To illustrate a social trend that amounted to a spectacular crime drop in the Big Apple throughout the 1990s and 2000s—as described by Franklin E. Zimring in the August 2011 issue of Scientific American, we mapped two decades of New York Police Department crime statistics precinct by precinct.

In the first map below, we combined data on the six most serious crime categories into one index. The other maps (click to unroll) illustrate the incidence of each type of crime.

(Find details about individual New York City precincts here.)

How the weighted crime index was derived
In the first series of panels relating to total crime, the color of each precinct represents a combined crime index. This index is the weighted total of the six crime types under study (homicide, rape, assault, robbery, burglary and car theft) reported in that precinct.

"Weighted" means that we didn't simply add up the raw numbers. Instead, for each crime type all numbers were rescaled (multiplied by a common factor) first. In the case of the 1990 statistics, the rescaling was the same as expressing the numbers as percentages of the citywide totals for that crime.

Thus, if in 1990 precinct X had 4 percent of all homicides, 4 percent of all rapes, and so on for each of the other four types of crime, the combined crime index for that precinct would be:

4 + 4 + 4 + 4 + 4 + 4 = 24

(Note that the combined crime index of all precincts totals 600, because the percentage for each crime type adds up to 100 percent citywide, and we are combining six categories.)

For subsequent years the number of crimes of each type in a precinct was divided by the 1990 citywide total for that crime instead of that year's citywide total.

Consequently, if in 2000 precinct X still reported exactly 4 percent of each type of crime committed in the city, but the actual numbers in that precinct had all dropped by half, its combined crime index for 2000 would be 50 percent lower than in 1990:

2 + 2 + 2 + 2 + 2 + 2 = 12

We used this combined crime index rather than simply adding raw numbers for all types of crimes because it provides a clearer comparison of how the numbers in each district stack up relative to the 1990 baseline. Simple addition of the incidence of each crime in a district could have hidden profound variations from precinct to precinct or from year to year, because numbers are vastly different for different crime types (thousands in the case of homicides, for instance, tens of thousands in the case of burglaries). If a precinct went from 10 homicides and 90 car thefts in 1990, say, to zero homicides and 100 car thefts in 2000, the simple addition would yield the same result for both years—but most people would consider the area to have become much safer.

— Davide Catelvecchi


Maps by XNR Productions; SOURCE: NYPD (raw data by precinct)