that a) you must take into account injuries, not just deaths, in calculating safety; b) just because officers don't die at higher rates doesn't mean their job is safe--it probably means their technology and training allows them to avoid injury and death; and c) the statistics do not show the whole picture because officers have a wide variety of jobs--some work at a desk, some patrol the streets, some work in vice, etc. Therefore, a truly accurate assessment of risk requires even more detailed statistics, because a properly done "danger analysis" (my words, not hers) depends on the particular city (Santa Cruz or Detroit?), the kind of job (vice squad or filing reports?), etc. In the end, she says, it's hard to generalize. (I also realized that national numbers would not be as accurate or reliable as local numbers. An officer in Saratoga, CA is probably much safer than an officer in Detroit, MI, regardless of nationwide stats.)

My sister's argument can be summarized as follows:

1. To analyze a job's safety, you must analyze both injuries and deaths.

2. Absent reliable data on both injuries and deaths within a particular profession, you cannot use the word "safe" to describe any particular profession.

3. The data you cite--both FBI and BLS--is unreliable because the stats relating to injuries within the police profession are incomplete and/or not sufficiently detailed to be reliable.

4. Therefore, you cannot argue, using FBI or BLS stats, that any police job is safe or unsafe.

My sister advised me to be softer in my language and not use the word "safe" when discussing police officers' occupational hazards. She said it would be better for me to say, "Because of great technology and training, officer deaths are actually rare compared to what is commonly believed." She said on any sensitive topic, I have a better chance of engaging people and getting them to change their minds if I don't try to be controversial.

I've always known my sister was smarter than me. I'm glad I finally got to post an example that supports my belief.

My sister's arguments are different from other people's arguments, where people would basically say, "I have personally seen officers get injured, and/or based on my own experience, I know they have a dangerous job; therefore, your statistics are crap." When attacking statistics or data, you must allege that the data is objectively unreliable or irrelevant or being applied improperly. Here, anecdotal evidence is particularly ineffective, because the statistics I cited presumably include injuries and/or deaths people have seen or heard about. My sister, in contrast, did not rely on anecdotal evidence--she alleged the data I was using was unreliable because it may under-report actual incidences of injury. (She was reversing my own argument--whereas I said the injury stats are unreliable because there could be no guarantee that recordable injuries included only serious injuries, my sister said the injury stats are unreliable because there is no way to guarantee that all officers would report all injuries. Moreover, just because an attack did not create a recordable injury does not mean an attack wasn't dangerous--it may mean that the only reason an injury did not occur is because of an officer's training and weapons.) [My response would be that the FBI statistics list all assaults against officers, including whether the assaults caused a recordable injury; therefore, there is no need to speculate about the total number of assaults against officers.]

Other people (not my sister) alleged that the mere fact that officers wore bulletproof vests and carried guns and tasers proved their job was dangerous. I really disliked this particular argument. If KPMG decided tomorrow that all of its accountants will be assigned bulletproof vests and given guns and tasers, does that necessarily make accounting jobs more dangerous or more safe? Of course not. What if the government's stimulus package allowed local police officers to buy machine guns and F-22s? Does that mean their job is extra-dangerous now?

I suppose what people meant is that the only rational reason officers would be required to wear vests and have guns and tasers is because rational people would not decide to wear or buy vests unless they believed the vests and guns would come in handy. I still don't think that argument passes muster. It's the old chicken-and-egg problem. What if officers and the public over-estimate the actual job-related danger and decide to go overboard just to be extra-safe? There's nothing wrong with supplying officers with vests and guns because we think these items will help them stay safer, but the act of giving them vests and guns doesn't mean there is a statistically higher chance that an officer is actually in danger.

So, in conclusion, most people should be able to agree with the following: one, patrol officers probably have a higher risk of getting injured than other professions, but we cannot know for certain until the injury stats begin to list officer injuries in more detail and by specific position; and two, because of great technology and training, officer deaths nationwide are thankfully rare compared to what is commonly believed.

Update: John Seiler has an interesting book review here, where he lists the most dangerous jobs, as complied by the Bureau of Labor Statistics:

1. Fishing-related workers.
2. Logging workers
3. Pilots and flight-related workers
4. Iron and steel workers
5. Taxi cab drivers
6. Construction workers
7. Farmers and ranchers
8. Roofers
9. Electrical power workers
10. Truck drivers and sales-related drivers
11. Garbage collectors
12. Law enforcement

Bonus: here is another link/article on police officer fatalities and risks.

Update on June 14, 2012: according to the United States Census, in 2010, there were 946,196 police protection jobs in local and state governments.  This figure does not include federal police jobs, such as the FBI.  More here:

Update on July 20, 2012: from Washington Times article, July 19, 2012: according to the National Law Enforcement Memorial Fund, a "total of 53 officers across the country have been killed since January [2012]." According to the same article, "Among those killed, the leading cause of death came from traffic-related incidents; 18 officers died in crashes during pursuits or routine patrols, while three were struck by traffic during a stop." Also, "Approximately 800,000 law enforcement officers currently serve across the country, according to the Fund. In 2010, the FBI reported that about 53,000 officers were assaulted in some way, being threatened or injured in the course of their official duties."  

No comments: