Crime Isn’t Rising, Except Where It’s Rising

The National Average Crime Rate May Not Reflect Average Recent Experiences of Crime

UPDATE: This story has gotten even more attention than I expected, and I expected a lot. Hello, reader! Welcome to my humble blog! A few notes: 1. I am not claiming that a meaningful number of states have higher crime rates than in the 1990s. They don’t. Crime is lower virtually everywhere. This post is about showing that *crime perceptions* may have *real foundations* that don’t correlate to *average crime rates*. 2. I am not arguing for or against any crime policy, politician, party, or anything like that. I just want this conversation about crime to have the appropriate degree of complexity and nuance it deserves. 3. Yes, I’m as nerdy as this post makes me sound. 4. Many people have suggested states are poor units of analysis. I agree. Counties would be better, or major media markets for local news networks would be ideal. Sadly, local crime data is VERY spotty, and I don’t have a crosswalk for police agency data to counties, or counties to media markets. I STRONGLY encourage somebody else to apply this method to a more rational set of geographies! Please do it! You don’t even have to cite me! I get no compensation or benefit of any kind from this writing, ask for no credit, and expect no citation. I just want society to have richer and more productive conversations. This post has flaws but, hopefully, it will incite people who are real experts in crime to explore this angle even more thoroughly.

UPDATE 2: I’ve been accused of being a “Trump apologist” for writing this. I’ll take that as a chance to explain why I care about this issue; it’s not because of any political allegiance. Rather, it’s a cultural-geographic allegiance. I’m from Kentucky and I have many friends and family in Appalachia. I have been perplexed for a long time how I could square their experience of crime with falling national crime. This kind of analysis helps me frame what I perceive as going on: why my DC neighborhood is safer than ever, while my wife’s hometown of Ashland isn’t. This is as much personal reflection as political argument.

UPDATE 3: An appendix at the bottom includes tables responding to some questions/criticisms, most notably about my use of UCR rather than National Crime Victimization Survey data. I have recalculated tables with NCVS data, and my point gets *even stronger* under that specification!

President Trump has said many times that crime rates are very high, portraying the current situation in the U.S. as one of relative lawlessness. Many commentators have responded by pointing out that, although the last 1–3 years may have seen a moderate rise in crime, overall average crime rates remain very low. This has been used as evidence that the President and his supporters don’t care about the facts.

But this has troubled me for a long time. I started talking about this last October:

And have brought it up several times since then. I’ve been dimly aware there are ways to test this, but have never really sat down to do the work. Recently, I finally sat down and did it some of it, at least a first pass.

So let’s talk about crime, and why I think that Trump voters who think crime is rising aren’t necessarily correct, but they’re not necessarily wrong either. All of my crime data will come from the Uniform Crime Reporting system, and I track nothing more recent than 2014.


National Average Crime Is Very Low

Here’s a graph of crime rates over time, from the Uniform Crime Reporting system:

As you can see, violent crime is low. I don’t want to get bogged down in if maybe crime has risen in the last 2 years, the “Ferguson effect,” or whatever. Let’s stick with the core reality that crime really is pretty low, on average.

But hold on. That’s violent victimizations over total population. Some people may be victims of multiple crimes. But they only cast one vote nonetheless, so we shouldn’t see declining number of crimes as politically meaningful if the number of people victimized is the same.

We could also look at the “prevalence rate,” which shows the % of population to have been victimized by crime, without regard to how many victimizations. That would give us a better indicator of how many people experience crime. If crime falls among repeat-victims but rises among single-victims, then total crime could fall, but prevalence could rise.

Spoiler: that’s not what happened. I don’t show a graph here because it’s super boring; it turns out prevalence and victimization have pretty similar trends. So that’s a no go.

But hold on. Most people who have moved to a law-and-order position aren’t there because they personally experience crime, but because they perceive that crime is a risk, perhaps because they observe or hear about others being victimized. How can we measure that?


The Geography of Falling Crime

One simple way is to get crime rates by some local geography, and seeing how many people live in areas of rising or falling crime. Easy-peasy.

Well, kind of. County-level data is kind of hard to come by in a reliable and available format. But state level data is readily available, going back to 1960. From the state level data, we can use basic population estimates to say what percent of the national population resides in a state where crime is increasing. That’s a blunt tool, but it’s a neat way to get a guesstimate of how observed crime may be changing. If more people live among rising crime, then perceptions of crime will probably rise. This is possible even if total crime rate falls if declines are concentrated in very high-crime areas.

So what does the historical share of the population living amidst rising crime look like? Here you go:

And BOOM! While the share of the population residing amidst rising crime was low in the late 1990s, it has been higher since then! This is based on annual data, but we can do this another way too. Let’s look at 5-year changes for smoother results:

We see the same results: more people live amidst rising crime today than in the late 1990s. Now, the late 2000s were the real problem time, but it’s possible than from 2014, the last year in my data, to today in 2017, these numbers have risen. Given that murder rates have risen especially in Illinois which is not marked as one of my crime-rising states, it seems likely that this figure did rise in 2015, 2016, and will rise in 2017. In other words, by the time the 2016 elections occurred, the share of the population living amidst rising crime was probably very far above late-1990s levels.

Which states have experienced rising crime? Here’s change in violent crime rates since 2010:

It’s a mixed bag, politically speaking. But here’s a scatterplot of change in GOP 2-party vote share 2012–2016 vs. change in crime rate 2010–2014:

The outlier is Utah, where Evan McMullin seriously damaged the GOP vote share.

When we drop Utah, we find a 0.4 correlation between GOP vote change and change in crime rate. That’s very high for a totally uncontrolled relationship!

In other words, while average crime isn’t rising, it is rising in some places, and in those places, it boosted the GOP share of the vote. The example states for this would be Indiana, Maine, Iowa, Mississippi, Montana, Nebraska, New Hampshire, New Mexico, North Dakota, South Dakota, West Virginia, and Wisconsin. Any of those sound familiar to Democratic strategists? Oh, I’m thinking maybe of Wisconsin and Iowa, as well as northern Maine.

The oddities here are Pennsylvania and Michigan: why did they flip red (3.1 and 4.9% GOP 2-party vote share increase, respectively) when their crime rates fell 14.3% and 13.3%?

Well, let’s look at the distribution of crime within Pennsylvania and Michigan!


Crime Is Rising in Low Crime Areas

The scatterplot below shows the change in crime rate from the 2007–2010 average vs. 2011–2014 average for every police department in Michigan and Pennsylvania for which I have sufficient information to measure a change.

As you can see, although there are outliers, broadly speaking crime has fallen the most in areas that had the highest crime beforehand, while crime has risen the most in areas of generally lower crime. The result is that the standard deviation of crime rates has actually narrowed within Ohio and Pennsylvania: there’s less of a gap between lower- and higher-crime places! However, crime as risen in low crime areas.

We can also calculate the same “percent of population experiencing rising crime” for Michigan and Pennsylvania areas with sufficient data. Here it is:

Huh. So, okay, this chart does not show any evidence of rising crime experience. That’s weird. There is, of course, a big spike in the late 2000s, but within MI and PA, the share living amidst rising crime is near historically low levels.

So Michigan and Pennsylvania remain a bit mysterious in terms of the crime-voting nexus. I never claimed crime is the single, sole explanation for voting behavior. Certainly there are other factors too.

To be clear, other states give more consistent examples. Take Wisconsin, where crime has risen 16.6% since 2010.

Wisconsin has seen a rise in the population share living amidst rising crime. No surprise given that crime has risen.

But what we can see is that:

  1. At the state level, the share of the population living amidst rising crime is not at historic lows, and is likely rising.
  2. At the state level, changes in crime rates are associated with changes in GOP 2-party vote share, suggesting Trump voters may have been motivated by increased perceptions of crime.
  3. This trend does not hold for all states, and may not hold for all sub-state geographies. More work is needed to figure out what drives public perception of crime.
  4. Policymakers and commentators need to think harder about the terms and statistics they use to discuss crime. This is a case where sloppy application of national averages has led commentators to systematically miss a story where a growing number of Americans are experiencing rising crime!

Conclusion

Is the United States bathed in a blood-tide of rape, murder, and chaos? No. Are total crime rates particularly high? No. Is there any reason to think they are about to become historically high? No. Are most people living amidst rising crime? No.

But crime is rising in many states where Trump outperformed Romney. And a growing share of Americans do live amidst rising crime, even if it isn’t yet a majority.

I leave you with a final graph. Here’s the population share by state that lives amidst rising state-level crime rates, graphed against Gallup crime perceptions. Reach your own conclusions. From this article and the graph I’ve made, you can guess mine.

APPENDIX:

I will add more charts here as I make them in response to questions. First off, I was asked what Kentucky’s trends look like internally. So, here’s Kentucky’s trends in people residing in rising crime areas. However, I should note here: I don’t have full coverage! The areas with sufficient data for my sample range from 17% to 42% of the total population of Kentucky. Urban areas are more completely covered than rural areas. So there’s a huge nonresponse bias here focused in rural areas. My suspicion is that these are areas that have seen even more rising crime. The same caveat holds for the WI, PA, MI data above.

We can also use a different data source, the National Crime Victimization Survey (NCVS). NCVS samples the whole population with rigorous, standardized questions, instead of the administrative, incomplete, more haphazard method of the UCR. Here’s NCVS data for population residing in states with rising crime:

As you can see, at least as of 2012, a majority of Americans lived in states with NCVS-measured rising crime! It should also be noted that NCVS is generally regarded as a more accurate survey for experiences of crime than UCR.

NCVS also provides geographic data for metro areas with over 1 million people. Restricting to just that sample population, which is about 1/2 of the U.S. total population, I find that many people again live amid rising crime:

If we add in our Gallup survey data and the UCR statistics, we can see a few key things. First of all, the Gallip data does not correlate very well with NCVS data, which is peculiar. However, from 2000–2009, NCVS and UCR seem fairly correlated. But in 2010–2012, NCVS skyrockets while UCR declines.

If you believe the NCVS data (and I’m hesitant, as it involves some significant imputation that can be fraught with peril when it comes to identifying specific geographic trends), then Americans in 2010–2012 were experiencing the most widespread increases in violent crime since the 1980s! Has that trend continued? It’s hard to say, but it certainly seems possible.

I’ll continue to update here as I get relevant data. But thus far, it seems to me that my presentation of the data is suggesting that crime-trends-as-people-experience-them may have almost nothing to do with crime-trends-as-national-statisticians-observe-them.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.