Threshold Analysis: Measuring Crime Data

Current Z SCPD Scores

I recently asked Santa Cruz Crime Analyst Jon Mitchell to present crime data in a visual format by crime type and produce a monthly report to inform police commanders of changes or variations in data, tipping them to changing trends. Presentations such as this information identify burgeoning problems or concerning crime fluctuations. Crime data can fluctuate broadly, so reducing the white noise and honing in on trends can have a tangible impact on public safety. The quicker we can identify crime problems and address them proactively, the more crime officers can prevent. 

A Police Commander’s primary responsibility is designing a strategy to prevent and reduce crime.  Commanders create an operational strategy to combat crime through Problem Oriented Policing, Hot Spot Policing, and Situational Crime Prevention among others.  To enable the commanders to identify crime patterns, SCPD employes a process to give them information rapidly and accurately. We examine data by using a Threshold Analysis. This examination of the Standard Deviation enables SCPD to prioritize and respond proactively to crime problems as they arise.

I asked Jon to explain our process in simple terms. Enjoy.

In 2007 Don Chamberlayne, Principal Crime Analyst for the Worcester, Massachusetts Police Department won the International Association of Crime Analysts (IACA) “Innovations in Crime Analysis” award for his modern approach to automated threshold analysis. Don developed an automated system to quickly identify potential patterns of crime within his agency. Coming from a statistical background, he developed an automated system using Threshold Analysis to try and help the command staff identify problems much earlier in the process.

The first step is to break our reports into eight major groups (Homicide, Rape, Robbery, Aggravated Assault, Burglary, Larceny, Auto Theft, and Arson). Next, we tag and tally reports that fall into these groups.

We then look at the numbers of reports –a problem arises. If we compare numbers month to month, we see a significant variation in crime. Our City population is around 65,000 people, but that number doesn’t correspond to the number of people the police department protects. Our city has an additional 3- 4 million tourists who stay at 55 hotels, visit the Boardwalk, and relax on our incredible beaches. On top of that, add in 20,000 UCSC students.

When one uses factors such as these, the population of Santa Cruz sways significantly from month to month. Coupled with populations swings and the natural ebb and flow of seasonal crime numbers alone are insufficient. Commanders are interested in the rate of change compared to our historical “normal” values. This Threshold Analysis is vital to understanding crime rates and addressing crime problems.

To do this, we use a mathematical model to calculate and make the comparisons to determine what is outside the Threshold. We give this number the term Z-Score and use it to compare the crime stats monthly with other months over time. The benefit of using this system is that it is easy to read, reduces the natural “noise” and normal statistical swings leaving a number we can compare month to month.

The Z-Score is a number that for “the normal amount” of monthly crime stays at 0, and the further it gets from 0 either in a positive or negative direction indicates the amount the crime pattern has changed. In other words, has crime risen or fallen? We set the threshold at -1.25 to +1.25, so anything outside of this range is flagged and analyzed by staff. Based on the analysis commanders implement strategies to reduces these spikes or crime deviations.

Method (Detailed)

What a Z-Score means in its purest form to us, is the number of standard deviations from the Mean that our crime stats are.

To calculate that we do the following:-

1.      Work out the Mean (the simple average of the numbers)

2.      For each number subtract the mean and square the result

3.      Then work out the Mean of THOSE squared numbers

4.      Take the square root of that, and this will give us the standard deviation.

5.      Then for the latest data, we take the (Amount of crime – the Mean) divided by the standard deviation this gives us our Z-Score

6.      Repeat for each crime group.

Below are an example of charts are given to police commanders to alert them to problems as they arise. When the data surpassed the red line, they are responsible for designing a strategy to address the issues. This strategy is reported out at weekly command staff meetings. We below the green line we look for what worked and try to replicate it.