InfoMullet: Violence in America – FBI Perspective 1960-2016

Spread the love

TLDRUpFront: This begins a new series by the InfoMullet. The aim of this series is to present clear data from credible sources on the historical patterns and current incidences of violence in the United States, and around the world, from trusted sources. No policy arguments are made on either side of this data. Only notes of method and caveats are provided. In this post, the FBI’s Uniform Crime Report (UCR) data is presented. Discussion limited to methodologies and notes is included. Future posts will include perspectives from health care agencies like the CDC or may focus on global rates of violence.


FullContextintheBack:   The FBI began an effort to create uniform criminal statistics beginning in 1929. By 1930, the Uniform Crime Report (UCR) had been established and it has been published annually ever since. It represents one of two primary criminal statistics reports for US law enforcement. The other is the National Crime Victims Survey (NCVS) conducted by the Bureau of Justice Statistics.  The two reports take two different methods to answer the same question: what are the statistical occurrences of crime in the United States?

The method of the UCR is to use incident counts, voluntarily submitted by police departments from around the country. The method of the NCVS is to conduct surveys of between 45,000-75,000 individuals, and use inferential statistics to derive at rates of crime. These two methods both have their strengths and weaknesses, so understanding crime is actually usually an effort of making comparisons between the two. As the UCR relies on incident counts it can be regarded as “hard data.”  However, it is only hard data on cases that make it to the stage where they are reported to police – and the police then subsequently report that data to the FBI.  This voluntary reporting regime and reliance on police reports, rather than victim experiences, are the biggest weaknesses of the UCR. For many reasons victims may be reluctant to report incidents to police, nor is it always assured that police are providing to the FBI all the cases that they are involved in.  For example it has been clear for years police under reported their own violent encounters with citizens and what data they did report could be considered compromised because there’s a conflict of interest in how they present themselves. Generally the UCR is a good data source, but one has to be careful to understand these nuances and they will be called out where appropriate.

The NCVS aims to resolve this discrepancy by means of inference. Surveying a large population and from their responses inferring the actual levels of crime, reported and unreported.  It’s weaknesses derive from this design. Whereas UCR’s are simple counts – the NCVS has to undergo sophisticated statistical processing to ensure data is normalized to region, gender, ethnicity and a host of other factors because they are only dealing with a sample.

In this article, the InfoMullet presents data from the UCR violence category which includes data on crimes such as criminal homicides, rapes, robberies, and aggravated assaults. The data presented will be segmented by violent crime type and depict both the actual count of crimes, and a per capita number.  The per capita number is an important figure because it normalizes the incident counts for a growing population. What this means is even though there were more homicides – the rate of violence has actually decreased.

Data on property crimes is not presented in this post. The UCR also has numerous extended tables which segment criminal data on all manner of potential items of interest: the race or gender of the victim, of the criminal, where it occurred, what weapons were used, what other crimes occurred. For purposes of this article however the only data presented from these “extended tables” is the weapon type used in murders.  Given the heightened public debates about firearm violence the belief is this will be useful data to know.  A second data break down displayed is the one level of segmentation on the national criminal homicide statistics.  This is by State.  Unfortunately the detail at the police department level has numerous gaps, and the UCR does not have a function to report by city or metropolitan statistical area (MSA) such as the CDC does. So state level segmentation is as good as it gets.  Even so it shows that violence remains a highly localized phenomena, and the risk levels matter both greatly on where you live – and who you are.  Future articles, especially using CDC data, will provide more granular detail to demonstrate this effect.

The intention of the series is to update these charts and repost them when the data is included. When the FBI releases the full UCR for 2017 the charts below will be updated and reposted.

 

Criminal Homicides 1960-2016

Criminal Homicides

Note that the criminal homicide data excludes the deaths of the 9/11 and Oklahoma City Bombing terrorist attacks. FBI UCR data excludes by definition suicides from criminal homicides. Also the per capita number is more meaningful here as the population of the US is 140m fewer people than it has today.  This is why although the actual number of criminal homicides appear to be increasing (red) the actual per capita rate remains at historically low levels. This can cause confusion when newspapers and online magazines don’t clarify what version they are referring too.  There’s also a caution about interpreting too much information from any 2-3 year period, such as the recent spike up or the odd-behavior between 1984-1986.  A host of factors contribute to homicides and often ‘crime waves’ in individual cities can substantially effect this report. The interim 2017 UCR (released in June with only half a year’s data) showed significant drops in criminal homicide rates in the first part of the year. When that is published these charts will be updated.

 

Murder by Means (all types)

The UCR includes data on what was used to commit criminal homicides and this represented below in two ways. The first is a chart that shows the breakdown across all types of weapons for 2016 exclusively. The second shows a distribution of weapons among the category of “firearms” as a percentage of all criminal homicides committed with a firearm.

 

Note the caveats below. 

A quick note on the funnel chart above. The FBI warns that UCR data is not appropriate to create “rankings” of city or location in terms of safety. Everyone does it anyway. But the reason why is the uncertainty of reporting in any given location. A city may have a stellar police department, and report their data, compared to a city with a more questionable force that selectively reports or forgets to report data. The better city might “look” more dangerous simply because they’re being honest.  This caveat wasn’t explicitly applied to ranking homicides by weapon type, probably because no one was crazy enough to try that. So just assume there are some pretty hefty caveats on this data and use it for notional and illustrative purposes.  The “Firearm Not Stated” category is dealt with below.

 

Murder by Means (Firearms Only) 1996-2016

Note caveats below on “Unstated Firearms.”

Given the concern and frequent public discussion of firearms, the InfoMullet thought breaking out criminal homicides by the type of weapon used might be useful. Note that this is only a subset of all homicides – the ones that used a firearm. A clear take away is that the pistol is doing the yeoman’s work of bloodletting in the United States, and it has for a very long time.  Overall rates of violence by type of firearm have more to do with factors such as accessibility, cost, and ease of concealment than they do the firing mechanics of the weapon itself.  It’s interesting to note the generational-growth of “Unstated Firearms.”  This is different than “Other Firearms” which are firearms that aren’t handguns, rifles or shotguns (blunderbuss perhaps?) But “Unstated Firearms” is a field marked “Firearms Type not Stated” in the expanded table 8 of the homicide data. Ostensibly this that an officer reporting a crime either doesn’t state the type because they weren’t old or they don’t know.  But to a data analyst, seeing a very clear growth rate over 20 years could speak to a different conclusion. We can see from the data that use of handguns is declining over the same period of time, but that decline isn’t represented by an uptick in any other weapon. It’s possible that “firearms unstated” is just what it means – or that meaningful data on the type of firearm used in a crime is not being entered correctly for ideological reasons. One of the problems of this data. It’s not above being manipulated. If there’s a less tin-foil reason on this discrepancy published we’ll update. For now though it’s just “interesting” to see that increase in what is essentially a “we don’t know” category that matches the decline of the dominant means of murder.

Update: Based on requests from our Facebook page the “Murder by Means” data has been expanded into three additional charts. All of these charts track homicides by the type of weapon used as a count, rather than a percentage. The first plots all types on a logarithmic scale with a base of 2.  Data analysts use log scales when comparing items with very dissimilar patterns by scale. In this case handguns are so much more a common implement in a homicide than anything else, that if we plotted them as counts all on the same chart it would dominate at the top. All other causes would be “smooshed” at the bottom losing fidelity of detail to notice patterns that might be relevant between say shotguns and rifles.  With a log scale the chart increases the Y-axis exponentially by counting it’s Base. So a base 2 doubles at every interval. A base 10 increases ten fold.

 

Third Warning; Logarithmic scale on Y-axis, Base 2. See above for explanation.

 

However, many are not familiar with log charts and may mistake the close proximity of shotguns to handguns as indicating they are used in similar ways. (Even though they are actually thousands of cases apart.)  For that reason two additional charts follow the log chart. These two charts are more traditional “count” charts, but segmented into two groups to allow meaningful resolution. The first group is of handguns and “firearm not stated” which each account for several thousand firearm homicides.  The second group, and third chart, plots shotguns, rifles, and “other” types of firearm. Since these are only ever in a few hundred per year, they are in effect a separate category of firearm homicides than handguns given usage rates.

Chart 1 of 2 focusing on high-use firearms.

Chart 2 of 2 focusing on low-use firearms in homicides.

Criminal Homicides by State Segmentation

The FBI repeatedly warns not to use it’s data in “ranking” locations because of potential discrepancies in reporting. But it’s interesting information and with that caveat in mind here I am about to do what the FBI tells me not to do. Again.

Segmentation of Homicides by State

Given the FBI’s warning, the InfoMullet does not advise planning a move on this data.  However it is useful to understand that criminal violence in the United States is not equally distributed. It varies heavily based on location and other personal factors such as race, gender, economic status etc.  Having gone through the CDC’s WISQARS data which reaches down to the county and city level – we can confirm that this fractal behavior does not stop at the state level. Since we tend to overstate statistically rare things – such as violence – it’s easy to see a headline number of a per-capita homicide rate and believe that applies to where you live. It probably won’t. You may be much luckier, or worse off – than the average result based on where you live.

 

Rape (Legacy & Revised)

The FBI revised it’s reporting procedures on how rape was calculated in 2013.  This was a clear case of the NCVS data diverging from UCR given the known reluctance of victims of sexual assault and rape to wish to interact with police. And a culture of police departments in the past not taking these claims seriously. Both the legacy data, extending back to 1960, and the revised methods are presented below by way of comparison.

Rape incidents reported in legacy fashion. 

The revised means of calculating rape incidents is a combination of actual reporting and statistical inference based on what is known by research of unreported rape.

Rape reported using revised means.

It’s worth noting that regardless of the method used, rape reports have been increasing over the entire period of reporting available online from the UCR. This stands in contrast to every other form of violent crime over the last several decades which show declines. It’s not immediately clear, and the UCR doesn’t comment, on why this may be. It may represent an actual increase in incidence rates, a greater willingness to report such incidents, or both.

 

Aggravated Assaults

Aggravated Assaults

 

This is another situation where comparing incident rates (red) versus per capita rates (blue) are useful. Looking just at the red line doesn’t convey that the rate of violence remains dropping as a trend.

 

Robberies

Historical Trend of Robberies

Robberies are the fourth category of violent crime reported in the UCR. It’s not immediately clear if these overlap with any other crimes, such as rape or criminal homicide committed during a robbery.

 

Future Posts & Acknowledgements

The InfoMullet doesn’t have a complete list of where it’d like to go next with these articles, but here are several ideas.

  •  National Crime Victim Survey Results
  • CDC Mortality and Fatal Injury Data (WISQARS).
  • Private-efforts to track police shootings (Washington Post & Guardian).
  • Global Terrorism Database (GTD) Terrorist incidents on US soil.

What else would you like to see?  Let us know on Facebook or Twitter.

Also I’d like to acknowledge the work of the team at CalculatedRiskBlog.com.  They began this idea of taking publicly published data and making it easier and more accessible to understand in simple to read charts. If you ever need to understand economic information or behaviors over time – we highly recommend them.

 

Leave a Reply