Do police reduce crime? The objective of this short paper is to understand this
Do police reduce crime? The objective of this short paper is to understand this question using the crime.dta. It has information on various types of crime in various cities of the US states over years. It also has the corresponding data on the size of police force. In writing your paper, consider doing the following to make it substantial. 1. Pick a state that has a few cities in the data and explore the scatterplot of the two variables under consideration. You may also find it helpful to fit a trend across the data to understand the correlation. 2. Consider looking into specific cities rather than plotting the data for all cities. Does it look different from the earlier picture? 3. Run a regression of robberies on lagged police per capita. Initiate your first regression without any fixed effects or clustering, then gradually add various types of fixed effects and clustering (and any controls that you may want to add) to clearly see how your estimates are changing as you add them. Your paper should have an introduction, econometric model explaining the various control variables that you will use in your regression, and the results with tables and graphs. In the introduction, motivate the research question -i.e. tell the readers why it is important to know the answer to this causal question, then provide a very short literature review (maybe one or two), and then discussion what you have found in your paper in a few sentence. Since this is a short paper, it should not exceed 5-6 pages. You should submit your .do file and .log file with your paper.
So this is what I would suggest to get a good paper.
1. Write an intro on the data including summary statistics on the data points you want to take into consideration on how crime is effected by police force
You can find summary statistic with the code:
sum (followed by the variables you wish to look at)
2. Find a study and do a short review of what they found on this question and write about it
3. Create some scatter plots with a regression fit on some smaller entities than the entire country. Maybe pick two states to see what their trends are or even smaller, some cities?
4. Run multiple regressions seeing how different control variables, fixed effects, and clustering change your coefficient for police. I suggest just looking at a couple states.
The code for regression is
reg (y variable) (all x variables), r cluster(whatever you want to cluster)
The easiest way to get fixed effects as variables in my opinion is
i.(variable you wish to use as a fixed effect)
For example: i.year
5. Interpret your regression result and draw up your conclusions
Leave a Reply