Data quality matters

I was reading yesterday on HN: Subway crime plummets as ridership jumps significantly in congestion pricing era. The facts: there were 147 reported crimes on NYC’s subway in a month after congestion pricing was implemented in Manhattan, which is lower than it has been in previous months. However, the top two comments interpret this fact differently:

1. This number is very small given MTA ridership. Media coverage of crime on public transit is overblown. TL;DR: Cities are livable.
2. This number is likely wrong given that crime is often underreported, as well as personal experience of crimes observed on public transit. TL;DR: Cities are unlivable.

Some of my preliminary beliefs (note that I no longer live in any of the cities mentioned below):

• Out of all light rail public transit systems in the United States, NYC’s subway feels like the safest per capita (I don’t know if it actually is). This is in contrast to things like the BART or Portland’s MAX.
• The vast majority of crime goes unreported. This is a non-exhaustive list for the factors that I see:
  • Cops aren’t actually present when crime happens.
  • Citizens lack faith in the justice system to close on the problem.
  • Paired with a lack of faith, contacting law enforcement takes too much time. 911 holds, having to make a statement for cops, etc. If people think that the process of reporting won’t lead to a good outcome, they won’t waste their time on it. And the amount of time to report crime is non-negligible.
  • A population with a tendency towards libertarianism, and a fear of retaliation against criticism, has no collective power to stop anti-social behavior (e.g. playing music out loud on public transit. Let the guy continue doing it? Tell him off and get no response? Tell the cops and get laughed at? Throw his phone out the door and get shanked or arrested?). Our population is jaded by bad actors, so people have learned that the only way to survive is to blot them out as opposed to expect authority or any collective of citizens to solve the problem.
• An increase in population density should decrease crime per-capita. As Jane Jacobs would say, there will be more “eyes on the street” to prevent such behavior.
  • I tend to think that there is a bell curve of density on the X axis and crime per capita on the Y axis, where once you reach some critical density, the amount of crime per-capita begins to drop (although I don’t think it drops to zero).
• I would feel (and have felt) safe on NYC’s subway system. I have felt safe in that I don’t expect to be the victim of, or witness of a violent crime. But I also feel safe in that I don’t often expect to witness anti-social, irrational, or degenerate behavior by other riders like playing music on a speaker, muttering incoherent and angry statements to themselves, or relieving themselves in the corner¹. I might not feel this way if, say, I’m waiting on a platform in outer Harlem or Queens at 2am. But if I’m riding in Manhattan anywhere between 6am and 8pm, I feel safe.
• I wouldn’t define any major public transit system in the United States as “livable.” NYC’s subway is the most livable out of all of them given its efficiency compared to other cities. But crimes still happen on the trains, there are still gross smells all throughout the underground, and the subway tiled environment isn’t a pleasant one. We can, and should, do better.

I think many people expect more from public transit in this country, and they rightfully should. At the same time, there are many people who look at certain statistics, assume that they’re infallible, and come to conclusions that don’t reflect the peoples’ lived experience. This amounts to those people burying their heads in the sand, claiming that they know a truth that could be built on lies.

There is a point at which you have to simply trust a data source as correct. We can’t be critical of every source of info that we’re using to discern truth – that way lies flat-earthism and vaccine skepticism. I’m not suggesting that we actively look for problems in how the data were created. I’m suggesting that we at least try to understand the ways that our data is collected as opposed to assuming that they reflect reality. Through that understanding, we can provide rational critiques on how to collect data better.

Statistics matter, because without them, we’ll all be living off of vibes which leads to horrible outcomes like the most recent US election. However, it isn’t wrong to understand and criticize how data are collected. You ought to try and understand how they’re collected, and note any problems with the process. Data analysis is how we collectively discern truth in an emotionally discharged way. If the dataset is tainted, then the data won’t be the finger pointing at the moon – it will be pointing at nothing.