Hacker News Reader

Comment by slg | Hacker News Reader

I don't live in San Francisco. I can't tell you exactly what is going on in the ground, so I don't know who is right here.

However, this comment is falling into the exact situation that OP was trying to warn about. You are using an anecdote to dismiss statistics because the anecdote is a powerful emotional narrative that feels true. It is after all your own personal experience. However, that experience might not be an accurate representation of reality. OP warned to keep that in mind. Your response is effectively saying "no, my emotional narrative feels true so I'm sticking with it."

Maybe you are right and your emotional narrative is true. As I said, I don't know. But I'll tell you if one side of a debate has statistics and the other side has feelings, I think it is smart to side with the side that has statistics.

View on HN

City-wide statistics seem almost entirely meaningless in these contexts. Most violent crime is concentrated in specific parts of the city where the people who participate in these debates never go. A 10% decrease in violent crime in the most dangerous part of a city could cancel out a 10x increase in the areas where most HN people would likely be living in.

Don't forget, there's different kinds of crimes with different visibilities.

So there could be a 50% decrease in violent crime in the most dangerous part of a city, along with a 500% increase in non-violent or petty crimes everywhere else: shoplifting, parked car smash-and-grabs, urinating in the train stations, pooping on the sidewalks, etc. So the tech crowd isn't going to notice fewer people getting murdered in the worst section of town, but they'll definitely see the other stuff.

I can't tell what point you are trying to make. Are you suggesting that while crime overall is down, this new situation is actually worse because the victims are now likely to have more money than past victims?

He is putting more emphasis on his direct experience than in statistics that are undoubtedly cherry-picked and potentially unreliable (by both political extremes). That's not buying into an emotional narrative; that's doing a reality check.

The direct experience of a single person has almost no value when assessing a problem as big as crime rates. It would be like saying cancer rates are up because two people you know got cancer. Direct experience shouldn't be given that much weight.

Direct experience is colored by emotion. I was mugged in the Mission 10 years ago, and it felt shitty, and it took several years for me to feel comfortable walking in some areas of the city again, especially at night. But my feelings were utterly irrelevant to what actual crime rates have been in the city since then.

I agree with you that crime stats can be politicized, but it's equally sketchy to consider n=1 anecdotes as more reliable indicators. Certainly if you yourself experience or are the victim of crime at some particular rate and intensity, then that's suggestive to you, personally. It's entirely logical to make decisions about your own life -- like moving to a safer neighborhood, or avoiding areas in the city where you've experienced crime -- but it's not particularly useful when talking about the city as a whole, or in making general recommendations to residents on how to be safe.

Aggregated year-over-year statistics are cherrypicked but his emotional experience isn't?

Ok, here's my not cherrypicked experience: The issue is massively overblown. It's nowhere near as bad as the sour-grapes living elsewhere would have you believe.

It is a kind of reality check, but as an individual you just see a very small slice of reality, so generalizing from that in space or time is often misleading. And some people cite their personal experience as a way of shutting down debate, ie 'don't tell me about the data on X, I've seen it with my own eyes', and making lengthy impassioned speeches to sideline other points of view.

> You are using an anecdote to dismiss statistics because the anecdote is a powerful emotional narrative that feels true

Statistics can be useful, not useful, or misleading. Nothing about them is inherently meaningful or valuable--it depends entirely on the question and process generating the data and statistic. GIGO applies.

Despite the meme that an anecdote isn't data, it actually is no different from data from the Bayesian point of view. It is an n=1 posterior with a prior of one's past life experiences. And many anecdotes together can be thought of as a multi-level model (if you don't believe this, just see what the methodology of many behavioral/observational studies looks like--it's collating anecdotes into "data"--including the crime statistics that you are after).

> the anecdote is a powerful emotional narrative that feels true

And researchers can absolutely be emotional! A study biased by the emotions and beliefs of a researcher will produce biased results. Statistics isn't an escape hatch from human bias; it actually compounds whatever bias exists in the first place.

Yes, statistics can be misleading and/or biased. Yes, multiple anecdotes become data when combined (which is even more susceptible to being misleading and/or biased). Do you think either of these things are happening in this instance? If so, make that argument. I don't think there is much value in arguing that anecdotes are generally more valuable than statistics because that obviously isn't true overall.

"Has statistics" does a lot of work here. When we weight visible dilapidation of the city against a set of figures showing some metric decreasing - a reasonable question would be how well this statistic reflects the reality on the ground. If we aggregate all crime over all the territory and get this metric to go down - does it mean we are doing great, or does it mean the crime became more concentrated in some places and now we have safe remote havens for the rich and a criminal hellscape for the less fortunate? Or does it mean we stopped reporting some things as "crime" either because the law does not care anymore or because citizens gave up on reporting it because it's useless anyway? I don't say it necessarily means that - I am just saying that you can not consider isolated metrics alone, you should always also consider how well these metrics reflect the underlying reality.

The problem is that official statistics are worse than useless because most victims of minor crimes never file police reports. And I don't think you'll find any statistics at all on the intensity of odors in public transit stations.

In that case, statistics in all cities should be equally as error prone.

Statistics also don't collect how many annoying people are blasting shitty music on subways or how many morons are rolling coal in lifted trucks. There are unpleasant people everywhere.

>most victims of minor crimes never file police reports

Minor crimes like murder?

> The problem is that official statistics are worse than useless because most victims of minor crimes never file police reports.

In that case, statistics should still be useful to observe trends, if they can't be used to determine accurate absolute numbers. We should expect the same percentage of people to not report crime when there are 1,000 car break-ins per month as when there are 100. Say that's 20%; seeing that number change from 200 to 20 is still teaches you about how crime rates change.

Besides, we're talking about murder, here, and I would suspect that pretty much every murder ends up accounted for in the statistics. Contrary to what mafia movies would like us to believe, it's not that easy to hide a body indefinitely.

> The problem is that official statistics are worse than useless because most victims of minor crimes never file police reports.

While urine smells certainly affect quality of life, I'm not convinced this particular statistic is all that relevant when talking about crime or public safety.

> We should expect the same percentage of people to not report crime when there are 1,000 car break-ins per month as when there are 100

This ignores changes in police enforcement and prosecution of crimes. If there is very little overall crime and the police have the time to find the perpetrator, and the prosecutor agrees to bring charges, you're more likely to report the crime. If there is high crime, police are too swamped to deal with yet another break-in, and the DA is too busy dropping felonies to misdemeanors or not bringing charges at all, then people aren't going to bother reporting crime as much anymore.

> In that case, statistics should still be useful to observe trends, if they can't be used to determine accurate absolute numbers. We should expect the same percentage of people to not report crime when there are 1,000 car break-ins per month as when there are 100. Say that's 20%; seeing that number change from 200 to 20 is still teaches you about how crime rates change.

That's definitely not true for car break-ins. In California pretty much every one I knew who lived there experienced some form of car break-in, and of course nobody got their stuff back. If I experienced one there I'd just shrug it off as a fact of life since the process of filing a police report takes time and effort, and the reward is expected to be zero.

In other places with fewer break-ins, I will most likely respond differently.

Reply on news.ycombinator.com