News & Updates

RACIAL, GENDER, AND COUNTY DISPARITIES FOR SIMPLE DRUG POSSESSION CONVICTIONS IN WASHINGTON STATE
September 10, 2021 | Alison Pagan

Robin Hardwick | Washington Defender Association
July 7th, 2021

 

INTRODUCTION

This post analyzes court data for individuals charged with simple drug possession (RCW 69.50.4013) in Washington State in a direct response to the recent State vs. Blake Washington State Supreme Court decision, 197 Wn.2d 170, 481 P.3d 521 (2021). In State vs. Blake, the Court decided that both retroactively and prospectively, it is unconstitutional to convict people on possession of a controlled substance. Such a groundbreaking decision has important ramifications for the tens of thousands of individuals convicted of a simple drug possession. These individuals are now eligible for re-sentencing and repayment of legal financial obligations (LFOs) expended on court fees for their simple drug possession charge. The Washington State Department of Corrections estimates the total number of affected individuals in their database (only including people who have been incarcerated) to be 13,516 people—the majority of which are currently incarcerated individuals. However, data from Anthony Powers with the Seattle Clemency Project finds that 126,173 individuals have been convicted of a simple drug possession in Washington State since 1999, when their database begins.

While the State vs. Blake decision is a huge step forward for the defense community, it also begs the question of who the individuals impacted by this decision are. For instance, do we see a disparity in the race, gender, and county of people convicted after controlling for population size? What demographic factors predisposed someone to be convicted of drug possession, and how does this further prove the argument of the unconstitutional nature of 69.50.4013? Who should be prioritized in the process of re-sentencing and repaying LFOs?

In short, this report takes a close look at how the criminalization of drug possession has harmed Washington State residents, especially the most vulnerable populations, shedding light on the racist police and court systems of our state.

DATA, METHODS, AND LABELS

My data analysis relies on information provided by Anthony Powers with the Seattle Clemency Project. The data is in spreadsheet form and aggregated to show the raw counts and percentage of individuals with a drug possession conviction by county and race. Additionally, the spreadsheet includes the raw counts and percent of racial populations by county, as measured in the 2010 US Census. When I later calculate proportion of each racial group impacted and disparity by racial group, I use these provided population measurements.

It is also important to say a brief note about the data’s categories. Race is coded as African American, Asian, Native American, Other, and white. Notably, this measurement excludes folks who may identify as two or more races and does not allow us to understand how Hispanic/Latinx individuals are impacted. The category for gender is binary, thus missing any individuals outside of the gender binary, as well as glossing over the differing experiences of binary transgender people. Lastly, there is no way to determine if the individual convicted was indigent, or of low/working-class socioeconomic status.

The nuance beyond these categories is important and should not be left unexplored or discussed. However, Anthony Powers’ database is constricted to what information is available, and we are constricted to these same missing categories. While we are unable to analyze these factors and the impact they have on drug possession convictions, it is still crucial to call attention to those excluded from this data and advocate for more representative future data collection.

PROPORTION OF EACH RACE AFFECTED

The following section details the proportion of each race’s population impacted by simple drug possession charges. By looking at proportions, I control for the different size of each racial population, allowing us to draw important conclusions about which races are more targeted. For instance, if the same number of Black/African American and white individuals were convicted of drug possession, this would provide strong support for a racial bias in drug possession convictions. Why is this the case if the same number of people are convicted? The answer lies in population demographics, as white individuals make up 69.3% of the state’s population while Black individuals make up 3.7% of the state’s population. Thus, even though it may seem as though there is no racial bias by the number of individuals convicted, we know that a higher proportion of the Black community is impacted, since they are a twentieth of the size of the white population in Washington State.

Additionally, by introducing variables such as gender and county, we are able to analyze how the proportion of each race differs by other important intersections of marginalization.

By Gender

The above graph is our first insight into which racial populations are impacted the most by drug possession convictions. For the two genders recorded, we see that the percent of white and Asian individuals convicted is the smallest. African American women and men of other, African American, and Native American races are impacted the most harshly. Large proportions of these populations have been convicted, with over 12% of men of other races convicted, over 6% of African American men, and about 4.5% of Native American men and African American women. These are large portions of these marginalized communities to have been convicted, especially in comparison to white people, where just 3% of white men and under 1.5% of white women are impacted. Less than 1% of Asian people for both genders are impacted.

Clearly, this graph indicates that there is a racial bias in who cops are searching for drugs, along with who the court system ends up convicting.

By Gender and County (Tables)


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.

 

Next in our journey to uncover more information about the proportion of each race convicted of drug possession in Washington, we begin to look at the difference by county level for women. The above table is arranged for a viewer to see the exact percent of the racial community affected in each county and colored to draw attention to where a large majority certain races are impacted.

Stand out counties here are Adams, Chelan, Columbia, Cowlitz, and Lewis, with the African American community targeted the most consistently. However, in counties such as Adams and Chelan, members of other races face extreme targeting at 17% and 8% of their communities, respectively. It is worth noting here that the percentage of the white community convicted never goes higher than 4% and never over 1% for Asian women—as opposed to the much higher numbers in the other racial columns.


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.

 

This next table also compares race and county, but for men. Here we observe numbers that are much larger than those shown on the previous table, indicating that men are targeted more for drug possession than women. Similar counties stand out in this table, though others emerge as problematic as well—especially if we keep in mind that 8% of a race convicted in a county was high in the last table. As the footnote points out, Adams is exceedingly problematic, convicting more people of other races than they even have in their county. Chelan has convicted 51% of other races, Douglas 23%, Franklin 43%, Grant 42%, and Yakima 37%. These numbers are even more concerning when looking at the percentages for different races in that county, which are much lower. There is clear selection bias when it comes to men of other races.

In addition to men of other races, African American men and Native American men are prejudicially targeted as well. Counties such as Cowlitz, Garfield, and so many others are extremely problematic in their convictions. The gross misrepresentation of minority races in the court rooms here is frightening, and points to a much larger systemic issue than just simple drug possession.

By Gender and County (Maps)

The above map displays the information shown above in a different format. The tiles are colored by the average percent (across racial groups) of that county’s women convicted of a drug possession. This allows us to see what counties are convicting the largest proportion of all their residents. As the map shows, Columbia County has the highest average conviction percent, due to how they have convicted 100% of their African American population. Other counties that stand out as problematic is Lewis, Adams, and Cowlitz.

The above map is the same as before, but this time showing the average conviction percent by county for men. Just as we observed with the earlier tables and bar plot, men are targeted far more for drug possessions, as evidenced by the darker shading across more counties. The same counties emerge as problematic that were identified in the table, including Garfield, Adams, Cowlitz, Chelan, and countless others.

Chi Square Test of Significance

The goal of the following section is to determine if the trends we saw above are, in fact, statistically significant by answering the question: are certain races convicted of drug possession in Washington State at a higher rate than others?

Women

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

Pearson’s Chi-squared test

The above statistical tests prove that there is in fact a statistically significant relationship between race and women who are convicted of drug possession in Washington state. The chi-square test (with a test statistic (𝜒2) of 2,073) proves that it is extremely unlikely that there is not a relationship between race and drug convictions. In fact, the likelihood that there is no relationship between race and who is convicted is essentially zero (, or 0.00000000000000022). Social scientists often reject the null hypothesis when this value is below about 0.10, meaning that this relationship is absolutely significant.

Furthermore, in calculating the effect size (Cramer’s V), we find that 17.90% of the variation observed drug possessions convictions in Washington State can be explained by race. For social scientists, explaining this much of all variation using one variable (race) is a huge finding, as the world is a complicated place that is not easily explained with just one variable. For one variable to explain more than about 5% of how the court system operates—especially given that the court system is founded upon the principles of fair trials—indicates extensive support for a racial bias in the court system wherein people are more likely to be convicted due to the color of their skin.

In short, the relationship between race and drug possession convictions is absolutely statistically significant; please see the table above to look at the breakdown by race. For instance, people who are white, Asian, or of other races are convicted at a lower rate than expected. However, African Americans and Native Americans are convicted much more than expected based on the percent of population they make up (as measured by the census), African American women especially, as they are expected to have about 323.49 convictions considering the total number of convictions and their population percentage. Instead, we see that 2,338 African American women were convicted. This is over seven times the expected amount.

Men

Race Men Actually Convicted Men 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

Pearson’s Chi-squared test

A chi-square test proves the significance of race on men’s drug convictions in Washington state. With a 𝜒2 value of 9,922.8 and a p-value of approximately zero, we find that probability of race not impacting men’s drug convictions is next to nothing. The effect size rises for men, in which race explains 23% of the variation observed in the men who were convicted. Due to the large sample size, the effect size calculation provides a “check” on our results, ensuring that we are getting a small p-value and large test statistic for valid reasons. Such a high Cramer’s V reaffirms this.

Some of the relationships in the above table between expected and observed convictions by race are different for men than for women. For instance, while white and Asian men are the least affected, Native American men also are not convicted at a higher rate than expected given their population size. However, men from other races and African American men are convicted at much higher rates than they should considering the proportion of the population they make up. 6,684 men from other races are convicted when only 3,752.96 should have been convicted if the convictions reflected each race’s population proportion. Additionally, we see that 12,540 African American men have been convicted, while 1,876.48 were expected to be convicted based on their percentage of the population in Washington (2%).

This difference is nearly seven times what it is expected to be. While it is of smaller magnitude than the relationship between observed and expected for African American women, the number of African American men impacted is much larger, indicating a large bias in Washington State for drug convictions by race.

Multiple Linear Regression

A chi-square test of significance is helpful for determining the relationship between one to two categorical variables, as demonstrated above with race. However, when considering the proportion of each race affected, we’ve interrogated not just the effects of race on drug possession convictions, but also the significance of a person’s gender and county. A linear regression allows for these relationships and their statistical significance and impacts on each other in one model. To do this, I use a modeling strategy from Schneiberg and Bartley Table 1 (2001), which looks at how a set of x variables (county, race, and gender) impacts a y variable (percentage of the population convicted).

View the multiple linear regression table for this model in the Appendix (Table 1A). Reading regression tables is a hard skill that takes practice, so do not worry if the table is hard to parse. I will lay out some key observations and tips for reading the table below.

  1. Follow the stars: as the key at the end of the table indicates, all variables that are statistically significant are marked with stars. The more stars, the more significant that variable is.
  • Using this logic, which counties are the most significant? Which races? Is gender significant?
    • Answer: Adams, Chelan, Columbia, Cowlitz, and Garfield Counties, African American and other races, and gender are all statistically significant.
  1. How to read the coefficients: since all of the x variables included are categorical, they are transformed into dummy variables so they can be used in the regression’s numerical equation. In the regression model, the variable corresponding to Adams County would be zero if the county is not Adams, but one if it was. Additionally, one of the categories is always omitted from the regression as a comparison point. For instance, King County, the white racial category, and women are omitted. This means that to read the coefficients, we interpret them as an average of difference in y between the omitted category and the category. Confused? Don’t worry, I’ll write out some interpretations of the coefficients in sentence form below!
  • In Model 1, Adams County convicts on average 11.66% more residents of drug possession than King County, which is a statistically significant difference on the p < 0.01 (probability under 1% of the variables having no relationship) level.
  • In Model 2, African Americans have 3.816% more of their population convicted of drug possession than white people when controlling for county differences, which is a statistically significant difference on the p < 0.01 (probability under 1% of the variables having no relationship) level.
  • In Model 3, men have 3.533% more of their population convicted of drug possession than women when controlling for both county and race, which is a statistically significant difference on the p < 0.01 (probability under 1% of the variables having no relationship) level.

RACIAL DISPARITY OF CONVICTIONS

So far, this report has looked closely at the proportion of each race impacted by drug possession convictions in Washington State. In this next section, I turn to analyzing the disparity of drug convictions by comparing the difference in the percentage of drug possession convictions by race using population percentages.

To explain this relationship more fully, I turn back to an earlier hypothetical, where the same number of Black and white individuals were convicted of drug possession. We determined that this would in fact indicate support of racial bias despite each race having the same number of individuals convicted because white individuals make up 69.3% of the state’s population while Black individuals make up 3.7% of the state’s population. In this hypothetical, 50% of the convictions are of white individuals and 50% of Black individuals. Disparity here is calculated by subtracting 69.3% (percentage of the population that is white) by the 50% of convictions that were white. Thus, the disparity for white people here would be 50 − 69.3 = -19.3. For the Black community, disparity would subtract 3.7% from 50%, becoming 50 − 3.7 = 46.3.

Essentially, the more positive the magnitude of disparity, the more people from that racial group are being convicted than they should when considering the percentage of the population they make up. This allows us to closely analyze trends that may have been hidden in the last section, such as counties who only convict a very small proportion of their population across all racial groups.

By Gender

The above bar plot provides the first glimpse for how disparity can provide an alternative way to visualize difference in conviction rates by race. As we see on the above graph, white men are convicted over 13% less than they should be based on the percentage of the population they make up. People of other races and African Americans pick up the difference, being convicted over 12% and 5% more than they should be, respectfully. Additionally, men see a higher disparity overall all racial categories, while women have less disparity by race. There is still disparity, however, with less Asian women being convicted while African American and Native American women are being convicted more than they should be to make up for the difference in percentage.

By Gender and County (Bar Plot)

This next bar plot no longer allows us to see the difference by race, instead adding together the disparity for all races in each county, allowing us to make comparisons on the county level. Here I took the magnitude (or absolute value) of disparity instead of finding the average, because values are both positive and negative and run the risk of canceling each other out otherwise. Generally, the same counties are the worst across both genders, with King, Adams, and Yakima Counties with the worst results. Looking at magnitude of disparity allows for us to further identify problematic counties, with some being the same ones proven to target racial groups by their population proportion, while some counties emerge as newly worrying—such as King County.

By Gender and County (Tables)


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.

 

While the previous section gave an overview of how magnitude disparity differs by county, this table breaks down by county and race the magnitude disparity women of different racial categories face for drug possession convictions. The darker a cell, the more that county targets that race and has unfairly convicted its members of simple drug possession. Adams County stands out, as white folks convicted 19% less than they should at the expense of other races, while King County convicts 26% more African American individuals than their percentage of the population, and 17% less Asian folks than their percentage of the population. King County’s disparity is important yet startling, as it was invisible when only analyzing the proportion of each racial population convicted, as King County does not convict enough of its larger population to make these trends visible. Instead, we have to take a closer look at the difference in the percentages of racial categories who were convicted to each race’s population percentages for these trends to become visible.


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.

 

Just as above, we see the same counties with large disparities targeting African Americans and members of other races, while convicting fewer Asian people and white people as a result. Additionally, disparity across all counties is greater with men, as the range of percentages increased from this table from the previous, with far more counties reporting disparities larger than 20%.

Analyzing the racial disparity is highly useful, as it allows us to see racist patterns in counties who convict just a small proportion of their entire population, thus masking how conviction unfairly target Native Americans, African Americans, and people of other races. Contributing to this issue are counties such as King, Douglas, and Franklin Counties.

By Gender and County (Maps)

The above map reflects the information shown in the previous table for women by county, though it is shaded by the magnitude disparity. This means that I took the absolute value of the disparity for each race and added it together. For counties whose convictions rates match the percentage of each race in their population the magnitude disparity will be low. However, counties who unfairly convict any race more or less than their prevalence in the population will have higher magnitude disparities. As with the tables, King and Adams County stand out the most, with Yakima and Franklin close behind.

This map shows the magnitude of disparity by county, but for men. As shown by the legend’s increase in magnitude disparity from the previous map, racial disparity is worse for men than women. The same four counties are again the most problematic (Adams, Yakima, King, and Franklin Counties), though much of central Washington also has high magnitudes of disparity. It is important to note that racial disparity habits tend to clump geographically, with western and easternmost Washington colored lighter, while central and northern Washington consistently exhibit more racist conviction patterns.

Multiple Linear Regression

Just as with the proportion of the population convicted, it is crucial to conduct statistical tests to ensure the significance of the relationships observed for magnitude disparity. As the variable of interest here (disparity) is in percentage form, we cannot run a chi-square test, as it would violate one of the test’s assumptions (that data correspond to raw counts and are not transformed into percentages or proportions). Thus, to prove the statistical significance of disparity by county, race, and gender, I run a multiple linear regression with all of these variables. The regression table is printed in the Appendix (Table 2A).

A couple of changes have been made in this table from the previous regression table. First of all, the y variable is no longer percentage convicted of drug possession, but rather the magnitude disparity (absolute value) for each race on the county level. Additionally, since King County is no longer the best county to make comparisons toward due to being one of the worst performing counties in terms of disparity, the omitted county is Asotin due to its small magnitude disparity for all races. White and female are still the other omitted variables.

In summary, Model 1 of the regression proves that certain counties target specific racial groups when convicting their residents of simple drug possession. This includes Adams, Chelan, Douglas, Franklin, Grant, King, Okanogan, Pierce, Snohomish, Whatcom, and Yakima Counties.

Model 2 of the regression confirms that all of the races are statistically significant in terms of disparity while controlling for differences by county. All of the coefficients are negative, as the y variable is the absolute value of disparity—meaning that while all races have lower magnitude disparities on average than white folks, this is because white people have the highest and most consistent negative disparities. In other words, white folks are targeted less than their percentage of the population indicates they should be (affirmed in the above two tables by gender). This gives them the largest magnitude disparity for all of the races.

The regression’s Model 3 confirms what the pattern between the tables and maps have hinted at: men have larger disparities than women on average by 1.518% across all counties and race divisions, significant at the p < 0.05 level. This relationship is held up even when controlling for the different disparities by county and race.

CONCLUSION

In short, this post provides support for a racial, gender, and county bias when considering both the percentage of people convicted of simple drug possession and the disparity between a category’s actual percentage convicted compared to their percentage in the population. By breaking down 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, especially those this report has pointed out as statistically significant in their targeted drug conviction habits by race.

Additionally, while simple drug possession convictions have been made unconstitutional, it is still important for the courts and us as Washington State residents to understand the ways in which communities have been disproportionately impacted by these charges. With this in mind, it is crucial that members of the African American community, Native American individuals, and those of other races are prioritized in re-sentencing, clearing of criminal records, and distribution of LFOs. It is also worth noting that individuals sentenced in Adams, Chelan, Douglas, Franklin, Grant, King, Okanogan, Pierce, Snohomish, Whatcom, and Yakima Counties alongside men were disproportionately impacted and deserve swift and thorough retributions. These steps are necessary in order to begin the long process of restoring and repairing the harm done to them by the 69.50.4013 law.

 

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. Anthony Powers is a formerly incarcerated individual and is currently working alongside a large team of folks to compile a database of information on all cases Washington State courts process, taking care to record demographic information such as gender, race, county, age, etc. This is part of a larger project of making data more accessible in the field of law, while also serving as a strong tool to advocate for much needed change in the court systems.

Please reach out to the author at robhardwi@reed.edu with any questions you may have.

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

 

APPENDIX

Table 1A. Multiple Linear Regression on Percent of Population Convicted of Simple Drug Possession

VARIABLES Model 1: County Model 2: Race Model 3: Gender
Adams 11.66*** 11.66*** 11.66***
(4.198) (4.071) (3.990)
Asotin 2.064 2.064 2.064
(4.198) (4.071) (3.990)
Benton 2.564 2.564 2.564
(4.198) (4.071) (3.990)
Chelan 6.970* 6.970* 6.970*
(4.198) (4.071) (3.990)
Clallam 0.553 0.553 0.553
(4.198) (4.071) (3.990)
Clark 0.669 0.669 0.669
(4.198) (4.071) (3.990)
Columbia 10.54** 10.54** 10.54***
(4.198) (4.071) (3.990)
Cowlitz 6.628 6.628 6.628*
(4.198) (4.071) (3.990)
Douglas 2.067 2.067 2.067
(4.198) (4.071) (3.990)
Ferry -0.660 -0.660 -0.660
(4.198) (4.071) (3.990)
Franklin 5.679 5.679 5.679
(4.198) (4.071) (3.990)
Garfield 9.511** 10.05** 10.05**
(4.452) (4.324) (4.238)
Grant 5.115 5.115 5.115
(4.198) (4.071) (3.990)
Grays Harbor 1.723 1.723 1.723
(4.198) (4.071) (3.990)
Island -0.410 -0.410 -0.410
(4.198) (4.071) (3.990)
Jefferson 0.108 0.108 0.108
(4.198) (4.071) (3.990)
King Omitted Omitted Omitted
(0) (0) (0)
Kitsap 0.944 0.944 0.944
(4.198) (4.071) (3.990)
Kittitas 2.351 2.351 2.351
(4.198) (4.071) (3.990)
Klickitat 2.470 2.470 2.470
(4.198) (4.071) (3.990)
Lewis 4.496 4.496 4.496
(4.198) (4.071) (3.990)
Lincoln 2.406 2.406 2.406
(4.198) (4.071) (3.990)
Mason 1.654 1.654 1.654
(4.198) (4.071) (3.990)
Okanogan 2.780 2.780 2.780
(4.198) (4.071) (3.990)
Pacific 0.921 0.921 0.921
(4.198) (4.071) (3.990)
Pend Oreille -0.513 -0.513 -0.513
(4.198) (4.071) (3.990)
Pierce 1.737 1.737 1.737
(4.198) (4.071) (3.990)
San Juan -0.829 -0.829 -0.829
(4.198) (4.071) (3.990)
Skagit 2.276 2.276 2.276
(4.198) (4.071) (3.990)
Skamania 2.057 2.057 2.057
(4.198) (4.071) (3.990)
Snohomish 0.0352 0.0352 0.0352
(4.198) (4.071) (3.990)
Spokane 2.500 2.500 2.500
(4.198) (4.071) (3.990)
Stevens 0.212 0.212 0.212
(4.198) (4.071) (3.990)
Thurston 1.175 1.175 1.175
(4.198) (4.071) (3.990)
Wahkiakum 0.173 0.173 0.173
(4.198) (4.071) (3.990)
Walla Walla 0.919 0.919 0.919
(4.198) (4.071) (3.990)
Whatcom 2.354 2.354 2.354
(4.198) (4.071) (3.990)
Whitman 0.262 0.262 0.262
(4.198) (4.071) (3.990)
Yakima 4.663 4.663 4.663
(4.198) (4.071) (3.990)
African American 3.816*** 3.816***
(1.470) (1.440)
Asian -1.439 -1.439
(1.458) (1.428)
Native American 0.958 0.958
(1.458) (1.428)
Other 4.873*** 4.873***
(1.458) (1.428)
White Omitted Omitted Omitted
(0) (0) (0)
Female Omitted Omitted Omitted
(0) (0) (0)
Male 3.533***
(0.906)
Constant 1.076 -0.565 -2.332
(2.968) (3.023) (2.997)
Observations 388 388 388
R-squared 0.100 0.163 0.199

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

 

Table 2A. Multiple Linear Regression on Magnitude Disparity

VARIABLES Model 1: County Model 2: Race Model 3: Gender
Adams 13.20*** 13.20*** 13.20***
(2.717) (2.470) (2.447)
Asotin Omitted Omitted Omitted
(0) (0) (0)
Benton 2.100 2.100 2.100
(2.717) (2.470) (2.447)
Chelan 6.400** 6.400*** 6.400***
(2.717) (2.470) (2.447)
Clallam 1.300 1.300 1.300
(2.717) (2.470) (2.447)
Clark 2.500 2.500 2.500
(2.717) (2.470) (2.447)
Columbia 3.500 3.500 3.500
(2.717) (2.470) (2.447)
Cowlitz 1.100 1.100 1.100
(2.717) (2.470) (2.447)
Douglas 4.900* 4.900** 4.900**
(2.717) (2.470) (2.447)
Ferry 4.400 4.400* 4.400*
(2.717) (2.470) (2.447)
Franklin 10.90*** 10.90*** 10.90***
(2.717) (2.470) (2.447)
Garfield 2.867 2.680 2.595
(2.791) (2.538) (2.515)
Grant 6.800** 6.800*** 6.800***
(2.717) (2.470) (2.447)
Grays Harbor 1.300 1.300 1.300
(2.717) (2.470) (2.447)
Island 2.400 2.400 2.400
(2.717) (2.470) (2.447)
Jefferson 1.800 1.800 1.800
(2.717) (2.470) (2.447)
King 11.70*** 11.70*** 11.70***
(2.717) (2.470) (2.447)
Kitsap 3.000 3.000 3.000
(2.717) (2.470) (2.447)
Kittitas 1.400 1.400 1.400
(2.717) (2.470) (2.447)
Klickitat 4.400 4.400* 4.400*
(2.717) (2.470) (2.447)
Lewis 0.900 0.900 0.900
(2.717) (2.470) (2.447)
Lincoln 2.800 2.800 2.800
(2.717) (2.470) (2.447)
Mason 0.600 0.600 0.600
(2.717) (2.470) (2.447)
Okanogan 7.100*** 7.100*** 7.100***
(2.717) (2.470) (2.447)
Pacific 1.200 1.200 1.200
(2.717) (2.470) (2.447)
Pend Oreille 1.600 1.600 1.600
(2.717) (2.470) (2.447)
Pierce 5.400** 5.400** 5.400**
(2.717) (2.470) (2.447)
San Juan 1.300 1.300 1.300
(2.717) (2.470) (2.447)
Skagit 3.900 3.900 3.900
(2.717) (2.470) (2.447)
Skamania 1.400 1.400 1.400
(2.717) (2.470) (2.447)
Snohomish 4.900* 4.900** 4.900**
(2.717) (2.470) (2.447)
Spokane 2.900 2.900 2.900
(2.717) (2.470) (2.447)
Stevens 2.200 2.200 2.200
(2.717) (2.470) (2.447)
Thurston 2.600 2.600 2.600
(2.717) (2.470) (2.447)
Wahkiakum 1.100 1.100 1.100
(2.717) (2.470) (2.447)
Walla Walla 1.300 1.300 1.300
(2.717) (2.470) (2.447)
Whatcom 4.800* 4.800* 4.800*
(2.717) (2.470) (2.447)
Whitman 3.900 3.900 3.900
(2.717) (2.470) (2.447)
Yakima 11.60*** 11.60*** 11.60***
(2.717) (2.470) (2.447)
African American -5.618*** -5.629***
(0.887) (0.879)
Asian -5.744*** -5.744***
(0.884) (0.876)
Native American -6.244*** -6.244***
(0.884) (0.876)
Other -2.077** -2.077**
(0.884) (0.876)
White Omitted Omitted Omitted
(0) (0) (0)
Female Omitted Omitted Omitted
(0) (0) (0)
Male 1.518***
(0.555)
Constant 0.800 4.736** 3.979**
(1.921) (1.834) (1.838)
Observations 389 389 389
R-squared 0.241 0.380 0.393

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1