APRIL 14, 2012 10:17AM

Florida’s “Stand Your Ground” Law: A Statistical Analysis

Rate: 6 Flag

Florida Statute 776.013 (3):

"A person who is not engaged in an unlawful activity and who is attacked in any other place where he or she has a right to be has no duty to retreat and has the right to stand his or her ground and meet force with force, including deadly force, if he or she reasonably believes it is necessary to do so to prevent death or great bodily harm to himself or herself or another or to prevent the commission of a forcible felony."

The Trayvon Martin tragedy has certainly struck a nerve with the American public, but among the myriad of newspaper articles and blog posts about the terrible death of Travyon, very few have actually discussed the "Stand Your Ground" law that initially allowed George Zimmerman to go free.

I was curious about the effects of the law myself, so I did some research. What I found is that the Tampa Bay Times (here) and the Washington Post (see here and here) have recently reported that the number of deaths associated with "self-defense/justifiable homicides" cases in Florida has significantly increased following the introduction of "Stand Your Ground" in 2005.

As expected, some people have been quick to dispute the data reported in these two newspapers. One of them, Mr. Walter Olson who is a senior fellow at the Cato Institute, took issue with the apparent increase in the number of deaths in an opinion piece he penned for The Daily Caller. His piece also contains several errors, but we'll get to that in a minute.

As most of you know by now, I'm in a line of work that requires analyzing the effects of interventions based on observational data (e.g., deaths, motor vehicle crashes, sick people, etc.). Currently, this analysis is performed with a methodological approach known as the before-and-after study, which is favored over other approaches because the observational data are analyzed for the same location, but at different points in time. The before-after study also allows for a better control of external factors that may affect the frequency of events (other than the intervention itself).

Since this is a blog post and not an academic paper, I won't bore you with the details related to the theory behind before-after studies (which I helped developed). I included a list of resources at the end for those interested.

For this kind of study, the primary goal consists of predicting the number of events in the after period if the intervention had not been implemented. Then, the secondary objective aims at comparing this predicted value with the observed number of events for the entities under study. Depending on the availability of the data, the before-after study considers the length of the time periods (before and after), changes in exposure (such as the prevalence of firearms for example) as well as local and regional time-trend effects if data can be collected for what we call the "control group." As with any methodology, there are limitations, but those won't affect the results discussed below.

Using the data provided by the Tampa Bay Times, I conducted a "simple" before-after analysis to specifically determine whether the number of permissible homicides has increased after the introduction of the SYG law (Florida Statute 776.013 (3)). The raw data are illustrated in the figure below:

Figure 1
 

The figure above shows that, on average, Florida experienced about 34 justifiable homicides per year for a total of 206 deaths in 6 years. In the after period, the annual death rate was about 85, which means that for the first four years, 333 justifiable killings were recorded (note: I didn't use the 2010 data because it wasn't complete). Even without using sophisticated analysis, there's a clear pattern that emerges from the raw data alone: there is an obvious increase in the number of deaths after 2005.

So, we definitely have more deaths. The important question now becomes, "Is the observed increase statistically significant or is it just a fluke?"

For that, I used the before-after methodology I described previously. I present the results in the table below:

Table 1
 

The results shown in the table above illustrate the following:

  • There has been a very strong statistically significant increase (t-value=6.64) in the number of deaths.
  • The increase in justifiable homicides increased 2.5 times.
  • The introduction of the law has caused close to 200 more homicides following its enactment in 2005.

The results are actually better illustrated in the figure below:

Figure 3
 
 
The figure above shows that if the SYG was not enacted, we would have expected to see about 137 justifiable homicides during the first 4 years. However, 333 deaths were observed for the same time period.

In short, the table and figure clearly illustrate that the "Stand Your Ground" law has had a significant effect on the number of deaths at the 1% significant level. (Note: this means that there is only 1% chance that the increase observed in the number of deaths WAS NOT caused by the SYG law; the difference is statistically significant because the confidence intervals don't touch each other.) 

In his Daily Caller piece, Mr. Olson indicated that the analyses reported in the WP contained serious methodological flaws. First, Mr. Olson claimed that the increase noted in data does not reflect the overall reduction in homicides rates for the entire state that has been observed between 2000 and 2010. Unbeknownst to Mr. Olson, if he had incorporated the general state-wide reduction in the analysis, the increase observed in the after period would be even larger than those presented above. In other words, the outcome would the complete opposite what he's claiming.

Secondly, Mr. Olson reported that the original data described above included police and other related shootings. The revised data lowered the annual average to 12 and 36 homicides in the before and after periods, respectively (also documented in one of the two WP articles). However, even if these revised numbers are used, the overall outcome of the before-after study remains the same. The difference is still statistically significant at the 1% level (I checked).

Finally, according to Mr. Olson, the characteristics of the data changed in the after period, in which several justifiable homicides may have been erroneously attributed to the newly introduced law. This may be a valid point. Nonetheless, although it is very difficult to properly validate each observation, removing a third of the deaths (from the 333) would still show a statistically significant increase in the number of homicides at the 1% significant level (to get a 5% significance level, about 50% would need to be taken away). One would have to remove more than 60% of the data after 2005 before the difference between the before and after periods becomes murky.

The bottom line is that if I were given this dataset without knowing its sources, I would positively conclude that what happened in 2005 significantly increased the number of events. I'm certain that if the data and subsequent analysis represented, say, the effects on car crash deaths from increasing the legal Blood Alcohol Content level from 0.08% to 0.10%, very few would be questioning the results.

Additional resources:

Cook, R., Wei, W., 2002. Selection effects in randomized trials with count data. Statistics in Medicine 21, 515-531.

Davis, G., 2000. Accident reduction factors and causal inference in traffic safety studies: a review. Accident Analysis & Prevention 32, 95-109.

Hauer, E., 1997. Observational Before-After Studies in Road Safety. Pergamon Publications, London.

Lord, D., Kuo, P.-F., 2012. Examining the effects of site selection criteria for evaluating the effectiveness of traffic safety improvement countermeasures. Accident Analysis & Prevention 47 , 52-63.

Park, E., Park, J., Lomax, T., 2010. A fully Bayesian multivariate approach to before- after safety evaluation. Accident Analysis & Prevention 42, 1118-1127.

Persaud B., Lyon, C., 2007. Empirical Bayes before-after studies: lessons learned from two decades of experience and future directions. Accident Analysis and Prevention 39, 546-555.

Update (April 18):

I was able to find crime data for the entire State of Florida and I will write a new post within the next few days. This post will include all murders and homicides involving a firearm between 200 and 2010. The results will be very surprising. Stay tuned! 

Here it is:

Added Homicides: More evidence against "Stand Your Ground"?

Update (April 30):

A new post on this subject: "Stand Your Ground": How does it stand up to history?

This figure is discussed in more details:

Figure 5
 

 

tumblr visitor stats

Your tags:

TIP:

Enter the amount, and click "Tip" to submit!
Recipient's email address:
Personal message (optional):

Your email address:

Comments

Type your comment below:
Good and interesting analysis, Kanuk,

It needed saying and you said it well. Your debunking of the Cato guy is easy to understand and the charts show such a significant and sudden increase it can't be explained away.
As an extended analysis based only on logic, and as a pondering of the unavailable intimate details of the individual acts, I would allow the following considerations.

What of the rise could be attributed to the details (or lack of same) of the law? The police/DA's inability to charge a crime they previously would have.

What part might be attributed to increased public awareness of the law? Does the knowledge of the law encourage those prone to encouragement to act irrationally, either misinterpreting vicious intent or trying to create it or something resembling an attack? "Go ahead, punk. Make my day."

Those curiosities would be difficult to document and what the public thinks isn't that relevant, which is to say the data already shows this to be a law that works against its purpose, as the odds are stacked against a sudden surge in "truly" justifiable homicides. The relevant difference has to be that what was once prosecuted, in general, no longer can be. So it's really not about "justifiable homicide, but "allowable" murder/manslaughter, at least in a such a significant way it shouldn't be a law.
Thanks, Paul!

Yes, you made very good points. I agree about the complexities associated with this law. In the end, no matter how complex it is, it has made things worse.
I always enjoy seeing stuff like this worked through, especially because you show your work and discuss your assumptions and reservations. So let me add some reservations of my own:

What I really want to see year over year data that shows not just justified homicides but all deaths in which another human is implicated, and then a comparison of the trends in deaths with the trends in justification.

As far as I can tell from the above data, I can't tell if just as many deaths happened but fewer people went to jail, or if more deaths happened. I suspect that the number of deaths went up. And you'd expect some of that.

One factor playing into this is the decision-making. For some number of deaths, the ones that would be justified under common law, people are going to kill anyway in true self-defense. But the question is the reasoning process: how many “elective deaths” if you were. It would be useful to see, too, how many of these don't go to trial.

And to be really fair, you'd want to plot some attempt at deciding which deaths might have been the other way around. That is, in some cases someone stood their ground in a situation where the aggressor would have turned violent. I bet that's hard to show for any given case, but the point is that if you see a drop in, say, homeowners getting killed, you need to at least take note of what you're buying. It's a too-easy point to say “this makes homes safer” if the cost is “walking on the street becomes less safe.” What are we to do? All stay home and buy guns? If deaths in the home drop from 100 to 10 while deaths in the street go up from 10 to 20, that might be argued, at least, to be a good trade. But if deaths in the home drop from 100 to 10 while deaths in the street go from 10 to 300, one might reasonably question the efficacy of the trade.

I'm not pushing any particular line of Good vs. Bad here, but rather saying that too often we reason with too little data. I like the analysis here, but I still lament (both here and in the discussion more broadly) the lack of assembled data offering useful visualizations of what all these things are.
Glad to see the statistical view Kanuk. It would be shocking if, after having made it easier to get away with "justifiable" homicide, there was no such increase.
I wonder if a good portion of this discussion is merely about the labels attached to certain deaths; in this case, the label "justifiable homicide." If the over-all number of deaths stayed within the expected bounds but some portion of them, now re-labeled "justifiable", would have previously shown up under another label, then I would see no problem from this law.

I would also like to know about any other factors that might give rise to more justifiable deaths such as the rate of increase in overall crimes. If there were, for instance, a greater than expected rise in overall crime, that would have a serious effect on an increase in justifiable deaths as more people sought to defend themselves and their property from this increase in criminal activity.

A perceived increase in crime rates might also have an effect on people's willingness to use deadly force to protect themselves also. Something, apparently unrelated, such as a change of editor at the major local newspaper, where the new editor gives more 'ink' to crime reporting, can make people think that they are in more danger than they thought previously.

While I appreciate your work here, I wonder if your method is just too simple for the complexities of the situation.
.
But that almost has to be the case, because that was the point in the sense that there was concern that some people were being convicted of a crime that was not a crime, i.e. that it was self-defense. The overall interactions in that case would be more important in the sense of what was the impact of the law on crime, if one could of course tease that out.
It's worth noting that at English Common Law, and in most Eastern states, one had a duty to retreat. To be fair, that sometimes resulted in people being convicted for things that didn't on its face seem just. Of course, if people became less willing to deescalate disputes... then overall killings would go up. But the main point I see from what you have here is that if you make it harder to convict for a given even becoming a crime, then yes, the law increased justifiable homicides, which might be a good thing or not. Depends on your model of what would have been happening otherwise.
Brilliant, Kanuck, we need a wider outlet for work like this. I, too, read the Tampa Bay Times article, noting the increase from a baseline of about 34 to a new plateau at around 100. Your analysis pins down an essential part of the equation with regard to statistical certainty.

Skypixeo adds a valid question, but I think we know the answer. I'll let you tackle it since you are so deep into this.

The Cato guys are always dishing nonsense; that's their business. But answering it significant substance is invaluable. Thanks for that.

The causes of the increase in justifiable homicide come at two critical junctures, police who refuse to charge, and judges who throw the charges out in a preliminary hearing on the basis of Stand Your Ground .

Paul is definitely on to something with his reference of the the confluence of justifiable homicide and "'allowable' murder/manslaughter." My read is that the definition of "justifiable" has changed.

The Times link included in your piece is a must-read, with its stories of real people behind the numbers. The story of Billy Kuch is heart-rending. Until we these laws are repealed we can expect to meet many more Billy Kuchs--and Trayvon Martins in the future.
Kent: Thanks! You also raised good points.

I believe some of your questions could technically be answered, such as whether such laws could have prevented certain kinds of deaths or homicides. In the end, if we change one homicide for another, say the victim who would have been killed, but now kills the perpetrator, it still one more (unnecessary) death. In my line of work, we often face this kind of challenges. For instance, the deployment of red-light running cameras will usually reduce right-angle collisions, but will increase rear-end crashes. However, the overall benefit is that most severe collisions (right-angle) are reduced, hence lowering the societal costs.

The most important question with the data I used above is whether the SYG law lead to someone using deadly force, but would not have use such force if did not exist. In this regards, Paul raised valid and interesting scenarios, which should be examined more closely.

Abrawang: Yes, I agree. The hypothesis which states that the number of homicides would remain the same (or becomes lower) after its introduction is, in my opinion, erroneous. I’m certain that, for some cases, the homicide would still have happened, but the person behind the killing can now claim self-defense. This would add more noise to the data. Perhaps Zimmerman could tell us more about this issue (whether he would still have shot Travyon if the law did not exist). Hopefully, one day, he will.
Skypixio: Thanks for commenting. As I discussed above, you are probably right that some homicides may have been reclassified from “unjustified,”so to speak, to justifiable homicides. However, I’m inclined to believe that many homicides may have been avoided, if this kind of law was not enacted. From what I read in various forums, an increase seemed to have been observed in every state where such laws were enacted. Given the very large number of homicides in the U.S., “justifiable homicides” still probably account for a minuscule proportion of all murders. Although this small proportion can be difficult to detect, more sophisticated techniques are still available to overcome these difficulties. Obviously, I only had access to a simplistic dataset, but it still provides useful information. If I wanted to publish the results, I would need more detailed data. I would be willing to consider it if any other researchers who work with crime data are interested…
Steve: Excellent points, especially the one about the police and judges. Indeed, the Tampa Bay Times article provides good examples about the negative effects of this law. As discussed in the article, the person who mistakenly shot Kuch would probably not have done it before 2005.
Interesting analysis, but it doesn't answer the real question. Does this law increase violence in Florida?
perdidochas: Very good question. Unfortunately, the data I extracted cannot be used to answer that question. One can probably get this kind of data from law enforcement agencies or maybe the CDC.
I don't see the relevant issue as whether SYG increases violence, but if it justifies what would have been and should be prosecutable violence. The increase in justifiable violence seems to be well beyond what could be anticipated by far more subtle influences. The most proximate cause of that is likely the shifting of burden of proof, especially in cases where one of the 2 witnesses is dead. Instead of the perp proving it was justified, the prosecution "proving" the perp wasn't justifiably frightened. This is a problem caused by extended Castle (Home) Doctrine immunity, where you have a right to be and to defend against intruders, to anywhere you have a right to be, which also applies to the "intruder," as they have a right to be there as well. Too often immunity applies to the one still breathing.

Between relaxing standards for concealed carry and eliminating a duty to retreat, you end up with a smaller % of responsible gun carriers and more who are irresponsible and prone to using deadly force under an overly-relaxed standard encouraged by the law. In the long term, it's very likely to increase violence overall.
Paul,

I agree that the effects of enacting the SYG law on the level of violence (all kinds) would be very difficult to quantify. I believe the SYG law mainly affects violence involving firearms, which are easier to analyze. Maybe we could add a few cases involving white blades and martial arts (bare hands), but these are most likely few and far between.
It’s good to see that another study, using a different approach, supports the analyses presented above and the second post:

Study: ‘Stand your ground’ laws result in an additional 4 to 7 killings per month
Dona,

Thank you and I'm glad you found it useful.