Forecasting crime is a dicey business. The range of potential outcomes is vast because crime and violence are influenced by so many variables, including macro factors like the health of the economy, the political environment, the state of policing, and the criminal justice system.

There are scads of ways to forecast, but one promising technique is to look at how past trends are correlated since past trends were pulled by the same factors that influence current trends. If, for instance, two trends have typically moved together, but have been pulled apart in recent years, a reasonable forecast might predict that those two trends will start to move together again. That, I believe, is exactly the situation we find ourselves in today.

What I will try and demonstrate below is that robbery and homicide rates have historically trended along the same path. That long-term relationship began to change about 15 years ago but remained relatively predictable. COVID, however, completely disrupted that relationship. But, if we have reached the end of COVIDs effects on crime, and I believe we have, then we can use a return to a more stable relationship between homicide and robbery to make a prediction about 2025.

Share

The Relationship Between Robbery and Homicide

First, so we are all on the same page, a robbery is the taking of property by force or the threat of force, which often includes stickups with a gun. Defined this way, it is easy to see how every robbery has some probability of becoming a homicide. If someone points a gun, demands a wallet, gets nervous, and squeezes the trigger a little, boom, homicide.

Robbery and homicide are also related in the sense that if more guns are used in street crime, both robbery and homicide might increase. And conversely, if policing and sentencing are ratcheted up and there is greater general deterrence of gun crimes, or if social programs lead to less gun carrying and aggression, both homicide and robbery may decline. The point is homicide and robbery are related.

And they are uniquely related. I don’t want to go off-topic, but you can see how homicide and robbery trends (orange line) move together and then look at how homicide and aggravated assault (grey line) do not.

My observation that robbery and homicide are cousins is hardly novel. When dissecting the rise of crime in the 1980s and the fall of crime in the 1990s, Al Blumstein observed that for almost two decades, robbery and homicide moved in lockstep. In fact, from 1980 to 1998, there were about 25 robberies per homicide almost every year.

In fact, if you extend this trend out another decade, that stable relationship between robbery and homicide holds fast. And it didn’t matter if violence was high or low. In 1980 there were 10.6 homicides per 100,000 people in the US. and the ratio was 24.6:1. In 2010 there were 4.4 homicides per 100,000 and the ratio was 24.8:1. So whether homicide was high or low, the ratio remained the same. That’s pretty impressive evidence of a stable relationship. But then, the ratio changes.

The Declining Robbery-Homicide Ratio

Beginning around 2008, homicide rates leveled off, but robbery rates continued to decline. As a result, the homicide-robbery ratio started to decline—slowly at first and then quickly, as shown in the ratio graphic above.

There’s a lot going on in those statements. First, there is the question of why this stable long-term ratio starts to change. And then there is a question about why it changes much faster during the pandemic.

The intuitive explanation is that a lot of robberies become a homicide, so in some sense robbery “causes” homicide. However, only about 25 percent of homicides are thought to result from a robbery, so while there is a direct relationship, most of the incredible stability of the ratio must derive from something else.

A more compelling idea is that there is some other force that is acting equally on both homicide and robbery. As noted in the introduction, there are a number of macro forces that could reasonably be expected to affect both kinds of violence. But the 30 years of stability between the two kinds of violence appears to limit the power of many of those explanations. The ratio between robbery and homicide was stable through very good economic times and deep recessions. It was stable both in periods when relatively few people were incarcerated and periods of mass incarceration. And it was stable regardless of which political party was in power.

However, one factor that affects both homicide and robbery has changed substantially since about 2008: the availability of guns. To be clear, the reason why a change in gun availability would affect the ratio of homicides and robberies is that it affects one more than the other, thus changing the ratio. Specifically, the more guns on the street there are, the easier it is for someone who intends to kill someone else to acquire a gun. That seems to be an important part of the story: that more homicides are premeditated than in the past.

So, here’s the gun story, starting with data from The Trace.

If you start on the far-left side of the graphic, you will note that gun sales were quite stable in the 2000s. And the ratio of homicides to robbery, as Al Blumstein predicted, is stable as well. But then America started buying a lot more guns, particularly handguns. In particular, between 2008 and 2016 the number of handguns purchased annually about doubled.

And that’s about the time the robbery-to-homicide ratio starts to decline (meaning that there are fewer robberies per homicide).

Now, this raises all kinds of interesting questions, but they are not germane to the forecast for 2024, so I want to set them aside for the moment. It is clear that the robbery-to-homicide ratio changes dramatically around the time gun sales have their first peak and then changes dramatically again after gun sales spike in 2020.

Is this related to gun sales? Trends in gun sales and trends in the robbery-homicide ratio are inversely correlated.

The idea here is simply to point out that rapid increases in gun sales create an exogenous shock to the robbery-homicide ratio. And in forecasting 2025 what matters most is whether there will be another shock. Last year was the fourth straight year of declining handgun sales. And 2025 is not a presidential election year, which historically tends to predict higher gun sales. So, all else being equal, it seems reasonable to predict another small decline in handgun sales for 2025.

Now, the Forecast

Let me put the long-term robbery and homicide data back up in front of you and say a little about what this all means. When you put all of that together, the key ideas are:

1. The long-term robbery trend shows a steady, decades-long decline. Robbery rates have been essentially flat from 2021-2024. In the past, during the long decline, when robbery rates stabilized (2006-2008 and 2014-2016) it was followed by a continuation of the long-term trend. Thus, all else being equal, robbery should decline in 2025, a bit, about 4 percent.

2. The homicide decline has been accelerating, with 2024 showing even bigger declines than the record-breaking declines in 2023. There is no historical analog to this, so there really is nothing to point to within homicide alone to make a prediction.

3. But! We do know something about the ratio of robbery-homicide. That, after all, is the point of this essay. And what we know is that without a massive shock, there is likely some equilibrium rate of robbery-to-homicide.

4. Finally, historic trends predict only small changes in gun sales in 2025, and absent the pressure of greater gun sales, the robbery-to-homicide ratio should return to equilibrium.

Now, we have all the ingredients for a forecast, save one. We need to make some predictions about homicide. And we can use the robbery-homicide ratio to make it. And I will not inflict any math on you to do it.

Since we are talking about a ratio, it is important to point out that the ratio can change if either the numerator or denominator changes. We expect the numerator (robberies) to decline a little. We do not have a solid basis of prediction for the denominator (the change in homicide). But, we can solve for the denominator if we know what the equilibrium ratio will be. So, what do we expect to happen to the ratio?

The chart of the ratio does not seem to give us any clear guidance. It was at equilibrium (25:1) for almost 40 years but has declined steadily since, before rising recently. With that said, there are three ways to look at this trend that all point to the same answer.

First, as a guidepost, I want to note that the ratio in 2024 is about 14 robberies to a homicide in 2024, and it was 16 in 2019, before the pandemic.

• The ratio has risen three years in a row (2022-2024). Twice in the past (1968-1970 and 1978-1980) the ratio rose rapidly for three years and then levelled off. That supports the idea of a stable ratio around 14.

• Over the last five years, the ratio has been sensitive to gun sales, declining rapidly when handgun sales rose and then increasing when they decline. Since I expect handgun sales to be relatively steady in 2025, that also supports a leveling off of the robbery-to-homicide ratio about where it is, at 14:1.

• Both robbery rates and homicide rates in 2024 are at same levels as 1964-1965. The robbery-to-homicide ratio was 14 in those two years (13.96 in 1964 and 14.06 in 1965).

So, if the robbery-to-homicide ratio is predicted to be around 14 in 2025, and robberies are expected to decline a moderate amount, then homicide would also be expected to decline a little, about 4%.

One important note. I have written several times that post-COVID federal investments in state and local government were important drivers of the decline in crime since 2022, and I was delighted to see that the Brookings Institution released a report last month that makes a similar argument. That federal funding will dry up over the next year and simply hoping that local economies have strengthened enough to replace those dollars with local revenue is likely not enough.

There is substantial momentum toward less crime and more public safety in America. But it will not be sustained with intentional effort.

Leave a comment

Musical Interlude

I lost my most faithful reader in November. I miss her. She would be appalled that I mentioned it here. But I think she liked this song.

The sky is falling

I’ll see you in a while