Economics 201, Spring 2009 Cottrell
Analysis of Crime Rates
This is the final piece of graded work for the half-semester. It is due on Friday, March 6. Only printed
copy will be accepted!
1 Outline of the assignment
Download the data file crime2.gdt. This is available in the docs folder on the ECN 201 home page.
The full URL is
http://ricardo.ecn.wfu.edu/˜cottrell/ecn201/docs/crime2.gdt
The idea is to analyse these date with a view to producing a model that explains as much as possible
of the variation in property crime rates across US cities. The basic dependent variable is prop2005,
the rate of property crime per 100,000 population in the year 2005. The data are from 244 US cities
and “places” with populations of 100,000 or more. Please read the description (menu item /Data/Read
info in gretl) for more details.
2 Comments on the data
You will find that the dataset includes several possible explanatory variables — plus some “raw”
variables that were used in constructing variables of interest, but that probably should not be used in
their own right, e.g., genexpend and policefrac, which were used to construct policepc, per-capita
expenditure on police protection.
Note that some of the potential explanatory variables are alternatives rather than complements. For
example medhhinc and pcincome are alternative measures of income levels; and fampovpc,totpovpc
and povpcu18 are alternative measures of poverty rates. In these cases you should determine which
variant has the greatest explanatory power over crime.
A comment on the education-related variables: nohischool is the percentage of the population aged
25 and over who do not have a high school diploma, and hischool is the percentage with a high
school diploma but no higher educational attainment.
And a comment on the age-structure variables such as pop18_24 (population between the ages of 18
and 24): these are in “raw” numerical form: to get the percentage in each of the age groups you’d need
to divide by population.
3 “Explore” the data
The first thing to do is explore the data, getting a sense of the numbers involved, and the distribu-
tions. This might also tell you whether there are any “odd cases” among the cities that might need
special treatment. Your basic tools here are summary statistics (means, medians, standard deviations,
etc.) and graphical methods (boxplots, frequency plots, pairwise X-Y scatter plots). You’ll have to be
selective here, homing in on what you reckon might be most important: I don’t want 20 pages of plots
and summary stats! Also in this context: you might find it helpful to look at some sorted tables (as
1
