Comment by lambdaphagy | Hacker News Reader

Comment by lambdaphagy | Hacker News Reader

Parent Full thread

It's not my method, it's the calculation of the percentage increase, which is the way of comparing 2021 and 2014's homicide rates that you originally proposed.

The y-axis of the chart you linked records the national homicide rate in units of homicides / 100k capita / yr, the standard units for reporting homicide rates. Those aren't percentages-- they have exactly the dimensions that the chart, and I, are talking about. So it's not correct to refer to "6.81%" in any way, because it's not a percentage. Subtracting the 2021 rate from the 2014 rate gives you a difference of 2.41 homicides / 100k / yr, which is also not a percentage. In order to express that as a percentage increase, you need to divide by the rate against which you're comparing (4.4 homicides / 100k / yr), which gives you ~1.5, i.e. a 50% increase.

We can discuss why that might have happened and whether you consider it notable, but we really can't debate a "2% increase" because that's just mathematically incorrect.

To your broader point, when we make diachronic comparisons of crime there's no particular reason to start the clock in 1990. The homicide rate is still higher than it was in 1950: https://www.statista.com/statistics/187592/death-rate-from-h....

And, crucially, there were many changes between 1950 and 2023 that made it easier to prevent homicides: material technology got way better in the form of smartphones, ubiquitous cameras, DNA evidence, shock trauma units (that turn would-be murders into attempted murders), and an aging population that's less prone to homicide in general. If you took all of those things away and tried to fight 2023's homicide rate with 1950's tech, you'd find we have quite a few thumbs on the scale today.

Yup, sorry for the bad math. Brain fart.

>And, crucially, there were many changes between 1950 and 2023 that made it easier to prevent homicides: material technology got way better in the form of smartphones, ubiquitous cameras, DNA evidence

These are good at solving crimes, but not preventing them. As BLM rioters (not protesters) knew and Jan 6 people didn't is that a simple surgical mask will foil most camera system identification.

>shock trauma units (that turn would-be murders into attempted murders)

This is true, but you still have to get the patient into the hospital in time. Gunshots and stab woulds turn fatal really fast.

>and an aging population that's less prone to homicide in general.

We are aging, but our largest demographic currently is between 19-40 (prime time age).

>If you took all of those things away and tried to fight 2023's homicide rate with 1950's tech, you'd find we have quite a few thumbs on the scale today.

Perhaps a little.

> These are good at solving crimes, but not preventing them.

In the long run, solving crimes is preventing them because crime (especially serious violent crime) is Pareto-distributed.

> This is true, but you still have to get the patient into the hospital in time. Gunshots and stab woulds turn fatal really fast.

The major innovation of shock trauma units just was getting patients to high-end hospitals really fast. In the 1970's, Baltimore began using helicopters to airlift critical patients to a city-wide shock center, which was considered outlandish at the time but paid off: https://en.wikipedia.org/wiki/R_Adams_Cowley_Shock_Trauma_Ce.... The system works well enough that Baltimore now has roughly two non-fatal shootings for every fatal one. That's great for the survivors but tends to mask the scale of the violence, as a lot fewer of those victims would have survived without Dauphin helicopters.

> We are aging, but our largest demographic currently is between 19-40 (prime time age).

This sounds like a statement about one possible way to divide the population into histogram bins in 2023 rather than a comparison of population structure between 1950 and 2023. Since 1950, the most murder-prone fraction of the population (men in their 20's) has decreased by about 10%. (https://www.populationpyramid.net/united-states-of-america/2...). There are various reasonably ways of making that comparison but they all give the same qualitative answer supporting the common-sense observation that the US population pyramid has aged.

>In the long run, solving crimes is preventing them because crime (especially serious violent crime) is Pareto-distributed.

How does that work? If you are saying because the violent criminals are now in jail, that's completely discounting new generations.

>That's great for the survivors but tends to mask the scale of the violence, as a lot fewer of those victims would have survived without Dauphin helicopters.

Ya, I'll buy that.

>This sounds like a statement about one possible way to divide the population into histogram bins in 2023 rather than a comparison of population structure between 1950 and 2023.

I linked both population trees somewhere.

>Since 1950, the most murder-prone fraction of the population (men in their 20's) has decreased by about 10%.

You need to provide some data for that, I don't see how that's possible since it's not accounting for baby boomers or for population growth (unless you mean as a percentage). GenX was quite a small generation, but the Millennia's are quite large.

Great conversation though!

Reply on news.ycombinator.com