Posted 1/25/18


Citywide crime statistics are ripe for misuse

     By Julius (Jay) Wachtel. Part I ended on a perhaps surprising note. Poverty and crime may be deeply interconnected, but our analysis of New York City crime data revealed that neither the city’s 2016 total major crime rate nor its change since 2000 were significantly related to the proportion of residents living in poverty.

     NYPD tracks seven categories of major crime: murder, rape, robbery, felony assault, grand larceny, and grand larceny of motor vehicle. Their sum yields an eight measure, “total major crime.” (See table in Part I, below. NYPD reports yearly frequencies and percentage changes. Instead of raw numbers we used population data to generate rates per 100,000 residents.)

     When total major crime didn’t yield the anticipated results we turned to one of its components, felony assault. Its 2016 rate per 100,000 pop. ranged from 0.5 (112th. and 123rd. precincts) to 8.1 (40th. pct.) (Precincts 14, 22 and 41 were excluded from analysis. See Part I). As expected, the mean rates of the ten lowest-felony assault rate districts (0.7) and the ten highest-rate districts (5.8) were significantly different (t=-4.9, p <.001). They also differed markedly as to poverty. That difference was in the expected direction: persons living in poverty comprise 15.8 percent of the population in low felony assault districts and 26 percent in the high rate districts (t=-3.7, p <.002, statistically significant).

     Correlation analysis was used to test the aggregate relationship between felony assault and poverty for all 73 precincts in this study. That revealed a statistically significant relationship in the “positive” direction, meaning that poverty and felony assault increased and decreased in unison (r=.54, p <.000). Here’s the graph (each precinct is a dot):

FAssault Pov

Statistically significant findings were also produced when we tested the relationships between poverty and the remaining violent crimes: robbery (r=.53, p <.000), rape (r=.46, p <.000) and murder (r=.48, p <.000). Poverty and all forms of violent crime went up and down together. There was also a significant positive relationship, of slightly lesser magnitude, between poverty and grand larceny of a motor vehicle (r=.31, p <.007; see comment below). In contrast, ordinary grand larceny (not of a vehicle) had a “negative” relationship with poverty: as one increased, the other decreased (r=-.43, p <.000, statistically significant). Here’s that graph:

GLarceny Pov

We concluded that this was the reason why there was no observable relationship between total major crime and poverty. In New York, larceny of the “grand” kind requires a loss exceeding $1,000. These are presumably more common in affluent areas. As by far the most common form of serious crime, grand larceny’s strong negative relationship with poverty apparently countered the influence of the other factors. (Incidentally, the positive relationship between grand theft of a motor vehicle and poverty is likely caused by the fact that in New York, the theft of any vehicle valued at $100 or more – that’s two zeroes – is “grand.”)

     Clearly, aggregate measures such as total major crime should be used with great caution. Fine. So, just how were the benefits of New York City’s crime drop distributed? Let’s compare crime rates for the ten poorest and ten most well-off precincts at two points in time: 2000 and 2016. (Precincts #14 and #22 were excluded for methodological reasons, and #41 for trustworthiness. See Part I.) We’ll begin with felony assault:

FAssault 10L
FAssault 10H

     These graphs dramatically depict income’s differential effects. In 2016 the mean felony assault rate in the high-poverty precincts was nearly three times that of their well-off counterparts (474.5 v. 162.4, t=4.3, p <.001, a statistically significant difference.) Note that in both sets of precincts, scores clustered in observable groups. Felony assault rates in all but one of the low-poverty precincts topped out at 235.5. Add nearly two-hundred points to that and you’ll reach the lowest score (425.7) in a group of eight high-poverty precincts.

     Poverty-stricken precincts had more lousy news. Excluding the besieged 40th., where the felony assault rate increased 15.8 percent between 2000-2016, its group’s mean decrease of 19.2 percent was less than half the 41.4 percent decrease enjoyed by the low-poverty group. That old saw about “the rich getting richer” seems to apply to felony assaults in the Big Apple.

     Let’s look at the graphs for robbery:


In 2016 the mean robbery rate of the high-poverty precincts was slightly more than twice that of their low-poverty counterparts (333.4 v. 154.1, t=3.5, p <.003, difference statistically significant.) Except for the 18th. (rate=301.5) low-poverty precincts clustered at the lower end of the scale, topping out with the 9th.’s 198.8. One-hundred points later we encounter the trailing edge of a loose group of eight high-poverty precincts, with rates ranging from the 52nd.’s 325.9 to the 4oth.’s skyscraper-worthy 580.3.

     Between 2000-2016 robbery rates declined 66.9 percent overall in low-poverty precincts and 44.5 percent in the high-poverty group. While both trends seem substantial, so was their difference (t=-4.2, p <.001, statistically significant). Rates were also distinctly dispersed: narrowly within low poverty  (range 53.8 to 77.6 percent) and broadly within high poverty (19.9 to 66.8 percent.) Why this difference between differences we don’t know, but such volatility inevitably reminds us of tendencies at NYPD and elsewhere to fudge the numbers (see Part I).

     And then we arrive at murder. This time we’ll begin with the high-poverty precincts:


Let’s skip rates and talk actual counts. In 2016 the range for the high-poverty group was from one murder in the 66th. to twenty-three in the 75th. These two precincts also had the extreme scores in 2000, when there were three killings in the 66th. and forty in the 75th. By 2016 murder receded in all high-poverty precincts but two, the 40th. and 73rd. In both killings ticked up a bit, going from thirteen to fourteen. Murders otherwise fell, most markedly in the 44th. (25-13), the 46th. (23-14), and especially, the 52nd., which plunged from twenty-five in 2000 to only three in 2016. (However, this precinct had twelve murders each in 2013 and 2015, so its numbers are volatile.)

     We won’t sweat the details: for lots (but not all) poor New Yorkers, the murder news seems at least somewhat favorable. Now consider the horrors the wealthier set faced:


Six of the ten low-poverty precincts had zero murders (thus, zero rates) in 2016. Scores for the other four ranged from one killing in the 24th. to five in the 9th. Only two precincts, the 6th. and 78th., scored zero murders in 2000. Others ranged from one killing in the 18th. to four in the 76th. (note that a relatively low population of 43,643 lends its rate an inflated appearance.) Murders during the 2000-2016 period increased in only one low poverty precinct, the 9th., which went from three to four.

Click here for the complete collection of crime & punishment essays

     Glancing at the charts, does it seem that the rich get to ride up front, crime-wise, while the poor are consigned to the caboose? If so, that’s hardly unique to Gotham. Consider Los Angeles. In “Location, Location, Location” we mused about our hometown. Between 2002-2015 murders fell from 656 (rate=17.3 per 100,000) to 279 (rate=7.3), a stunning drop of fifty-seven percent. Now consider two of the dozens of communities that comprise the “City of Angels”: poverty-stricken Florence, pop. 49001, and upscale Westwood, pop. 51485. During 2002-2015 murder in Florence dropped from an appalling twenty-five killings (rate=51.0/100,000) to a merely deplorable eighteen (rate=36.7). Kind of like…New York City’s 44th.! Meanwhile murder in Westwood went up: from zero in 2002 to (yawn) one in 2015, a rate of 1.9. And that resembles…NYC’s 24th!

     Back to New York. Our chart in Part I indicates that between 2000-2016 murders in Gotham fell from 673 (rate 8.4/100,000 pop.) to 335 (rate 3.9.) But let’s look within. In both the downtrodden 40th. (2016 pop. 79,762, poverty 28.2 percent) and the equally challenged 73rd. (pop. 86,468, poverty 28.6 pct.) killings ticked up from twelve to thirteen, yielding rates of 15.3 and 16.2, four times the citywide rate. Meanwhile, in the affluent 18th. (pop. 54,066, poverty 10.3 pct.), murders declined from one to zero (rate of zero) while in the large and fabulously rich 19th. (pop. 208,259, poverty 7.1 pct.) they fell from three to two, generating a rate of, um, one.

     That’s our “point.” New Yorks’ citywide poverty rate is 19.9 percent. As long as it has a sufficient proportion of well-off residents, it can use summary statistics to brag about “great crime drops” until the cows come home. Except that unlike citywide numbers, people aren’t composites. Can we assume that residents of the 40th. and 73rd. precincts feel – or truly are – as well served as those who live in the more fortunate 18th. and 19th.? What do poorer citizens think when they hear Mayor de Blasio boast that his administration has turned crime around? Are they as reassured about things as their wealthier cousins?

     As we suggested in “Location,” it really is about neighborhoods. Aggregating seventy-six precincts because they’re located within a single political boundary, then acting as though the total truly reflects the sum of its parts, is intrinsically deceptive. Actually, when it comes to measuring crime and figuring out what to do about it, the 40th., the 73rd. and a host of other New York City precincts really aren’t in the Big Apple. They’re a part of that other America – you know, the one where the inhabitants of L.A.’s beleaguered Florence district also reside.

Did you enjoy this post?  Be sure to explore the homepage and topical index!

Home   Top   Permalink     Print/Save     Feedback     


Why do Cops Lie?     Be Careful (Part I)     Location, Location, Location     Role Reversal

Good Guy, Bad Guy, Black Guy     Cooking the Books     Liars Figure     Too Much of a Good Thing?

Posted 1/15/18


Is the Big Apple’s extended crime drop all it seems to be?

Five city comparo

     By Julius (Jay) Wachtel. Remember the “Great Crime Drop” of the nineties? Observers trace its origin to the end of a decade-long crack epidemic that burdened America’s poverty-stricken inner cities with unprecedented levels of violence. Once the crack wars subsided the gunplay and body count eased. But the news didn’t remain positive everywhere. In “Location, Location, Location” we identified a number of less-prosperous burgs (e.g., Chicago, St. Louis, Baltimore, Detroit, Newark, Cleveland and Oakland) that have experienced recent increases in violence. Murder in Chicago, for example, soared from 422 to 771 between 2013-2016 (it backed off a bit last year, but only to 650.)

     In some lucky places, though, the crime drop continued. Few have crowed about it as much as New York City, which happily reports that its streets keep getting safer even as lawsuits and Federal intervention have forced cops to curtail the use of aggressive crime-fighting strategies such as stop-and-frisk.

     Indeed, New York City’s numbers look very good. As the above graph shows, its 2016 murder rate of 5.7 per 100,000 pop. was the lowest of America’s five largest cities and just a tick above the U.S. composite rate of 5.3. (Los Angeles was in second place at 7.3. Then came Houston, at 12.9 and Philadelphia, at 18.1. Chicago, with a deplorable 765 murders, brought up the end at 28.1.) Even better, it’s not only killings that are down in the Big Apple: every major crime category has been on a downtrend, reaching levels substantially lower – some far lower – than at the turn of the century:

NYC major cr table

      Year 2016 precinct crime rates were computed using population estimates on the NYPD
      precinct map. Year 2000 crime rates were computed by adjusting for estimated population
      changes in each Borough. For population data sources click here and here.

     What’s responsible for the persistent progress? New York City’s freshly-reelected Mayor and his police commissioner credit innovative law enforcement strategies and improved community relations. But in a recent interview, Franklin Zimring, whose 2011 book “The City That Became Safe” praised  NYPD for reducing crime, called the reasons for its continued decline “utterly mysterious.”

     Causes aside, when it comes to measuring crime, complications abound. Even “winners” may not be all that they seem. As we discussed in “Cooking the Books” and “Liars Figure,” lots of agencies – yes, including NYPD – managed to look good, or better than they should, by creating crime drops with tricks such as downgrading aggravated assaults (which appear in yearly FBI statistics) to simple assaults (which don’t). That problem has apparently not gone away.

NYPD 40 41 42 Fel Aslt

     This graph uses the NYPD’s own data to display 2000-2016 felony assault trends in three highly crime-impacted precincts, the 40th., 41st. and 42nd., all in the Bronx. Just look at that pronounced “U” curve. Soon after cops outed NYPD for fudging stat’s (that happened in 2010) each precinct’s trends reversed. But the 41st.’s return to presumably more accurate reporting was only brief. Between 2013 and 2014 felony assaults in “Fort Apache” plunged from 732 to 353, an inexplicable one-year drop of fifty-two percent. And the good news kept coming, with 347 felony assaults in 2015, 293 in 2016 and a measly 265 in 2017.

Click here for the complete collection of crime & punishment essays

     There is plenty of reason to be wary of NYPD’s numbers. Still, assuming that the 41st.’s recent shenanigans are unusual – we couldn’t find another example nearly as extreme – the city’s post-2000 gains against crime seem compelling. But assuming that they’re (mostly) true, how have they been distributed? Has every citizen of the Big Apple been a winner? Let the quest begin!

     NYPD has seventy-six precincts. Our main data source was NYPD’s 2000-2016 online crime report. (We excluded precincts #14, Times Square and #22, Central Park, for methodological reasons, and #41 because its recent numbers seem untrustworthy.) We also coded each precinct for its official poverty rate by overlaying the city’s 2011-2015 poverty map on NYPD’s precinct map. (For how NYC measures poverty click here.)

     We’ll start with the total major crime category, which combines the seven major offenses. Its 2016 rate per 100,000 pop. ranged from 3.1 (123rd. pct.) to 45.6 (18th. pct., Broadway/show district.) Comparing the means for total major crime of the ten lowest-rate districts (6.25) with the means of the ten highest-rate districts (24.13) yields a statistically significant difference (t=-7.36, sig .000). So these groups’ total major crime levels are different. But their proportion of residents living in poverty is not substantially dissimilar. Actually, the raw results were opposite to what one might expect: the mean poverty rate was higher in the low major crime than the high major crime precincts (19.3 & 15.9, difference statistically non-significant.)

     Similar results were obtained when comparing the 2000-2016 change in the major crime rate of the ten most improved precincts (mean reduction, 62.05%) with the ten least improved precincts (mean reduction, 14.69%). While the magnitude of these groups’ crime decline was significantly different (t=14.37, sig .000), the difference between the proportion of their residents who lived in poverty was slight and statistically non-significant (poverty mean for most improved, 19.28 pct.; for least improved, 21.31 pct.)

     We then (by this point, somewhat unsteadily) ran the numbers the other way, comparing total major crime and its improvement over time between the ten high and ten low poverty precincts. Our central finding didn’t change: poverty wasn’t a significant factor. With all seventy-three precincts in the mix we also tested for relationships between total major crime rate and poverty, and between 2000-2016 changes in the major crime rate and poverty, using the r coefficient. Again, neither total major crime nor its change over time seemed significantly related to poverty.

     So poverty doesn’t matter? New Yorkers are equally likely to benefit from the crime drop – or not – regardless of their place on the pecking order? As it turns out, not exactly. But that’s enough for now. We’ll deliver “the rest of the story” in Part II!

Did you enjoy this post?  Be sure to explore the homepage and topical index!

Home   Top   Permalink     Print/Save     Feedback     


Why do Cops Lie?     Be Careful (II)     Is Crime Up or Down?     Location, Location, Location

Role Reversal     Good Guy, Bad Guy, Black Guy     Cooking the Books     An Inconvenient Truth

Liars Figure     Too Much of a Good Thing?


  Home               About               Index               Links              For educators             Contact          Novel