LAH spends $100k+ to dubious effect

-

It was really great to open the Town Crier and see some actual crime data plotted out for all to see. Critical reporting on crime data is vital for all of us to know what's going on, so we can make good choices.

Unfortunately, Los Altos Hills city manager Peter Pirnejad came to some incorrect conclusions from the data. We took a look at some of these claims back in October. In particular, it still seems that he is claiming it is impossible to know if preventative measures have been effective, which is hogwash of course.

Now Pirnejad is saying that crime is down this year, and that it's probably because of the increased funding they've put on policing. 

Spoiler: it seems like LAH has increased funding by at least $100k with no discernable impact on burglary.

The city manager's main point is summarized in the graph in the article, which I'll reproduce here:

credit: Christina Casillas, Town Crier

There are a couple of things going on here. First of all, the data is plotted for fiscal year, not calendar year, presumably to make it easier to compare to budget levels. We don't get total expenditure for each fiscal year in the article. It seems that FY2023 is September 2022 through August 2023.

Second, we don't see any data for the months March through August. Of course, that data doesn't exist yet for FY24, but we lose some important context when we don't see the FY23 data.

The idea, it seems, is that burglary was very high in February 2023, and much lower in February 2024.

But look how small the absolute numbers are! As any sports fan knows, small sample sizes can be deceiving. A variation of just a couple events can make big jumps.

There are statistical ways of dealing with this. The most conventional way is with a something called a Poisson Means Test. This test asks: If it were really true that the crime rate were unchanged, how unlikely is it that we would measure these two values (37 for 2023, and 25 for 2024)? If it's really unlikely, then we have to doubt the notion that crime is unchanged. This number, called a p-value, is usually required to be less than 5% (0.05) in order to reject the idea that crime is constant.

When we use a nice python routine to calculate the p-value, we get 0.13, or 13%. Since that is larger that 0.05, it means this result is not statistically significant.*

That is, when we apply some statistics to the reported data, we can't say that the crime rate is down.

That's entirely consistent with what we saw in the previous post: crime bounces around a bit over the last 5 years, but it's neither up nor down.

What we would like to see is a clear correlation between the amount of money LAH spends on policing, and the crime rate. In fact, what we see is a big increase in spending, and no clear effect on burglary.

Conclusion

In order to evaluate the effectiveness of LAH's program, we need to know:
1. How much difference it made
2. How much it cost

Using the department's data, we find that the difference in burglary rates between 2023 and 2024 was not statistically significant.

We don't have clear numbers for each year, but it looks like the budget was increased by at least $100k in 2024.

The increased expenditure did not lead to a statistically significant change. And that's consistent with the data for the last 6 years, when property crime rates in Los Altos Hills have been fairly constant.

Bottom line: that $100k+ budget increase did not lead to any significant decrease in burglary.

A tangent on data transparency

At council meetings and public forums, police departments have to simplify the data they present to the public. However, if they don't provide a way to get that raw data, it's very difficult to verify their conclusions. The Santa Clara County Sheriffs Department (along with the Los Altos Police Department) does upload their incident data to CityProtect quarterly, which is much better than most departments in the Bay Area. However, their incident records include only a block address ("200 Block Main St") without the city. It takes a fair amount of work to use those addresses to determine a city, and then pull out the Los Altos Hills incidents, and the process isn't perfect.

It sure would be helpful if LAH would release all the data they are using for these results.

Furthermore, it's hard to know exactly what they are looking at. Is this data burglary, exclusive of theft and robbery? How do the trends change if we include all property crime?


Notes

*If you want to try it yourself, use 
res = stats.poisson_means_test(37, 1, 25, 1)
res.statistic, res.pvalue




Comments

Post a Comment

Popular posts from this blog

What's going on in Sunnyvale?

-
The big news on local TV today is the "59% rise in violent crime" in Sunnyvale. That caught my attention, because a) it sounds like another bogus use of statistics and b) as we've seen in Los Altos next door, crime has been generally low and not shooting up. The news report alluded to "California DOJ" data, but failed to actually point to it. Have no fear, gentle reader, for we have trod where KTVU fears to go: straight to the data. The DOJ does, in fact, provide this data, although it takes a little work to get to. You can download it here, if you are interested: https://openjustice.doj.ca.gov/exploration/crime-statistics   Next you have to filter that down to only show Sunnyvale. Excel, Google Sheets, and LibreOffice all have a way of doing that, using "filters." Once you set them up, you get a little pull-down menu for each column, and you can select only the values you want. That's what I did, for each of the 5 cities shown below. Then I plott...
Read more

RIPA Update for Los Altos

-
Image
 The helpful folks at the city recently provided me with the RIPA data through August, so it's a good time to check in on the State of Los Altos Policing for the first half of the year. It's been a fairly quiet summer on the police policy front, with a few high-visibility arrests using ALPRs, and discussion going on about purchasing drones for the police. Here's what we are seeing on the data front -- after all, this is a data blog! RIPA Stops As you all know by now, every police department in California is required to collect data on every single stop that they make. This data, suitably anonymized, is then public. You can get this by submitting a public records request to the City; I've also put it in the github repo, along with all the analysis code, if you'd like to look at it yourself. Here are the stops per month since Los Altos started collecting RIPA data. As you can see, the vast majority of stops are for traffic violations. There are lots of ways you might ...
Read more

Is there bias in our policing?

-
Image
 It was the summer of George Floyd that really shook my confidence in the police. Like most white kids, I was raised on the image of the friendly policeman -- if you ever have trouble, find a policeman (they were always men back then) and he will help you. But as more and more stories emerged... as more and more departments were revealed to have covered up, condoned, and even rewarded the most grotesque racism... and as I really listened to people of color talking about their own experiences, my childhood image was shaken. I respect all of those lived experiences. They are real and true. At the same time, the thing that I can offer to help is data analysis. That's not a very splashy gift; sorry about that. But it's what I have, so it's what I bring. One thing about data analysis is that you really can't do it in a vacuum, without context. It's not a machine, where you pour numbers in at the top and get crisp and unambiguous answers out the bottom. You have to know t...
Read more