# What does data tell us?

You’re probably getting tired of my blog posts about violence in Baltimore. Well, tough. That’s where my “Thesis 2.0”, as it were, is headed, so you’re going to have to live with it.

Anyway, I got a little curious last night at the rate of homicides per day in Baltimore and whether or not there was an increase after the Freddie Gray “unrest.” As it turns out, there was.

The red dot on the graph is the day in which Freddie Gray was buried and when Baltimore experienced “unrest” in the form of an all-out riot. (No, I’m not at this time having the discussion of whether or not it was a riot.) From January 1 to abut the beginning of April, the rate of homicides per day was going down. From about April 20 onward to yesterday, the rate of homicides per day has increased to almost one homicide per day.

For these calculations, I used the number of cumulative homicides and divided the number  on any given day by the number of day of the year. For example, by January 15, there were 10 homicides. Ten divided by 15 is 0.67, or about two homicides every three days. The pace certainly picked up in April and has kept trending upward since.

Mayor Rawlings-Blake has said that other cities are seeing an increase in homicides this year. Her and others often cite Chicago as an example, so I went and looked at Chicago data. (For Baltimore, I used Cham Green’s page and the Baltimore Sun’s page.) I did the same calculation as I did on Baltimore data and then plotted the two together:

As you can see, both cities experienced a similar rise in the number of homicides per day, with Chicago having a larger homicide-per-day rate. So does this mean that Chicago is going through a similar phenomenon as Baltimore is?

No, it doesn’t. It doesn’t because Baltimore has about 622,000 residents. Chicago has 2.7 million residents. In order to compare these two cities, you need to take into account their differences in number of residents. You also need to take other things into account, but the very basic thing is the population.

So, when you account for the differences in population, this is what we see:

As you can see, there was an increase in the number of homicides per day per 100,000 residents for both cities, but it was much more pronounced for Baltimore. Furthermore, the magnitude of rate is flipped from what we saw earlier: Baltimore has a higher homicide-per-day-per-population rate throughout. Also, it started increasing in April and just never stopped in Baltimore. In Chicago, it started increasing around March but has since pretty much leveled off.

But does this happen every year in the spring? Do the number of homicides per day just increase as the year goes on, in Baltimore or anywhere else?

To answer that question I created two graphs. The first is the homicide-per-day rate since 2010:

Like the other graphs, there’s an initial acceleration before a stabilization of the rate. That’s because the first few days may have a lot of homicides relative to the number of days that have passed. (You see this same effect in the other graphs.) The rate goes up and down, but then takes a sharp increase at the red dot which — like in the other graphs — is the day of the Freddie Gray riots.

But, again, is there a cyclical nature to this increase in the homicide rate per day? For this next graph, I restarted the cumulative count of homicides every year. (There are gaps where the were no homicides at the beginning of years.)

Like in the other graphs, you see the initial acceleration/deceleration in the rate because of the cumulative number of homicides compared to the number of days that have passed. Notice that the apparent cyclical nature of the rate of homicides per day is not there. There are years when the rate stays flat throughout. There are years when it slowly rises. For 2015, however, it just spikes after April.

So what does the data tell us?

For me, it tells me that something definitely happened at a macrosocial level in Baltimore in April. What exactly happened is up for a lot of discussion and a lot of more advanced analysis than what I can whip up in Microsoft Excel in a couple of hours. (When I get more time, I’ll run this through a program to see if I can fix those artifacts and do something more biostatistical to see if the trend seen in 2015 stands up to chance.)

One thing that the data do not dispute is that things are bad in The Charmed City.