NYPD stop-and-frisk numbers drop, racial bias remains

[jsoma]

NYPD stop-and-frisk numbers drop, racial bias remains

[jsoma]

Can we get any more data before Jan 2015? I feel like it could tell a better story with more time, like: http://www.nyclu.org/content/stop-and-frisk-data (and who cares if it's already been done, theirs isn't very nice looking.)

I know there's not much time, but if you happen to have the race breakdown I might give a shot at doing it on a race basis, since it's very much a race issue.

[gcgruen]

I have started on a racial breakdown, but the charts were even more ugly, that's why I haven't posted them yet. But yeah, that's totally up on the table

[gcgruen]

NYPD stop-and-frisk numbers drop, racial bias remains [STORY]

Arbitrary and racially biased: The NYPD's stop-and-frisk practice received a lot of criticism in the past. Data show a decline in practice, however the racial bias prevails.

After heavy criticism in 2011 and a 2013 ruling declaring NYPD's stop-and-frisk as unconstitutional, data proves a sharp decline of on the practice:

However, the racial bias - one of the critics' main argument - remains unchanged: Still black and hispanic people are stopped disproportionally more often than white people.

On top of that, after being stopped, black and hispanic people are also frisked more often than white people.

Each data-entry is assigned a "crime code description", suggesting what a person was stopped for. The five most frequent ones in 2015 are 'criminal possession of a weapon' (30% of stops), 'robbery' (14% of stops), 'grand larceny auto' (10% of stops), 'bulgary' (9% of stops), and 'assault' (6% of stops).

The data records do not detail if stops are based suspicion of these crimes or based on evidence.

However, 2015 data suggest that out of 100 people stopped, only 17 would be arrested, whereas only 2 would be summonsed -- proving the point of the practice still being arbitrary. One possible reason might be that not yet all officers are aware of the reform, analysts argue.

If you are wondering what offenses people got arrested or summonsed for -- you unfortunately have to keep wondering: data on reasons for arrests and summons are gathered non-systematically.

Data Source: NYPD

Story issue checklist

My pitch was (use the number): #37

• My pitch has been approved (see PITCHING.md)
• My story issue links to my pitch issue
• I have included an update of my visualization/story in a comment
• I have received two comments of peer feedback
• I have received editorial feedback

[gcgruen]

Moved checklist to initial post

[raschuetz]

I love it! You did a great job with this. I'm trying to think of ways you can deal with some of the data you mention in the last few paragraphs, but nothing helpful comes to mind. I'll think on it.

Purely visually, as I read the graph, at first I thought you didn't list the data source, but then I realized it was at the top. Personally, I'd expect them to be close to one another—but that's just one person.

Also, I think the pastel palette you're using for the first two graphics works so nicely that I almost don't want to deviate from it with the last one. One workaround I could think of is drawing the outlines of the pie chart in the pastel color, leaving the unfrisked slice unfilled, and shading in the frisked slice with the pastel color? Depends on your aesthetic, but then it'd have the added benefit of lacking black outlines, which neither of your first graphs have. Additionally, for the pie charts, your descriptions under the pie charts are so clear, I didn't need a key—and I think you've shaded the wrong part of the pie for white people. If it's 44%, the small slice should be shaded, which would make the message pretty striking.

[gcgruen]

Dear Rebecca, that you very much for your thoughtful feedback!

I have a follow-up-question: You say you would expect the data source "close to one another" -- close to what/where would you expect it to be? I can easily change that, just couldn't find a spot where it didn't look weird to me ;)

Thanks for liking my colour palette! I was in fact unsure, whether it is too "happy" for such a serious topic. The palette is actually the same for all graphics, for the second one, the opacity is set to 60 percent -- reocurring on the third graphic, in the non-frisked part.

I like that you bring the coloring issue up with the pies -- I thought a lot about it and am not happy with the current solution either.
You are right, the black lines are out of scheme, I replaced them with white.
With the actual filling: the idea was to take the colour from graph 2 equalling the total amount -- and highlighting the share of that that's being frisked. That's how I ended up with the two different shadings. If I make them the same shading it looks as if in graph 2 the bar would equal people frisked, not all searched. Which is also is not correct. But I reworked it and hope this is a good compromise!

BTW the key was in response to editorial feedback I got in lab from Stephan ;)

[barjacks]

I really like the way you have approached the story and the variety of charts you have chosen. Clear message and tidy layout. I do, however, have a little problem with your colours. To me they do seem a little too 'happy'. And I would try choosing exactly the same colour in the pie charts as in the bar graph.

Looking more closely at the pie charts they also seem a little out of proportion. There are approx. 5 x more black people stopped than white people. But the pie looks a lot more than 5x bigger.
But maybe it's correct. Volume comparisons such a tricky thing to read sometimes, I think.

I also had to go on a little bit of a hunt for the data source. Why did you decide separate your byline from the source?

[gcgruen]

Dear barjacks! Thanks a lot for your comment!

Added source and byline in the same place -- thank you :)

Also, I found a mistake in graphic 2! I ignored "all others" in the stacked bar chart

Pie chart: YOU WERE SO RIGHT about the size! Thanks for highlighting that! Corrected:

With the colors of the pie: lots of dicussion on this one.
I cannot make them exactly the same color/opacity, because it could be misread as referring to the same number of people in the second and third chart.

However, the highlighted part in the third graphic is only a share of the people of in the second graph -- example: blue part in graphic 2 == all black people being stopped. in graphic three: whole bubble == all black people being stopped. highlighted part == those being not only stopped, but also frisked (vs the non-highlighted part of those being stopped, but not frisked)

That's why the color of second and third has to somehow be different, but also related, racewise.

[snajmabadi]

@gcgruen, I really enjoyed looking at these graphics, and I like how much information is contained in the three of them together. Just a few suggestions: I would consider reordering the legend in the second graphic to have the same color order as the bar chart. Someone might read your graphic from the top down (headline, to caption, to the light blue color, to the yellow color, to the red color, to the navy color), so it could make it faster if the legend mimicked the order on the rest of the chart.

Also, you might want to consider using a similar or neutral color for the line graph -- since the three graphs are all part of one package, a matching color scheme could help it look more cohesive. (For the line graph, I wouldn't use the same navy/red/yellow colors you use for the race breakdowns, as that might be confusing. Maybe a neutral color, or a complementary one from https://color.adobe.com/create/color-wheel/ ??). Great job!

[gcgruen]

Dear @snajmabadi Thanks for your feedback! The colour of the line graph is part of the color palette used in the other graphs.
palette.pdf
But maybe I should think of setting the opacity back to 60 percent as done in graph 2.

It would look like this then:

Also changed the legend, thanks for the hint!

[jsoma]

Beautiful!

One last tweak I'd make is to put a filled black circle where the annotation lines hit the graphed line, just to really solidify where it's touching down. You might also want to center the years under the bars, and experiment with pushing the y-axis up against the bars, with the same amount of spacing as between each bar (1 pixel?).

[gcgruen]

Updated both charts, thanks for the input!

[jsoma]

Merged PR #62

[playfairbot]

Closing since pull request #62 has been accepted