Figured it is a reasonable amount of time to see how that alleged trend of crime rate, especially murders, being “up, way up” has progressed since last checked.
A bad last week for murders but even so the last 28 day period has 3.2% fewer murders than the same 28 day period in 2014 and 18% fewer shooting victims. Year to date shooting victims is roughly the same to 2014 (under 2% difference, 728 to 716) and down over 24% over 5 years. Murders ytd down about 35% over 5 years. The major crimes all together are overall down over 6%.
Murders are up by 10% year over year though. But yes, I agree that with crime overall still dropping, it seems opportunistic for the police to be going after de Blasio.
So lets look at the numbers that went into the report. There were 54 murders this year as compared to 45 last year at this time. If we assume a Null hypothsis that the murder rate is unchanged and that the number of murders assumes a Poisson distribution then we find that there is about a 42% likelihood that we would see this degree of difference between one year and the next. The shooting come a little closer, with 149 vs 126, having a 18.4% chance of occurring just due to random variation. But both of these actually probably over estimate the significance since murders can come in clusters increasing the variance and making it more likely that there is a large change from year to year. It also doesn’t take into account the possibility of cherry picking of data, other factors that might lead to an increase, regression to the mean after a trend of decreasing murder rates, etc.
But St. Louis is not New York is not Houston is not Chicago is not San Francisco et al. If someone can show what one thing all those cities had in common that changed in all of them, and was followed by an increase in the murder rate, and then explain why it affected just the murder rate and not necessarily the rates for other violent crimes, I’d be willing to listen.
You are using numbers from the first two months of the year only, which provoked the breathless conclusions of a trend of up way up (!). The link I provided gives through the week of July 13 to 19, a bit more than half of the year.
Not a great stats person myself but for murders it is 186 to 169 and for shooting victims 728 to 716.
I was trying to counter the up way up argument directly. Your numbers make my point even more strongly.
For murders the probability of getting such a difference or greater by chance is 39.6%, for shootings its 77.2%. Basically these differences are pure noise.
(For all you stat junkies out there, the p-values reported are two sided, based on a binomial distribution with p=0.5, which is the same as joint distribution of two Poissons conditioned on the number of murders observed.)
A rule of thumb is that, for social trends, you should consider at least five years of statistics. When things are changing in society (including crime), it’s rare that the change will be consistently up or down from one year to the next. When you look at a graph of any social phenomenon, it’s unlikely that it will be a consistently upward or downward line. It’s more likely to be a jagged line. It will be necessary to look at five to ten years of statistics to be able to tell what the trend is. There are mathematical methods which allow you to say more precisely how significant an upward or downward movement over a shorter period is, but if you don’t want to have to do the math, you can just use this rule of thumb.
Because in percentage terms, time series of small numbers bounce around like a major-league knuckleball in a fair breeze. A change from one year to the next is often both large in percentage terms, and meaningless by any non-P.R. measure.
Thank you president Trump! MAGA!!! Or maybe they figured out how to police communities without acting like criminality is color coded. Seriously, as much as I hate De Blasio, stop & frisk basically expanded reasonable suspicion to include young black and Hispanic males.
