Crime Statistics

I went to the Detroit Public Library last weekend, to gather data from the Detroit City Directories. These directories go back to the early 1800's, and list all the people in the city (children usually excluded), and also lists all the people who live on each street. My intent was to target several blocks, like Maple Street, Antietam Street, Field Street, Baldwin Street, Wilshire Street, and Kilbourne Street. For each target block I was planning to gather turnover data for each block from the beginning of time up through the 1970's. Then I could compare the earlier turnover to the turnover from the white flight of the 1970's. However, when I got to the library, I realized many individual years are missing from the library, and parts of some of the books are unusable. Then I called into question the validity of the directories, after seeing that my Dad was still listed in the 1974 directory, even though he had left town in 1971. So I decided against the turnover project. However, I did collect some interesting data which I hope to report on at a later time.

My Uncle Ken tells me there is one simple reason for everybody leaving Detroit in the 1970's - crime. He said the casualty rate among store owners on Mack Avenue was higher than troops in Vietnam, then he proceeded to rattle off a bunch of names of people who had been killed. He says he got tired of having to carry guns with him to/from work, and that type of thing. He says that race had nothing to do with it. I can personally think of several reasons to support, as well as to refute, his argument. But I really like having data, so I went to the internet to look for Detroit Crime Statistics.

This isn't my area of expertise, but I am interested in this subject. And I have been well-trained in statistics and research methods.

For starters, I found a site from the Wayne State University's Center for Urban Studies, and a report entitled: "Detroit Crime Barometer: May 2007." This report is very short and has several charts. Unlike most academic work, the author did not cite the sources of his data, which is a bit unusual. But I find it highly unlikely the data can't be trusted, coming from Wayne State. I have attached two of the charts to this posting. Click to enlarge.

One of these charts compares the murder rate to the unemployment rate, from 1970 to 2006. The other chart compares the robbery rate to the unemployment rate, from 1970 to 2005. We can see that on both charts, the trends track each other 'fairly' well. There is a notable exception to this tracking, during the crack epidemic. Apparently, during that time, there were many drug-related murders attributed to the characteristics of crack cocaine, rather than the usual socio-economic factors. Also, we can see that unlike the murder rate, the robbery rate tends to lag the unemployment rate. This finding suggests to me that there is a learning curve, or an apprenticeship, associated with robbers that is not there with murderers.

Correlation is a necessary, but not sufficient, condition to determine cause. A relation between X and Y can be because X causes Y, Y causes X, because the variable that causes X to vary also causes Y to vary, or is simply an empirical coincidence with no conceptual link. Even with the ability to establish temporal precedence, it is nearly impossible to determine "for sure" which variable causes the other. This analysis is further complicated because there is probably also a recursive effect: that even though unemployment may 'cause' crime, crime also probably 'causes' unemployment, with jobs moving to lower-crime areas. But let's ignore that for aspect for now.

One of the problems we have with this analysis is that we have population-level data to model crime. Ideally, we'd like to have individual-level data on a large sample of people, and then analyze the data with the dependent variable being whether or not the individual committed a crime or not. But this kind of study is very difficult and very expensive. And do you think people would respond honestly about their crimes? I don't think so.

I doubt that all crime is caused by unemployed people. The unemployment rate can cause economic insecurity for people who have a job, too, if they are worried about getting laid off, etc. So even if the measured group-level independent variable is the unemployment rate, the underlying conceptual variable is probably something like individual-level perceptions of economic insecurity, even if we aggregated it to the group level. But it is difficult to obtain data of economic insecurity perceptions, especially after the fact, so we use unemployment rate data instead.

Likewise, the measured crime rates probably vary somewhat, too. Some crime is unreported, for various reasons. Some murders may be misidentified as natural or accidental deaths. Aggressive law enforcement can cause an increase in the number of criminals behind bars, and this would affect the unemployment rate to crime relationship.

This analysis doesn't break down the types of robberies, or types of murders. Armed robberies of stores may have different antecedents than personal robberies, breaking and entering, etc. Likewise, drug deal murders likely differ from spousal murders, random shootings, etc. But it is likely the unemployment rate has some effect on these crimes.

We can't ignore the effect of heat, either. Like I said earlier, crime statistics is not my area of expertise. But you never hear about riots happening January! Prison riots, either. So I think if this data were broken down by month instead of year we might see an even stronger effect. Or to statistically adjust the crime rates to account for the effect of temperature. While we are at it, we should throw in population density and average age as well.

Oh, and let's not forget that Detroit neighborhoods differ significantly from one another. Some neighborhoods have much more crime than other neighborhoods. To do this right, we really should break these data down to neighborhood-level data.

We might also want to compare the unemployment rate to crime relationship of Detroit to other cities and communities, too.

Having said all that, it is pretty clear there is a relation between unemployment rate and crime, to show such a strong relation even with the likely errors in the data. If, as my Uncle Ken says, crime is what caused everyone to leave Detroit, what caused the spike in crime? Was it simply that black people moved into the neighborhood? Was it, like in my Uncle Dan's case, too many hillbillies moved into his neighborhood on Corbett? Or was it due to the increase in unemployment in the city, caused by automation in the factories and the moving of factories down South in pursuit of lower taxes/utilities/labor costs?

What do you think? Post a comment and let us all know!