Do Casinos Increase Crime?

Casino gambling was illegal until 1931 when Nevada first legalized it and it grew later after 1988 when the United States allowed gaming to be regulated on Indian owned land, known as the Indian Gaming Regulatory Act. Since then, states have allowed gambling and have created casinos to generate jobs and revenue for social welfare, education or other development initiatives. Today there are over 1,000 casinos, but many believe casinos bring crime and other destructive socioeconomic impacts.

Since casinos bring gambling, it is no surprise that it also brings compulsive gambling issues. The more severe the problem is, the more likely they will turn to crime. Approximately one half of compulsive gamblers commit crime. Typically, their motivation is financial and non-violent to either collect more money to gamble or repay debts.

There have been many studies on this research question and others have found either low increase or a gradual increase after 5 years from when the casino was built. Earl L. Grinols and David B. Mustard conducted a study on this topic that was noteworthy due to their large scope. They published a paper called "Casinos, Crime, and Community Costs" where they ultimately concluded that higher crime rates were caused by existing casinos and that it was only in the fourth or fifth year after the casino was built that the crimes were highest.

To find out myself about how crime is related to casinos, I will use the Horseshow Casino in Baltimore to map against crime. The horseshoe was constructed August 26, 2014. I will use crime data in 2014 showing pre-conditions and 5 years after construction, 2019, to show post conditions.

My research question is simple: Does the presence of the Horseshoe Casino Baltimore increase crime?

To begin I downloaded the crime data from the City of Baltimore Open Portal and uploaded it to my project. It is important to note, that crime data is innately uncertain due to over policing, under policing, and preferential policing. It should be used as a general

This data contained all arrests since the data was made available. In order to analyze 2014 and 2019, I had to export those timeframes into a new dataset. Next I aggregated this crime data (points) into census tracts. I explored this data through histograms, scatter plots and distributions.

Total crime count for 2014 was 42,620.

Total crime count for 2019 was 1,638,600.

In order to normalize my crime data, I calculated crime per capita (total population/crime) for each tract. In the map below, you can compare 2014 with 2019 by sliding the bar.

It seems fairly obvious that is more crime in 2019 than 2014. But are there statistically significant areas where crime occurs or don't occur? To answer this question I have use the Hot Spot and Cluster and Outlier Analysis tools.

The hot spot analysis showed in 2014 only one tract that had a significant hot spot of crimes. When looking at 2014 the same hot spot remains and a few hots spots with 90% confidence were added, closer to the Horseshoe casino.

When looking at this map, you can see four distinct colors (red, light red, blue, light blue) highlighting statistically significant tracts with outliers tracts represented by red (high values) and blue (low values). It also shows clusters represented by light red (high values) and light blue (low values).

What's interesting here are the red tracts surrounded by light blue tracts in the north area. This means those red tracts have a statistically higher crime rate than its surrounding neighborhood. In the south area, we can see the opposite, where the blue tracts have statistically lower crime rates than its surrounding neighborhoods.

Although we are looking at crime rates before and after the Horseshoe Casino was constructed. The crime data includes all levels of crime that may or may not have anything to do with the casino. As mentioned earlier, usually gamblers will commit non-violent crimes to get money. I wanted to find out what types of crimes are being committed in these areas and are they reasonable to think it might be because of increased gambling.

I chose to run multivariate clustering to find out how the crimes would be grouped. Please check the legend to see the groups. They are not intended to match from 2014 to 2019. Below are are the resulting box plots for each year.

I decided to use Robbery - Street crime as my explanatory variable against all crime (my dependent variable) to better see if street crime had any significant impact near the casinos where there would be an increase of gambling. See final map below.

Although street crime robbery did increase, the local R-Squared values (0.91 to 0.95) remained similar in the end. More investigations into other crime types would be recommended before concluding any real results.