Back in 1948, the Chicago Tribune published a newspaper declaring “Dewey Defeats Truman.” There was only one problem: Truman won. Here is an AI summary of the event.
The 1948 U.S. presidential election is a landmark event in polling history, famously illustrated by the erroneous headline “DEWEY DEFEATS TRUMAN” published by the Chicago Daily Tribune. This headline was based on pre-election polls, including those by Gallup, which predicted a decisive victory for Republican candidate Thomas E. Dewey over incumbent President Harry S. Truman. These polls used quota sampling, a method that aimed to ensure the sample reflected the population’s demographics by targeting specific groups in predetermined proportions. However, this method was flawed because it relied on telephone interviews, and in 1948, wealthier Dewey supporters were significantly more likely to have telephones than less affluent Truman supporters, leading to a biased sample. Despite the polls favoring Dewey, Truman won the election with 303 electoral votes to Dewey’s 189, and a 4.4-percentage point margin in the popular vote. The discrepancy highlighted the critical importance of using probability sampling, where individuals are randomly selected, to achieve an accurate representation of the population.
When we talk about Defensive Gun Uses (DGUs), the findings usually depend on the sample we look at. Here is the AI generated summary of the sample technique of the National Firearm Survey:
The 2021 National Firearms Survey, conducted between February 17 and March 23, 2021, used an online survey administered to a representative sample of approximately 54,000 U.S. residents aged 18 and over, identifying 16,708 gun owners for in-depth questioning about firearm ownership and use. This survey, carried out by the professional firm Centiment, utilized a large sample size described as the largest of any gun survey study to date, surpassing prior well-respected surveys in scale. The sampling methodology was approved by the Georgetown University Institutional Review Board.
If you are a gun owner and you want to look at what the breakdown would be of a typical DGU (assuming the future will resemble the past), the National Firearm Survey would be the best place to go.
However, what if the National Firearms Survey didn’t ask the questions you wanted answered? For example, let’s assume you wanted details about DGUs where the victim fired their weapon at the offender? Well, you would want to conduct another National Firearm Survey, making any modifications to improve the sample selections, and then ask the questions you want answers to. Unfortunately, to rerun a survey of that magnitude would be cost-prohibitive.
One approach taken by Claude Werner was to analyze the NRA’s Armed Citizen column and compile statistics. I don’t know what the process is for the NRA to select a situation to publish in an Armed Citizen column, but I suspect it is nothing like the National Firearms Survey. Hence, there will be some unknown sampling bias that impacts the statistics.
Why am I talking about this situation? Well, in the GUNS magazine September 2025 issue, I ran across two different quotes in two different articles. This one is by Lew Gosnell:
[…] that a quality revolver loaded with effective ammunition is entirely adequate for 99% of typical self-defense requirements.
Then, in a different article by Massad Ayoob we have this:
Speaking of Tom Givens, one of the very best in the armed self-defense teaching business, he also teaches with the same type of gun he carries – a full-size 9mm service pistol backed up with two spare mags because that’s what his intensive analysis of civilian gunfights suggests is the best setup.
What a disparity! Yet, both of these authors are highly experienced with law enforcement, violent crime, and accurate shooting. This made me curious about Tom Given’s analysis. Here is the AI summary.
Tom Givens, a firearms instructor with over 40 years of experience, has conducted a detailed analysis of civilian gunfights based on data from his students. His research, derived from over 60 documented defensive encounters involving his students, reveals consistent patterns in distance, duration, and the number of shots fired. Most of these incidents occurred at a distance of three to seven yards, which is roughly the length of a car, with the majority happening between three and five yards. This distance range is critical for training, as it represents the most likely scenario for a civilian self-defense encounter.
The analysis shows that these gunfights are typically very brief, lasting approximately three seconds, and involve the firing of about three to four shots. The data indicates that the vast majority of incidents are financially motivated, such as armed robberies, carjackings, or home invasions, and occur in public places like convenience stores, parking lots, and shopping malls, rather than at home. Givens emphasizes that the success of a defensive encounter hinges on the ability to draw the firearm quickly and deliver accurate, well-placed shots from concealment. His students, who were armed, won all 63 to 68 documented encounters, with no losses, and most sustained only minor injuries or none at all.
Givens’ analysis also highlights the importance of mindset and preparation. He stresses that the use of deadly force is justified only to terminate an immediate deadly threat to oneself or someone for whom one is responsible. He advocates for training that prioritizes fast, reliable presentation of the handgun from concealment, two-handed shooting for stability and accuracy, and the ability to keep the gun running through reloading and malfunction clearance. The data underscores the necessity of carrying a firearm consistently, as the three students who died in robberies were unarmed at the time. This real-world data, drawn from his students’ experiences in Memphis, a high-violence city, forms the foundation of his training philosophy, which differs significantly from military or law enforcement paradigms.
Givens has excellent data. The problem comes when people want to extrapolate it and say this is what a DGU will be. To see what the problem is, I’ll give a couple of examples:
- You survey people who have a graduate degree in piano performance, asking them what their favorite songs are. After tabulating the results, you indicate that America likes these songs the best.
- You survey people who have $200,000+ sports cars, asking them what they think about State Troopers. After tabulating the results, you indicate that America has a problem with State Troopers.
As you can see, if you select from students of your self-defense classes who live in Memphis, you are not selecting a random sample of individuals. You may be selecting people who are more likely to have dangerous jobs, work in dangerous environments, live in dangerous areas, or work at dangerous locations. Also, they may adopt your specific training ideas and use them in the field. This means the data you get can be self-reinforcing.
I continue to believe the National Firearm Survey to have our best data about DGUs. However, even the National Firearm Survey can’t tell us everything. Things like:
- A DGU where the victim was killed
- A DGU where the victim was charged with a crime and imprisoned
- A DGU where the victim was killed by responding police officers believing he was a threat.
These are important to know, as well as other harms of concealed carry.
All in all, I endorse the position of Lew Gosnell about the advantages of a double-action revolver for an ordinary civilian.
