What the Justice Department's own studies found about the 1994 ban

Here is something rare in Washington: a law that was built with an instruction to be studied, and then was studied, by researchers the government itself paid. The results were not what the law's champions had promised. They were not what its critics had predicted either, which is part of why they are worth reading.

The law is the federal assault-weapons ban of 1994. It expired in 2004. Every few years since, someone proposes to bring it back. Before the country decides whether to do that, it is fair to ask what the best available evidence said the first time. The evidence comes from two reports by a criminologist named Christopher Koper and his colleagues, written for the National Institute of Justice, the research arm of the Department of Justice. I will lay out what the ban covered, what the reports found, what the strongest argument against my reading is, and then what I think a law should have to show before it limits a right.

What the ban actually covered

The Public Safety and Recreational Firearms Use Protection Act was part of the Violent Crime Control and Law Enforcement Act of 1994, signed on September 13 of that year. It had a sunset clause. It expired on September 13, 2004, and Congress did not renew it.

Three things about its design matter.

It named models. The act listed particular firearms by name, including the Colt AR-15 and a group of other semi-automatic rifles, pistols and shotguns, such as the AK-47 variants, the Uzi and the TEC-9. The text referred to the named guns and "copies or duplicates" of them.

It defined "semiautomatic assault weapons" by cosmetic features. A semi-automatic rifle that could accept a detachable magazine was covered if it had at least two of a list of features: a folding or telescoping stock, a pistol grip that protrudes conspicuously below the action, a bayonet mount, a flash suppressor or a threaded barrel to accept one, and a grenade launcher. Similar lists applied to pistols and shotguns. Note what is not on the list. None of these features changes how the rifle fires. A pistol grip does not make a bullet faster. A bayonet lug has, so far as I know, played no role in an American crime of note.

It banned "large-capacity" magazines. The act barred the manufacture and import of magazines that hold more than ten rounds.

And the key to everything that follows: it grandfathered. The ban applied to manufacture and transfer after the date of the law. Firearms and magazines lawfully possessed on September 13, 1994 stayed legal to own, sell and transfer. The country already held a large stock of guns and of magazines of more than ten rounds, and the law left them where they were.

The workaround

Gun makers did what anyone would do. They redesigned. A rifle that had a pistol grip and a bayonet mount could be sold with the bayonet mount removed and a different stock, and it was no longer a covered assault weapon. The functional parts, the action and the caliber, were unchanged. The law had written its definition in terms of looks, and the market answered in kind. Critics of the ban called this a loophole and supporters called it a flaw in drafting. I call it the predictable result of banning a thing by describing what it looks like.

What the studies found

The first report was by Jeffrey Roth and Christopher Koper, prepared through the Urban Institute, "Impact Evaluation of the Public Safety and Recreational Firearms Use Protection Act of 1994," delivered in 1997 under a Justice Department grant. The second was an update by Koper, with Daniel Woods and Jeffrey Roth, a report to the National Institute of Justice dated July 2004, a couple of months before the sunset.

Here, in plain terms, is what they found. I am summarizing the reports' conclusions, not quoting them, and you should read them yourself; they are public.

First, the weapons were a small part of gun crime to begin with. Before the ban, the firearms it targeted by name or feature were used in a small fraction of gun crimes. The researchers' range, from various studies of police records, put them at a low single-digit share of crime guns, with a ceiling somewhere under one in ten. Rifles of any kind are a minority of guns used in murders, a point I make in The weapons of war argument against the AR-15 fails on its own terms.

Second, the share of crime guns that were banned models fell during the ban. The 1997 report found that the proportion of the banned guns among guns used in crime declined after the law took effect. That is the result supporters point to, and it is genuine. A decline in a small share is still a decline in a small share.

Third, the researchers could not find a clear effect on gun violence overall. The 2004 update concluded that the ban's effects on gun violence were likely to be small at best, and perhaps too small for reliable measurement. The reasons were the ones above. The banned guns were used in few crimes. The grandfathered supply was large. Offenders could substitute other guns, including post-ban variants and guns that fired the same ammunition. And the ban's covered features had little connection to lethality.

Fourth, magazines were a different matter, and the researchers said so. Magazines holding more than ten rounds were involved in a larger share of gun crimes than the named assault weapons were, accounting, by the 2004 report's count, for 14 to 26 percent of guns used in crime before the ban. Some of the reduction in firepower, the authors reasoned, might follow if the supply of those magazines dried up. But the grandfather clause meant the supply did not dry up. Magazines made before 1994 stayed legal, and millions more were imported under the grandfathering exemption. The authors observed that any effect on gun violence would be visible only after the pre-ban stock was used up or lost, which would take longer than the ten years the law lasted.

I would not overread any of this. These are two reports, not the last word of social science, and the authors were careful in their caveats. They did not say the ban could never work. They said that the one that Congress passed, with its sunset and its grandfathering, did not produce effects they could measure.

The arithmetic, with a hypothetical

Let me put some arithmetic on this, labeled as a hypothetical so no one mistakes it for a statistic.

Suppose a city records 1,000 gun crimes in a year. Suppose banned weapons were used in 2 percent of them. That is 20 crimes.

Now suppose the ban works better than the evidence suggests and cuts the use of banned guns in crime by half over a decade as the old stock wears out. That is 10 fewer crimes involving a banned gun.

But some of those offenders will pick up another firearm. If half of them do, which is a generous assumption, then five crimes are averted entirely and five are committed with something else. Five fewer crimes out of 1,000 is half of one percent.

That is a hypothetical, but notice that the arithmetic is driven by the first number. If the targeted weapon is used in 2 percent of crimes, even a heroic success against it can produce only a small fraction of a percent in total. This is how the reports' phrase, "too small for reliable measurement," arises. A change of that size is smaller than the noise in the crime data from year to year, and nobody can honestly say what produced it. A change in the murder rate of a city due to economic conditions, policing, or the weather in a given summer will dwarf it.

The strongest counterargument

The best case on the other side is about magazines, and it comes in three parts.

A shooter with a 30-round magazine can fire thirty rounds without stopping. A shooter limited to ten must reload, and each reload takes a few seconds, during which potential victims can run, hide or fight back. In a mass shooting, those seconds can mean lives. The 1994 ban failed to show an effect because it grandfathered the stock, covered cosmetics and lasted only ten years. A better-designed law, one that restricted magazine capacity across the board and had time to work, could show an effect. Researchers, Koper among them, have since argued that limiting magazine capacity may reduce deaths in mass shootings.

I take that seriously, and I want to give it its due. It is not a silly argument. It is a reasoned conjecture about a mechanism, and the mechanism is plausible.

Here are the answers.

1. The mechanism is a conjecture about rare events. Mass shootings, however horrifying and however much coverage they get, are a small share of gun deaths, which means any effect would show up in a small number of incidents, in which it is difficult to disentangle the effect of a magazine from the many other variables.
2. A reload is a skill, not a barrier. A person who has planned an attack will practice swapping magazines, and will carry several. The seconds the argument counts on are the seconds of an untrained shooter, and a person bent on murder is not a casual user. Meanwhile the family at home, the person the law claims to protect, is the one for whom limiting capacity is a real constraint, since in a stressful encounter people miss.
3. A ten-round limit has costs to the lawful. The ordinary owner is the one who loses, and the ordinary owner is the majority of owners. Under Heller, a firearm and its ordinary parts in common use for lawful purposes are protected, and magazines holding more than ten rounds are standard equipment on many of the most popular firearms in the country. Standard equipment in the hands of millions of lawful owners is not an unusual weapon.
4. The case for the right does not depend on the case for the crime. Even if a magazine limit saved a few lives, the question is whether the government may take a common lawful item from millions of people who have done nothing wrong, and Bruen answered that question with a test that asks for historical support, not a cost-benefit tally.

The question I would ask of any law

Which brings me to the question at the end of this exercise. What must a law show before it limits a right?

I would propose a four-part test, and I think any fair person, left or right, would accept it for any right they care about.

1. A real problem. The harm the law addresses must be shown, with numbers, by the people proposing it.
2. A fit. The thing banned must be a significant cause of the harm, so that removing it would reduce the harm by an amount that could be seen. The 1994 ban fails here: its target was a small share of crime guns, defined by features that do not change what the gun does.
3. No better alternative. If enforcing existing law would reduce the harm with less intrusion, the less intrusive route comes first. I take up that kind of alternative in Lying on Form 4473 is a felony, so why is it so rarely charged.
4. A willingness to be measured. The law should include a review, and its sponsors should accept the results. The 1994 ban was reviewed and the results were published, and anyone proposing a revival owes the public an account of what those results say.

We would apply this test to a law restricting speech without hesitation. The Constitution does not rank some rights above others, and the reasoning holds for the one in the Second Amendment. For the history of what came before, see ninety years of federal gun law in the dates that changed the rules, and for the vocabulary that shapes the debate, assault weapon, ghost gun, arsenal, how news vocabulary does the arguing.

A word about humility

None of this means that no one should ever propose a gun law. It means the proposal should come with a theory of the case and the evidence to match, and that the evidence the government itself paid for ought to be on the table. That is not a partisan standard. A person who cares about reducing violence should want laws that work, and a law that cannot show an effect, however well-intentioned, spends a right and a measure of public trust and buys nothing.

I say this as someone who wants the debate to be a better one. The opposing side in this argument has often been right about the stakes and wrong about the instrument. The instrument, in 1994, was a list of features. The stakes were people's lives. A country that takes lives seriously should be willing to ask whether the instrument worked.

The case in brief

1. The 1994 ban named models, defined "assault weapons" by cosmetic features, barred magazines over ten rounds, and grandfathered everything already owned; it expired on September 13, 2004.
2. Justice Department research by Koper and colleagues in 1997 and 2004 found the targeted guns used in a small share of gun crime and the ban's effect on gun violence too small to measure reliably.
3. The strongest counterargument concerns magazines and mass shootings, which is a plausible conjecture about rare events and not a demonstrated result.
4. Even a heroic success against a weapon used in 2 percent of crimes yields less than a percentage point, as the arithmetic shows.
5. A law that limits a right should show a real problem, a fit, no better alternative and a willingness to be measured.

What you can do: search for the National Institute of Justice reports on the 1994 ban, read the summary of the 2004 update, and bring its findings to any town hall where a revival is proposed.