Are we circling the drain?

8 min readMay 31, 2023

Crime in central Indiana isn’t changing the way you’re told it’s changing.

It’s a common refrain (daily for some): Central Indiana is falling apart. Crime is running rampant and getting worse all the time. This is bleated the loudest whenever some public office is up for election. The tactic works. Fear mobilizes people to turn their minds over to whomever promises the most loudly to install safety. It also plays well to point fingers and lay blame. After all, if crime is always spiraling upward, then somebody has to be held responsible. If you’re a challenger, you can blame the incumbent. If you’re the incumbent, you can claim that the “other side” is interfering in your crusade, so you have to get reelected or the “other side” will make things even worse.

Funny thing, it’s easy to bleat about how crime is always getting worse because our minds are programmed (mostly) to act like scared monkeys. In a world that can be pretty nasty, it can be useful to remember scary things most vividly. That way, you would be more likely to be primed to run if a lioness actually does pop up to eat your monkey butt. But what if some really clever monkey decides that he can convince you that you have to let him tell you what to do? Well, he can just keep hooting about the ever-present threat of lionesses. So, you’ll troop behind him like a good little monkey, even if the threat of lioness rampage is actually going down!

In a monkey-shell, Is crime really as bad as we are told by politicians, and is it getting worse every year?

So, why am I writing all this? My idle curiosity isn’t that consuming. Truth be told, it is, but I have another reason. Someone I know is running for public office, and she asked a public question: Is crime in her town as bad as people keep claiming?

It just so happens that the FBI has a collection of reported police incidents that includes incidents by town and city. It’s called the Uniform Crime Reporting Program (dum da-da dum!). So, I got data for the Indianapolis Metropolitan Area (aka the “Indianapolis-Anderson-Carmel IN MSA”, to give its government name). This is a cozy little village of about 2,000,000 inhabitants, all of whom spend a great deal of energy putting down everyone else who shares the same MSA but lives in a different town — nobody fights harder than family.

The area is centered, of course, on Indianapolis and has three anchor cities four smaller cities with 50k to 100k people, 16 with 10k to 50k people, and a bunch of towns with less than 10k people. I could have looked at every town and village in the area, and I did, but that’s a different article. This time, I was more interested in towns that are on the larger size.

When you have historical data, you can emphasize totals or you can emphasize trends. You can talk about both, but that muddies the water. I find trends far more interesting. Is a city getting better or worse and how does that compare to other cities. Now, when you look at historical trends, it can get wacky if you take it waaaaaaaaay far back. So, I looked at the trends from 2016 to 2021 (which is the latest year that data is currently available for). The other thing I did was create crime rates by dividing crime counts by population. This is far more realistic at presenting risk.

Anyway, I got the data and did a quick analysis — details are at the end. My results:

Total Crime

Looking at total crime, we get the following picture:

What’s this mean? It means that, of the cities we looked at, They have different totals (dots) and trends (lines). So what? So, it looks like total crime has been decreasing steadily for several cities since 2016. So much for the bleating of idiot politicians. However, a pretty picture can hide a few details. This is why I have the following table.

So what does that mean? You can figure out that “City” is a city name. “% Improvement” means that total crime rate dropped by that percentage per year from the period 2016–2021. Easy stuff. The tricky thing is “Group”. It can be useful to have a rule to tell you if a difference is worthwhile. One way to do this is to build statistical groups. In a nutshell, if two cities share a Group symbol, then they should be considered the same in terms of their improvement rates. What we can see is that the city that has been improving the most is Franklin, and it’s ahead of everyone else. Then there is a group of Greenfield/Greenwood/Plainfield, which is in second for rate of improvement. Anderson comes in third. The rest, roughly speaking, all flop into one nebulous fuzzy group. Before I forget, the +/- thing is the “95% confidence interval” for each city.

Property Crime

Of course, not all crime is the same. It’s common (and sensible) to group crime into property crime and violent crime. When I looked at property crime, this is what I saw.

Looks pretty similar to the total crime chart. This is because, no matter what news sources or your online browsing might say, property crime always is the majority of crime. Total crime trends are generally driven by property crime. This one has a table, too.

Well, well, looks like Greenwood leads the pack in reduction of property crime, but it’s in a four-city group with Franklin, Greenfield, and Zionsville. Oddly enough, Fishers, Westfield and Carmel all saw an increase in property crime over 2016–2021. If you’re sharp, you’ll see that this doesn’t appear to match the trendlines. This is because the analysis was weighted (more on that later) to give more emphasis to recent years. Likewise, total crime numbers are fairly small for most of the cities in the “wrong direction” group, meaning that any difference gets magnified on “percent change”.

Violent Crime

Here’s the big boy, the scary monster, the Really Bad Thing that gets most attention on news and political potboiling. “IS YOUR FAMILY SAFE?????”

Okay, nobody is surprised that Indy had the most, but Shelbyville? Shelbyville has got to get a hobby. Anyway, we’re interested in rates of change, so here we go:

Well, looks like Fishers, Westfield, and Carmel are as good at Indianapolis in reducing violent crime — opsie? Zionsville WTF? INCREASE? (That’s what a negative improvement means). Don’t get hysterical. Z-land has a very low rate to begin with, so any wiggle could send up upwards, furthermore its 95% confidence interval is gigantic, meaning that the “real” number could be zero. Anyway, it looks like Franklin, Greenfield, Shelbyville, Greenwood, and Plainfield are leading the way in reduction of violent crime from 2016–2021.

What does that mean?

The take-home is that a higher crime rate is not the same as things getting worse. It’s possible for a city to be improving and still not perfect. What’s more important is that the politicians who keep bleating that crime is getting worse all the time are just flat-out lying — or they have special secret data that you’re not allowed to check for accuracy. Would you trust that?

You may not want to believe me, but you are free to look up the data yourself and do your own analyses! I do not hide my date. I don’t hide my methods. You can check my work. Do politicians let you see all their data and all their work? Of course, you can also say that the FBI is a giant poopyhead fibber who makes stuff up just to make politicians look bad. Okay, can I sell you a bridge in New York or swampland in Florida, then?

Making Sausage

I began by downloading data from UCRP. It can be tedious. Then I decided that looking at the whole USA was just silly. Looking at al of Indiana was just silly. It made more sense to look at cities similar to each other. This meant cities in the I-C-A SMR. There are several, but since there is a big difference between tiny little villages and full-fledge cities, I simplified matters by looking at the larger (>10,000 people) towns and cities. However, these cities didn’t always report data. Soooooo, I excluded cities that lacked data for 2021, for 2016, or for any two years in the period 2016–2021. This was to avoid problems that come from missing too much data or from missing data at either end of a time series. This left me with 12 cities. I got the annual population data for each from the US Census.

My interest was in calculating risk trends. A common way to do this is to use what’s called “logistic regression”, so that’s what I used. I reformatted the data to be appropriate for logistic regression (which automatically takes population into account) and ran the model “Crime ~ City * Year”, where “Crime” could be total crime, property crime, or violent crime. Individual data points were exponentially weighted to emphasize more recent years over past years. Weighting used α = 0.230258482192819, which was selected to have had data from 2011 (if I had gone back that far) to be weighted at 0.1 while data from 2021 was weighted at 1. This is based on the reasonable presumption that the further back you go, the less conditions would resemble the present day. Whether or not to weight time series data is an important question and partly depends on whether you intend to analyze the data to find causes (inference) for what you see or to describe it with an eye to the future (prediction). If you are wanting to describe and predict, and testing ideas about specific causes isn’t a big deal for you, then weighting is the way to go.

From the models, I estimated marginal trends by City over the period and calculated statistical groups using Benjamini-Hochberg false discovery rate adjusted p values. With these I constructed the tables. The models were also used to draw the trendlines. ANOVA for the models are as follows: