Racial and Gender Disparities for Simple Drug Possession Convictions in Washington State

Robin Hardwick, Washington Defender Association

June 29, 2021

Introduction

Our presentation pulls data from individuals charged with simple drug possession (RCW 69.50.4013)
In light of the Blake decision, this presentation is an investigation through data into which counties and communities were and are the most impacted by the criminalization of drug possession.
- The answer to this question also shows us which communities should be prioritized in re-sentencing and LFO efforts.
126,173 individuals have been convicted of a simple drug possession in Washington since 1999
Data taken from Anthony Powers at the Seattle Clemency Project
- Link to the data spreadsheet
- Link to the blog post version of this presentation

A Note About Labels

Race: coded as African American, Asian, Native American, Other, and white
Gender: coded as male or female
No coding of socioeconomic status (our indigent communities)

Part I: Proportion of Each Race Affected

The first part of this presentation controls for the different size of each county and race’s population by looking at proportions.
Example: If the same number of Black people are convicted as white people, this would provide strong support for a racial bias in drug possession convictions. Why is this the case if the same number of people from each racial group is convicted?
- Answer: Population demographics and disproportionate impacts!

Part I: Proportion of Race By Gender

Part I: Proportion of Race By Gender and County

Percent of Female Drug Possession Convictions by Race in Washington State
Data provided by Anthony Powers with the Seattle Clemency Project
County	Percent of Racial Group Convicted of Drug Possession¹
County	African American	Asian	Native American	Other	White
Adams	1%	0%	0%	17%	1%
Asotin	0%	1%	3%	1%	2%
Benton	3%	0%	1%	2%	2%
Chelan	1%	0%	2%	8%	2%
Clallam	1%	0%	2%	0%	1%
Clark	1%	0%	1%	0%	1%
Columbia	100%²	0%	0%	0%	1%
Cowlitz	8%	0%	1%	2%	4%
Douglas	0%	0%	1%	2%	1%
Ferry	0%	0%	1%	0%	1%
Franklin	5%	0%	1%	4%	1%
Garfield	Inf%³	0%	0%	0%	2%
Grant	1%	0%	2%	4%	1%
Grays Harbor	2%	0%	3%	0%	2%
Island	1%	0%	0%	0%	0%
Jefferson	2%	0%	0%	0%	1%
King	1%	0%	1%	0%	0%
Kitsap	2%	1%	2%	0%	1%
Kittitas	2%	0%	2%	1%	1%
Klickitat	4%	0%	4%	2%	1%
Lewis	10%	1%	3%	2%	3%
Lincoln	5%	0%	2%	1%	1%
Mason	5%	0%	3%	1%	2%
Okanogan	3%	0%	2%	1%	1%
Pacific	0%	0%	1%	1%	2%
Pend Oreille	0%	0%	0%	0%	1%
Pierce	2%	0%	3%	0%	1%
San Juan	0%	0%	0%	0%	0%
Skagit	4%	0%	2%	2%	1%
Skamania	0%	0%	2%	0%	1%
Snohomish	1%	0%	1%	0%	1%
Spokane	4%	0%	5%	0%	1%
Stevens	1%	0%	0%	0%	1%
Thurston	1%	0%	3%	0%	1%
Wahkiakum	0%	0%	3%	0%	1%
Walla Walla	1%	0%	1%	1%	1%
Whatcom	3%	0%	4%	1%	1%
Whitman	0%	0%	1%	0%	0%
Yakima	2%	0%	1%	4%	1%
¹ Cell values correspond to the percent of the racial community in that county (as measured by the census) that have been convicted of a drug possession. Thus, a more positive (darker blue) value indicates that women from that racial group are being targeted for drug possessions over other races. ² Only two residents of Columbia County were African American, as measured in the 2010 census—a number that is subject to change in the past ten years. Two African American folks have been convicted of drug possession, meaning the percent of African Americans convicted is 100%. ³ The census did not record any residents as African American during the 2010 census, meaning that we cannot be sure what percent of the race's population was convicted of drug possession, as there is no record of African Americans living in Garfield. This is quite telling, however, of how African Americans are targeted more in Garfield than other races.

Part I: Proportion of Race By Gender and County

Percent of Male Drug Possession Convictions by Race in Washington State
Data provided by Anthony Powers with the Seattle Clemency Project
County	Percent of Racial Group Convicted of Drug Possession¹
County	African American	Asian	Native American	Other	White
Adams	2%	2%	0%	103%²	2%
Asotin	12%	0%	6%	3%	4%
Benton	10%	1%	1%	14%	4%
Chelan	10%	1%	2%	51%	3%
Clallam	3%	0%	3%	2%	3%
Clark	8%	1%	1%	1%	3%
Columbia	0%	0%	3%	10%	2%
Cowlitz	31%	5%	3%	14%	9%
Douglas	2%	0%	1%	23%	2%
Ferry	0%	0%	0%	1%	2%
Franklin	11%	1%	1%	43%	2%
Garfield	NaN³	0%	60%	19%	4%
Grant	7%	1%	2%	42%	3%
Grays Harbor	3%	2%	4%	7%	5%
Island	2%	1%	0%	1%	1%
Jefferson	3%	1%	2%	1%	2%
King	6%	0%	1%	1%	1%
Kitsap	7%	2%	2%	1%	3%
Kittitas	11%	2%	4%	7%	3%
Klickitat	4%	1%	8%	10%	2%
Lewis	13%	2%	2%	14%	6%
Lincoln	12%	4%	7%	1%	1%
Mason	2%	1%	5%	5%	4%
Okanogan	3%	1%	5%	19%	2%
Pacific	1%	3%	2%	4%	5%
Pend Oreille	0%	0%	2%	0%	2%
Pierce	11%	2%	5%	1%	3%
San Juan	0%	0%	0%	2%	1%
Skagit	6%	1%	2%	14%	2%
Skamania	13%	0%	7%	5%	3%
Snohomish	4%	0%	2%	1%	2%
Spokane	14%	1%	7%	1%	3%
Stevens	6%	0%	1%	0%	2%
Thurston	6%	1%	4%	2%	4%
Wahkiakum	0%	0%	0%	5%	3%
Walla Walla	4%	1%	0%	9%	3%
Whatcom	11%	1%	6%	5%	2%
Whitman	3%	0%	5%	2%	1%
Yakima	8%	0%	2%	37%	2%
¹ Cell values correspond to the percent of the racial community in that county (as measured by the census) that have been convicted of a drug possession. Thus, a more positive (darker blue) value indicates that women from that racial group are being targeted for drug possessions over other races. ² In Adams, the 2010 census recorded 130 men who indicated their race as other. However, 134 drug convictions for males whose race is recorded as other have occurred in Adams County. This means that the entire population (give or take a few folks who may have moved to the county in the past ten years) has effectively been convicted of a drug possession. This is why the number is over 100% here—it is not a mistake, rather, it is a frightening reflection of the reality folks of other races have to experience every day in Adams County. ³ The census did not record any residents as African American during the 2010 census, and no drug convictions were given to male African Americans. However, dividing by a zero produces an error, which is why we see a 'NaN' value.

Part I: Chi Square Test of Significance

The goal of a chi square test is to determine if the relationship between categorical variable(s) are statistically significant.
- Chi square tests do this by comparing the expected counts of individuals based on their population proportion and the actual observed counts.
- Reports a p-value: the smaller the p-value, the more evidence we have for the relationship.
We can also run a function to get the Cramer’s V, which reports the proportion of variation in the observed values that can be explained by our categorical variable.

Chi Square for Women

Table of Observed and Expected Values for Women Convicted of Simple Drug Possession by Race

##                  women_actually_convicted women_expected_to_be_convicted
## African American                     2338                         323.49
## Asian                                 459                        1293.96
## Native American                      1282                         970.47
## Other                                 860                        1293.96
## White                               27410                       28467.12

Chi Square Test

## 
##  Pearson's Chi-squared test
## 
## data:  fem_all_table
## X-squared = 2073, df = 4, p-value < 2.2e-16

Cramer’s V

## [1] 0.1790015

Chi Square for Men

Table of Observed and Expected Values for Men Convicted of Simple Drug Possession by Race

##                  men_convicted men_expected_to_be_convicted
## African American         12540                      1876.48
## Asian                     1946                      2814.72
## Native American           2054                      2814.72
## Other                     6684                      3752.96
## White                    70600                     82565.12

Chi Square Test

## 
##  Pearson's Chi-squared test
## 
## data:  men_all_table
## X-squared = 9922.8, df = 4, p-value < 2.2e-16

Cramer’s V

## [1] 0.2299558

Part I: Multiple Linear Regression

Chi Square tests are helpful when looking at relationships between 1-2 categorical variables.
However, Part I of this presentation looked not just at race, but at gender and county.
- Questions: How do all of these marginalized identities interact with one another? What if we want to identify how much more certain populations are likely to be convicted of simple drug possession?
A multiple linear regression creates ONE MODEL to look at all of these variables together.
- To do this, I look at the effects of a set of x variables (county, race, and gender) on a y variable (percentage of the population convicted).

Part I: Regression Results

Reading regression tables is a hard skill that takes a lot of practice! If you want to learn how to read regression tables and see the ones made with this data, check out the blog post!
We find that:
- Adams, Chelan, Columbia, Cowlitz, and Garfield Counties are statistically significant.
- African American and other races are statistically significant.
- Gender is significant.

Part II: Racial Disparity of Convictions

This section turns to looking at the disparity of drug convictions by comparing the difference in the percentage of drug possession convictions of each race by the population percentage of that race.
To understand disparity, imagine that 50% of the convictions are white and 50% are Black people.
- Thus, the disparity for white people here would be 50 – 69.3= -19.3. For the Black community, disparity would subtract 3.7% from the 50%, becoming 50 − 3.7 = 46.3.
The more positive the magnitude of disparity, the more people from that racial group are being convicted than they should be.

Part II: Racial Disparity By Gender and Race

Part II: Racial Disparity By Gender and County

Disparity of Drug Convictions By Race (for Women) in Washington State
Data provided by Anthony Powers with the Seattle Clemency Project
County	Disparity by Race¹
County	African American	Asian	Native American	Other	White
Adams	−1%	−2%	−3%	25%	−19%
Asotin	−1%	0%	1%	−1%	2%
Benton	1%	−4%	0%	1%	2%
Chelan	−1%	−2%	0%	9%	−8%
Clallam	0%	−2%	4%	−4%	2%
Clark	0%	−5%	0%	−4%	9%
Columbia	13%	−2%	−2%	−3%	−5%
Cowlitz	1%	−2%	−1%	−2%	5%
Douglas	−1%	−1%	−1%	4%	−2%
Ferry	0%	−2%	−1%	−6%	9%
Franklin	8%	−2%	0%	12%	−18%
Garfield	4%	−2%	−1%	−2%	0%
Grant	0%	−1%	2%	8%	−10%
Grays Harbor	0%	−2%	4%	−4%	2%
Island	4%	−6%	−1%	−4%	7%
Jefferson	0%	−3%	−2%	−3%	8%
King	26%	−17%	2%	−5%	−6%
Kitsap	2%	−4%	0%	−6%	8%
Kittitas	1%	−2%	2%	0%	0%
Klickitat	2%	−1%	6%	1%	−7%
Lewis	1%	−2%	0%	−2%	2%
Lincoln	2%	−1%	3%	2%	−5%
Mason	1%	−2%	4%	−2%	−1%
Okanogan	1%	−2%	14%	1%	−14%
Pacific	−1%	−3%	−1%	−1%	6%
Pend Oreille	0%	−1%	−4%	−2%	7%
Pierce	8%	−7%	2%	−7%	4%
San Juan	0%	−2%	−1%	−3%	6%
Skagit	2%	−2%	3%	2%	−6%
Skamania	0%	−2%	1%	−2%	3%
Snohomish	2%	−11%	0%	−5%	13%
Spokane	3%	−3%	5%	−4%	−3%
Stevens	0%	−1%	−3%	−4%	8%
Thurston	0%	−6%	1%	−4%	9%
Wahkiakum	0%	−2%	3%	−5%	4%
Walla Walla	0%	−2%	1%	0%	3%
Whatcom	2%	−4%	12%	−1%	−9%
Whitman	−2%	−10%	2%	−2%	11%
Yakima	3%	−2%	2%	16%	−19%
¹ Cell values correspond to the difference between the percent of women convicted and the percent female population in that county. Thus, a positive (darker blue) value indicates that more women from that racial group are being convicted than they should be, considering the percentage of the population they represent in the county.

Part II: Racial Disparity By Gender and County

Disparity of Drug Convictions By Race (for Men) in Washington State
Data provided by Anthony Powers with the Seattle Clemency Project
County	Disparity by Race¹
County	African American	Asian	Native American	Other	White
Adams	−2%	0%	−4%	45%	−39%
Asotin	1%	−1%	0%	−1%	0%
Benton	3%	−2%	−1%	8%	−7%
Chelan	1%	−1%	−1%	25%	−24%
Clallam	−1%	−2%	1%	−2%	3%
Clark	3%	−4%	−1%	−3%	4%
Columbia	−1%	−2%	0%	9%	−6%
Cowlitz	2%	−1%	−1%	2%	−2%
Douglas	0%	−1%	−1%	24%	−22%
Ferry	−1%	−1%	−12%	−3%	17%
Franklin	7%	−2%	−1%	32%	−35%
Garfield	0%	−1%	7%	7%	−13%
Grant	2%	−1%	−1%	26%	−25%
Grays Harbor	−1%	−1%	−2%	2%	3%
Island	4%	−3%	−1%	−1%	1%
Jefferson	1%	−1%	−1%	−3%	4%
King	35%	−13%	0%	−2%	−19%
Kitsap	6%	−3%	−1%	−6%	2%
Kittitas	3%	−1%	1%	5%	−7%
Klickitat	0%	−1%	6%	11%	−17%
Lewis	1%	−1%	−1%	4%	−3%
Lincoln	4%	1%	6%	−1%	−11%
Mason	−1%	−1%	1%	1%	0%
Okanogan	0%	−1%	6%	17%	−23%
Pacific	−1%	0%	−2%	−1%	4%
Pend Oreille	−1%	−1%	−1%	−2%	5%
Pierce	17%	−4%	0%	−5%	−8%
San Juan	−1%	−1%	−1%	4%	−2%
Skagit	2%	−1%	−1%	15%	−13%
Skamania	2%	−1%	2%	3%	−6%
Snohomish	6%	−9%	−1%	−3%	7%
Spokane	8%	−2%	2%	−3%	−4%
Stevens	1%	−1%	−3%	−3%	6%
Thurston	2%	−5%	0%	−3%	4%
Wahkiakum	0%	−1%	−1%	2%	1%
Walla Walla	1%	−2%	−1%	6%	−5%
Whatcom	5%	−3%	5%	4%	−11%
Whitman	4%	−10%	3%	1%	2%
Yakima	3%	−2%	−2%	38%	−37%
¹ Cell values correspond to the difference between the percent of men convicted and the percent male population in that county. Thus, a positive (darker blue) value indicates that more men from that racial group are being convicted than they should be, considering the percentage of the population they represent in the county.

Part II: Racial Disparity By Gender and County

Part II: Multiple Linear Regressions

Percent disparity violates one of the assumptions for a Chi Square test (that values must be counts, not percents), meaning we skip right to a regression for determining significance.
Results show that:
- We are confident that Adams, Chelan, Douglas, Franklin, Grant, King, Okanogan, Pierce, Snohomish, Whatcom, and Yakima Counties target specific racial groups.
- All of the races are statistically significant in terms of disparity; white folks have the highest magnitude disparity.
- Gender is significant.

Conclusion

There are racial, gender, and county biases when looking at both ways to measure drug possession.
By interrogating how different races are treated in various counties, these data reaffirm the importance of holding counties accountable for their actions and the furthering of racist actions.
While simple drug possession convictions have been made unconstitutional, it is still important for the courts to understand the ways in which specific communities have been disproportionately impacted by these charges.

Bibliography

The author would like to thank Anthony Powers with the Seattle Clemency Project for providing the data that motivated and allowed for the creation of all the visualizations and statistical tests.
And XKCD for their lovely comics.

This study was supported in part by the Paul K. Richter & Evalyn Elizabeth Cook Richter Memorial Fund.