The likely study
Chris Leblanc's 2007 article in the Journal of Non-lethal Combatives argues strongly that the claim "most fights go to the ground" originates with the Gracie family, famous for popularizing Brazilian jiu-jitsu, a groundfighting art. I vaguely recall seeing Rorion and Helio repeating this claim in video, but cannot produce an example at this time. The claim is certainly widespread in mixed martial arts circles, particularly among Brazilian jiu-jitsu practitioners.
Leblanc argues further that the Gracies are referencing a specific LAPD study:
In 1991, Sergeant Dossey, an exercise physiologist with the LAPD, completed a comparative study of use of force incidents reported by LAPD for the year 1988. Sergeant Dossey looked at all 5,617 use of force incident narratives written by officers in 1988, and devised a method for codifying the information contained and analyzing it for what they identified as dominant altercation patterns. The study was replicated in 1992 by LAPD’s Training Review committee.
Notice that already, the study deviates from what one would normally consider "all fights". These are police/civilian use-of-force incidents, not civilian/civilian altercations or military/military violence. This distinction becomes clear when Leblanc excerpts the results:
Five scenario patterns accounted for 95% of the altercations...
- Subject pulls away from an officer’s attempt to control the subject’s
- Subject attempts to punch or kick the officer...
- Subject refuses to assume a searching position...
- Subject flees and officer pursues...
- Subject takes a combative posture, but does not attempt to strike the officer
Those are very distinct from common civilian self-defense scenarios, such as conflicts of ego, domestic violence, rape, road rage, murder, robbery, or kidnapping. Leblanc's summary of the study should put the nail in the coffin of the idea that the LAPD study has any relationship to a BJJ or wrestling-related claim that "95% of fights go to the ground" or "all fights end up on the ground":
The report concluded: “Nearly two thirds of the 1988 altercations (62%) ended with the officer and subject on the ground with the officer applying a joint lock and handcuffing the subject.” Given this, it is better put that the LAPD data says when officers physically fought with suspects (versus simply encountering minor resistance or non-compliance which required a minor use of force, but did not escalate into an altercation), 95% of the time those fights took one of five patterns, and 62% of those five types of altercations ended up with the officer and subject on the ground with the officer locking and handcuffing the suspect.
So if the Gracies are citing a particular study, and if the one Leblanc found is that study, they're pretty obviously dead wrong, or at least misrepresenting an unrelated set of statistics.
If, however, the Gracies (or anyone else) are just saying that generally fights go to the ground, or that a different study or set of evidence shows that most fights go to the ground, then the question is still open.
Self-defence blog reviews a study (available in e-book):
Bakari Akil II (Ph.D), who is an Assistant Professor of Communication at Middle Georgia College and a no-gi Brazilian Jiu Jitsu martial artist, conducted a study to find out what percentage of fights go to the ground. He viewed hours of CCTV footage examining a variety of street fights looking for patterns in the violence. In summary the study discovered the following:
- 42% of fights ended up with both people on the ground while 72% ended with at least one person on the ground.
- Where at least one person was grounded 57% went down due to a throw or a take down, 7% were pushed, 35% from being punched and 1% were downed from a kick.
- The person who hit the ground first lost the fight 57% of the time, while 33% were a draw (no discernable winner) and 8% went on to win the fight. These figures were reversed for those who hit the ground second or stayed standing.
Restricting one's research to closed-circuit-TV-captured fights puts a major restriction on the type of altercations captured, but this is a good start. We see fairly convincingly that throws and groundwork happen with frequency in unscripted fights. We can also conclude that we want to be the ones doing the throwing, and that if we do end up on the ground, we want to be able to respond effectively.
The conclusion I think one should draw from this study, and from similar surveys of street fights and instances of violence, is that the clinch (standing grappling) should take up the majority of one's self-defense training. Ground grappling, striking, and come-along joint locks should have a place as well, but the clinch is where most fights either end up naturally. (One can also initiate a clinch easily if it does not.) The clinch is also the place where it is decided if someone goes to the ground, and if so, which participant it will be.
More text of the study is available at judoforum.
You asked, "what reasons are there to think that fights usually end up on the ground?" The "usually" clause is what's tripping us up. Without somehow collecting data from violence of all causes--which is an absurdly long list--we're just lost. We simply cannot make any evidence-based conclusions without data, and I don't think the data currently exists.
There are two ways to proceed once we realize that we don't have the information we want.
- We can do the best we can by watching street fights on YouTube, studying patterns in mixed martial arts contests, putting on a mouthpiece and sparring as close to "no rules" as possible with people we trust, and collecting anecdotal data from our friends and coaches. This is an empirical approach.
- We can reason about how a variety of violent scenarios would play out. This is a theoretical approach.
As Kant noted(1), empiricism and idealism (theory) are both fundamentally flawed. We need to use (and recognize the drawbacks) of both. Theorizing about fights is good, but we should inform ourselves as well as possible before doing so.
Violence in our lives is not restricted to a few moments of brawling in the bar, nor to knife fights to the death in dark alleys. In evaluating the need for ground fighting skills, we also must take into account situations of...
- de-escalation, where we might need to wrestle with, say, a friend or a teenager without harming them
- sexual violence, where one might expect the situation to become violent while already in grappling range or on the ground
- being taken by surprise, or our striking or clinch skills failing us, leaving us on the ground despite our best efforts
Whether or not we conclude that ground-fighting happens in most fights, we cannot determine where the fight will happen. This is true even if we are skilled at a non-sparring art, or some form of striking--note how frequent clinches occur in boxing, or how frequently who people who want to punch each other in a brawl will fall to the ground together. Fights are chaotic.
This should, of course, be considered alongside broader questions about the goals of one's training. One's training time might be allocated differently depending on whether you are training for basic self-defense, or for comprehensive self-defense, or for self-improvement, military training, or another purpose.
(1) A serviceable summary of my point: "Kant responded to his predecessors by arguing against the Empiricists that the mind is not a blank slate that is written upon by the empirical world, and by rejecting the Rationalists’ notion that pure, a priori knowledge of a mind-independent world was possible. Reason itself is structured with forms of experience and categories that give a phenomenal and logical structure to any possible object of empirical experience."