The previous post looked at the police to population ratio, which raises the question: how do we measure the appropriateness of those ratios? One way to look at that is to see how much crime is in a State, see how many officers there are for the population, and compare that with other States’ outcomes. If we follow the “laboratory of Democracy” model, with each State coming to independent conclusions, a common optimal solution ought to be found over time, perhaps with some States getting it wrong.
Again, using the FBI numbers, we can see such a pattern emerging in a data visualization of the numbers. Indeed, forty-five of the Fifty States are clustered in a tight cloud, and therefore seem to have arrived at very similar outcomes, if not solutions. Five states plus DC are outliers. Errors could be found in three directions – too much crime, too little staffing, or too much staffing (few would argue for ‘too little crime’). These errors can occur independently on both axes. The numbers are only as good as the reporting methodology, but the best we have.
Oregon and South Dakota deserve special praise for having the most efficient policing solutions. Oregon for the lowest crime rate at the lowest-cost end of the group, and South Dakota for achieving the lowest crime rate and being below the median staffing levels.
To figure out which of these errors is occurring in a State, if any, we can draw a bounding box on the tight cluster of States (shown here in yellow). The attributes of outliers in either of the North, South, East, or West directions can be considered normal within those bounds. Those in the NW, NE, SE, SW directions can be considered errors. Note: we’ve already declared the South direction (too little crime) to not be an error, and none of the States make the error in the West direction (too little staffing as an outlier).
Where States are making multiple errors, to discern the primary error we can draw a linear regression trendline through the data set. Those above the trend line have an error primarily in the Y-axis measure; those below the trendline have primarily an X-axis error.
Given those criteria, we can draw the following conclusions about the outliers:
South Carolina: properly staffed, crime is too high. Primary problem: crime rate.
New York, New Jersey: over staffed, respectable crime rates. Primary problem: overstaffing.
Vermont: overstaffed, too high a crime rate. Primary problem: crime rate.
Louisiana: overstaffed, much too high a crime rate. Primary problem: crime rate.
D.C.: tremendously over-staffed, too high a crime rate. Primary problem: overstaffing.
The States with a primary problem as their crime rate need to look at the factors that are causing their crime rates. Are the police operating effectively? Is there corruption within the police force? Are the police over-policing and thus creating a reporting anomaly? Are there poverty problems caused by a lack of education or a high tax rate? Vermont and Louisiana should reduce their staffing levels to at least 2.9. South Carolina may want to increase its staffing levels towards 2.9 until the crime problem is solved.
The States (New York and New Jersey) and D.C., with staffing levels as their primary problem, should work to immediately reduce the cost to taxpayers of superfluous police forces. Returning that money to the productive economy will do more to decrease poverty than it’s doing with an inordinately high level of police staffing, which should have a positive impact (decrease) on crime rates. Target 2.9 as a first step, verify the crime rates remain consistent or decrease, then move towards the median level of 2.2 per thousand.
D.C. deserves special mention because it’s such a mess on both axes. It may face special challenges as being a city-district (future work should compare it with other cities specifically) but placing D.C’s crime rate on a straight line between South Carolina and Louisiana (which it lies between in terms of crime rate) excuses a staffing level of no more than 3.2.
Further work should look at factors such as population density and tax rates to see what impact they have on the crime rates and see if the outliers would be brought back into the ‘normal cloud’ if their numbers were adjusted for the Fifty-States trend for those two factors.