This story draft by @escholar has not been reviewed by an editor, YET.
Authors:
(1) Yuan Wang, University of Rochester (e-mail: [email protected]);
(2) Yangxin Fan, University of Rochester (e-mail: [email protected]).
WP fatal police shooting dataset insight
Fatal police shooting rate and victims race prediction
Since 2015, The Washington Post (WP) has created a database cataloging every fatal shooting nationwide by a police officer in the line of duty. There have been less than 1,000 people killed by police every year. The killed rate of African American people is disproportionally higher than any other race (use Black or B to distinguish with Asian or A in the following). The Figure-1 show the number of people killed by police shooting by year in national wide. Figure-2 shows the average proportion rate of each race killed by police shooting from WP’s website.
Admittedly, there is no doubt that Black people’s rate is high than any other group of people if we compare it with the population proportion. However, once we add the proportion of violent incidents offenders [12]to each racial group, we see the ratios have matched each other accordingly. See Figure-3 below.
We collected 2472 police shooting victims with known reported media and race from 2016 till now from KBP. We hold our null hypothesis that media reported news by each race should follow the real happened case distribution. We use the racial proportion of victims from WP Data as the ground truth. We selected media with over 100 fatal police shooting news reporting, which include one conservative media, FOX (318) and three liberal media: ABC (244), CBS (227), and NBC (135). The media’s political inclination is showed by Figure-4: Political bias of selected media. We excluded media with less than 100 reporting since most of them are local media whose news may be impacted by the local demographics. Figure-5 shows the comparison results: all four media have different deviations on reporting the truth. In general, Black victims were over-reported among all the media.
To exam the difference of deviation between four media, we defined the measurements and calculation methods:
1. Reporting deviation rate of media B regarding race A = R(B, A) = reported proportion of race A by media B – real proportion of race A in WP Data. If R(B, A) < 0, media B underreports race A victims. Else R(B, A) > 0, media B overreports race A victims.
2. Total absolute reporting deviation rate of media B = = PN i=1 |R(B, Ai)|, Ai is the i-th race and N is the number of races.
We then get Figure-6: Four media reporting deviations. FOX has the least deviation rate from the WP Data, and there are only –3.3% deviation for White, +5.6% for Black, -1.5% for Hispanic, -0.6% for Asian, and –0.6% for Native American, and +0.4% for Other. Nevertheless, ABC, CBS, and NBC have larger reporting deviations, which underreported 10% White victims while overreported Hispanic and African American victims. Specifically, NBC underreported White victims’ proportion by 17.4% and overreported Black victims’ proportion by 15.0%. Furthermore, it even reported more Black victims (41.5%) than whites (33.3%). ABC overreported Hispanic victims’ proportion by 13.4%. It reported Hispanic victims (32.0%) at the same level as White victims (36.1%). The Figure-7 shows four media total absolute deviation rate.
In terms of total absolute reporting proportion error, NBC has the largest reporting deviation rate (39.8%), followed by ABC (35.3%), and CBS (28.9%), while FOX has the least rate (12.0%) shown above.
This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license.