A better title for this website might be "Where are crimes reported in SF?" -- it does not help us understand or contextualize how "safe" SF is overall compared to anything else, or even understand the types of crimes reported and whether that impacts your personal safety.
I think the implied effect was to understand how safe a neighborhood in SF is. Partially because I moved here a week ago and didn't really understand what to avoid. This was a quick project, but yes, showing the type of crimes is important. Though generally, most crimes affect the safety of an area, which is the assumption I went with
Classic crime reporting pet peeve: Market & Powell is right in front of the Westfield. Obviously there's more property crime there.
When Police would over rely on heatmaps for their patrol routes, they always ended up spending too much time in front of Walmarts and not enough in neighborhoods.
Anecdotally, crime in SF is far worse than what's reported. Police won't even show up for your smashed windows.
Last I was there, a Burger King had smashed windows and was serving customers like nothing happened. Didn’t even clean the shards, just put some yellow tape in the area.
Something looks wrong with the month-by-month graph... I'd expect December and January to be close to each other, but they are radically discontinuous. That's a priori very surprising.
Two of the intersections mentioned (boardman & Bryant, eddy & jones) are where police stations are. Is there a problem with the data where they just use the location of the police station by default?
It could also be a very practical issue; if the city is known as having crime, it may deter some employees to come working here or they may defend more their right to work from home.
When I went to San Francisco, I was very surprised by the difference between reality, and the picture that the company shared with me.
(And before you say "A-ha, this is proof that the press does not censor itself!", did you know of this specific case before today? How has this not been widely discussed in society for the past seven years?)
Gentrification is also a contributor to the problem but it's hard to tell rich people that without them getting mad and blaming someone/something else.
That may be true but I don't think it has much to do with the reluctance to discuss the issue (or to actively play down the importance of it when discussed.)
The only solution for either side is to move away, and if the rich people refuse to pack up and leave the less fortunate sure as hell aren't going to buy themselves plane tickets and skedaddle.
So it kinda makes sense why people downplay it when the only alternative to moving somewhere else is perpetuating false victimhood.
I'm willing to give the mods the benefit of the doubt that this was flagged in the assumption it would turn into a low-signal mess but I don't agree with the decision given the comments so far don't seem particularly low value. People are discussing page design and data modeling issues. I think it's useful from a technical standpoint. Oh well.
This does not address the safety of SF relative to other areas, adjusted for population and other pertinent factors. SF is wildly safer than the zeitgeist seems to suggest. San Francisco's overall crime rate has been decreasing, not increasing. Sure, we can have a discussion about reporting rates and adjust for those. As far as I've read, homelessness, a separate but interdependent issue, does not necessarily correlate with crime with any r^2 worth mentioning.
> San Francisco's overall crime rate has been decreasing, not increasing.
How much of this is due to reduced reporting? I'm 100% willing to believe that overall crime is decreasing in reality and not just on paper, but there's a pretty convincing narrative that the decline on paper is a result of people just not filing police reports for crimes they know won't be investigated.
Anecdotal, but my neighborhood police chief (Ingleside) recently shared that thefts are down across the board in her reporting area. Seems backed up by aggregate reports: https://www.sfchronicle.com/crime/article/san-francisco-crim... (She attributed it to the plainclothes retail ops they've been doing leading to significant arrests, and the DA prosecuting a bunch of theft over the past year)
You can't know how safe SF is from the official data, because they release criminals without charges. Most recent high profile example: Golden Gate bridge "protesters".
Honest question: Why do so many good companies {OpenAI, Anthropic, etc.} continue to set up offices in the highest-crime neighborhoods of SF (most notably: Mission and the vinicity of Market)?
I feel like they would do their less-physically-tough employees (women, people with disabilities, physically smaller men) a HUGE favor by moving to Potrero Hill or some other safer neighborhood.
I shouldn't have to fear for my physical safety to get to/from work. I live in a very low-crime area for a reason.
A level of grittiness has been long sought after by innovators, artists, writers, and the like. Too much comfort is not good for creative endeavors because you quickly become out of touch.
At first I wanted to criticize this comment as being a gross oversimplification of the city’s crime rates. After some thought though, there is a lot to be desired from the author’s when it comes down to contextualizing the data.
I mean, can the notion of safety really be relayed by showing a person a bar graphs and telling them to avoid the tall cities?
Apparently this is a hobby project though, so I think I’m expecting too much.
I really like the layout of this page (and the design of the authors homepage -- https://www.hudzah.com/) . Feels like it could also be used on more positive content.
> it's clear that some areas experience significantly higher crime rates than others.
(emphasis mine)
"crime rates" — but then the data immediately after that is incident counts. The word "rate" means to measure some quantity against another, usually to find a frequency. And here, that would be per capita.
(The chart Y axis labels are also lopped off. E.g., the max on the one chart for me reads "0,000".)
They come to the conclusion that their crime numbers are reflecting underlying density differences, but then don't redo the analysis with a per-capita control? Seems like, without that, this isn't a very useful presentation.
Yes, they don't actually answer the question of how safe SF is. They say that some areas are more safe than others. Is the worst area in SF unsafe at all, or are the absolute numbers (and per capita) actually low? How unsafe? Who should be concerned about doing what, and when?
One thing to note is that SF's population swells greatly during the day [1] as people commute into the city for work (Less than before the pandemic, but still substantial). So per-capita crime might still not be the best metric to use because it would be misleading.
The list of crimes counted includes: ...Larceny Theft, Malicious Mischief, Assault, Burglary, Motor Vehicle Theft, Fraud, Lost Property, Warrant, Drug Offense, Robbery, Missing Person and more.
It doesn't seem like "Warrant" (someone with a warrant is apprehended?), "Missing Person", or "Drug Offense" are relevant to general safety.
The author concludes with:
We notice this more when analyzing the heatmap of incidents. Incidents across Mission are much more spread out than incidents across Tenderloin. Therefore, it's unfair to say that the neighborhoods are equally dangerous.
I am not sure why it is not fair to say the neighborhoods are equally dangerous! Notably, the author does not tell us what it is fair to say. What does the degree of dispersion tell us about which one is more dangerous?
Why? If the three only crime reported in a neighborhood is a bunch of arrest warrants, why is that less safe than a place that has a bunch of violent crimes reported?
I think you are implying that as neighborhood that has a bunch of criminals is inherently unsafe. But it is only unsafe if those people commit crimes there. There may be a relation, there may not be one.
At the level of a state, there is probably a close connection between warrants served there and crimes committed there, because most people spend most of their time in one state. At the level of neighborhoods, though, there probably isn't, because most people travel across numerous neighborhoods every day and warrants are served wherever it happens that the criminal is found by police or stopped for an unrelated reason.
My thoughts are to represent general crime around the areas represent it's safety. The best way to normalize the data is viewing the crimes per capita, in which case Tenderloin is the highest. It's not fair; simply because the Mission is much larger in area.
Either way, I will be making updates to the site and data with time, this was a quick 1 day project!