> Starting on March 19, 2024, Flock Safety began installing ALPR cameras in various strategic locations across San Francisco. This rollout is expected to take place over the next 90 days. Per 19B ALPR policy, the administration of the Flock ALPR system is the responsibility of the Investigations Bureau.
How did the Flock cameras cause two crime drops before their installation?
The article's note about 2018 is talking about extending backwards, not forwards. It's entirely accurate, and a direct quote from your link.
The chart is trending down by January 2020, changes directions (upwards) right around the March 2020 spot, and again around (down) the July 2023 spot.
The fact that they only have data going back to 2018 means it's hard to say if the pre-COVID stuff was the norm or unusual.
To be super-clear, here's the chart annotated to show that 90 day window (black rectangle) in which the cameras were installed. https://imgur.com/a/i00Gna0
"that drop is obviously in early 2020", to reemphasize, is several years before the cameras got installed.
I read this as 2020 was Covid related drop, it then returned to normal for 2 years, then began dropping again in late 2023. The covid blip is explained by what was going on at the time, nothing since 2023 has any explanation and could be flock
It looks like that rapid increase was a return to pre-covid normal. It never spikes above pre-covid. Given the world was returning to normal, this is precisely what you'd expect most trends to look like, something like in-restaurant dining probably looks similar.
That’s why I said “how I read this chart is ….” I don’t know what pre2018 looked like either. But on this chart, it was the precovid portion.
Nowhere on here am I seeing how covid caused a spike up, that’s what you said though and signifying our differences in reading the chart that was shown.
The data is open, and so we don't have to do the visual reasoning off an imperfect graph. SF Chronicle has done a pretty rare (but I think good journalistic practice) of specifying the source of the data: https://data.sfgov.org/Public-Safety/Police-Department-Incid...
First to match the graph you make sure you pick 'Larceny - From Vehicle' only (there are some others one might argue matter) and ensure you're only counting incidents once (many rows reference the same incident). That lets us recreate the original graph.
When looking at many things I like to look at seasonal effects just to see, and it doesn't look like they are significant here (but you can see the Mar 2020 drop to the next year quite easily which I like): https://wiki.roshangeorge.dev/w/images/2/2e/SFPD_Vehicle_Bre...
I also tried overlaying various line charts but that's useless for visually identifying the break.
I like PELT because it appeals to my sensibilities (you don't say ahead of time how many changepoints you want to find - you set an energy/cost param and let it roll) and it finds that one changepoint. You can have some fun with the other algos and changing the amount of breakpoints or changing the PELT cost function. And then you can have even more fun by excluding 2020 or excluding Mar 2020 onwards or replacing it by estimates from the previous years (quite suspect considering what we're trying to do but hey we're having fun - a bunch of algos all flag Nov 2023 as some moment of truth)
Anyway, anyone curious should download the data. It's pretty straightforward to use and if I goofed up with off-by-one or whatever, you can go see for yourself.
Your analysis also supports a covid trough, not a covid peak and certainly no covid effect. I agree with other commentators suggesting that flock cameras are not the full or even most of the story, but absolutely disagree with the GP that car break-ins are some identifiable covid phenom or that the decrease is merely a post-covid return to normalcy.
Hopefully there was nothing wrong with posting a news article with a graph instead of doing the data analysis myself.
I was avoiding getting into the specifics because rather than tea-leaf-reading a picture one can simply look at the numbers themselves and they cannot support anything but that the one year period immediately following the lockdowns was much lower than the surrounding years.
And I think it was great you shared the news article! For many others, analyses one does oneself are less believable. I prefer doing it myself to convince myself but I wouldn’t expect it to convince others. Here I did it because I wanted to know what the fact is and I always have trouble with picking change points on a bar graph without all the ticks marked.
I put it at this level because it feels supplemental to your link not because it’s a debunking of your comment or whatever though perhaps https://news.ycombinator.com/item?id=47690707 is the best place to do it.
In other words: https://www.youtube.com/watch?v=xSVqLHghLpw