University of Southern California Dissertations and Theses

An analysis of racial disparity in the distribution of alcohol licenses and retailers in Orange County, California

(USC Thesis Other)

An analysis of racial disparity in the distribution of alcohol licenses and retailers in Orange County, California

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

An Analysis of Racial Disparity in the Distribution of Alcohol Licenses
and Retailers in Orange County, California

by

Kelly Woody Gulledge

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

December 2020

Copyright © 2020 Kelly Woody Gulledge

ii

Dedication

To my wife Linda and mother-in-law Joyce.
With your support everything is possible.

iii

Acknowledgements
A very special thanks to Dr. Darren Ruddell for his advice and support. His tireless feedback and
encouragement kept me on track and made this project possible.

I would also like to thank all the USC GIST faculty and staff that I have had the pleasure to work
with and learn from. This has been a challenging and rewarding experience.

Fight On!

iv

TABLE OF CONTENTS
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Tables ................................................................................................................................. vi
List of Figures ................................................................................................................................. x
List of Abbreviations ................................................................................................................... xiii
Abstract ........................................................................................................................................ xiv
Chapter 1 Introduction .................................................................................................................... 1
1.1 Essential Concepts: Structural Racism and Disparate Distribution .....................................3
1.2 Structural Racism Persists After Racist Policies and Practices Have Ended .......................7
1.3 Racially Neutral Policies May Not Address Structural Racism ..........................................8
1.4 California Alcohol Retail Sales Licensing Regulations .....................................................13
1.5 Study Area: Orange County, California .............................................................................15
1.6 Thesis Objective and Research Questions .........................................................................17
1.7 Thesis Organization ...........................................................................................................18
Chapter 2 Related Work................................................................................................................ 19
2.1 Density of Alcohol Retailers and Disparate Distributions .................................................20
2.2 Spillover and MAUP: Influences of and on the Built Environment ..................................22
2.3 Improving Results and Avoiding MAUP: Two Areal Aggregation Scales .......................25
2.3.1. Scale 1: Census Tracts .............................................................................................25
2.3.2. Scale 2: Scaled Population Grid ..............................................................................28
Chapter 3 Data Sources and Methodology ................................................................................... 31
3.1 OC Alcohol Retailer License Data ....................................................................................34
3.1.1. Acquiring OC Alcohol Retailer License Data .........................................................34
3.1.2. Geocoding Alcohol Retailer License Data ..............................................................35

v

3.1.3. Categorizing Alcohol Retailer Types.......................................................................37
3.1.4. Orange County Alcohol Licenses and Retailer Summary .......................................43
3.2 Spatial Analysis: Two Areal Aggregation Units ...............................................................47
3.2.1. ACS Race/Ethnicity Estimates and Margins of Error..............................................47
3.2.2. Scale 1: Census Tracts .............................................................................................48
3.2.3. Scale 2: Scaled Population Grid ..............................................................................53
3.3 Quantifying Race Neutral and Disparate Distributions .....................................................58
Chapter 4 Results .......................................................................................................................... 62
4.1 Scale 1: Census Tract Analytical Results ..........................................................................62
4.1.1. Census Tract Alcohol License Summary Statistics .................................................63
4.1.2. Census Tract Alcohol License Density ....................................................................85
4.1.3. Census Tract Alcohol License Hot Spots: Getis-Ord Gi* Statistic .........................96
4.2 Scale 2: Scaled Population Grid Analytical Results ........................................................110
4.2.1. Scaled Population Grid Alcohol License Summary Statistics ...............................111
4.2.2. Scaled Population Grid Alcohol License Density .................................................133
4.2.3. Scaled Population Alcohol License Hot Spots: Getis-Ord Gi* Statistic ...............146
4.3 Quantification of Race Neutral and Disparate Distributions ...........................................154
4.4 Census Tracts Versus Scaled Population Grid: Outcome Variations ..............................157
Chapter 5 Conclusion .................................................................................................................. 159
5.1 Finding Disparate Distributions of Alcohol Licenses in Orange County ........................159
5.2 Limitations of the Study ...................................................................................................162
5.3 Areas for Future Study .....................................................................................................162
References ................................................................................................................................... 164
Appendix A: ABC License Types .............................................................................................. 169

vi

List of Tables
Table 1 Datasets and Sources ....................................................................................................... 33
Table 2 NAICS Code and Vendor Categories .............................................................................. 38
Table 3 Vendor Categories by SIC Code ...................................................................................... 39
Table 4 Temporary Classification of Unmatched Alcohol Retailers ............................................ 40
Table 5 Orange County License Summary ................................................................................... 44
Table 6 License Nearest Neighbor Statistics ................................................................................ 46
Table 7 Orange County Race/Ethnicity Summary ........................................................................ 48
Table 8 Orange County Summary Statistics of Census Tracts with Zero License ....................... 64
Table 9 Orange County Summary Statistics of Census Tracts with Alcohol Licenses ................ 66
Table 10 OC Summary Statistics of Census Tracts with Zero Type 20 Licenses ........................ 68
Table 11 OC Summary Statistics of Census Tracts with Type 20 Licenses ................................. 69
Table 12 OC Summary Statistics of Census Tracts with Zero Type 21 Licenses ........................ 70
Table 13 OC Summary Statistics of Census Tracts with Type 21 Licenses ................................. 70
Table 14 OC Summary Statistics of Census Tracts with Zero Liquor Stores .............................. 72
Table 15 OC Summary Statistics of Census Tracts with Liquor Stores ....................................... 72
Table 16 OC Summary Statistics of Census Tracts with Zero Wholesale Clubs ......................... 74
Table 17 OC Summary Statistics of Census Tracts with Wholesale Clubs .................................. 74
Table 18 OC Summary Statistics of Census Tracts with Zero Grocery Stores ............................ 75
Table 19 OC Summary Statistics of Census Tracts with Grocery Stores ..................................... 76
Table 20 OC Summary Statistics of Census Tracts with Zero Convenience Stores .................... 77
Table 21 OC Summary Statistics of Census Tracts with Convenience Stores ............................. 77
Table 22 OC Summary Statistics of Census Tracts with Zero Gas Stations ................................ 78
Table 23 OC Summary Statistics of Census Tracts with Gas Stations ......................................... 79
Table 24 OC Summary Statistics of Census Tracts with Zero Pharmacies .................................. 80

vii

Table 25 OC Summary Statistics of Census Tracts with Pharmacies .......................................... 80
Table 26 OC Summary Statistics of Census Tracts with Zero Department Stores ...................... 81
Table 27 OC Summary Statistics of Census Tracts with Department Stores ............................... 82
Table 28 OC Census Tracts with Licenses Population Summary ................................................ 83
Table 29 OC Census Tracts with Zero Licenses Population Summary ........................................ 84
Table 30 Census Tract Linear Regressions per Square Mile Trend Line Summary .................... 91
Table 31 Census Tract Linear Regressions per 1,000 People Trend Line Summary ................... 96
Table 32 OC All Licenses Optimized Hot Spots Summary Statistics .......................................... 99
Table 33 OC Type 21 Licenses Optimized Hot Spots Summary Statistics .................................. 99
Table 34 OC Type 20 Licenses Optimized Hot Spots Summary Statistics ................................ 100
Table 35 OC Liquor Stores Optimized Hot Spots Summary Statistics ...................................... 101
Table 36 OC Convenience Stores Optimized Hot Spots Summary Statistics ............................ 102
Table 37 OC All Licenses Observational Hot Spots Summary Statistics .................................. 104
Table 38 OC Type 20 Licenses Observational Hot Spots Summary Statistics .......................... 105
Table 39 Census Tract 524.08 Type 20 Hot Spots Summary Statistics ..................................... 107
Table 40 OC Liquor Stores Observational Hot Spots Summary Statistics ................................. 108
Table 41 OC Convenience Stores Observational Hot Spots Summary Statistics ....................... 109
Table 42 OC Gas Stations Observational Hot Spots Summary Statistics .................................. 109
Table 43 OC Summary Statistics of Cells with Zero Licenses ................................................... 112
Table 44 OC Summary Statistics of Cells with Alcohol Licenses ............................................. 113
Table 45 OC Summary Statistics of Cells with Zero Type 21 Licenses .................................... 114
Table 46 OC Summary Statistics of Cells with Type 21 Licenses ............................................. 115
Table 47 OC Summary Statistics of Cells with Zero Type 20 Licenses .................................... 116
Table 48 OC Summary Statistics of Cells with Type 20 Licenses ............................................. 116
Table 49 OC Summary Statistics of Cells with Zero Liquor Stores ........................................... 117

viii

Table 50 OC Summary Statistics of Cells with Liquor Stores ................................................... 118
Table 51 OC Summary Statistics of Cells with Zero Grocery Stores......................................... 119
Table 52 OC Summary Statistics of Cells with Grocery Stores ................................................. 119
Table 53 OC Summary Statistics of Cells with Zero Convenience Stores ................................. 121
Table 54 OC Summary Statistics of Cells with Convenience Stores ......................................... 122
Table 55 OC Summary Statistics of Cells with Zero Gas Stations............................................. 123
Table 56 OC Summary Statistics of Cells with Gas Stations ..................................................... 124
Table 57 OC Summary Statistics of Cells with Zero Pharmacies .............................................. 125
Table 58 OC Summary Statistics of Cells with Pharmacies ....................................................... 126
Table 59 OC Summary Statistics of Cells with Zero Department Stores ................................... 127
Table 60 OC Summary Statistics of Cells with Department Stores ........................................... 128
Table 61 OC Summary Statistics of Cells with Zero Wholesale Clubs ..................................... 129
Table 62 OC Summary Statistics of Cells with Wholesale Clubs .............................................. 130
Table 63 OC Cells with Alcohol Licenses Population Summary ............................................... 131
Table 64 OC Cells with Zero Licenses Population Summary .................................................... 132
Table 65 Licenses per Cell Linear Regressions Trend Line Polarity Summary ......................... 139
Table 66 Cell Linear Regressions per 1,000 People Trend Line Summary ................................ 145
Table 67 OC Cell Based All Licenses Optimized Hot Spot Summary Statistics ....................... 148
Table 68 OC Cell Based Type 21 Licenses Optimized Hot Spot Summary Statistics ............... 149
Table 69 OC Cell Based Type 20 Licenses Optimized Hot Spot Summary Statistics ............... 149
Table 70 OC Cell Based Liquor Stores Optimized Hot Spot Summary Statistics ..................... 150
Table 71 OC Cell Based Grocery Stores Optimized Hot Spot Summary Statistics ................... 151
Table 72 OC Cell Based Convenience Stores Optimized Hot Spot Summary Statistics ........... 151
Table 73 OC Cell Based Gas Stations Optimized Hot Spot Summary Statistics ....................... 152
Table 74 Summary of Distribution Evaluation Points ................................................................ 156

ix

Table 75 Comparison of Type 20 Licenses between Tracts and Cells ....................................... 157

x

List of Figures
Figure 1 Dominant Racial/Ethnic Group per Census Tract, Orange County, CA ........................ 16
Figure 2 Examples of Two Interrelated MAUP Issues (Jelinski and Wu 1996) .......................... 24
Figure 3 Example of MAUP and Spillover: ................................................................................. 27
Figure 4 Creating Atoms from Source and Target Boundaries .................................................... 30
Figure 5 Orange County with 5-Mile Buffer for Filtering Licenses ............................................. 36
Figure 6 Decision Tree for Categorization Process ...................................................................... 41
Figure 7 Alcohol Retailers within 5-Mile Buffer of Orange County............................................ 43
Figure 8 OC Alcohol Licenses by Type and Retailer per Census Tract ....................................... 45
Figure 9 OC Alcohol Licenses by Type and Retailer per Square Mile per Census Tract ............ 45
Figure 10 OC and Surrounding Counties...................................................................................... 49
Figure 11 OC Race and Ethniciy with Diversity Index Shading .................................................. 50
Figure 12 OC Census Tract Population Dot Map ......................................................................... 51
Figure 13 2017 ACS Table DP05 Estimates of Population Race/Ethnicity by Census Tract ...... 52
Figure 14 2017 ACS Table DP05 Estimates of Population Race/Ethnicity by Square Mile ....... 52
Figure 15 OC Census Tract Population Density ........................................................................... 53
Figure 16 LandScan Population Surface, 2018............................................................................. 55
Figure 17 OC Scaled Population Grid .......................................................................................... 56
Figure 18 OC Scaled Population Grid with Dominant Race/Ethnicity ........................................ 57
Figure 19 OC Census Tracts with Alcohol Licenses .................................................................... 63
Figure 20 OC Licenses/Retailers per Square Mile ....................................................................... 86
Figure 21 OC Linear Regressions on All Licenses per Square Mile ............................................ 87
Figure 22 OC Linear Regressions on Type 20 Licenses per Square Mile .................................... 87

xi

Figure 23 OC Linear Regressions on Type 21 Licenses per Square Mile .................................... 88
Figure 24 OC Linear Regressions on Liquor Stores per Square Mile .......................................... 88
Figure 25 OC Linear Regressions on Grocery Stores per Square Mile ........................................ 89
Figure 26 OC Linear Regressions on Convenience Stores per Square Mile ................................ 89
Figure 27 OC Linear Regressions on Gas Stations Stores per Square Mile ................................. 90
Figure 28 OC Linear Regressions on Pharmacies per Square Mile .............................................. 90
Figure 29 OC Licenses and Retailers per 1,000 People per Census Tract ................................... 92
Figure 30 OC Linear Regressions on Type 21 Licenses per 1,000 People ................................... 93
Figure 31 OC Linear Regressions on Liquor Stores per 1,000 People ......................................... 94
Figure 32 OC Linear Regressions on Grocery Stores per 1,000 People ....................................... 94
Figure 33 OC Linear Regressions on Pharmacies per 1,000 People ............................................ 95
Figure 34 OC Optimized Hot Spots Based Upon Census Tract Boundaries ................................ 97
Figure 35 OC Three Mile Observational Hot Spots Based Upon Census Tract Boundaries ..... 103
Figure 36 Observational Hot Spot of Type 20 Licenses Occurring at Census Tract 524.08 ...... 106
Figure 37 OC Licenses and Retailers Per Cell ............................................................................ 111
Figure 38 OC Cells with Wholesale Clubs ................................................................................. 129
Figure 39 OC Alcohol Licenses per Cell .................................................................................... 134
Figure 40 OC Linear Regressions on All Licenses per Cell ....................................................... 135
Figure 41 OC Linear Regressions on Type 21 Licenses per Cell ............................................... 136
Figure 42 OC Linear Regressions on Type 20 Licenses per Cell ............................................... 136
Figure 43 OC Linear Regressions on Liquor Stores per Cell ..................................................... 137
Figure 44 OC Linear Regressions on Grocery Stores per Cell ................................................... 137
Figure 45 OC Linear Regressions on Convenience Stores per Cell ........................................... 138

xii

Figure 46 OC Linear Regressions on Pharmacies per Cell ......................................................... 138
Figure 47 OC Alcohol Licenses per 1,000 Population per Cell ................................................. 140
Figure 48 OC Linear Regressions on All Licenses per 1,000 People per Cell ........................... 141
Figure 49 OC Linear Regressions on Type 21 Licenses per 1,000 People per Cell ................... 142
Figure 50 OC Linear Regressions on Type 20 Licenses per 1,000 People per Cell ................... 142
Figure 51 OC Linear Regressions on Liquor Stores per 1,000 People per Cell ......................... 143
Figure 52 OC Linear Regressions on Grocery Stores per 1,000 People per Cell ....................... 143
Figure 53 OC Linear Regressions on Convenience Stores per 1,000 People per Cell ............... 143
Figure 54 OC Linear Regressions on Gas Stations per 1,000 People per Cell ........................... 144
Figure 55 OC Linear Regressions on Pharmacies per 1,000 People per Cell ............................. 144
Figure 56 OC Optimized Alcohol License Hot Spots Based Upon Cell Boundaries ................. 146
Figure 57 OC Alcohol License Observational Hot Spots Based Upon Cell Boundaries ........... 153

xiii

List of Abbreviations
ABC California Department of Alcoholic Beverage Control
ACS American Community Survey
API Application Programming Interface
GIS Geographic Information System
GISci Geographic Information Science
HTTP Hypertext Transfer Protocol
LG18 LandScan Global 2018
MAUP Modifiable Area Unit Problem
OC Orange County, California
SSI Spatial Sciences Institute
U.S. United States
USC University of Southern California
USEPA United States Environmental Protection Agency
WGS World Geodetic System

xiv

Abstract
Systemic racism, institutional racism, structural racism: these are the terms used to describe
unequal minority participation in job markets, over representation in the criminal justice system,
and lack of access to and enjoyment of clean and safe neighborhoods. Studies in social justice
and environmental justice are now starting to quantify structural racism by utilizing Geographic
Information Systems and applying analytic methods of Geographic Information Science. One
area ripe for study of structural racism is whether race-neutral laws and regulations promote
race-neutral distributions in the built environment or perpetuate existing structural racism.
The distribution of alcohol retailers in Orange County, California, provided an
opportunity to explore how a race neutral regulation—in operation for over two decades—has
impacted the built environment. Exploring the distribution of alcohol retailers informs our
understanding of structural racism because a higher density of retailers has been correlated with
negative impacts on neighborhoods such as increased crime, negative health outcomes, and
poverty. Moreover, California’s alcohol licensing regulations are race-neutral and as such do not
consider race as a factor in determining the approval or rejection of a license application.
This study analyzed the February 18, 2020 inventory of active off-site retail sales alcohol
licenses in Orange County and compared the distribution of licenses with race/ethnicity across
the county. The comparison was repeated at two spatial scales: census tract and a scaled
population grid based on the Oak Ridge National Laboratory’s LandScan 2018 dataset with 30
arc-second cells (~ 0.5 miles). This study found that Hispanic populations were consistently
overrepresented in census tracts and cells where alcohol licenses were found. This result
suggested that requiring laws and regulations to avoid recognition of race is insufficient to ensure
race-neutral distributions of benefits and detriments in the built environment.

1

Chapter 1 Introduction
Few would argue that buying a six-pack of beer, a bottle of wine, or fifth of whiskey at the
corner market is a racist act. Take California for example; first, the buyer is required to meet
race-neutral age identification requirements and have legal tender. Second, in order to sell
alcohol a retailer must apply for and be granted a license based on race neutral criteria codified
in state law as Business and Professions Code § 23958 (AB 1994, BPC 2019). Notwithstanding
the race neutral context of buying/selling alcohol, could the density of alcohol retailers in the
built environment, which has already been found to have a negative impact on neighborhoods,
provide evidence of structural racism?
As Palma Strand explains in The Invisible Hands of Structural Racism in Housing: Our
Hands, Our Responsibility, structural racism arises when minority neighborhoods continue
experiencing unequal burdens compared to white neighborhoods (Strand 2019). So, to explore
structural racism in the built environment requires comparing the distribution of burdens found
in predominantly minority communities to nearby majority white communities. Moreover, we
can use the density of alcohol retailers as a proxy for burden because we know from studies such
as Alcohol Retail Density and Demographic Predictors of Health Disparities: A Geographic
Analysis, that a higher density of alcohol retailers in a community is a likely risk factor for
increased crime, negative health outcomes, and poverty (e.g. Berke et al. 2010; Halonen et al.
2013; Young, Macdonald, and Ellaway 2013; Dwivedi et al. 2019). In other words, if the
distribution of alcohol retailers in California is similar across the white and minority
communities, then we can have confidence that the application of race neutral licensing
requirements has facilitated a race neutral distribution of the burdens associated with higher
density of alcohol retailers; otherwise, an unequal distribution suggests that structural racism has

2

played a part in the unequal distribution of those burdens. This approach is similar to that used in
studies like Food Swamps Predict Obesity Rates Better Than Food Deserts in the United States
which explored the prevalence of obesity as a function of how the local built environment
introduces significant biases for individuals (and households) as to the potential choices for diet
and physical activity (Cooksey-Stowers et al. 2017; Liu et al. 2015; Hurvitz et al. 2009).
To be clear, structural racism does not require overt acts of racial animus or racial intent
to discriminate. It can be the result of many presumably race-neutral actions that when added
together over time perpetuate historical racial segregation, discrimination, and injustice as
reflected in the built environment. It can arise as emergent phenomena in novel ways, for
example a race-neutral data mining algorithm intended to identify patients in greatest need of
medical intervention instead reinforced minority unequal access to health care. It can be the
continuation of historical racial profiling, such as when historically redlined, predominantly
minority neighborhoods do not enjoy the same economic advantages as nearby white
neighborhoods. It can occur when a neighborhood council implements a crime-free, race-neutral
ordinance prohibiting landlords from renting to tenants who have had contact with the criminal
justice system, which in operation prevents minorities from accessing housing in that
neighborhood.
While some form of structural racism has existed for as long as segregated,
disadvantaged communities have developed in societies, the study of structural racism has only
become a topic of direct research in the last several decades (Groos et al. 2018). Much of the
initial structural racism research was in the form of social justice studies conducted by social
scientists looking at societal level effects, such as incarceration rates or negative health outcomes
of minorities. However, there is now an increasing trend of research in the growing field of

3

environmental justice, among other disciplines, to utilize Geographic Information Systems (GIS)
to develop and apply techniques and methods of Geographic Information Science (GISci) to
identify and quantify structural racism as part of the physical built environment in our
communities (Groos et al. 2018; Kim and Chun 2019).
One alarming theme that appears consistently across nearly all the social and
environmental justice studies is that minority communities continue to experience a
disproportionate share of negative social and economic outcomes, including poorer quality of
life, higher crime rates, higher concentration of toxic dumps, fewer banks, and fewer green
spaces. This leads to the question: How and why does the phenomenon of structural racism
continue when the United States has had robust anti-discrimination laws in place at every level
of government for at least the last sixty years? To begin to answer the question requires a deeper
discussion of the terms and concepts commonly used in identifying and describing structural
racism as well as understanding how the historical echoes of racism have contributed to the rise
and continuation of structural racism in American society (Pulido, Sidawi, and Vos 1996). This
background also provides context as to why exploring the distribution of alcohol retailers for
evidence of structural racism is important.
1.1 Essential Concepts: Structural Racism and Disparate Distribution
To begin to understand the importance of identifying and remedying structural racism in
our society requires establishment of a working definition and discussion of the term itself and
two additional related terms: racially neutral and disparate impact. First, the term “structural
racism” is typically defined as unconscious and implicit biases—as opposed to intentionally
discriminatory choices—within institutions, agencies, businesses, and society that continue the
status quo of socio-economic disadvantages experienced by racial, ethnic, or other minorities

4

(Strand 2019; Baroca and Selbst 2016). Structural racism is also generally synonymous and
interchangeable with the terms “institutional racism” and “systemic racism.” While some
individuals in a society, community, or institution may have a covert—or even overt—racist
agenda, that does not automatically give rise to structural, systemic, or institutional racism.
Although, if a racist agenda is left unaddressed, it can eventually manifest in various forms as
structural racism.
Next, the term “racially neutral” refers to the language of a law, policy, regulation, or
practice that is: 1) absent of all mention of race, or 2) has an included racial component which on
its face operates in a race-neutral fashion (e.g., tracking race for census purposes). In practice,
however, to be effectively race-neutral requires more than the mere absence of words denoting
race. This distinction is critical because racially neutral language when put into operation
through policies, practices, or algorithms can still produce unintended/unexpected detriments that
are born disproportionately by racial, ethnic, or other minorities (Archer 2019; Obermeyer et al.
2019; Kau, Fang, and Munneke 2018).
Another term for this result is “disparate impact.” A detrimental outcome resulting from a
race-neutral law, policy, or practice reflected as a racial or ethnic minority receiving unequal
treatment or an unintended detriment when compared to a non-minority (Fisher, Kelly, and
Romm 2006; Rolok 2011). For example, in An Unintended Consequence of Mortgage Financing
Regulation —A Racial Disparity, the authors showed empirically that race-neutral mortgage fair-
lending laws—prohibiting lenders from considering race and ethnicity—had the unintended
consequence of non-white borrowers paying more than whites over the life of their mortgages
(Kau, Fang, and Munneke 2018). The cause appeared to be related to different prepayment habits
exhibited between white and non-white borrowers not being taken into account when mortgage

5

contract prices were calculated. Prepayments operate as a risk to lenders because they reduce the
total value that a lender expects to receive when offering a loan. Thus, a lower likelihood of
prepayment is associated with a less risky loan and should result in a lower interest rate for the
borrower. This has led to an unintended consequence: the fact that non-whites generally do not
pay their mortgages off early as often as whites cannot be factored into determining loan risk
when calculating the loan contract price, which allows the accelerated pay off tendencies of
whites to inflate the mortgage contract prices for all borrowers. As a result, the operation of the
racially neutral law has precipitated a racially skewed detrimental economic outcome for
minorities.
A second, but distinct, related term is “disparate impact.” As Nancy Rolok makes clear in
New Methodology: Measuring Racial or Ethnic Disparities in Child Welfare—a study of the
unequal representation of minority children in the Illinois welfare population—the term
“disparate” is not the same as “disproportionate.” Specifically, the two terms represent different
evaluative concepts. Disparity requires an evaluation of equality and tends to be subjective; the
evaluator will likely consider multiple abstract factors such as decision points, access to services,
and absence of negative outcomes. Disproportionality frequently refers to objective evaluations,
often in the form of comparisons between percentages (Rolok 2011).
Here it should be noted that disparate impact as defined in this thesis is broader than the
current legal assessment of disparate impact as framed by the U.S. Supreme Court. In Texas
Department of Housing and Community Affairs, a case holding that plaintiffs may bring
disparate impact claims related to government allocation of tax credits in low income housing,
the Supreme Court specified that a disparate-impact is more than just a statistical disparity, there
must be a correlation between the racial imbalance and the policy or policies causing that

6

disparity (Texas Dept of Housing 2015). However, this legal definition whitewashes the fact that
there is almost always some pre-existing disparity in the environment affecting a minority
population such that it will be impossible to find a simple correlation between a policy and a
particular disparity. Although it is laudatory that the Supreme Court recognized the existence of
racial disparity, it is this type of simplistic, color-blind approach to disparate impact that allows
structural racism and segregation to occur, acquire legal approval, and continue.
Finally, this study introduces the term “disparate distribution,” which adds spatial and
temporal dimensions to the concept of disparate impact. Disparate distributions arise when
communities with predominantly non-white populations experience decreased access to positive
environmental conditions (banks, grocery stores, parks etc.) or increased exposure to detrimental
environmental conditions (dumps, toxic industry, pollution, etc.) as compared to areas with
predominantly white populations. Thus, for this thesis, the concept of disparate distribution
focuses on whether a facially neutral policy (regulation, algorithm, etc.) contributes to burdens
(or decreases positive environmental factors) on a predominantly non-white population, which is
experiencing disparate neighborhood impacts. In practice, to identify a disparate distribution
requires analysis of the impacted communities since laws, regulations, and policies are not
operating in a vacuum but rather in a complex web of dynamic, inter-dependent spatial,
temporal, political, and social factors.
An informative example of this approach can be found in Dissecting Racial Bias in an
Algorithm Used to Manage the Health of Populations where the study authors were attempting to
discern why a race neutral algorithm was under-selecting blacks for inclusion in a high-risk
patient intervention program (Obermeyer et al. 2019). They found that black patients, even
though they presented with the same or greater health risks than whites, generated less health

7

care costs which were assumed to be a primary predictor of patients most likely to benefit from
intervention. Thus, black patients—the ones who were actually most likely to benefit from
intervention—were under-selected by a race-neutral algorithm because they do not seek health
care as often as similarly situated white patients.
1.2 Structural Racism Persists After Racist Policies and Practices Have Ended
While neighborhood redlining—the practice of highlighting minority neighborhoods in
order to exclude those residents from favorable mortgage opportunities—began in the 1930s and
was banned by the late 1960s, the practice literally set in concrete (and asphalt) certain aspects of
today’s built environment that are at least partially responsible for poor and minority
communities experiencing exposure to higher land surface temperatures than surrounding
neighborhoods (Hoffman, Shandas, and Pendleton 2020; Strand 2019). For example, a recent
study compared the historically redlined neighborhoods with their surrounding non-redlined
neighborhoods and found that nearly 94% of the redlined areas experienced greater land surface
temperatures relative to the non-redlined areas (Hoffman, Shandas, and Pendleton 2020).
Basically, the study posits that the act of excluding residents of poor and minority neighborhoods
from access to the same mortgages offered to whites led to the lack of meaningful real estate
investment in those communities and subsequent depressed land values. Over time, those
depressed land values became favorable opportunities for impermeable surface land use projects
such as commercial tracts, industrial facilities, and freeways. Unfortunately for the residents in
those redline-impacted communities, impermeable surfaces are more effective at absorbing heat
and solar energy and later radiating heat back into the environment when the land surface
temperature should otherwise be cooling.

8

In a broader sense, it has been forcefully argued that today’s racial disparities continue to
exist because “the present is strongly tied to the past” (Johnson and Hoopes 2019, 5). To
disclaim the existence of institutional or structural racism by disconnecting historical events from
today’s social and political patterns is a “historical fallacy.”
Thus, while redlining did not create the heat pockets lingering over poor communities, it
may have laid the groundwork for a course of events leading to the disparate impact experienced
by poor communities of greater heat exposure than nearby affluent neighborhoods (Hoffman,
Shandas, and Pendleton 2020). Moreover, without concerted and focused commitment to remedy
these inequities, the resulting built environment often continues to reinforce the lower land
values—and attendant latent and patent disparities—of the affected neighborhoods (Strand
2019).
1.3 Racially Neutral Policies May Not Address Structural Racism
The Civil Rights Act of 1968 was passed to correct centuries of direct maltreatment of
minorities seeking fair housing in America. From that point in time forward, the national
mandate compels racial neutrality in housing laws, policies, and practices. As a practical matter,
this mandate prohibited race from being taken into account when calculating or distributing
housing related benefits, services, detriments, and entitlements. While the basic public policy is
sound, the reality of drawing a line at 1968, and requiring race neutrality from that moment
forward ignores the very real disparities that minorities were experiencing prior to 1968 because
of the centuries of direct and over discrimination (Strand 2019; Johnson and Hoopes 2019).
One simple analogy would be to imagine a track meet where minority runners are held at
the starting line at the beginning of the race and the event judge notices they have been held
back. But instead of restarting the race or making allowances for the distance already run by the

9

non-held runners, the judge declares all runners are now to be treated the same and allowed to
run freely from that moment forward. Thereafter, even though all runners now experience equal
treatment, the biased results of that initial race continue as factors in determining objective race
results which in turn determines the upcoming track meets runners may attend and lane selection
priority of the initial runners and future generations of their teammates. This is so because in
competitive track and field, a runner’s past wins give them an ongoing advantage in future
meets. In other words, the future is defined by the past. But alas, reality is much more complex
than the simple track meet analogy.
To illustrate these complexities, consider that many local municipalities and government
agencies have begun adopting racially neutral crime-free housing ordinances and programs;
these programs are portrayed as a race-neutral approach to address the laudable goal of reducing
crime and ensuring safe neighborhoods (Archer 2019). In a jurisdiction under this regime,
private landlords are required to screen their tenants and evict those having contact with criminal
legal system (regardless of the reason for the contact or how many years have passed since the
contact). Now, consider that African Americans experience disproportionately higher rates of
arrests and convictions compared to their proportion of the general population (Archer 2019;
Johnson and Hoopes 2019; NAACP 2015). According to the NAACP, African Americans make
up 33% of those incarcerated for drug offenses whereas they represent only 12.5% of drug users
(NAACP 2015). While reducing crime and promoting safety is an important community concern,
these ordinances in reality operate to exclude minorities from residing in the implementing
jurisdiction and force them into surrounding communities (Archer 2019). This occurs as
minorities (especially African Americans) are first removed from the crime-free community
through evictions and then new arrivals continue to be excluded from entry. Furthermore, it is

10

highly likely that the segregative effect is magnified because the evictions and resulting
exclusion of minorities occur at their higher rate of contact with the criminal system, not their
lower proportion of the general population.
Sadly, there is a potentially even darker disparate impact lurking beyond the direct
segregative effects of these racially neutral policies. A 2009 study found that in large urban
cities, the rate of violent crimes increased relative to the magnitude of segregation (Krivo,
Peterson, and Kuhl 2009). It further found that while the minority segregated communities bear
the brunt of this increased violence, all neighborhoods across highly segregated cities experience
greater violent crime than more integrated cities. Thus, in an ironic twist, a racially neutral
ordinance to reduce crime may actually increase crime and perpetuate systemic segregation.
“Recognizing these fundamental realities of the interconnections of race, place, and inequality is
required for understanding race-ethnic differences in a host of arenas, including in levels of
criminal violence” (Krivo, Peterson, and Kuhl 2009: 1766).
These disparate impacts extend far beyond the intersection of policies, crime, and
housing (Obermeyer et al. 2019; Groos et al. 2018). For example, a 2018 review of 165 scholarly
works studying structural racism, dated between January 1, 2007 through December 31, 2017, in
PubMed and Embase databases identified twenty original studies quantifying structural racism
related to population health, mortgage discrimination, and political participation in addition to
the traditional topics of crime and housing (Groos et al. 2018).
In general, unintentionally racially-biased outcomes may result when applying race-
neutral policies or algorithms to diverse populations with latent racial biases (Obermeyer et al.
2019; Barocas and Selbst 2016). Studies of big data and data driven processes provide examples
of this occurring while examining potential societal factors at play. For example, search engines

11

results include more ads for arrest records when search terms include black-identifying names
compared to terms with white-identifying names (Sweeney 2013). Minorities pay more over the
life of a mortgage because mortgage lenders calculate mortgage rates on the assumption that all
borrowers present the same prepayment risk, even though white borrowers prepay much more
frequently (Obermeyer et al. 2019). Blacks have unequal access to high-risk medical intervention
programs because insurance algorithm screening for candidate patients does not account for
lower health care costs of black populations even though they have highest health risk factors
(Barocas and Selbst 2016).
Looking more closely at race-neutral mortgage fair-lending laws (discussed above in
Section 1.1), a study revealed a disparate impact resulting from lenders using race-blind
prepayment risk formulas that cannot take into account race-based prepayment patterns (Kau,
Fang, and Munneke 2018). Prepayment occurs when a home is sold, refinanced, or outright paid
off and is a risk to the lender because it reduces the expected total value of the loan by the
amount of the lost interest over the contractual life of the loan. Lenders account for this risk by
adjusting various terms of the loan, one of which is to increase the loan interest rate (with a
subsequent increase in dollar amount of monthly payments). Because the prepayment risk
formula is based on the prepayment patterns of the entire pool of borrowers—whites and non-
whites—the subset of borrowers who do not prepay (mostly non-whites) are punished in the
form of higher monthly payments over their longer loan payment period.
There is an open question as to why whites prepay more frequently than non-whites, but
there are some anecdotal explanations. Generally, borrowers with better credit scores tend to
prepay more frequently (whites). Borrowers with higher value properties (whites) or properties
that rise in value more quickly (whites) tend to prepay more frequently. Borrowers with higher

12

education (whites) tend to prepay more frequently. On the other hand, borrowers with lower
credit scores (non-whites) tend to prepay less frequently, and also default more frequently.
Borrowers in less desirable locations (non-whites) tend to prepay less frequently. Borrowers with
mortgage provisions that penalize prepayment (non-whites) tend to prepay less frequently. While
the knee jerk response is to exclaim that everyone can prepay, that whitewashes the reality that
non-whites are generally starting at an economic disadvantage.
Moreover, not accounting for differing black/white population dynamics when applying
race neutral processes can literally be life threatening. For example, a study examining a race
neutral algorithm used by health insurance companies to direct high-risk patients into
intervention programs found that the algorithm under screened high-risk black patients into the
programs (Obermeyer et al. 2019). The researchers evaluating the algorithm found that while the
implementation used rational variable selections and risk criteria—prior year health costs and
biomarkers for health among other non-race variables—the underlying population had latent
racial dynamics. Specifically, black patients generated less medical expense per health status.
Since the algorithm based future risk on prior costs, and white patients generated higher costs per
health status, they had a higher representation at the highest risk scores.
Again, there is an open question as to why black patients generate lower costs per health
status, but there are anecdotal explanations. Compared to white patients, black patients have
more emergency visits and costs related to dialysis, but require less outpatient specialist costs
and fewer inpatient surgical costs. There may be other underlying socio-economic barriers to
seeking medical care such as lack of access to transportation, unavailability of child care, and
inability to take time off from work (Obermeyer et al. 2019). Finally, evidence suggests that lack
of common racial background between doctor and patient may result in patients seeking less care

13

or result in other negative health outcomes (Alsan, Garrick, and Graziani 2019; Obermeyer et al.
2019).
Just as the impact of decades of redlining did not end immediately in 1968 with the
passage of the Civil Rights Act, the impact of inherent racial disparities in social, political, and
spatial domains do not immediately disappear with the invocation of race neutral language.
Ultimately, inclusion and understanding of latent and patent racial disparities is necessary before
policies, programs, and algorithms can be tailored to operate in race neutral fashion.
1.4 California Alcohol Retail Sales Licensing Regulations
In the early 1990’s, California legislators deliberated on the link between alcohol,
poverty, and crime and determined that a statewide approach to alcohol licensing was needed
(AB 1994). As a result, in 1994 California passed racially neutral, statewide legislation—
Business and Professions Code § 23958—to curtail the issuance of alcohol sales licenses in areas
already experiencing high crime or over-concentration of alcohol retailers (AB 1994, BPC 2019).
The legislation also assigned the California Department of Alcoholic Beverage Control (ABC)
the sole responsibility for evaluating and issuing alcohol sales licenses across the state.
On its face, the California law specifying the terms and conditions for evaluating and
issuing alcohol sales licenses is racially neutral. Some of the major factors considered for issuing
a license are total county population, local crime, and census tract-level population (BPC 2019).
Moreover, the when, what, and where of locations of high crime rates and over-concentration of
licenses—as well as what constitutes public convenience and necessity—are constantly evolving
determinations. Among the many factors used to make these determinations, the ABC must
evaluate crime statistics compiled yearly by local law enforcement, census tract-level population
information based upon most recent U.S. decennial (or special) census, and annual county-level

14

population information compiled by the Demographic Research Unit of the California
Department of Finance (BPC 2019 § 23958.4). Moreover, a license applicant may petition to
establish that the census tract-level population has increased from value initially relied upon by
the ABC.
Additionally, the ABC must solicit and consider input from local governing entities (city
councils, administrative districts, city managers, etc.). These local entities may support the
issuance of licenses to local businesses for “Public Convenience or Necessity” in otherwise
proscribed locations (Drummond 2014, BPC 2019). For example, in 2004, the Yorba Linda City
Manager made a finding in support of a CVS Pharmacy receiving an alcohol sales license as it
“would afford city residents the ability to purchase alcoholic beverages … while shopping for
other convenience items” and “reduce the length and number of vehicular trips …, thereby
reducing traffic” (Drummond 2014). Moreover, a license applicant can also petition on its own
behalf for a public convenience or necessity exemption. However, while the legislative mandate
requires that the ABC consider factors “which may affect the public welfare and morals…,”
there is no mandate to evaluate disparate impacts or disparate distributions (BPC 2019 § 23958).
Finally, the ABC administers 54 unique license categories pertaining to alcohol sales or
transactions (see Appendix 1). This thesis focused on two license types—Type 20 Off-Sale Beer
& Wine and Type 21 Off Sale General. These are the only licenses that grant establishments
permission to transact retail sales of sealed containers of alcohol for off premise consumption,
i.e. grocery stores, liquor stores, convenience stores, gas stations, etc. Of the two licenses, type
20 is the more restrictive—authorizing only beer and wine sales—while type 21 permits the sale
of all packaged alcohol products. However, at no point in the license application process is race
of the licensee or racial composition of the community considered or a factor in granting or

15

rejecting a license, regardless of the type of license sought or the nature of the establishment
seeking a license.
1.5 Study Area: Orange County, California
Orange County (OC) was selected because of its size, demographic diversity, and range
of socioeconomic conditions. These factors make OC an ideal study area to explore structural
racism across a diverse urban landscape. OC is located on the southern coast of California,
between Los Angeles and San Diego Counties. The county covers 948 square miles, and with a
2010 Census population count of over 3 million, it ranks as the third-most populous county in
California and sixth-most in the nation (OCHS 2020). Figure 1 shows the ACS 2017 population
estimates (N=3,155,816) of OC broken down by census tract, using three colors to highlight the
predominant race in each tract. According to the 2017 American Community Survey (ACS)
Demographic and Housing Estimates, the majority of OC’s population was non-Hispanic whites
(~41%) followed closely by Hispanics of any race (~34%), and Asians (~20%) (U.S. Census
Bureau 2017). In the figure, the “Dominant Population” was determined by comparing the ACS
race and ethnicity variables for each census tract and selecting whichever race/ethnicity had the
highest estimate value. Because the figure is intended as an aid for visualizing the general OC
population, the margins of errors were not evaluated to determine if any particular best- or worst-
case margin of error scenarios would result in a different dominant census tract population being
displayed.

16

OC also has a diverse economy base that includes tourism (Disneyland), Fortune 500
companies (Broadcom, Western Digital, and First American Corporation), and fashion (Oakley,
Inc., Hurley International, and Vans). According to the U.S. Census Bureau, between 2014 to
2018, 65.5% of the total population over 16 years old was employed in the civilian labor force
(U.S. Census Bureau 2020). In that same period, 85% of the population 25 and older had a high
school diploma and 39.9% had a Bachelor’s degree or higher. However, the U.S. Census Bureau
also estimated that in 2018, 10.5% of the population was in poverty.
Figure 1 Dominant Racial/Ethnic Group per Census Tract, Orange County, CA

17

Geographically, the northern portion of the county tended to have a greater proportion of
minorities, older developed neighborhoods, and denser population which resulted in areally
smaller census tracts. The southern portion of the county was more recently developed and
tended to have fewer minorities and areally larger census tracts. The coastal side of the county
had the greatest concentration of wealth, while the eastern portion of the county was the least
developed. These geographic variations in distributions of populations, minorities, and wealth
provided an excellent landscape for exploring whether a race neutral regulation manifested a race
neutral distribution in the built environment.
1.6 Thesis Objective and Research Questions
This thesis set out to evaluate the potential existence of structural racism in the form of
disparate distributions of alcohol licenses and retailers and attempted to answer the following
questions:
• What is the relationship between the presence and absence of alcohol licenses/retailers
and the relative percentages of associated Asian/Hispanic/White populations?
• How does the relationship of race/ethnicity (percentage Asian/Hispanic/White) and the
distribution of alcohol licenses/retailers manifest at different spatial scales of areal
aggregation?
• Does the disparate distribution assessment vary depending upon the choice of spatial
scale and aggregation?
Specifically, this thesis looked at how the application of a racially neutral alcohol sales licensing
framework has combined with private market forces to influence the spatial distribution of
alcohol retailers in OC and compare the distribution of alcohol retailers between predominantly
racial minority communities and majority white communities in order to quantify if a disparate

18

distribution has occurred. This was accomplished by utilizing a GIS to aggregate and analyze
census tract and scaled population grid cell populations with and without alcohol point of sale
retailers to measure variations in racial/ethnic proportions.
The ultimate goal of this thesis is to provide an analytical framework for spatially
evaluating racially neutral zoning/licensing policies and their unexpected consequences in the
form of disparate distributions. In other words, it builds upon the idea that correcting institutional
or structural racism to realize actual race-neutrality requires recognition that in America the
playing field for minorities is not inherently level to begin with and that blind insistence on race-
neutral language may inadvertently reinforce systemic and structural disparities and segregation
(Krivo, Peterson, and Kuhl 2009; Kim and Chun 2019; Johnson and Hoopes 2019).
1.7 Thesis Organization
The next chapter, Chapter Two, details the recent research within the U.S. for exploring
disparate impacts of environmental burdens and benefits on various racial minority communities
compared to majority white communities. This related work sets the framework for the alcohol
retailer density distribution comparison methods that were used in this thesis. Chapter Three
details the methodology for how this study was conducted and describes in detail the data
sources and spatial analysis that was performed. Chapter Four presents the analytic results and
discusses whether there is support for the hypothesis that structural racism is occurring and being
reinforced in the built environment. Finally, Chapter Five discusses the significance of these
results and the how they relate to refining racially neutral regulation in future legislation or
public policy so that historical racial disparities can be corrected.

19

Chapter 2 Related Work
Minorities—whether racial, ethnic, or disadvantaged economic groups—in urban settings tend to
experience some form of disparity more frequently than white populations (Krivo, Peterson, and
Kuhl 2009; Kim and Chun 2019; Kubrin and Hipp 2016; Fisher, Kelly, and Romm 2006;
Hoffman, Shandas, and Pendleton 2020). However, studies looking into issues of disparity or
inequity may not use those terms but instead frame the issue as a lack or imbalance of
environmental or social justice. Ultimately, these concepts are intertwined: disparity and inequity
arise when there is a lack of environmental or social justice and vice versa. Referring to United
States Environmental Protection Agency’s (USEPA) definition of environmental justice helps
make this clear: environmental justice is “the fair treatment and meaningful involvement of all
people regardless of race, color, national origin, or income, with respect to the development,
implementation, and enforcement of environmental laws, regulations, and policies” (USEPA
2018).
Regardless whether framed in the negative (disparity) or the positive (justice), empirical
evidence continues to show that environmental risks are not equitably distributed across racial
groups (Kim and Chun 2019; Unger et al. 2020). This is important because, as many observers
have come to recognize, disparity contributes to the creation or continuation of communities or
populations experiencing greater exposure to crime or pollution; likewise, disparity is often a
contributing factor that accounts for decreased access to green spaces and healthy food choices
(Krivo, Peterson, and Kuhl 2009; Kim and Chun 2019, Cooksey-Stowers et al. 2017). Moreover,
disparity also occurs in less easily detectable forms such as increased incarceration rates,
increased environmental heat exposure, and greater negative health outcomes (Johnson and
Hoopes 2019; Hoffman, Shandas, and Pendleton 2020; Dwivedi et al. 2019).

20

It is easy to imagine that problems of causation and correlation may arise because there
are numerous factors that may contribute to the rise or continuation of disparity for any particular
group or population and because disparity can occur in so many patent and latent forms.
Moreover, when considering disparate impact, it is important to consider that there may be
lingering echoes of historic decisions, practices, and policies on today’s population and built
environment (Pulido, Sidawi, and Vos 1996; Johnson and Hoopes 2019). Furthermore, just as the
built environment is not the result of a single decision or building, disparity is not the result of a
single factor. Rather, both are functions of the totality of the circumstances at the nexus of the
moment of observation of a population in situ with the built environment (Pulido, Sidawi, and
Vos 1996). The following sections review recent works examining these issues and shed light on
how disparate impacts on poor and minority communities and populations may arise as a result
of race neutral policy influencing the built environment through race neutral decisions such as
where to locate a liquor store.
2.1 Density of Alcohol Retailers and Disparate Distributions
As a general premise, the local built environment introduces significant biases for
individuals (and households) as to the potential choices for diet, physical activity, entertainment,
transportation, employment, etc. (Drewnowski et al. 2019; Cooksey-Stowers et al. 2017; Hurvitz
et al. 2009). Moreover, one of the seminal studies on the topic of alcohol retailer density,
“Alcohol Retail Density and Demographic Predictors of Health Disparities: A Geographic
Analysis,” confirmed that urban alcohol retailer density correlated with poverty and minority
populations, among other things, at the national level (Berke et al. 2010). Similarly, studies have
correlated that increased access to alcohol at the local level, by way of greater density of alcohol
retailers in a community, is a likely risk factor for increased crime, negative health outcomes,

21

and poverty (e.g. Berke et al. 2010, 1967; Halonen et al. 2013, 295; Young, Macdonald, and
Ellaway 2013, 124-125; Dwivedi et al. 2019, 105742).
Studies examining similar issues are often expressed in terms of environmental justice—a
field of research examining and quantifying degrees to which all people enjoy healthy
environments, protection from health hazards, and access to community decision-making
processes—a framework particularly suited for investigating disparate distributions (USEPA
2018; Kim and Chun 2019). Specifically, these studies focus on quantifying environmental
inequity by observing the differential distributional outcomes of environmental risks that are
borne by different social populations (Kim and Chun 2019). Broadly speaking, an environmental
risk is the chance that a stressor—any chemical, biological, or physical entity that may trigger an
adverse outcome—produces a harmful effect to health or the community. Some recent examples
of stressors from an environmental justice perspective include pollution, proximity to fringe
lenders and unlicensed cannabis dispensaries, and lack of access to green space (Kim and Chun
2019; Kubrin and Hipp 2016; Unger et al. 2020; Wolch, Byrne, and Newell 2014). Thus, the
concentration of alcohol retailers can be evaluated for environmental inequity by analyzing the
distribution like other dissociative environmental risks, such as pollution sources, payday
lenders, cannabis dispensaries, and parking lots or associative environmental benefits such as
grocery stores and parks.
However, while the density of alcohol retailers in a community correlates with negative
social impacts, there is an open question as to whether reducing that density would result in a
corresponding reduction of negative outcomes (Hippensteel et al. 2018). Moreover, care should
be taken to ensure that the selected spatial unit of analysis is appropriate for the topic of study

22

and that data underlying the study supports the planned methods of analysis (Montello and
Sutton 2013).
2.2 Spillover and MAUP: Influences of and on the Built Environment
The built environment can be described as the socio-physical environment resulting from
both current and historical human influences through patterns of activity, land use choices, and
human-made structures and infrastructure (Popkin, Duffey, and Gordon-Larsen 2005).
Moreover, the built environment evolves over decades from many physical, legal, and policy
factors including health, safety, cost, traffic patterns, preserving historic architecture, etc. (Texas
Dept of Housing 2015; Popkin, Duffey, and Gordon-Larsen 2005). Simultaneously, the local
built environment introduces myriad options and limitations for individuals (and households) as
to choices for diet, physical activity, entertainment, transportation, employment, etc.
Importantly, the local built environment has been shown to be a powerful predictor of the
health of the local population (Drewnowski et al. 2019). Likewise, various micro-level physical
characteristics of the built environment have been found to induce or deter violent crime (He,
Páez, and Liu 2017). For example, structures in the built environment—such as bridges—
provide shelter for homeless, and thus the structure locations can be used to predict homeless
populations and homeless-related crime (Yoo and Wheeler 2019). Even fringe banks—payday
lenders and check cashers—appear to be spatially concentrated in low-income and minority
population neighborhoods, and robbery hot spots are often found within a block of a fringe bank
location (Kubrin and Hipp 2016).
On the other hand, the racial composition of a local population may also correlate highly
with certain characteristics of the built environment. For example, race alone can predict the
likelihood of hazardous waste sites (Kramar et al. 2018). Likewise, at the census tract level, the

23

risk of pollution exposure increases with increases in minority populations (Kim and Chun
2018). However, without careful consideration of the spatial effects on the accuracy, precision,
and efficacy of analytical methods, built-environment-to-population and population-to-built-
environment influences are at best general associations and do not necessarily allow for causal
inferences to be made (Drewnowski et al. 2019; Kim and Chun 2018; Kogure and Takasaki
2019, Oka and Wong 2016).
One important consideration is the potential influence of spatial spillover effects: spatial
linking between neighborhoods and/or the spatial proximity between points of observations or
measurements (Oka and Wong 2016; Tobler 1970). For example, census tracts (a common
enumeration unit) are often treated as discrete information pools where the population or a
population characteristic is monolithic across the entire area. While this may be an acceptable
assumption for inhabitants at the census tract center or immobile populations, it may not
accurately reflect how people live or make decisions for those near the edge or on the boundary
(Oka and Wong 2016). In other words, population mobility or exposures from outside the
enumeration unit undermines the assumption that the enumeration unit alone fully depicts the
environment.
One remedy for spillover issues is to derive spatial weighted variables based on the areal
data (Oka and Wong 2016). While there are many ways to generate spatially weighted variables,
the various methods tend to fall within one of two schemes: binary weighting and spatial kernels.
Under binary weighting approaches, the derived variables are computed based on adjacent or
contiguous enumeration units. On the other hand, spatial kernels generate a derived variable
based on all the enumeration units within a defined distance and apply a distance decay
multiplier to reduce the contribution of more distant enumeration units.

24

Likewise, choice of spatial scale and boundaries can profoundly impact the significance
of observations and the accuracy and precision of spatial and statistical analysis results (Jelinski
and Wu 1996; Dark and Bram 2007; Smith and Sandoval 2018). These two interrelated issues—
often described as scale and zone effects—are widely known as the modifiable areal unit
problem (MAUP). The scale effect recognizes that aggregating smaller areal units into larger
units results in a loss of data variation. Figure 2 (a-c) shows the classic MAUP example of scale
effects: as data is aggregated from (a) into larger units (b and c) the mean value does not change,
but the variance declines. Figure 2 (d-f) also shows the zone effect: the choice of zone
boundary—even when holding the number of spatial units constant (b and d; c, e, and f)—can
impact both the mean and variance (Jelinski and Wu 1996; Dark and Bram 2007).

Figure 2 Examples of Two Interrelated MAUP Issues (Jelinski and Wu 1996)

25

Basically, as spatial data is aggregated into new larger spatial units, there are likely
unintended smoothing or filtering functions occurring (Jelinski and Wu 1996). This occurs as
part of the transformation function, loss due to over generalization, or loss when differing
smaller units are recharacterized and combined to a single category over a larger unit.
2.3 Improving Results and Avoiding MAUP: Two Areal Aggregation Scales
One way to reduce MAUP, improve accuracy of results, and increase insight into
relationships and patterns is to utilize multiple spatial scales (Smith and Sandoval 2018).
Moreover, incorporating at least one finer-scale geographic unit can provide additional insight
into patterns that arise at different scales.
2.3.1. Scale 1: Census Tracts
A frequently relied upon resource for socio-economic and demographic data is the U.S.
Census which acquires and maintains multiple datasets relevant to performing spatial analysis on
the American landscape and population at various administrative boundary levels, including
county and census tract levels. Census tracts, like all political and administrative demarcations—
i.e. counties, school districts, congressional districts, etc.—are artificial geographic constructs
with shapes and sizes chosen for specific political and administrative purposes without regard to
spatial analysis (Smith and Sandoval 2019; Berke et al. 2010; Fisher et al. 2006). However,
census tract boundaries are not entirely arbitrary demarcations; they are delineated by a
committee of local demographers and data users based upon boundary and demographic criteria
established by the Census Bureau (U.S. Census Bureau 1994).
First, a census tract boundary must be easily identifiable in the field, often following
visible, permanent features, including roads, highways, canals, railroads, and power lines. Next,
census tracts should enclose populations of 2,500 to 8,000 individuals and include 1,000 to 3,000

26

housing units, and averaging all census tracts in a county should result in an average census tract
population of approximately 4,000 people and 1,500 housing units. Finally, the census tract
should enclose a population with similar housing and socio-economic characteristics. Thus,
although census tracts do not have an areal boundary definition, the uniform application of
Census Bureau guidelines makes census tracts reliable as enumeration units for data aggregation
for the Census Bureau and other researchers (Oka and Wong 2016).
However, issues regarding the areal variability of census tracts must still be addressed
before performing spatial analysis. The first issue relates to the spillover concept, which looks at
whether the population or variables under study are confined completely within census tracts or
may exist, influence, or be influenced by factors outside the tract boundaries (Oka and Wong
2016; Berke et al. 2010). For example, a pollution source in one census tract may impact a
community that is just across the street in a different census tract. Likewise, individuals are
mobile and will often work and shop in locations completely outside their home census tract.
The second issue, MAUP, relates to how the choice of areal unit may lead to distinct
analytic outcomes (Smith and Sandoval 2019; Oka and Wong 2016). MAUP relates to the
alternative outcomes that arise—depending upon the choice of scale and boundaries—when data
is aggregated to larger areal units or disaggregated to smaller areal units. Unless these factors are
accounted for, results may be misleading due to underestimations or misspecifications of study
area characteristics.
Figure 3 provides an example of MAUP and spillover issues with OC census tract
boundaries and license locations. First, census tracts are the result of aggregating two or more
census block groups (shown in light blue outline), which in turn were the aggregation of two
more census blocks (not shown). Each aggregation has the potential to introduce data loss or

27

uncertainty. Moreover, census tracts presume continuity of the underlying population(s), but that
presumption is hard to reconcile with irregular sizes and shapes of census tract boundaries such
as between census tracts 637.01, 637.02, and 636.05. Likewise, licenses for this study were
aggregated at the census tract level, and many licenses were within 100 feet or less of a census
tract boundary while the tract interior was often devoid of licenses. A slight change in boundary
location would completely alter the census tract license counts and distribution analyses.

Figure 3 Example of MAUP and Spillover:
Irregular Census Tract Boundaries and License Locations

28

Second, the close proximity between license locations and the census tract boundaries
implicates spillover issues. For example, would residents in the western half of census tract
638.08 be more inclined to visit the license locations on the census tract’s eastern border, or the
license location in the northwest corner of census tract 637.01?
2.3.2. Scale 2: Scaled Population Grid
Using a combination of spatial scales can improve the identification of patterns and
increase the accuracy and precision of analytic results (Smith and Sandoval 2019). When data
does not exist at a chosen scale, spatial interpolation methods can be used to create new spatial
units at different scales. For example, Risk Terrain Modelling (RTM) is often employed in
analyzing crime at finer scales and relies upon spatial interpolation to create uniform grids from
larger aggregated spatial units such as census tracts (Smith and Sandoval 2019; Youngmin and
Wheeler 2019).
Moreover, the accuracy and precision of spatial interpolation methods can be improved
with the inclusion of ancillary data (Ruther, Leyk, and Buttenfield 2015). This is often termed
dasymetric mapping and is used to refine spatial estimates when the underlying data was
aggregated into spatial units defined for convenience of enumeration rather than data
aggregation. For example, land cover information or other remotely sensed data can be used to
refine population density estimates in census tracts, counties, or other administrative boundaries
(Leyk et al. 2019).
The Oak Ridge National Laboratory’s LandScan Global 2018 (LG18) is an example of an
ancillary dataset of global population distribution that can be used for dasymetric refinement
(Rose et al. 2019, Leyk et al. 2019). The LG18 utilizes multiple spatial and population modeling
approaches to create an interpolated grid surface composed of 30 arc-second cells (~ 0.5 miles)

29

based on the World Geodetic System (WGS) 84 datum. Each cell is assigned an integer value
representing the ambient population count estimate of the earth’s surface covered by the cell.
However, the LG18 is a gross population estimate only, it does not include any information
regarding race/ethnicity or other socio-economic demographics data.
Datasets such as LG18 are often used for the binary dasymetric refinement of target areas
to include only populated areas (Ruther, Leyk, and Buttenfield 2015). Moreover, the ancillary
grid cells themselves may become the analytic spatial unit after interpolation of the source data
and ancillary data (Leyk et al. 2019). However, care should be taken that the target grid scale
represents an appropriate proxy for the intended analysis and does not introduce increased local
imprecision.
One of the oldest and simplest areal interpolation methods for generating a new spatial
unit is based on areally weighting the data between the source and target spatial units (Ruther,
Leyk, and Buttenfield 2015). The process works by first creating “atoms,” which are an areal
quantum resulting from the spatial intersection of source and target boundaries (see Figure 4).
Fractions of the source data are then assigned to the atoms based on the proportion of the source
area encompassed by the atom. The process is repeated for all intersections of the source and
target boundaries. Atoms are then reassembled into target zones by summing the individual atom
values of the atoms that are encompassed by the target boundaries:
𝑦 𝑡 ̂ = ∑
𝐴 𝑠𝑡
𝐴 𝑠 𝑦 𝑠 , 𝑠 (1)
where 𝑦 ̂
𝑡 is the estimated population count in target zone (t), 𝑦 𝑠 is the source (s) population
count, 𝐴 𝑠 is source area, and 𝐴 𝑠𝑡
is the atom area.

30

Figure 4 Creating Atoms from Source and Target Boundaries

31

Chapter 3 Data Sources and Methodology
At the heart of this study is the simple premise that a race-neutral alcohol retail sales license
regulation should produce a distribution of retailers where the aggregate of local race/ethnicity
population proportions near the retailers follows the county-wide race/ethnicity proportions. In
other words, the absence of a race-neutral distribution function is presumed if the aggregate
race/ethnicity population proportions near alcohol retailers deviates from county-wide
proportions by something more than nominal differences and margins of error.
This premise can be applied to the study of license distributions in OC because the
locations of licensed alcohol retailers are known and the county is mostly composed of three
uniquely identified racial/ethnic subpopulations which are quantified by ACS Table DP05
race/ethnicity population estimates (U.S. Census Bureau 2017). Asian and Hispanic (all races)
are two minority subpopulations and White (non-Hispanic) is the majority subpopulation. The
remaining ACS populations estimates include Black and multiple race/ethnicity combination
subpopulations, but these remain subpopulations are so small that ACS estimates are often
overshadowed by margins of error.
While the premise is simple, issues with MAUP and spillover effects, if not addressed,
could likely introduce errors that attenuate or exaggerate the population proportions calling into
question whether the observations are reliable. There are also different possible distribution
patterns for the various types of retailers (i.e. grocery stores versus gas stations). In order to
mitigate these spatial and license/retailer issues, two different areal aggregation systems—census
tracts and a scaled population grid based on LandScan cells—were selected to provide
complimentary areal coverage and the license holders were categorized into one of seven
categories using standard business data analytic classification systems.

32

This work relied upon several types of spatial, categorical, and population data, from
multiple sources. Moreover, none of the data were directly useable for analysis in their original
format. Table 1 details the data sets and sources relied upon for this study. The following
sections discuss the study area and the purpose, source, and transformations associated with each
data set.

33
Table 1 Datasets and Sources
Dataset Description Format Data Type Spatial Scale
Reporting
Period
Source
Alcohol
Licenses
License Type and
Addresses of California
alcohol licensees
.csv Point data in the
form of street
addresses in text
fields
Point locations in
census tracts
2019
Department of
Alcoholic
Beverage Control
Business
Vendor Type
Information
Business name, SIC codes,
and address of retailers
within 5 miles of Orange
County
.csv Point data in the
form of street
addresses in text
fields
Point locations in
census tracts
2020
ReferenceUSA
(Infogroup, Inc.)
County and
Census Tract
Boundaries
Administrative boundaries
for Orange County,
surrounding counties, and
census tracts
.shp Vector data
(polygon)
OC and Census
tracts of various
areal sizes
2018
U.S. Census
Bureau
TIGER/Line files
Race/Ethnicity Dataset reporting Non-
Hispanic, Hispanic, and
other race/ethnicity
population estimates
.csv Data in the form of
aggregated census
tract population
estimates in text
fields
Census tracts of
various areal sizes
2017
U.S. Census
Bureau ACS Table
DP05
LandScan
Global 2018*
Ambient population
distribution
ESRI
GRID
Raster data
(integer value:
number of people)
30 arc-second cells,
approx. 0.25 sq mi
2018
Oak Ridge
National
Laboratory

* "This product was made utilizing the LandScan 2018™ High Resolution Global Population Data Set copyrighted by UT-Battelle, LLC, operator of Oak Ridge National
Laboratory under Contract No. DE-AC05-00OR22725 with the United States Department of Energy. The United States Government has certain rights in this Data Set. Neither
UT-BATTELLE, LLC NOR THE UNITED STATES DEPARTMENT OF ENERGY, NOR ANY OF THEIR EMPLOYEES, MAKES ANY WARRANTY, EXPRESS OR
IMPLIED, OR ASSUMES ANY LEGAL LIABILITY OR RESPONSIBILITY FOR THE ACCURACY, COMPLETENESS, OR USEFULNESS OF THE DATA SET."

34

3.1 OC Alcohol Retailer License Data
As indicated above in Section 1.4, in California, there are two types of license that pertain
to retail alcohol sales for consumption off premises: Type 20 for beer and wine only sales and
Type 21 for all types of alcohol beverage sales; either of these licenses can be obtained by any
type of vendor permitted to sell alcohol, including grocery stores, drug stores, gas stations, liquor
stores, and warehouse clubs (i.e. Costco and Sam’s Club). The ABC allocates these licenses and
maintains a database of the license holders. Because the type of retailer and spatial location of
the business was not included in the ABC license data, additional processing steps were required
to obtain and add that information.
3.1.1. Acquiring OC Alcohol Retailer License Data
Obtaining California alcohol retailer license data required accessing the ABC website to
generate a license information report (ABC 2020). The license data obtained from the ABC was
plain text in a comma separated file format where the information was stored in row and column
format. In the file, each row was a unique record, and each column represented a field (i.e. name,
address, license type, etc.) of information for the record. Moreover, each license was assigned a
unique numeric file number value and there was only one record for each license. However,
some vendors had multiple licenses for multiple locations, in which case there were multiple
records with the same vendor. Likewise, each license type was unique to each location (i.e. there
is no unique address that will have two Type 21 licenses); however, locations may have multiple
types of licenses. Finally, the license location information was simply the street address of the
location in plain text and required subsequent geocoding in order to perform geospatial analysis.

35

3.1.2. Geocoding Alcohol Retailer License Data
In its raw format, the only spatial information in the licenses data was in the form of
several text fields for street addresses. To perform spatial analysis required utilizing a geocoding
service in order to convert the street addresses into geographic coordinates—latitude and
longitude pairs. For this study, the Google Maps Platform Geocoding Service was chosen to
perform the conversion.
The Google Geocoding Application Programming Interface (API) uses a hypertext
transfer protocol (HTTP) interface for submitting a street address and receiving a geocoded
response (Google 2020). To use the interface, the requester formats an HTTP string with the
necessary street address information and sends the string to the google server. After receiving the
request, the Google Geocoder responds with an HTTP message that indicates either an error
message regarding problems with the request or a geocoded response with additional diagnostic
information regarding the accuracy of the response. This sequence of submitting requests and
receiving a response was repeated for each address that requires geocoding.
The initial ABC dataset consisted of all licenses (N=122,043) within California, on
February 18, 2020. A filter was applied to select only “Active” licenses, of Type 20 or 21, and
within Orange, Los Angeles, San Bernardino, Riverside, and San Diego counties (N=12,571).
This set of licenses was then geocoded using the Google Map Geocoding API in preparation for
additional filtering.
The Google Map Geocoding API appended additional details regarding the success and
status of each record submitted for geocoding. Out of the 12,571 records submitted for
geocoding, only 1 record returned with an error. However, the record was for a license with a
Riverside county address, so it was removed from further evaluation. The remaining geocoded
addresses were then filtered several times to further refine the license relevant to this study.

36

First, a filter was applied to select only licenses (N=3,070) within Orange County or
cities bordering Orange County. Next, a spatial filter was applied to select only licenses
(N=2,469) within a 5-mile buffer of Orange County (see Figure 5). The geocoding results of
these license were then re-evaluated for precision and accuracy.
At the time of this study, the Google Map Geocoding API did not have a direct
quantification of precision or accuracy, but did provide text values in two fields that provided an

Figure 5 Orange County with 5-Mile Buffer for Filtering Licenses

37

indirect indicium of precision and accuracy. First, a “location_type” field provided an indication
of the level of precision; of the 2,469 addresses in the 5-mile buffer, 2,390 were reported by the
Google Map Geocoding API as having a “Rooftop” level of precision. The remaining licenses
(N=79) were reported as having “Range_Interpolated” (N=53) or “Geometric_Center” (N=26).
Second, a “types” field provided an indication of the accuracy with simple descriptors such as
“premise,” “street_address,” “subpremise,” etc. After geocoding, all geocoded addresses were
reported as having an accuracy of at least “premise.” Thus, while not a direct measure of spatial
precision, review of the various combination of text values across all the results suggested the
geocoded coordinates were spatially within the bounds of a street block, parking lot, or center of
a collection of buildings related to the geocoded addresses. The spatial data was then projected
using the California State Plane VI FIPS 0406 coordinate system for further spatial analysis.
3.1.3. Categorizing Alcohol Retailer Types
This study required classifying the vendors into several categories in order to determine if
there was any variance in the distributions of the licenses and retailers based on the types of
vendors. Table 2 lists the seven broad categories chosen for this study, based upon the North
American Industry Classification System (NAICS) codes, of vendor types that potentially sell
alcohol for off premises consumption in OC.

38

Table 2 NAICS Code and Vendor Categories
Code Category
445110 Supermarkets and Other Grocery (except Convenience) Stores
445120 Convenience Stores
4452xx Specialty Food Stores
445310 Beer, Wine, and Liquor Stores
446110 Pharmacies and Drug Stores
4471xx Gasoline Stations
4523xx General Merchandise Stores, including Warehouse Clubs and Supercenters
In its original format, the license data was a plain text file, and the only vendor
information was the business name. In order to classify the business entities into vendor
categories, it was necessary to cross reference the vendors in the license data with a business
information source to identify and categorize by vendor category. The ReferenceUSA business
analytics database was utilized to generate a cross reference list of OC and surrounding county
businesses using the NAICS codes applicable to this study (Infogroup 2020).
Notwithstanding the use of NAICS codes to select the businesses, the data obtained from
ReferenceUSA used the older Standard Industrial Classification (SIC) codes. Fortunately, the
SIC codes and business classifications translated to equivalent NAICS codes and classifications
selected for this study. Table 3 lists the SIC codes, SIC Category, and a description of the
businesses typical of the categories.

39

Table 3 Vendor Categories by SIC Code
SIC Code Category Typical Business Examples
592102 Liquor Stores Stores selling primarily
alcohol or with Liquor in
the name
Frontier Liquor, Food
Mart Liquor, Happy’s
Liquor
554101 Service Stations Gas stations Mobil, Union 76, Circle K
541103 Convenience Stores Smaller markets selling
groceries and other
conveniences
Circle K, 7 Eleven,
Shop-n-Go
531110 Wholesale Clubs Membership-based stores
selling groceries and
other consumer goods.
Costco, Sam’s Club
531102 Department Stores Department stores selling
groceries and other
consumer goods.
Target, Kmart, Walmart
591205 Pharmacies Pharmacies Rite Aid, Walgreens, CVS
541105 Grocery Stores Full size grocery stores Vons, Ralphs, Food 4
Less
The business data obtained from ReferenceUSA was in plain text file with comma
separated fields where the information is stored in row and column format. In the file, each row
was a unique business record, and each column represented a field (i.e. name, address, SIC code,
etc.) of information for the record. However, some businesses had multiple records depending
upon whether the business had multiple locations and/or provided one or more goods or services
at each location. For example, Vons (grocery store) had multiple records, some of which were
for a number of different locations and others for the goods and services (such as groceries,
bakery, and pharmacy) it provided at each location. Moreover, because there was no consistent
business name and street address conventions between the ReferenceUSA data and the license
data (or even within the datasets themselves), a combination of manual and programmatic
matching schemes using street address and business names was employed to match the records
between the two datasets.

40

Of the 2,469 alcohol licenses, 1,928 were matched with a corresponding business
analytics record. The majority of the matched records were coded with one of the seven SIC
codes selected for this study. However, 230 of the records had SIC codes that did not match the
expected SIC codes and the remaining licenses (N=541) were unmatched to any business
records.
Of the 230 records with unrecognized SIC codes, 75 were found to have codes that were
subcategories of the expected study codes and were reclassified to the corresponding primary
code. The remaining 155 records were identified as businesses which had been licensed by the
ABC as off-site alcohol retailers but do not typically sell retail packaged alcohol products for
off-site consumption; rather, the sales at those businesses were akin to prepared food vendors,
catering or party service providers, or otherwise do not sell directly to the general shopping
public. These 155 records were coded with temporary business classifications (see Table 4).
Table 4 Temporary Classification of Unmatched Alcohol Retailers
Category Example of Retailer/Business N
Residential Review of address images indicated a
residential property (private home)
3
Specialty Sales An establishment for wine tasting or
entertainment that included wine sales
15
Secondary Sales A gift shop that sells baskets that may include
wine
23
Food Service Bakeries and delis 28
Internet Sales ABC website indicates Internet Sales Only 72
Hotel Hotel gift shops and a nuDist colony 14
TOTAL 155
Finally, a manual process of name evaluation and review of images from Google Maps
was used to categorize any remaining un matched licenses (see Figure 6). The goal of the process

41

was to ensure a consistent outcome to three instances of subjective and objective classification
decisions which arose when distinguishing between 1) liquor stores and convenience stores, 2)
convenience stores and grocery stores, and 3) convenience stores and gas stations.

Figure 6 Decision Tree for Categorization Process

42

For example, the decision between convenience stores and gas stations was primarily an
objective decision based upon the name (i.e. Shell, Chevron, Fuel, Oil) or gas pumps evident in
an image of the business address. The decision between liquor and convenience stores was a
more interesting case as almost 15% of the convenience stores (N=51) had a Type 21 license,
while several stores initially identified as liquor stores using the business analytics data had a
Type 20 license. The resolution was to bin business chains traditionally associated with
convenience stores (i.e. 7 Eleven, Circle K, Alta Dena) as convenience stores even if they had a
Type 21 license, while recategorizing all stores identified as liquor stores as convenience stores
if they had a Type 20 license.
Finally, while most of the categorization of convenience and grocery stores occurred by
matching business analytics data or name identification of chains, a small number of business
required a subjective evaluation of the business image at the license address. In this case, images
at the license address were reviewed and smaller retailers in strip malls or corner markets were
classified as convenience stores while larger markets and mall anchor stores were classified as
grocery stores. Figure 7 shows the final results of geocoding and categorization process.

43

3.1.4. Orange County Alcohol Licenses and Retailer Summary
After geocoding and categorizing the alcohol licenses, the license data set was loaded
into ArcGIS Pro to select only the licenses within the Orange County boundary as defined by the
Census TIGER/Line shapefile. Table 5 summarizes the records (N=1,805) that were selected by
this process. As the table indicates, 1,672 licenses records were deemed appropriate for study

Figure 7 Alcohol Retailers within 5-Mile Buffer of Orange County

44

after excluding 133 licenses from the OC study area based upon a combination of subjective and
objective factors (see Section 3.1.3). A license was objectively excluded if it was identified as
“Internet Only Sales” on the ABC website; whereas, it was subjectively excluded if there was
evidence that the retailer did not sell packaged alcohol to the general public for off premise
consumption (i.e. hotel gift shops, residential homes, caterers, bakeries, delis, and a nudist
colony). Figure 8 provides summary statistics of the count of alcohol licenses included for this
study by type and retailer in each census tract. Figure 9 normalizes the alcohol license statistics
by the square-mile-area of each census tract.
Table 5 Orange County License Summary

Orange
County
Type 21 Type 20
Total Licenses 1,805 (100%) 993 (55%) 812 (45%)
Excluded from Study 133 (7.37%) 16 (0.89%) 117 (6.48%)
Study Licenses
1,672
(92.63%)
977 (54.13%) 695 (38.5%)
Liquor Stores 418 (23.16%) 418 (23.16%) 0
Grocery Stores 412 (22.83%) 300 (16.62%) 112 (6.2%)
Convenience Stores 347 (19.22%) 51 (2.82%) 296 (16.4%)
Gas Stations 270 (14.96%) 15 (0.83%) 255 (14.33%)
Pharmacies 159 (8.81%) 135 (7.48%) 24 (1.33%)
Department Stores 50 (2.77%) 42 (2.33%) 8 (0.44%)
Wholesale Clubs 16 (0.89%) 16 (0.89%) 0

45

Figure 8 OC Alcohol Licenses by Type and Retailer per Census Tract

Figure 9 OC Alcohol Licenses by Type and Retailer per Square Mile per Census Tract

46

Next, a Nearest Neighbor analysis was performed to determine if the distributions of the
licenses were spatially clustered, dispersed, or random. The test was performed for all retailer
licenses in Orange County as a single group, the Type 20 licenses, the Type 21 licenses, and then
all licenses by each category of retailer. Each Nearest Neighbor test was repeated twice, first
using Euclidean distance and then Manhattan distance (see Table 6).
Table 6 License Nearest Neighbor Statistics

Licenses
Expected
Mean
Method Mean
z
score
p
value
Spatial
Distribution
All 1672 1,825
Euclidean 790 -44.33 0.00 Clustered
1

Manhattan 967 -36.77 0.00 Clustered
1

Type 20 695 2,831
Euclidean 1,521 -23.34 0.00 Clustered
1

Manhattan 1,839 -17.67 0.00 Clustered
1

Type 21 977 2,387
Euclidean 1,213 -29.4 0.00 Clustered
1

Manhattan 1,488 -22.52 0.00 Clustered
1

Liquor
Stores
418 3,650
Euclidean 2,360 -13.82 0.00 Clustered
1

Manhattan 2,878 -8.27 0.00 Clustered
1

Convenience
Stores
347 4,006
Euclidean 2,813 -10.61 0.00 Clustered
1

Manhattan 3,476 -4.72 0.00 Clustered
1

Gas Stations 270 4,541
Euclidean 3,118 -9.85 0.00 Clustered
1

Manhattan 3,755 -5.44 0.00 Clustered
1

Department
Stores
50 10,553
Euclidean 8,440 -2.71 0.006 Clustered
1

Manhattan 10,900 0.44 0.66 Random
Pharmacies 159 5,918
Euclidean 4,539 -5.62 0.00 Clustered
1

Manhattan 5,396 -2.13 0.03 Clustered
2

Wholesale
Clubs
16 18,656
Euclidean 17,048 -0.65 0.65 Random
Manhattan 20,769 0.87 0.03 Random
Grocers 412 3,676
Euclidean 2,328 -14.24 0.00 Clustered
1

Manhattan 2,867 -8.55 0.00 Clustered
1

1
Less than 1% likelihood that pattern could be result of random chance

2
Less than 5% likelihood that pattern could be result of random chance

The results of the Nearest Neighbor tests supported the general premise that alcohol
licenses are not randomly distributed in the built environment, especially in the case of all
licenses. One reason for spatial clustering is that retailer site selection is generally confined to the

47

commercial and business zones of the built environment. On the other hand, the distribution of
licenses categorized as Department Stores (i.e. Target and K-Mart) and Wholesale Clubs (i.e.
Costco and Walmart) suggested different distribution functions operated based on the type of
retailer as these two categories produced random (Euclidean) and dispersed (Manhattan)
distributions. Alternatively, these distributions could simply have been the result of the smaller
number of retailers (Dept Stores: N=50 and Wholesale Clubs: N=16) in these two categories.
3.2 Spatial Analysis: Two Areal Aggregation Units
An integral data requirement for exploring structural racism is the relevant demographic
data of the population under study. For this study, race/ethnicity population estimates at the
census tract level and the administrative boundaries for OC and OC census tracts meet that
requirement. Also, in order to mitigate MAUP and spillover issues, this study utilized the Oak
Ridge National Laboratory’s LandScan Global 2018 population dataset to create a scaled
population grid. The following sections describe these data sources in greater detail.
3.2.1. ACS Race/Ethnicity Estimates and Margins of Error
The U.S. Census, through the American Community Survey (ACS), maintains
demographic summaries and statistics for multiple administrative units, including counties and
census tracts; these demographic summaries are available in table format as text files. ACS 2017
5-Year Estimates Table DP05 Demographic and Housing Estimates contained the relevant race
and ethnicity data required for this study (U.S. Census Bureau 2017). This study utilized the
ACS race/ethnicity classification that included a Hispanic/Latino category to accommodate the
significant portion of OC population that identifies as Hispanic or Latino (see Table 7).
Likewise, there are significant OC population segments that racially identify as Asian or White,
which are provided as single race estimates in Table DP05. While Table DP05 estimates that the

48

OC Black population is exceedingly small (1.6% of the population), it was included as a
category for analysis in this study, while the remaining racial categories in Table DP05 were
small fractions of a percent of the population and were combined as a single Other category for
analysis.
Table 7 Orange County Race/Ethnicity Summary
2017 ACS Table DP05 Estimate Margin of Error Percent
Total Population 3,155,816 * 100%
Hispanic (of any race) 1,079,172 * 34.2%
White alone 1,306,398 +/- 790 41.4%
Asian alone 615,659 +/- 2,831 19.5%
Black alone 49,590 +/- 1,181 1.6%
Other (some other race
alone or two or more races)
105,027 +/- 3,580 3.3%

*Estimate is controlled, margin of error treated as zero

Moreover, the ACS race/ethnicity values are estimates based on survey data, and each
estimate has a corresponding margin of error. Aggregating ACS data, and transforming it to new
spatial scales, requires additional processing to derive new margins of error. These new margins
of error were calculated from original ACS county and census tract margins of error using
guidelines and formulas published by the U.S. Census (U.S. Census Bureau 2018).
3.2.2. Scale 1: Census Tracts
For this study, the necessary administrative units (counties and census tracts) and their
boundaries were all available from the U.S. Census as TIGER/Line shapefiles. Figure 10 shows
the TIGER/Line county boundaries for the study area and four surrounding counties (Los

49

Angeles, San Bernardino, Riverside, and San Diego). Notably, the north-western part of the
county was composed primarily of small regularly shaped census tracts with significant roadway
infrastructure, while the south-eastern part of the county was composed of irregularly shaped
large census tracts with less infrastructure. Frequently, census tract boundaries followed the
transportation infrastructure creating the regular grid patterns in the north and the irregular
shapes in the south.

Figure 10 OC and Surrounding Counties

50

The ACS 2017 5-Year race/ethnicity estimates in Table DP05 were linked to the OC
census tracts. OC has a total of 583 census tracts; however, one census tract was removed before
analysis in this study. Census tract 9901 was removed because it has no land area and zero
population. On the other hand, census tract 9800—with an estimated population of only 27
Hispanics and a margin of error of +/-18—was retained even though it covers the Disneyland
resort complex which is an area that is mostly compromised of theme parks, restaurants, and
commercial and hotel properties related to the tourism industry. Figure 11 depicts census tract
dominant race and ethnicity with diversity indicated by applying shading based upon a diversity
index.

Figure 11 OC Race and Ethniciy with Diversity Index Shading

51

The shading algorithm is based on Simpson’s Diversity Index—a method to quantify
whether a community is dominated by a single group versus having multiple groups with similar
populations—as an aid for visualizing census tract diversity (Barcelona Field Studies Centre
2020). In general terms, the diversity index produces a range of values from 0 to 1, where 0
represents no diversity and 1 represents infinite diversity. For OC, the index ranged from 0.072
(almost no diversity, darker shades) to 0.725 (fairly diverse, lighter shades), with 0.52 being the
average index value across the census tracts. Figure 12 provides an alternative representation of
density and diversity based upon ACS census tract data where each dot represents 500 people.

Figure 12 OC Census Tract Population Dot Map

52

Figure 13 provides a box plot summary of the racial/ethnic population estimates for OC
census tracts while Figure 14 normalizes the data by square mile per census tract.

Figure 13 2017 ACS Table DP05 Estimates of Population Race/Ethnicity by Census Tract

Figure 14 2017 ACS Table DP05 Estimates of Population Race/Ethnicity by Square Mile

53

3.2.3. Scale 2: Scaled Population Grid
The spatial distribution of humans in OC presents potential issues with using census
tracts directly for spatial analysis. For example, the northern part of the county was densely
populated with small areal census tracts, compared to the southern portion where there were
large areal but sparsely populated census tracts (Figure 15). Moreover, there were multiple
census tracts with large areas that were completely devoid of housing; these areas include

Figure 15 OC Census Tract Population Density

54

Disneyland, Naval Weapons Station Seal Beach, and numerous city, county, state, and national
parks.
To address these population variation issues, this study utilized the ORNL LandScan
Global 2018 (LG18) dataset, which provided an ambient population distribution raster with
approximately 30 arcsecond (~ 0.5 mile) resolution. Figure 16 shows the LG18 raster
superimposed over the study area. In the figure, the darker squares indicate higher population,
whereas the white and tan areas indicate zero population. For this study, the LG18 raster was
used as both a grid to spatially redistribute the census tract population counts into 30 arcsecond
grid cells and to scale the underlying census tract populations.
This was necessary for three reasons. First, the LG18 dataset does not have any racial
data, and second, some portions of the census tracts have areas with no population. Second, by
scaling the census tract data areally to the LG18 population counts, the overall census tract racial
populations remain the same, while the areal population distribution more closely resembles the
built environment. Third, the cell samples the local population effectively no further than
approximately a half mile from an alcohol license street address.

55

An areally weighted interpolation process was used to scale and redistribute the census
tract population estimates into the grid cells. While the scaling and redistribution process
required multiple steps, the process can be summarized simply. Where an LG18 cell intersects
more than one census tract, split the LG18 population into each census tract by proportion of the
LG18 cell covered by each census tract. Next, divide the census tract populations proportionally
by area into a new grid LG18 based grid (scalar 1). Where multiple census tracts intersect a cell,
divide the census tract populations proportionally by their proportional area within the cell, then

Figure 16 LandScan Population Surface, 2018

56

split the new LG18 cells by the census tract boundaries and calculate areal differences (scalar 2).
Next scale the total and racial population values using scalar 1, scalar 2, and the original LG18
cell population values (scalar 3). Finally, create the final grid by summarizing all the scaled
population values from step Four into an LG18 based grid.
The result was a scaled population grid with 3,097 cells with each cell approximately
0.28 square miles in size and 0.57 miles on a side. Moreover, of the 3,097 cells, 2,172 were
identified as having some population. Figure 17 presents the final redistributed population grid
for OC.

Figure 17 OC Scaled Population Grid

57

Furthermore, each cell has the ACS racial populations based upon the underlying census
tracts scaled locally by the census tract composite values of the LG18 cells (see Figure 18).
However, while the grid provided higher population fidelity for areas with minimum and
maximum local population distributions, the allocation of the racial populations to each grid cell
is a proportional (fractional) distribution across the landscape—a potentially unlikely scenario in
the real world, especially in large areal census tracts. While this did introduce a question as to
whether the race/ethnicity dynamics in the census tracts were amenable to proportional allocation
in the cells, there are two factors that suggest those concerns were minimal in Orange County.

Figure 18 OC Scaled Population Grid with Dominant Race/Ethnicity

58

First, the denser areas of Orange County have smaller areal census tracts which were
scaled into similarly sized LandScan cells covering the same general area. Thus, any localized
race/ethnicity dynamics would likely be dispersed or concentrated in no more than two to four
scaled cells within close proximity to the origin census tracts. Second, the larger census tracts in
Orange County tended to have less racial diversity while also having significant open space
where there was little to no population. Overall, the previously inaccurate areal dispersal of the
census tract population was more accurately concentrated in cells that had been identified by the
LandScan data as having discernible populated areas.
3.3 Quantifying Race Neutral and Disparate Distributions
While this study’s premise that race-neutral regulations should result in distributions of
retailers where the populations near the retailers are representative of county-wide populations is
straightforward, observing such distributions in the spatial reality of the built environment is
more complicated. First, individual census tracts or other areal units are unlikely to have
population proportions matching the county-wide proportions. Second, no single sampling
scheme or analytic method can prove the absence or presence of race-neutrality under all
conditions arising in and from the built environment.
Thus, multiple complimentary analytical methods applied to both census tracts and the
scaled population grid cells ensured that there were sufficient robust observations to support the
study’s conclusions. One set of methods analyzed summary statistics regarding the presence and
absence of licenses and retailers in the aggregate, while a second set used simple linear
regressions to analyze license densities versus population proportions. After each analysis, non-

59

valid results were discarded and additional thresholds applied to account for margins of error and
other dynamic variations in the built environment.
For the first method, summary population proportions were created for both the presence
and absence of all licenses, each license type, and all retailers based on the aggregate
race/ethnicity populations of census tracts and scaled population grid cells. Population
proportions were similarly created for each bin of Getis Ord Gi* hot spot analyses performed on
both the census tract and scaled population grid cell data. The differences between the observed
population proportions from county-wide proportions were then calculated and adjusted using
margins of error values (from both the county-wide and observed variables) that would produce
the least difference between the county-wide and observed proportions. The least difference
adjustment was chosen in order to bias the results towards an outcome inline with a best-case
scenario of the ACS estimates being accurate, whereas a greatest difference adjustment would
have biased the results towards a worst-case scenario of the ACS estimates being inaccurate.
Thereafter, differences within the margin of error (designated “E” in Dist columns in summary
tables) were excluded from further evaluation.
Finally, the remaining differences were evaluated for disparate distributions using tiered
cutoffs to account for dynamic built environment variations; these evaluations were captured in
Dist columns of summary tables. If there was a difference of less than 5%, the observation was
deemed a race-neutral distribution (“N”). If the difference was between 5% and 10%, the
observation was deemed a race-correlated distribution (“C”). Finally, if the difference was 10%
or greater, the observation was deemed a disparate distribution (“D”). These cutoffs were chosen
to account for random population variations while recognizing that to qualify as a disparate
distribution needed to be more than a nominal difference from the county-wide average.

60

For the second method, the license/retailer density per square mile per census tract and
density per 1,000 people per census tract were analyzed using a simple linear regression:
𝑌 𝑙 = 𝑏 0
+ 𝑏 1
𝑋 𝑝 (2)
Where independent variable 𝑋 𝑝 is the Asian/Hispanic/White percent of population per tract and
dependent variable 𝑌 𝑙 is the license or retailer density per tract. Similar regressions were
performed on the scaled population grid cells.
If a regression result p-value was greater than 0.05, the result was discarded as not
statistically significant; otherwise, the trend line polarities (signs) of the race/ethnicity
populations were compared. Results where polarities were the same were deemed race neutral
distributions (“N”), while opposite polarities were deemed race-correlated distributions (“C”).
The slope polarity provides a simple metric that indicates a positive or negative correlation
between the dependent and independent variables, with the assumption that a race neutral
distribution would occur when all races/ethnicities exhibit the same slope polarity. Arguably,
comparing the slope magnitudes would provide greater certainty, but there are currently no
benchmarks for analyzing what magnitudes would be significant for each set of race/ethnicity
and license/retailer density scenarios.
After performing the analyses outlined above, the totality of the results for the Asian,
Hispanic, and White populations for each scenario was assessed. An occurrence of two or more
D’s was deemed a disparate distribution, while a single D or the occurrence of two or more C’s
was deemed a race-correlated distribution. Any other combination was deemed race neutral;
these assessments were documented in summary tables in Dist columns. This provided a
consistent framework for evaluating whether race-neutral or disparate distributions were
occurring for all combinations of race/ethnicities and licenses/retailers at both the census tract

61

and scaled population grid cell level. Additionally, the results for the Black and Other
populations were also generated and included for anecdotal review, but were not factors in the
final distribution assessments.

62

Chapter 4 Results
As the works cited in Chapter 2 suggests, excessive access to alcohol has been associated with
negative outcomes for individuals and communities. Moreover, minority communities tend to
experience greater negative environmental burdens compared to nearby white majority
neighborhoods. Yet, as a general premise, the concentration of retail outlets selling package
alcoholic beverages in a community should be largely uniform and not correlated with the racial
composition of the community. In California, legislation has existed since the early 1990’s that
mandates just such a race neutral alcohol retailer licensing scheme. This study set out to
determine if the race neutral licensing scheme has resulted in a race neutral distribution of
alcohol retailers in Orange County, California.
Analyzing the spatial distribution of alcohol licenses in Orange County, California
entailed a multi-step process. First, the license data was obtained from the ABC and geocoded to
geographic coordinates. Next the license data was matched against a business analytics database
to facilitate classifying the licenses into retailer type (liquor store, grocery store, gas station, etc.)
for analysis. Finally, the spatial distributions of the licenses were analyzed at two scales—census
tract and a scaled population grid of ~ 0.25 square mile cells—using multiple analytical
techniques. This chapter details the results.
4.1 Scale 1: Census Tract Analytical Results
The distributions of alcohol licenses at the census tract level were analyzed using
summary statistics, simple linear regression trend line slope analysis, and Getis Ord Gi* hot spot
analysis. The following sections examine the results of those analyses at the census tract level.

63

4.1.1. Census Tract Alcohol License Summary Statistics
If a race-neutral function controls the distribution of alcohol retailers in Orange County
census tracts, then both the presence and absence of alcohol retailers should generally follow the
demographic profiles of the census tracts. Thus, the first step in analyzing the distribution of
alcohol licenses at the census tract level was to explore the percentage of the Orange County
census tract populations that do and do not have alcohol licenses (see Figure 19).

Figure 19 OC Census Tracts with Alcohol Licenses

64

Out of 582 census tracts, 101 did not have any licensed alcohol licenses. These 101
census tracts represented 15.3% of the OC population. As Table 8 illustrates, an expected
population distribution (% Expected column) was created by scaling the county census tract
population percentages by 15.3% to allow comparison with the actual aggregated population
percentages (% Actual column) of those census tracts with no licenses. Next, the differences
between the county (Pop % County column) and tract (Pop % Tracts column) percentages and
the differences between the expected and actual population distributions were evaluated to
determine if there were any disparate distributions (Dist columns).
Table 8 Orange County Summary Statistics of Census Tracts with Zero License
Tracts with Zero Alcohol Licenses: 101 / 15.3% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.55%
(±0.48)
C 2.98%
3.14%
(±0.07)
N
Hispanic (any race)
34.2%
(*)
18.75%
(±0.56)
D 5.22%
2.86%
(±0.08)
D
White Alone
41.4%
(±.1)
55.18%
(±0.67)
D 6.32%
8.43%
(±0.1)
D
Black Alone
1.6%
(±.1)
1.51%
(±0.23)
D 0.24%
0.23%
(±0.04)
E
All Other Race(s)
3.3%
(±.1)
4.01%
(±0.26)
D 0.51%
0.61%
(±0.04)
D
Totals
100% 100%

15.3% 15.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

However, when comparing the census tract actual percentages to the expected values, it
must be kept in mind that the actual values are not linear, but rather aggregations of discrete
values determined by the population of each individual census tract. For example, the census
tract with the highest population in Orange County contributes 0.76% to the total population of

65

Orange County while the second most populated census tract contributes 0.62%. Anecdotally,
both these census tracts also happen to have majority Asian populations. While inclusion or
exclusion of highly populated census tracts like these could bias the aggregated percentages to a
particular race/ethnicity, the assumption is that the aggregation of more than twenty census tracts
will sufficiently render that particular bias small enough to be considered negligible.
The results in Table 8 suggest that Asians and Whites tended to have greater
representation in no alcohol licenses census tracts than expected, even accounting for ACS
margins of error. On the other hand, Hispanics tended to be underrepresented in those census
tracts. These results were further bolstered by examining the population distribution within the
target no alcohol census tracts (Pop % Tracts column) and comparing it with the general county
distribution (Pop % County). Again, Whites and Asians had greater representation in no alcohol
retailer census tracts compared to their county-wide populations, while Hispanics were
significantly less represented compared to their county-wide population.
Moreover, the quantity of D values in the Dist columns suggested disparate distributions
were occurring at multiple evaluation points. While these results were not conclusive of racial
disparity in the distribution of alcohol licenses, they suggested a disparate distribution for the
absence of alcohol licenses in those census tracts.
A similar approach was applied to analyze the census tracts with alcohol licenses. Out of
582 census tracts, 481 had one or more licensed alcohol retailers within their boundaries.
Moreover, these 481 census tracts represented 84.7% of the OC population. As Table 9
illustrates, an expected population distribution (% Expected column) was created by scaling the
county census tract totals by 84.7% to allow comparison with the actual aggregated population
percentages ( % Actual column) of those census tracts with retailers. Next, the differences

66

between the county (Pop % County column) and tract (Pop % Tracts column) percentages and
the differences between the expected and actual population distributions were evaluated to
determine if there were any disparate distributions (Dist columns).
Table 9 Orange County Summary Statistics of Census Tracts with Alcohol Licenses
Tracts with Alcohol Licenses: 481 / 84.7% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.32%
(±0.23)
E
19.75%
(±0.13)
E 16.53%
16.37%
(±0.19)
E
Hispanic (any race)
34.2%
(*)
36.98%
(±0.34)
C
39.07%
(±0.2)
D 28.98%
31.33%
(±0.29)
C
White Alone
41.4%
(±.1)
38.91%
(±0.28)
C
36.35%
(±0.15)
D 35.08%
32.97%
(±0.24)
C
Black Alone
1.6%
(±.1)
1.58%
(±0.1)
D
1.65%
(±0.05)
C 1.33%
1.34%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.21%
(±0.13)
N
3.18%
(±0.07)
E 2.82%
2.72%
(±0.11)
E
Totals
100%
100%

100%

84.7% 84.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

The results in Table 9 suggested that Whites tend to have lower representation in alcohol
retailer census tracts than expected, even accounting for ACS margins of error. On the other
hand, Hispanics tended to be overrepresented in those census tracts and the expected percentage
of Asians was within the ACS margin of error to their actual percentage. These results were
further bolstered by examining the population distribution within the aggregated census tracts
(Pop % Tracts column) and comparing it with the general county distribution (Pop % County
column). Again, Whites were underrepresented compared to their county-wide populations,
while Hispanics were significantly overrepresented and Asians were within the margin of error.

67

However, as Figure 19 indicated, most census tracts had more than two alcohol licenses
and this initial analysis did not account for the number of alcohol licenses in the census tracts. To
evaluate the impact of multiple licenses in the census tracts, each census tract population value
was multiplied by the number of licenses in the census tract and the resulting values were then
aggregated to calculate new population proportions. this license scaled population proportion is
displayed in the (Pop x L) % Tracts column.
As this column shows, after scaling the census populations by the number of alcohol
licenses, the Hispanic population’s overrepresentation had increased; suggesting that majority
Hispanic census tracts had more retailers than would have occurred if a race neutral function was
in operation. On the other hand, the White population showed greater underrepresentation after
scaling compared to the White county-wide population suggesting the opposite, while the Asian
population was still within the margin of error of its county-wide proportion. Finally, the
combination of results in the three Dist columns suggests the absence of a race-neutral function
in the distribution of alcohol licenses in Orange County.
As the above two tables indicate, the majority White population was both
overrepresented in census tracts without alcohol retailers and underrepresented in census tracts
with alcohol retailers. Likewise, the Hispanic population was both overrepresented in census
tracts with alcohol retailers and underrepresented in census tracts without alcohol retailers.
Moreover, while the Asian population somewhat tracked the White population in both
categories, its over/underrepresentation was much closer to, if not within, the margins of error.
Next, of the 582 census tracts, 218 did not have any Type 20 licensed alcohol retailers.
These 218 census tracts represented 35% of the OC population. Table 10 compared the expected
35% population estimates with the county-level percentages in those census tracts with No Type

68

20 licenses. The No Type-20 licenses statistics closely tracked the no licenses of any type
statistics, with Hispanics underrepresented, White overrepresented, and Asian within the margins
of error.
Table 10 OC Summary Statistics of Census Tracts with Zero Type 20 Licenses
Tracts with Zero Type 20 Alcohol Licenses: 218 / 35.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.21%
(±0.33)
N 6.84%
6.73%
(±0.12)
E
Hispanic (any race)
34.2%
(*)
23.66%
(±0.43)
D 11.98%
8.29%
(±0.15)
D
White Alone
41.4%
(±.1)
51.78%
(±0.46)
D 14.5%
18.14%
(±0.16)
D
Black Alone
1.6%
(±.1)
1.5%
(±0.15)
D 0.55%
0.53%
(±0.05)
E
All Other Race(s)
3.3%
(±.1)
3.83%
(±0.19)
D 1.17%
1.34%
(±0.06)
C
Totals
100% 100%

35.0% 35.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Turning to Type 20 licenses, these licenses occurred in 364 census tracts, representing
65% of the OC population. Table 11 presents the Type 20 licenses, which followed the same
pattern as all Alcohol Licenses: Hispanics overrepresented, Whites underrepresented, and Asians
nearly within the margins of error. Moreover, comparing the values in the % Actual and Pop %
Tracts columns suggested that Type 20 licenses tended to be more prevalent in Hispanic
dominant census tracts. Overall, multiple observations surpassed the 10% threshold for the
difference between county-wide values and observations to be deemed disparate distributions.

69

Table 11 OC Summary Statistics of Census Tracts with Type 20 Licenses
Tracts with Type 20 Alcohol Licenses: 364 / 65.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.67%
(±0.26)
E
18.99%
(±0.19)
N 12.67%
12.78%
(±0.17)
E
Hispanic (any race)
34.2%
(*)
39.87%
(±0.41)
D
43.04%
(±0.31)
D 22.22%
25.9%
(±0.26)
D
White Alone
41.4%
(±.1)
35.79%
(±0.31)
D
33.38%
(±0.22)
D 26.9%
23.25%
(±0.2)
D
Black Alone
1.6%
(±.1)
1.61%
(±0.11)
D
1.66%
(±0.08)
A 1.02%
1.04%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.06%
(±0.14)
E
2.93%
(±0.1)
N 2.16%
1.99%
(±0.09)
E
Totals
100%
100%

100%

65.0% 65.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Type 21 licenses were found in 424 census tracts and represented 76.3% of the
population; the remaining 158 census tracts do not have Type 21 licenses and represent 23.7%
population. Table 12 provides the summary statistics for census tracts with no Type 21 licenses
and Table 13 the summary for census tracts with Type 21 licenses. These two tables show that
while the Hispanic population was underrepresented in census tracts with zero Type 21 licenses,
they appeared nominally race neutral unless license scaling was factored. On the other hand, the
majority White population continued to manifest overrepresentation in the zero Type 21 license
tracts and nominally race neutral representation in the Type 21 tracts. These values showed a
different distribution profile than that which occurred in the Type 20 White and Hispanic
populations.

70

Table 12 OC Summary Statistics of Census Tracts with Zero Type 21 Licenses
Tracts with Zero Type 21 Alcohol Licenses: 158 / 23.7% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.21%
(±0.33)
N 4.63%
4.38%
(±0.09)
N
Hispanic (any race)
34.2%
(*)
23.66%
(±0.43)
D 8.12%
7.23%
(±0.12)
C
White Alone
41.4%
(±.1)
51.78%
(±0.46)
D 9.83%
10.95%
(±0.12)
D
Black Alone
1.6%
(±.1)
1.5%
(±0.15)
D 0.37%
0.35%
(±0.04)
E
All Other Race(s)
3.3%
(±.1)
3.83%
(±0.19)
D 0.79%
0.83%
(±0.05)
E
Totals
100% 100%

23.7% 23.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 13 OC Summary Statistics of Census Tracts with Type 21 Licenses
Tracts with Type 21 Alcohol Licenses: 424 / 76.3% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.84%
(±0.25)
E
20.28%
(±0.17)
N 14.88%
15.13%
(±0.19)
E
Hispanic (any race)
34.2%
(*)
35.36%
(±0.37)
N
36.34%
(±0.26)
C 26.08%
26.96%
(±0.28)
N
White Alone
41.4%
(±.1)
39.93%
(±0.3)
N
38.39%
(±0.2)
C 31.57%
30.45%
(±0.23)
N
Black Alone
1.6%
(±.1)
1.6%
(±0.1)
D
1.65%
(±0.07)
C 1.2%
1.22%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.27%
(±0.13)
C
3.34%
(±0.09)
N 2.54%
2.5%
(±0.1)
E
Totals
100%
100% 100%

76.3% 76.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

71

Moreover, comparing the values in the Pop % Tracts and the (Pop x L) % Tracts columns
between this and the Type 20 scenario suggested that Type 21 licenses tended to be more
prevalent in White dominant census tracts whereas Type 20 licenses were more prevalent in
Hispanic dominant census tracts. On the other hand, the Asian population values were close to
expected for a race neutral function or too close to the margins of errors. Overall, the values of
the (Pop x L) % Tracts made the distribution more than race neutral, but also did not pass the
disparate threshold.
The next step was to analyze each type of retailer, starting with the Liquor Store
category. First, there were more tracts without liquor stores (N=297) than tracts with liquor
stores (N=285). However, the percent of the population living in tracts without liquor stores was
48.5% compared to 51.5% living in tracts with liquor stores. Moreover, this category presented a
unique case since it had the greatest number of retailers (N=418), and all Liquor Store retailers
only had Type 21 licenses. Table 14 presents the no Liquor Store license summary statistics and
Table 15 the Liquor Store summary statistics.

72

Table 14 OC Summary Statistics of Census Tracts with Zero Liquor Stores
Tracts with Zero Liquor Store Retailers: 297 / 48.5% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.45%
(±0.3)
C 9.46%
9.92%
(±0.14)
N
Hispanic (any race)
34.2%
(*)
27.76%
(±0.36)
D 16.59%
13.46%
(±0.18)
D
White Alone
41.4%
(±.1)
46.67%
(±0.37)
D 20.08%
22.64%
(±0.18)
D
Black Alone
1.6%
(±.1)
1.49%
(±0.13)
D 0.76%
0.72%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.64%
(±0.18)
D 1.62%
1.76%
(±0.09)
E
Totals
100% 100%

48.5% 48.5%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 15 OC Summary Statistics of Census Tracts with Liquor Stores
Tracts with Liquor Store Retailers: 285 / 51.5% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.62%
(±0.29)
N
18.03%
(±0.23)
C 10.05%
9.59%
(±0.15)
N
Hispanic (any race)
34.2%
(*)
40.26%
(±0.48)
D
40.87%
(±0.39)
D 17.61%
20.73%
(±0.25)
D
White Alone
41.4%
(±.1)
36.43%
(±0.35)
D
36.45%
(±0.29)
D 21.32%
18.76%
(±0.18)
D
Black Alone
1.6%
(±.1)
1.65%
(±0.12)
D
1.64%
(±0.1)
A 0.81%
0.85%
(±0.06)
E
All Other Race(s)
3.3%
(±.1)
3.04%
(±0.14)
E
3.01%
(±0.12)
E 1.71%
1.56%
(±0.07)
N
Totals
100%
100%

100%

51.5% 51.5%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

73

Hispanics fared worse with liquor stores compared to their Type 21 statistics. They were
even more underrepresented in no liquor census tracts (27.76%: liquor vs 30.46%: Type 21) and
likewise further overrepresented with regards to population scaled by the number of Type 21
licenses compared to number of liquor stores (40.87%: liquor vs 36.34%: Type 21). Whites were
nearly unchanged in census tracts without liquor stores compared to Type 21 licenses, but Asian
have increased representation (20.45% liquor vs 18.43% Type 21). On the other hand, both
Whites and Asians each represented nearly 2% less population for liquor stores compared to the
Type 21 licenses. These statistics suggested that liquor stores may be more concentrated in
Hispanic dominated census tracts compared to the other Type 21 retailers. The significant
quantity of Ds in the Dist columns of both tables further suggested Liquor Stores were
disparately distributed.
One other retailer category, Wholesale Clubs (i.e. Costco and Sam’s Club), was
comprised solely of retailers with Type 21 licenses. However, because only 2.8% of the county
population was present in the fifteen census tracts where those retailers (N=16) were located, this
sample was deemed too small to make a meaningful assessment and the results were excluded
from distribution assessment (see Table 17). On the other hand, 567 census tracts do not have a
Wholesale Club with an alcohol license and those tracts represent 97.2% of the county; this large
sample size did allow for an inference of a race-neutral function operating for their absence (see
Table 16).

74

Table 16 OC Summary Statistics of Census Tracts with Zero Wholesale Clubs
Tracts with Zero Wholesale Club Retailers: 567 / 97.2% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.28%
(±0.21)
N 18.96%
18.74%
(±0.2)
E
Hispanic (any race)
34.2%
(*)
34.32%
(±0.3)
N 33.24%
33.36%
(±0.29)
E
White Alone
41.4%
(±.1)
41.53%
(±0.26)
N 40.24%
40.36%
(±0.25)
E
Black Alone
1.6%
(±.1)
1.56%
(±0.09)
D 1.53%
1.51%
(±0.09)
E
All Other Race(s)
3.3%
(±.1)
3.32%
(±0.12)
C 3.24%
3.22%
(±0.11)
E
Totals
100% 100%

97.2% 97.2%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 17 OC Summary Statistics of Census Tracts with Wholesale Clubs
Tracts with Wholesale Club Retailers: 15 / 2.8% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
27.53%
(±1.77)
D
26.19%
(±1.68)
D 0.55%
0.77%
(±0.05)
D
Hispanic (any race)
34.2%
(*)
29.82%
(±2.76)
E
32.51%
(±2.65)
E 0.96%
0.84%
(±0.08)
E
White Alone
41.4%
(±.1)
36.88%
(±1.83)
C
35.65%
(±1.74)
C 1.16%
1.03%
(±0.05)
C
Black Alone
1.6%
(±.1)
2.08%
(±0.77)
E
2.09%
(±0.73)
E 0.04%
0.06%
(±0.02)
E
All Other Race(s)
3.3%
(±.1)
3.69%
(±0.7)
E
3.56%
(±0.66)
E 0.09%
0.1%
(±0.02)
E
Totals
100%
100%

100%

2.8% 2.8%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

75

Grocery Stores (N=412) represented the second largest retailer category behind Liquor
Stores (N=418) in the OC built environment. Although Grocery Stores may hold either a Type
20 or Type 21 license, the majority (N=300) operated with a Type 21 license like Liquor Stores.
In Orange County, there were 292 census tracts with 45.5% of the population that did not have a
grocery store with an alcohol license compared to 290 tracts with 54.5% of the population that
did (see Table 18 and Table 19). Table 18 shows that the Hispanic population is
underrepresented in census tracts without grocery stores holding an alcohol license, but overall,
the table Dist values did not cross the threshold for disparate distribution. Table 19 Dist column
values, on the other hand, suggested disparate distributions were occurring in the census tracts
with grocery stores.
Table 18 OC Summary Statistics of Census Tracts with Zero Grocery Stores
Tracts with Zero Grocery Store Retailers: 292 / 45.5% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.04%
(±0.28)
N 8.88%
8.67%
(±0.13)
E
Hispanic (any race)
34.2%
(*)
31.09%
(±0.42)
D 15.57%
14.16%
(±0.19)
C
White Alone
41.4%
(±.1)
44.86%
(±0.37)
C 18.85%
20.43%
(±0.17)
C
Black Alone
1.6%
(±.1)
1.58%
(±0.13)
D 0.71%
0.72%
(±0.06)
E
All Other Race(s)
3.3%
(±.1)
3.43%
(±0.15)
D 1.52%
1.56%
(±0.07)
E
Totals
100% 100%

45.5% 45.5%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

76

Table 19 OC Summary Statistics of Census Tracts with Grocery Stores
Tracts with Grocery Store Retailers: 290 / 54.5% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.9%
(±0.3)
E
20.5%
(±0.26)
N 10.63%
10.84%
(±0.16)
E
Hispanic (any race)
34.2%
(*)
36.79%
(±0.43)
C
38.3%
(±0.39)
D 18.63%
20.04%
(±0.23)
C
White Alone
41.4%
(±.1)
38.5%
(±0.36)
C
36.41%
(±0.3)
D 22.55%
20.97%
(±0.19)
C
Black Alone
1.6%
(±.1)
1.56%
(±0.13)
D
1.54%
(±0.1)
C 0.86%
0.85%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.24%
(±0.17)
C
3.24%
(±0.14)
C 1.81%
1.77%
(±0.09)
E
Totals
100%
100%

100%

54.5% 54.5%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Like, retailers in the Grocery Store category, retailers in the remaining categories may
hold either a Type 20 or Type 21 license. The third most prevalent retailer category was
Convenience Store (N=347). The majority of retailers in this category held a Type 20 license
(N=296), while the rest held a Type 21 (N=51). There were 332 census tracts representing 54.4%
of the population that did not have a retailer in the Convenience Store category (see Table 20),
while 250 tracts with 45.6% of the population did (see Table 21).

77

Table 20 OC Summary Statistics of Census Tracts with Zero Convenience Stores
Tracts with Zero Convenience Store Retailers: 332 / 54.4% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.9%
(±0.28)
C 10.6%
11.36%
(±0.15)
C
Hispanic (any race)
34.2%
(*)
26.25%
(±0.36)
D 18.59%
14.27%
(±0.2)
D
White Alone
41.4%
(±.1)
47.73%
(±0.36)
D 22.5%
25.94%
(±0.2)
D
Black Alone
1.6%
(±.1)
1.43%
(±0.11)
D 0.85%
0.78%
(±0.06)
E
All Other Race(s)
3.3%
(±.1)
3.7%
(±0.17)
D 1.81%
2.01%
(±0.09)
C
Totals
100% 100%

54.4% 54.4%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 21 OC Summary Statistics of Census Tracts with Convenience Stores
Tracts with Convenience Store Retailers: 250 / 45.6% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
17.85%
(±0.31)
C
17.8%
(±0.26)
C 8.91%
8.15%
(±0.14)
C
Hispanic (any race)
34.2%
(*)
43.66%
(±0.51)
D
45.54%
(±0.43)
D 15.61%
19.93%
(±0.23)
D
White Alone
41.4%
(±.1)
33.86%
(±0.37)
D
32.2%
(±0.3)
D 18.9%
15.45%
(±0.17)
D
Black Alone
1.6%
(±.1)
1.74%
(±0.15)
E
1.7%
(±0.12)
E 0.72%
0.8%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
2.89%
(±0.15)
N
2.76%
(±0.13)
C 1.52%
1.32%
(±0.07)
C
Totals
100%
100%

100%

45.7% 45.6%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

78

The values in the tables indicated that a large percentage of the Hispanic population had
access to convenience stores and are overrepresented compared to Whites and Asians. On the
other hand, the Asian values were slightly outside the margins of error and more closely track the
White population than the previously examined categories. Finally, Dist column values in both
tables suggested disparate distributions were occurring.
The fourth most prevalent retailer category was Gas Stations (N=270). Like Convenience
Stores, the majority of retailers in this category held a Type 20 license (N=255), while the rest
held a Type 21 (N=15). There were 372 census tracts representing 62.0% of the population that
did not have a retailer in the Gas Station category (see Table 22), while 210 tracts with 38.0% of
the population did (see Table 23). As the values in the Dist columns in both tables show, there
appeared to be a mix of race-neutral and absence of race-neutral distributions occurring with Gas
Stations. The Zero Gas Stations scenario appeared to be nearly race-neutral.
Table 22 OC Summary Statistics of Census Tracts with Zero Gas Stations
Tracts with Zero Gas Station Retailers: 372 / 62.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.47%
(±0.24)
C 12.1%
11.45%
(±0.15)
N
Hispanic (any race)
34.2%
(*)
33.29%
(±0.36)
N 21.21%
20.65%
(±0.22)
N
White Alone
41.4%
(±.1)
43.45%
(±0.32)
C 25.68%
26.95%
(±0.2)
N
Black Alone
1.6%
(±.1)
1.48%
(±0.12)
D 0.97%
0.92%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.32%
(±0.13)
C 2.07%
2.06%
(±0.08)
E
Totals
100% 100%

62.0% 62.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

79

Table 23 OC Summary Statistics of Census Tracts with Gas Stations
Tracts with Gas Station Retailers: 210 / 38.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
21.21%
(±0.54)
C
22.19%
(±0.35)
D 7.41%
8.05%
(±0.14)
C
Hispanic (any race)
34.2%
(*)
35.68%
(±0.42)
N
34.88%
(±0.5)
E 12.99%
13.55%
(±0.21)
N
White Alone
41.4%
(±.1)
38.04%
(±0.14)
C
37.73%
(±0.38)
C 15.72%
14.45%
(±0.16)
C
Black Alone
1.6%
(±.1)
1.72%
(±0.21)
E
1.79%
(±0.13)
E 0.6%
0.65%
(±0.05)
E
All Other Race(s)
3.3%
(±.1)
3.34%
(±0.21)
C
3.41%
(±0.2)
E 1.26%
1.27%
(±0.08)
E
Totals
100%
100%

100%

38.0% 38.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

There were 159 Pharmacies licensed to sell alcohol in Orange County; the majority of
which had Type 21 licenses (N=135). The pharmacies were spread among 140 census tracts
containing 26.1% of the population leaving 442 census tracts with 73.9% of the population
without pharmacies (see Table 24 and Table 25). Notably, Pharmacy was the only retailer
category where, although within the race neutral threshold, the White population was
overrepresented in the census tracts with the retailer.

80

Table 24 OC Summary Statistics of Census Tracts with Zero Pharmacies
Tracts with Zero Pharmacy Retailers: 442 / 73.9% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.4%
(±0.23)
N 14.41%
14.33%
(±0.17)
E
Hispanic (any race)
34.2%
(*)
35.21%
(±0.35)
N 25.26%
26.01%
(±0.26)
N
White Alone
41.4%
(±.1)
40.61%
(±0.29)
N 30.58%
30.0%
(±0.21)
N
Black Alone
1.6%
(±.1)
1.57%
(±0.1)
D 1.16%
1.16%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.21%
(±0.12)
C 2.46%
2.37%
(±0.09)
E
Totals
100% 100%

73.9% 73.9%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 25 OC Summary Statistics of Census Tracts with Pharmacies
Tracts with Pharmacy Retailers: 140 / 26.1% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.83%
(±0.63)
E
19.52%
(±0.43)
N 5.1%
5.18%
(±0.12)
E
Hispanic (any race)
34.2%
(*)
31.33%
(±0.56)
C
32.63%
(±0.59)
N 8.94%
8.19%
(±0.16)
C
White Alone
41.4%
(±.1)
43.61%
(±0.2)
N
42.53%
(±0.52)
E 10.82%
11.4%
(±0.15)
N
Black Alone
1.6%
(±.1)
1.58%
(±0.27)
D
1.58%
(±0.18)
D 0.41%
0.41%
(±0.05)
E
All Other Race(s)
3.3%
(±.1)
3.65%
(±0.27)
E
3.74%
(±0.28)
E 0.87%
0.96%
(±0.07)
E
Totals
100%
100%

100%

26.1% 26.1%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

81

Overall, the Pharmacy category Dist column values, like the Wholesale Club category,
suggested that a race neutral function operated for the absence of pharmacies in the built
environment. On the other hand, where pharmacies occurred barely passed the threshold for the
absence of a race neutral distribution.
Department Stores was the final category. As Table 26 shows, Department Stores (N=50)
were absent in 536 census tracts and the Dist column values showed a race-neutral distribution.
Likewise, Table 27 indicates they were present in 46 census tracts and the Dist column values
show a disparate distribution; moreover, the Asian population appeared overrepresented in the
census tracts where department stores were present.
Table 26 OC Summary Statistics of Census Tracts with Zero Department Stores
Tracts with Zero Department Store Retailers: 536 / 91.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.91%
(±0.21)
N 17.75%
17.21%
(±0.19)
N
Hispanic (any race)
34.2%
(*)
34.19%
(±0.31)
N 31.12%
31.11%
(±0.28)
E
White Alone
41.4%
(±.1)
42.04%
(±0.27)
N 37.67%
38.25%
(±0.25)
N
Black Alone
1.6%
(±.1)
1.55%
(±0.09)
D 1.43%
1.41%
(±0.09)
E
All Other Race(s)
3.3%
(±.1)
3.31%
(±0.11)
C 3.03%
3.01%
(±0.1)
E
Totals
100% 100%

91.0% 91.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

82

Table 27 OC Summary Statistics of Census Tracts with Department Stores
Tracts with Department Store Retailers: 46 / 9.0% of OC Population

Pop %
County
Pop %
Tracts
Dist
(Pop x L)
% Tracts
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
25.54%
(±1.26)
D
25.31%
(±0.9)
D 1.76%
2.3%
(±0.08)
D
Hispanic (any race)
34.2%
(*)
34.28%
(±0.85)
E
33.51%
(±1.28)
E 3.08%
3.09%
(±0.11)
E
White Alone
41.4%
(±.1)
34.89%
(±0.31)
D
35.83%
(±0.83)
D 3.73%
3.15%
(±0.08)
D
Black Alone
1.6%
(±.1)
1.74%
(±0.4)
E
1.74%
(±0.31)
E 0.14%
0.16%
(±0.03)
E
All Other Race(s)
3.3%
(±.1)
3.54%
(±0.4)
E
3.61%
(±0.38)
E 0.3%
0.32%
(±0.04)
E
Totals
100%
100%

100%

9.0% 9.0%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 28 summarizes the (Pop x L) % Tracts column values and overall distribution
assessments of the license/retailers for all the census tract scenarios. As the table indicates, the
Hispanic-dominated communities had the most overrepresentation scenarios. Also, the majority
of scenarios exceeded the disparate distribution threshold for the Hispanic population. Two
anecdotal observations were also made, first although the Black population accounted for less
than 2% of the Orange County total population and was not part of the distribution assessments,
there was a positive correlation for the Black population with all but two of the scenarios,
Pharmacies and Grocery Stores. Second, there did not appear to be a consistent positive or
negative correlation between the Other population and the presence of alcohol licenses or
retailers; this was likely due to the fact that the Other population was made up of multiple small
sub-populations.

83

Table 28 OC Census Tracts with Licenses Population Summary
Asian Hispanic White Black Other Dist
Orange County
19.5%
(±0.1)
34.2%
*
41.4%
(±0.1)
1.6%
(±0.1)
3.3%
(±0.1)
N/A
All
19.75%
(±0.13)
39.07%
(±0.2)
36.35%
(±0.15)
1.65%
(±0.05)
3.18%
(±0.07)
D
Type 21
20.28%
(±0.17)
36.34%
(±0.26)
38.39%
(±0.2)
1.65%
(±0.07)
3.34%
(±0.09)
C
Type 20
18.99%
(±0.19)
43.04%
(±0.31)
33.38%
(±0.22)
1.66%
(±0.08)
2.93%
(±0.1)
D
Liquor Stores
18.03%
(±0.23)
40.87%
(±0.39)
36.45%
(±0.29)
1.64%
(±0.1)
3.01%
(±0.12)
D
Grocery Stores
20.5%
(±0.26)
38.3%
(±0.39)
36.41%
(±0.3)
1.54%
(±0.1)
3.24%
(±0.14)
D
Convenience Stores
17.8%
(±0.26)
45.54%
(±0.43)
32.2%
(±0.3)
1.7%
(±0.12)
2.76%
(±0.13)
D
Gas Stations
22.19%
(±0.35)
34.88%
(±0.5)
37.73%
(±0.38)
1.79%
(±0.13)
3.41%
(±0.2)
C
Pharmacies
19.52%
(±0.43)
32.63%
(±0.59)
42.53%
(±0.52)
1.58%
(±0.18)
3.74%
(±0.28)
C
Department Stores
25.31%
(±0.9)
33.51%
(±1.28)
35.83%
(±0.83)
1.74%
(±0.31)
3.61%
(±0.38)
D
Wholesale Clubs
26.19%
(±1.68)
32.51%
(±2.65)
35.65%
(±1.74)
2.09%
(±0.73)
3.56%
(±0.66)
X

*Estimate is controlled, margin of error treated as zero

C: Race Correlated | D: Disparate Distribution | N/A: Not Applicable | X: Exclude

Table 29 summarizes the Pop % Tracts values and overall distribution assessments of the
zero license/retailers for all the census tract scenarios. As the table indicates, the Hispanic
population was underrepresented in the census tracts without licenses, while the White

84

population was overrepresented in nearly every census tract without licenses, except for the
Pharmacies scenario.
Table 29 OC Census Tracts with Zero Licenses Population Summary
Asian Hispanic White Black Other Dist
Orange County
19.5%
(±0.1)
34.2%
*
41.4%
(±0.1)
1.6%
(±0.1)
3.3%
(±0.1)
N/A
All
20.55%
(±0.48)
18.75%
(±0.56)
55.18%
(±0.67)
1.51%
(±0.23)
4.01%
(±0.26)
D
Type 21
18.43%
(±0.37)
30.46%
(±0.5)
46.12%
(±0.49)
1.48%
(±0.18)
3.5%
(±0.21)
D
Type 20
19.21%
(±0.33)
23.66%
(±0.43)
51.78%
(±0.46)
1.5%
(±0.15)
3.83%
(±0.19)
D
Liquor Stores
20.45%
(±0.3)
27.76%
(±0.36)
46.67%
(±0.37)
1.49%
(±0.13)
3.64%
(±0.18)
D
Grocery Stores
19.04%
(±0.28)
31.09%
(±0.42)
44.86%
(±0.37)
1.58%
(±0.13)
3.43%
(±0.15)
C
Convenience Stores
20.9%
(±0.28)
26.25%
(±0.36)
47.73%
(±0.36)
1.43%
(±0.11)
3.7%
(±0.17)
D
Gas Stations
18.47%
(±0.24)
33.29%
(±0.36)
43.45%
(±0.32)
1.48%
(±0.12)
3.32%
(±0.13)
C
Pharmacies
19.4%
(±0.23)
35.21%
(±0.35)
40.61%
(±0.29)
1.57%
(±0.1)
3.21%
(±0.12)
N
Department Stores
18.91%
(±0.21)
34.19%
(±0.31)
42.04%
(±0.27)
1.55%
(±0.09)
3.31%
(±0.11)
N
Wholesale Clubs
19.28%
(±0.21)
34.32%
(±0.3)
41.53%
(±0.26)
1.56%
(±0.09)
3.32%
(±0.12)
N

*Estimate is controlled, margin of error treated as zero

C: Race Correlated | D: Disparate Distribution | N/A: Not Applicable | X: Exclude

Individually, these various results suggested that multiple factors and functions
influenced both the presence and absence of licenses and retailers in the built environment.
However, there did appear to be support for correlations between race and the density of alcohol
licenses, which suggested disparate distributions. For example, the overrepresentation of

85

Hispanic populations in census tracts with alcohol licenses, or the overrepresentation of White
populations in census tracts without alcohol licenses. But a conclusion beyond those generalized
observations would not be supported by the data and analysis of this study.
4.1.2. Census Tract Alcohol License Density
The next step in analyzing the likelihood of race-neutral distributions of alcohol licenses
at the census tract level was to assess how race/ethnicity correlated with the license density per
square mile in each census tract. Figure 20 provides a visual representation of the census tract
license density per square mile for Orange County. The slope polarity (sign) of a linear
regression trend line was used to indicate a positive or negative correlation between a dependent
variable (licenses per square mile per census tract) and an independent variable (race/ethnicity
percent population per census tract).
First, scatter plots were generated for each race/ethnicity census tract percentage versus
the census tract licenses or retailers per square mile. Next, linear regressions were performed
with the licenses and retailers per square mile as the dependent variable for each scatter plot and
the resulting trend lines were coded red if the race/ethnicity indicated a positive correlation with
increasing population percentage and green if the race/ethnicity indicated a negative correlation
with increasing population percentage. Finally, a regression result was rejected for further
analysis if the p-value was greater than 0.05. Out of ten regression scenarios, the Department
Stores and Wholesale Club categories were rejected for further analysis since their regression p-
values were greater than 0.05.
For the regressions with p-values less than 0.05, if the slope polarities between two or
more populations were inconsistent, then the likelihood of a race-neutral distribution function
was rejected and assumed to be a race-correlated distribution. To be clear, only the differences in

86

the slope polarities between the populations were assessed. This methodology allowed for a
quick visual inspection to determine the potential existence of race-neutral versus race-correlated
distributions in the distributions per square mile per census tract of the licenses/retailers.

Figure 20 OC Licenses/Retailers per Square Mile

87

Figure 21 shows the scatter plots and trend lines for All Licenses per square mile per
census tract. This figure presented a positive slope with increasing Hispanic population
percentage and increasing areal alcohol license density correlation, while the White and Asian
populations manifested a negative slope and decreasing population percentage correlations. This
result was interpreted as a race-correlated distribution. Similar results occurred for the Type 20
licenses (see Figure 22).

Before assessing correlations for the Type 21 licenses and Liquor Stores, the Asian
regressions had to be rejected for p-values greater than 0.05 (see Figure 23 and Figure 24). .
Notwithstanding rejected of the Asian results, both these scenarios produced opposite slope

Figure 21 OC Linear Regressions on All Licenses per Square Mile

Figure 22 OC Linear Regressions on Type 20 Licenses per Square Mile

88

polarities between the Hispanic and White populations, this study’s criteria for a race-correlated
distribution.

Grocery Stores and Convenience Stores were the next categories to be analyzed. The
Grocery Store category was the second largest retailer category (N=412) and the majority of
retailers (N=300) held Type 21 licenses, while Convenience Stores was the third largest (N=347)
with a majority of retailers (N=296) holding Type 20 licenses (see Figure 25 and Figure 26).
Both scenarios presented one race with opposite trend line polarities to the other two races and
were deemed race-correlated distributions.

Figure 23 OC Linear Regressions on Type 21 Licenses per Square Mile

Figure 24 OC Linear Regressions on Liquor Stores per Square Mile

89

Gas Stations and Pharmacies were the last two categories that were analyzed. Gas
Stations (N=270) were composed primarily of Type 20 license holders (N=255), while
Pharmacies (N=159) were primarily Type 21 licenses (N=135). Before observing the trend line
polarities, the Asian regressions were rejected for having p-values greater than 0.05. The non-
rejected trend lines had opposite polarities, meeting the criteria for race-correlated distributions.

Figure 25 OC Linear Regressions on Grocery Stores per Square Mile

Figure 26 OC Linear Regressions on Convenience Stores per Square Mile

90

Table 30 summarizes the results for the percent population versus licenses per square
mile regressions. For all the non-rejected results, the trend line slope polarities for the Hispanic
population were always opposite to the White population. These results met the study’s threshold
for the presence of race-correlated distribution functions in the built environment, at least as
between White and Hispanic populations. A similar result appeared likely as between Asian and
Hispanic populations, although six Asian observations had to be rejected as inconclusive due to
p-values greater than 0.05.

Figure 27 OC Linear Regressions on Gas Stations Stores per Square Mile

Figure 28 OC Linear Regressions on Pharmacies per Square Mile

91

Table 30 Census Tract Linear Regressions per Square Mile Trend Line Summary
Asian Hispanic White Black Other Dist
All
─ + ─
X
─
C
Type 21 X
+ ─
X
─
C
Type 20
─ + ─
X
─
C
Liquor Stores X
+ ─
X X C
Grocery Stores
─ + ─
X
─
C
Convenience
Stores
─ + ─
X
─
C
Gas Stations X
+ ─
X
─
C
Pharmacies X
+ ─
X X C
Department
Stores
X X X X X X
Wholesale Clubs X X X X X X
─ Negative Slope +Positive Slope X: Excluded C: Race Correlated

Another metric evaluated for race/ethnicity correlation was the distribution of alcohol
licenses based on the density of licenses per 1,000 people (see Figure 29). This metric was
concerned with the population density where alcohol licenses are present, whereas licenses per
square mile evaluated the areal density of those licenses.

92

First, scatter plots were generated for each race/ethnicity census tract percentage versus
the census tract for each license type and all the retailer categories per 1,000 people in each
census tract. Next, linear regressions were performed with the licenses/retailers per 1,000 people
as the dependent variable for each scatter plot. Again, the resulting trend lines were coded red if
the race/ethnicity indicated a positive correlation with increasing population percentage and

Figure 29 OC Licenses and Retailers per 1,000 People per Census Tract

93

green if the race/ethnicity indicated a negative correlation with increasing population percentage.
Finally, regression results with p-values greater than 0.05 were rejected for further analysis. As a
result, six of the ten license/retailer scenarios were rejected for having two or more p-values
greater than 0.05: All Licenses, Type 20 licenses, Convenience Stores, Gas Stations, Department
Stores, and Wholesale Clubs. For the regressions with p-values less than 0.05, if the slope
polarities between two or more populations were inconsistent, then the likelihood of a race-
neutral distribution function was rejected and assumed to be a race-correlated distribution.
Figure 30 shows the regression results for Type 21 scenario and Figure 31 the results for
Liquor Stores which was also composed entirely of retailers with Type 21 licenses While the
Type 21 Hispanic regression result was rejected for its p-value being too large, the Asian and
White regressions p-values were under 0.05 and the exhibited opposite polarity slopes. The
Liquor Store regressions were all valid, and the Asian and Hispanic trend lines exhibited
opposite slope polarities to the White trend line. Thus, these scenarios met the criteria for race-
correlated distributions

Figure 30 OC Linear Regressions on Type 21 Licenses per 1,000 People

94

The Hispanic regression results in the remaining two categories, Grocery Store (Figure
32) and Pharmacy (Figure 33), were also rejected for having p-values greater than 0.05.
However, the White and Asian p-values were below 0.05 and manifested opposite polarity
slopes, satisfying the criteria for race-correlated distributions.

Figure 31 OC Linear Regressions on Liquor Stores per 1,000 People

Figure 32 OC Linear Regressions on Grocery Stores per 1,000 People

95

Table 31 summarizes the results for the percent population versus licenses per 1,000
people regressions. Overall, there were a number of rejected results, however, for non-rejected
results the slope polarities for the Asian population were opposite to the White population. These
results met the study’s threshold for the presence of race-correlated distribution functions in the
built environment, at least as between White and Asian populations. A similar result appeared
likely as between White and Hispanic populations, although nine Hispanic observations had to
be rejected as inconclusive due to p-values greater than 0.05.

Figure 33 OC Linear Regressions on Pharmacies per 1,000 People

96

Table 31 Census Tract Linear Regressions per 1,000 People Trend Line Summary
Asian Hispanic White Black Other Dist
All
─
X X X
+
X
Type 21
─
X
+
X X C
Type 20
─
X X X
+
X
Liquor Stores
─ ─ +
X X C
Grocery Stores
─
X
+
X X C
Convenience
Stores
X X
+
X X X
Gas Stations X X
+ +
X X
Pharmacies
─
X
+
X X C
Department
Stores
X X X X X X
Wholesale Clubs X X X X X X
─ Negative Slope +Positive Slope X: Excluded C: Race Correlated

4.1.3. Census Tract Alcohol License Hot Spots: Getis-Ord Gi* Statistic
A Getis-Ord Gi* statistic was utilized to determine the presence of statistically significant
clustering of alcohol licenses. Specifically, the Optimized Hot Spot Analysis tool in ArcGIS Pro
was configured to assess the optimal parameters for aggregating all study area licenses into
bounding polygons defined by the census tract boundaries in OC. The tools output was then
reviewed to identify the neighborhood distance that was identified by the run (~6.4 miles). The
statistic was then run for all retailer licenses in Orange County as a single group, the Type 21
licenses, the Type 20 licenses, and for licenses by each category of retailer using the same

97

parameter for each run to create the Optimized Hot Spot Analysis. However, the Hot Spot
Analysis failed to provide results for Wholesale Clubs (N=16) because the statistic requires a
minimum of 30 data points to generate valid results. Figure 34 shows the Optimized Hot Spots
results.

Figure 34 OC Optimized Hot Spots Based Upon Census Tract Boundaries

98

As Figure 34 reveals, there appeared to be a consistent distribution bias of license hot
spots for the northern portion of the county and license cold spots for the southern portion.
Surprisingly, the Type 21 licenses hot and cold spot distributions were attenuated in distribution
and statistical confidence. Moreover, the Optimized Hot Spot analysis for Grocery Stores and
Gas Stations registered just a few hot spots, while Pharmacies and Department Stores showed no
statistically significant hot or cold spots. Thus, those categories were excluded from further
analysis.
After generating the Optimized Hot Spots statistics, the race/ethnicity summary statistics
were generated for the confidence levels for All Licenses, Type21, Type 20, Liquor Stores, and
Convenience Stores. As previously discussed, inclusion or exclusion of highly populated census
tracts could bias estimates made with small numbers of census tracts; however, the assumption in
this study is that the aggregation of more than twenty census tracts would sufficiently render that
bias negligible. That is not to say that aggregations of less than twenty tracts cannot produce
valid and meaningful results, only that those results were not evaluated for disparate
distributions.
Table 32 shows the summary statistics for the All Licenses Optimized Hot Spot Analysis;
there appeared to be converse racial representations in each confidence level of the hot and cold
spot populations between Hispanic and White populations. On the other hand, the Asian
population was overrepresented in two of the three Hot bins and underrepresented in all the Cold
bins. Table 33 shows that while the Type 21 Licenses manifested only two statistically
significant bins—Hot 90% Confidence and Cold 90% Confidence—the Asian and Hispanic
populations were aligned in overrepresentation in Hot and underrepresentation in Cold, with the

99

White population produced the opposite representations. Finally, the Type 20 Licenses (Table
34) also showed converse representations between Hispanic and White populations.
Table 32 OC All Licenses Optimized Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
127
24.95%
(±0.11)
D
52.91%
(±0.18)
D
18.34%
(±0.09)
D
1.67%
(±0.04)
E
2.13%
(±0.04)
D
Hot 95%
Confidence
70
23.42%
(±0.07)
D
41.53%
(±0.11)
D
30.61%
(±0.08)
D
1.76%
(±0.03)
D
2.68%
(±0.03)
D
Hot 90%
Confidence
30
15.63%
(±0.04)
D
43.98%
(±0.07)
D
35.45%
(±0.05)
D
1.41%
(±0.02)
D
3.53%
(±0.02)
C
Cold 90%
Confidence
26
17.03%
(±0.05)
D
21.14%
(±0.06)
D
56.06%
(±0.07)
D
1.75%
(±0.03)
D
4.01%
(±0.03)
D
Cold 95%
Confidence
50
13.95%
(±0.05)
D
16.94%
(±0.06)
D
63.37%
(±0.08)
D
1.44%
(±0.03)
A
4.29%
(±0.03)
D
Cold 99%
Confidence
21
11.99%
(±0.03)
D
15.4%
(±0.05)
D
66.05%
(±0.06)
D
1.52%
(±0.02)
E
5.05%
(±0.03)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Table 33 OC Type 21 Licenses Optimized Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 90%
Confidence
126
23.64%
(±0.1)
D
41.84%
(±0.16)
D
29.62%
(±0.11)
D
1.94%
(±0.04)
D
2.97%
(±0.05)
D
Cold 90%
Confidence
74
13.58%
(±0.07)
D
18.05%
(±0.08)
D
62.39%
(±0.11)
D
1.54%
(±0.03)
E
4.44%
(±0.05)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

100

Table 34 OC Type 20 Licenses Optimized Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
147
21.86%
(±0.11)
D
54.49%
(±0.19)
D
20.06%
(±0.1)
D
1.58%
(±0.04)
E
2.02%
(±0.04)
D
Hot 95%
Confidence
56
22.06%
(±0.07)
D
43.96%
(±0.11)
D
29.36%
(±0.07)
D
1.73%
(±0.03)
D
2.89%
(±0.03)
D
Hot 90%
Confidence
16
18.21%
(±0.03)
X
47.54%
(±0.05)
X
29.77%
(±0.04)
X
1.81%
(±0.01)
X
2.67%
(±0.02)
X
Cold 90%
Confidence
9
14.7%
(±0.03)
X
18.35%
(±0.03)
X
60.43%
(±0.04)
X
1.37%
(±0.02)
X
5.15%
(±0.02)
X
Cold 95%
Confidence
15
16.04%
(±0.03)
X
16.54%
(±0.04)
X
61.06%
(±0.05)
X
1.3%
(±0.02)
X
5.06%
(±0.03)
X
Cold 99%
Confidence
7
14.89%
(±0.01)
X
13.44%
(±0.02)
X
66.44%
(±0.03)
X
1.89%
(±0.01)
X
3.35%
(±0.01)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Overall, the Dist columns of the tables indicated that the majority of the bins manifested
distributions that exceeded this study’s disparate distribution threshold of 10% difference from
the county-wide proportions. However, four bins from Type 20 licenses, although statistically
significant, were excluded from the disparate distribution analysis for having less than 20 census
tracts represented in the results.
Liquor Stores and Convenience Stores were the only retailer categories with at least one
bin containing more than 20 census tracts. The Liquor Store category, Table 35, indicated a
strong overrepresentation of the Hispanic population in hot spots and a strong overrepresentation
of the White population in cold spots. However, the Asian population had mixed over and under

101

representation in both hot and cold spots. Likewise, the Convenience Store category, Table 36,
also manifested overrepresentation of Hispanic populations in hot spots and overrepresentation
of White populations in cold spots.
Table 35 OC Liquor Stores Optimized Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
216
26.31%
(±0.13)
D
38.05%
(±0.2)
D
30.76%
(±0.14)
D
1.79%
(±0.05)
D
3.1%
(±0.06)
C
Hot 95%
Confidence
26
16.28%
(±0.04)
D
52.1%
(±0.06)
D
27.82%
(±0.04)
D
1.86%
(±0.02)
D
1.95%
(±0.01)
D
Hot 90%
Confidence
20
10.91%
(±0.03)
D
52.73%
(±0.06)
D
32.4%
(±0.04)
D
1.82%
(±0.01)
D
2.15%
(±0.01)
D
Cold 90%
Confidence
17
27.58%
(±0.05)
X
20.71%
(±0.04)
X
45.79%
(±0.06)
X
1.69%
(±0.02)
X
4.23%
(±0.05)
X
Cold 95%
Confidence
40
23.24%
(±0.07)
D
16.63%
(±0.09)
D
53.94%
(±0.09)
D
1.17%
(±0.02)
D
5.02%
(±0.04)
D
Cold 99%
Confidence
104
18.41%
(±0.1)
C
16.67%
(±0.1)
D
58.71%
(±0.13)
D
1.78%
(±0.05)
D
4.42%
(±0.05)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

102

Table 36 OC Convenience Stores Optimized Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
167
24.34%
(±0.12)
D
50.91%
(±0.2)
D
20.9%
(±0.11)
D
1.65%
(±0.05)
E
2.19%
(±0.05)
D
Hot 95%
Confidence
52
23.28%
(±0.06)
D
37.72%
(±0.09)
D
33.79%
(±0.06)
D
2.04%
(±0.03)
D
3.16%
(±0.03)
C
Hot 90%
Confidence
28
16.86%
(±0.04)
D
48.19%
(±0.07)
D
30.06%
(±0.05)
D
1.56%
(±0.02)
E
3.33%
(±0.02)
E
Cold 90%
Confidence
11
24.96%
(±0.04)
X
10.34%
(±0.03)
X
58.62%
(±0.05)
X
1.21%
(±0.01)
X
4.87%
(±0.04)
X
Cold 95%
Confidence
35
14.13%
(±0.04)
D
17.65%
(±0.06)
D
62.13%
(±0.07)
D
1.39%
(±0.02)
D
4.69%
(±0.04)
D
Cold 99%
Confidence
79
14.27%
(±0.07)
D
18.07%
(±0.09)
D
61.85%
(±0.11)
D
1.59%
(±0.04)
E
4.21%
(±0.04)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Overall, the majority of Optimized Hot Spot bins provided strong support for the
conclusion that the alcohol license and race/ethnicity population ratios in the hot/cold census
tracts exhibited environmental clustering differing significantly from the county norm.
Moreover, the Dist column values in the tables likewise exceeded this study’s disparate
distribution threshold of differences between observed and county-wide population proportions
greater than 10%.
The Hot Spot Analysis was also performed using a 3-mile distance band in order to
observe whether clustering also occurred at finer scale (see Figure 35). Three miles was chosen
to represent a reasonable distance an OC resident would travel to a retailer on a regular basis. As
Figure 35 reveals, there continued to be a distribution bias of license hot spots in the northern

103

portion of the county and a lesser distribution of cold spots in the southern portion. However,
while this observational method attenuated some of the hot and cold spot distributions in both
quantity and statistical confidence for several scenarios, new hot and cold spots were also
identified.

Figure 35 OC Three Mile Observational Hot Spots Based Upon Census Tract Boundaries

104

Table 37 exhibits the Observational Hot Spots summary statistics for All Licenses.
Examining the scenario, the number of hot and cold spots had diminished significantly, with only
one bin—Hot 90% Confidence—having a sufficiently large sample size (N=33) for disparate
analysis. That bin, Hot 90% Confidence, showed both significant overrepresentation for the
Hispanic population and underrepresentation of the White and Asian populations. Moreover, the
Dist column values exceeded the disparate distribution threshold.
Table 37 OC All Licenses Observational Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 95%
Confidence
11
9.46%
(±0.02)
X
70.1%
(±0.06)
X
16.33%
(±0.02)
X
2.15%
(±0.02)
X
1.97%
(±0.01)
X
Hot 90%
Confidence
33
17.1%
(±0.05)
D
58.8%
(±0.1)
D
19.32%
(±0.05)
D
2.3%
(±0.02)
D
2.47%
(±0.02)
D
Cold 90%
Confidence
19
19.44%
(±0.03)
X
15.87%
(±0.03)
X
59.44%
(±0.04)
X
1.75%
(±0.01)
X
3.5%
(±0.02)
X
Cold 95%
Confidence
2
17.21%
(±0.01)
X
8.61%
(±0.0)
X
68.13%
(±0.01)
X
0.33%
(±0.0)
X
5.72%
(±0.01)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Table 38 provides the summary statistics for Type 20 licenses (there were no hot or cold
spots for Type 21 licenses). The number of hot and cold bins and census tracts increased
compared to the combined licenses scenario, indicating that much of the attenuation of the All
Licenses Observational Hot Spots from the All Licenses Optimized was attributable to the Type
21 license distributions. For Type 20 licenses, four bins showed both significant
overrepresentation for the Hispanic population and underrepresentation of the White population

105

and mixed representations for the Asian population. Overall, the majority of bins surpassed this
study’s disparate distribution threshold.
Table 38 OC Type 20 Licenses Observational Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
27
9.75%
(±0.04)
D
70.77%
(±0.1)
D
16.12%
(±0.04)
D
1.52%
(±0.02)
E
1.83%
(±0.02)
D
Hot 95%
Confidence
49
11.74%
(±0.05)
D
69.11%
(±0.11)
D
15.52%
(±0.05)
D
1.8%
(±0.03)
D
1.83%
(±0.02)
D
Hot 90%
Confidence
26
20.63%
(±0.05)
C
53.11%
(±0.08)
D
22.74%
(±0.05)
D
1.53%
(±0.02)
E
1.98%
(±0.02)
D
Cold 90%
Confidence
28
23.03%
(±0.06)
D
15.1%
(±0.05)
D
56.46%
(±0.06)
D
1.08%
(±0.01)
D
4.33%
(±0.03)
D
Cold 95%
Confidence
10
13.64%
(±0.02)
X
10.38%
(±0.02)
X
71.78%
(±0.04)
X
1.52%
(±0.01)
X
2.67%
(±0.01)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Moreover, there was an interesting difference in the Type 20 Observational Hot Spots
from the Type 20 Optimized Hot Spots: the occurrence of a hot spot located at census tract
524.08 (see Figure 36). This hot spot illustrated the importance of carefully reviewing and
understanding Hot Spot parameters and results. Specifically, census tract 524.08 had zero Type
20 licenses while most of the surrounding census tracts within three miles of census tract 540.08
contained at least one. Thus, census tract 524.08 was like a Type 20 license free island in a sea of
census tracts with Type 20 licenses.

106

Even though census tract 524.08 did not have any Type 20 licenses, it was presumed that
the race/ethnicity population dynamics were representative of its neighbors within the distance
band value (3 miles). Table 39 provides the summary statistics for census tract 524.08, from
which it can be inferred that the tracts within three miles also likely have a greater proportion of
White population than the county-wide statistics. Moreover, this hot spot illustrated how the hot

Figure 36 Observational Hot Spot of Type 20 Licenses Occurring at Census Tract 524.08

107

spot analysis indirectly accounts for spillover effects because the Getis-Ord Gi* statistic takes
into account licenses occurring in nearby census tracts within the distance band value.
Table 39 Census Tract 524.08 Type 20 Hot Spots Summary Statistics
Tracts Asian Hispanic White Black Other
County
Statistics
582
19.5%
(±0.1)
34.2%
*
41.4%
(±0.1)
1.6%
(±0.1)
3.3%
(±0.1)
Hot 95%
Confidence
1
13.16%
(±0.01)
9.99%
(±0.01)
68.93%
(±0.01)
0.63%
(±0.0)
7.28%
(±0.01)

*Estimate is controlled, margin of error treated as zero

As mentioned, the observational Hot Spot analysis was performed using a three-mile
distance band; however, at some smaller value there would not have been a hot spot at census
tract 524.08. On the other hand, the hot spot may have grown or moved to the two or three
census tracts south of 524.08 where multiple Type 20 licenses occur if a larger band value was
used. Further increasing the distance band value would eventually result in the area becoming a
cold spot or not statistically significant (see Figure 34 where the distance band was ~6.4 miles).
Table 40 provides the summary statistics on Observational Hot Spots for Liquor Stores,
which are a subset of Type 21 licenses. The fact that there were hot spots with Liquor Stores and
not Type 21 licenses suggested that the other retailers with Type 21 licenses were more
diffusively distributed compared to Liquor Stores. Moreover, Table 40 shows that Liquor Store
Observational Hot Spots continued to trend heavily Hispanic while the local Cold Spots trended
heavily White. Overall, the majority of bins surpassed this study’s disparate distribution
threshold.

108

Table 40 OC Liquor Stores Observational Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 95%
Confidence
67
24.2%
(±0.08)
D
43.71%
(±0.13)
D
26.69%
(±0.08)
D
2.16%
(±0.03)
D
3.24%
(±0.04)
E
Hot 90%
Confidence
37
23.06%
(±0.05)
D
43.37%
(±0.08)
D
29.69%
(±0.05)
D
1.66%
(±0.02)
E
2.21%
(±0.02)
D
Cold 90%
Confidence
15
24.58%
(±0.04)
X
12.68%
(±0.03)
X
55.81%
(±0.05)
X
2.17%
(±0.02)
X
4.76%
(±0.02)
X
Cold 95%
Confidence
43
26.85%
(±0.07)
D
13.13%
(±0.06)
D
53.14%
(±0.08)
D
1.86%
(±0.03)
D
5.03%
(±0.03)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Convenience Stores and Gas Stations were the only two categories remaining with at
least one bin containing sufficient samples for disparate distribution analysis. Table 41 indicates
overrepresentation of Hispanic populations in Hot Spots and overrepresentation of White
populations in Cold Spots for Convenience Stores. Likewise, Table 42 shows a Hispanic
overrepresentation in the Hot 90% Confidence bin for Gas Stations. Overall, the majority of bins
surpassed this study’s disparate distribution threshold, although the Gas Stations scenario was
close to being excluded for having only 27 census tracts.

109

Table 41 OC Convenience Stores Observational Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
19
10.89%
(±0.03)
X
72.76%
(±0.08)
X
13.35%
(±0.03)
X
1.44%
(±0.01)
X
1.56%
(±0.01)
X
Hot 95%
Confidence
67
14.61%
(±0.07)
D
65.97%
(±0.14)
D
16.05%
(±0.07)
D
1.61%
(±0.03)
E
1.76%
(±0.03)
D
Hot 90%
Confidence
31
23.41%
(±0.05)
D
49.77%
(±0.08)
D
22.59%
(±0.05)
D
1.38%
(±0.02)
D
2.85%
(±0.03)
D
Cold 90%
Confidence
42
24.82%
(±0.07)
D
14.29%
(±0.06)
D
54.95%
(±0.08)
D
1.48%
(±0.02)
E
4.46%
(±0.03)
D
Cold 95%
Confidence
17
15.6%
(±0.02)
X
16.28%
(±0.03)
X
62.91%
(±0.04)
X
1.27%
(±0.01)
X
3.93%
(±0.02)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Table 42 OC Gas Stations Observational Hot Spots Summary Statistics
Tracts Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
582
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 90%
Confidence
27
13.14%
(±0.03)
D
54.2%
(±0.07)
D
28.32%
(±0.04)
D
1.94%
(±0.02)
D
2.4%
(±0.02)
D
Cold 90%
Confidence
3
2.96%
(±0.0)
X
94.45%
(±0.03)
X
2.27%
(±0.01)
X
0.08%
(±0.0)
X
0.24%
(±0.0)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Overall, whether OC census tracts were analyzed by license count, license per square
mile, license per 1,000 population, by Optimized Hot Spots, or by Observational Hot Spots, the
Hispanic population appeared to be overrepresented at the census tract level more often than

110

would be expected based upon the county-wide population statistics for nearly all license types
and retailer categories. The Asian population, on the other hand, showed mixed representation
results, with overrepresentation in some scenarios and underrepresentation in others.
Furthermore, the White population showed consistent overrepresentation in census tracts that do
not have alcohol licenses and often had many indicators suggesting underrepresentation in
census tracts with alcohol licenses, with possibly the exception of Pharmacy retailers. Finally,
the majority of scenario results exceeded this study’s disparate distribution thresholds.
4.2 Scale 2: Scaled Population Grid Analytical Results
The census tract level analysis of alcohol license distributions suggested that
race/ethnicity biases were in operation in Orange County. However, there was concern with
using census tracts as the basis for spatial analysis because of the potential introduction of
unknown issues in the form of modifiable areal unit problems (MAUP) due to the variable nature
of census tract boundaries. There was also the issue of spillover effects—unmeasured impacts in
adjacent census tracts—due to the fact that many retailers were right next to census tract
boundaries because census tract boundaries often run down the centerline of streets. While some
of these concerns were partially addressed by the Census Tract Hot Spot Analyses, another way
to address these concerns was to replace the random areas defined by census tract boundaries
with a consistently applied scaled population grid.
After creating the scaled population grid, the cells with no population were removed in
order to aggregate the distributions of alcohol licenses to cells with identified populations. These
scaled population cells were then used to perform the same simple summary statistics, linear
regression trend line slope analysis, and Getis Ord Gi* hot spot analysis performed in the
previous sections. The following sections examine the results of those analyses.

111

4.2.1. Scaled Population Grid Alcohol License Summary Statistics
If a race-neutral function controls the distribution of alcohol retailers in Orange County
built environment, then both the presence and absence of alcohol retailers should generally
follow the demographic profile of the county. Thus, the first step in analyzing the distribution of
alcohol licenses at the cell level was to explore the percentage of the cell populations that do and
do not have alcohol licenses for all the various licenses and retailer scenarios. The maps in
Figure 37 present the cell counts for all the license and retailer scenarios for Orange County,
except for Wholesale Clubs which was excluded due to the small sample size (N=16).

Figure 37 OC Licenses and Retailers Per Cell

112

Table 43 shows the zero licenses summary statistics, while Table 44 shows the summary
statistics for cells that contain licenses. These tables, with different county-wide population
denominators compared to the census tract zero licenses tables (cells: 47.8% and 52.2% vs tracts:
15.3% and 84.7%) still manifested very similar race/ethnicity dynamics to the census tract
versions. Moreover, because the cell sizes are approximately 0.28 square miles, they represented
the populations within roughly 0.5 miles of the alcohol retailers, compared to the random range
of distances when using census tracts.
Table 43 OC Summary Statistics of Cells with Zero Licenses
Cells with Zero Alcohol Licenses: 1,461 / 47.8% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.47%
(±0.31)
C 9.33%
9.79%
(±0.15)
N
Hispanic (any race)
34.2%
(*)
25.81%
(±0.39)
D 16.36%
12.35%
(±0.19)
D
White Alone
41.4%
(±.1)
48.49%
(±0.4)
D 19.8%
23.19%
(±0.19)
D
Black Alone
1.6%
(±.1)
1.52%
(±0.14)
D 0.75%
0.73%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.7%
(±0.18)
D 1.59%
1.77%
(±0.09)
C
Totals
100% 100%

47.8% 47.8%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

113

Table 44 OC Summary Statistics of Cells with Alcohol Licenses
Cells with Alcohol Licenses: 711 / 52.2% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.63%
(±0.27)
N
17.86%
(±0.17)
C 10.18%
9.72%
(±0.14)
N
Hispanic (any race)
34.2%
(*)
41.88%
(±0.46)
D
44.37%
(±0.29)
D 17.84%
21.85%
(±0.24)
D
White Alone
41.4%
(±.1)
34.89%
(±0.33)
D
33.26%
(±0.2)
D 21.6%
18.2%
(±0.17)
D
Black Alone
1.6%
(±.1)
1.62%
(±0.12)
D
1.62%
(±0.07)
C 0.82%
0.84%
(±0.06)
E
All Other Race(s)
3.3%
(±.1)
2.98%
(±0.14)
E
2.89%
(±0.09)
C 1.74%
1.56%
(±0.08)
C
Totals
100%
100%

100%

52.2% 52.2%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

In this scenario, the Hispanic population continued to be overrepresented in cells with
alcohol licenses and underrepresented in cells that did not have licenses. Likewise, the White
population continued to be overrepresented in cells without licenses and underrepresented in
cells with licenses. On the other hand, the Cells with Licenses scenario showed the Asian
population with underrepresentation compared to the census tract scenario Asian population.
Overall, while some race-neutral distribution was observed for the Asian population in this
scenario, the majority of evaluation points surpassed the disparate distribution threshold for this
study.
There were 1,603 cells containing Type 21 licenses representing 57.3% of the OC
population with no Type 21 licenses (Table 45). On the other hand, there were 569 cells
containing 42.7% of the OC population with Type 21 licenses (see Table 46). These tables also
had different county-wide population denominators compared to the census tract versions

114

(cells: 57.3% and 42.7% vs tracts: 23.7% and 76.3%), and also manifested the race/ethnicity
dynamics of overrepresentation of Hispanics in cells with Type 21 licenses compared to the
underrepresentation of Whites found in the census tract versions. However, in these cells, the
overrepresentation of Hispanics increased from 36.34% to 41.26%. Overall, while some race-
neutral distribution was observed for the Asian population in this scenario, the majority of
evaluation points exceeded the disparate distribution threshold for this study.
Table 45 OC Summary Statistics of Cells with Zero Type 21 Licenses
Cells with Zero Type 21 Alcohol Licenses: 1,603 / 57.3% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.96%
(±0.24)
N 11.17%
11.43%
(±0.16)
E
Hispanic (any race)
34.2%
(*)
29.39%
(±0.33)
D 19.58%
16.83%
(±0.21)
D
White Alone
41.4%
(±.1)
45.6%
(±0.31)
D 23.7%
26.11%
(±0.2)
C
Black Alone
1.6%
(±.1)
1.52%
(±0.11)
D 0.9%
0.87%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.52%
(±0.14)
D 1.91%
2.02%
(±0.09)
E
Totals
100% 100%

57.3% 57.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

115

Table 46 OC Summary Statistics of Cells with Type 21 Licenses
Cells with Type 21 Alcohol Licenses: 569 / 42.7% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.9%
(±0.31)
E
18.42%
(±0.22)
N 8.34%
8.08%
(±0.13)
E
Hispanic (any race)
34.2%
(*)
40.63%
(±0.5)
D
41.26%
(±0.38)
D 14.62%
17.37%
(±0.22)
D
White Alone
41.4%
(±.1)
35.76%
(±0.37)
D
35.6%
(±0.28)
D 17.7%
15.29%
(±0.16)
D
Black Alone
1.6%
(±.1)
1.63%
(±0.13)
D
1.64%
(±0.1)
C 0.67%
0.7%
(±0.06)
E
All Other Race(s)
3.3%
(±.1)
3.07%
(±0.16)
E
3.08%
(±0.12)
E 1.42%
1.31%
(±0.07)
E
Totals
100%
100%

100%

42.7% 42.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Next, Type 20 licenses were analyzed; as Table 47 indicates, 65.6% of the OC population
occurred within 1,722 cells, and Table 48 shows that 34.4% of the population occurred within
450 cells. These tables also had different population denominators than their census tract counter
parts (cells: 65.6% and 34.4% vs tracts: 35% and 65%), but followed the same basic trends for
Hispanics and Whites. Hispanics again were overrepresented in cells with licenses and
underrepresented in cells without them, while Whites were the converse of the Hispanic
population. Overall, the majority of evaluation points exceeded the disparate distribution
threshold for this study.

116

Table 47 OC Summary Statistics of Cells with Zero Type 20 Licenses
Cells with Zero Type 20 Alcohol Licenses: 1,722 / 65.6% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.41%
(±0.26)
C 12.8%
13.39%
(±0.17)
N
Hispanic (any race)
34.2%
(*)
28.44%
(±0.35)
D 22.43%
18.66%
(±0.23)
D
White Alone
41.4%
(±.1)
46.01%
(±0.33)
D 27.15%
30.18%
(±0.22)
D
Black Alone
1.6%
(±.1)
1.53%
(±0.11)
D 1.03%
1.%
(±0.07)
E
All Other Race(s)
3.3%
(±.1)
3.6%
(±0.15)
D 2.18%
2.36%
(±0.1)
E
Totals
100% 100%

65.6% 65.6%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 48 OC Summary Statistics of Cells with Type 20 Licenses
Cells with Type 20 Alcohol Licenses: 450 / 34.4% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
17.79%
(±0.33)
C
17.08%
(±0.25)
D 6.71%
6.12%
(±0.11)
C
Hispanic (any race)
34.2%
(*)
45.16%
(±0.58)
D
48.63%
(±0.47)
D 11.77%
15.54%
(±0.2)
D
White Alone
41.4%
(±.1)
32.6%
(±0.39)
D
30.03%
(±0.3)
D 14.25%
11.22%
(±0.14)
D
Black Alone
1.6%
(±.1)
1.65%
(±0.15)
D
1.58%
(±0.11)
D 0.54%
0.57%
(±0.05)
E
All Other Race(s)
3.3%
(±.1)
2.81%
(±0.17)
C
2.67%
(±0.13)
D 1.15%
0.97%
(±0.06)
D
Totals
100%
100%

100%

34.4% 34.4%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

117

The Liquor Store category had the most licenses (N=418) in Orange County. As Table 49
and Table 50 show, Liquor Stores followed the trend of the previous scenarios where Hispanics
were underrepresented in the absence of Liquor Stores and overrepresented in their presence and
the White population the converse. Moreover, the cell version of the statistics showed greater
correlation with the Hispanic population in the presence of Liquor Stores than the census tract
version. Overall, the absence of Liquor Stores did not surpass the disparate distribution
threshold, but was more than race-neutral. On the other hand, the presence of Liquor stores
surpassed the disparate distribution threshold, particularly between Hispanic and White
populations.
Table 49 OC Summary Statistics of Cells with Zero Liquor Stores
Cells with Zero Liquor Store Retailers: 1,835 / 73.4% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.7%
(±0.27)
N 14.31%
14.45%
(±0.18)
E
Hispanic (any race)
34.2%
(*)
31.3%
(±0.38)
C 25.09%
22.97%
(±0.25)
C
White Alone
41.4%
(±.1)
44.01%
(±0.34)
C 30.37%
32.29%
(±0.22)
C
Black Alone
1.6%
(±.1)
1.53%
(±0.12)
D 1.15%
1.12%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.46%
(±0.15)
D 2.44%
2.54%
(±0.1)
E
Totals
100% 100%

73.4% 73.4%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

118

Table 50 OC Summary Statistics of Cells with Liquor Stores
Cells with Liquor Store Retailers: 337 / 26.6% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.99%
(±0.38)
E
17.97%
(±0.33)
C 5.2%
5.06%
(±0.1)
E
Hispanic (any race)
34.2%
(*)
42.16%
(±0.65)
D
43.64%
(±0.59)
D 9.11%
11.23%
(±0.17)
D
White Alone
41.4%
(±.1)
34.19%
(±0.47)
D
33.78%
(±0.41)
D 11.03%
9.11%
(±0.12)
D
Black Alone
1.6%
(±.1)
1.69%
(±0.17)
D
1.69%
(±0.15)
D 0.42%
0.45%
(±0.05)
E
All Other Race(s)
3.3%
(±.1)
2.96%
(±0.2)
E
2.92%
(±0.17)
E 0.89%
0.79%
(±0.05)
E
Totals
100%
100%

100%

26.6% 26.6%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

There were 412 Grocery Stores with alcohol licenses in Orange County, second only to
Liquor Stores with 418. Moreover, Grocery Stores may hold either a Type 20 or Type 21 license,
but the majority (N=300) operated with a Type 21. Table 51 shows the sample size of cells with
zero Grocery Stores was quite large (N=1,856 out of 2172 cells) and produced a very similar
absence of retailer profile where Hispanics are underrepresented and Whites are overrepresented.
Table 52 summarized the cells containing Grocery Stores with alcohol licenses, which closely
tracked the Liquor Store scenario with Hispanics significantly overrepresented and Whites
underrepresented.

119

Table 51 OC Summary Statistics of Cells with Zero Grocery Stores
Cells with Zero Grocery Store Retailers: 1,856 / 74.7% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.06%
(±0.28)
N 14.57%
14.98%
(±0.18)
N
Hispanic (any race)
34.2%
(*)
31.46%
(±0.39)
C 25.53%
23.49%
(±0.25)
C
White Alone
41.4%
(±.1)
43.46%
(±0.35)
C 30.91%
32.44%
(±0.23)
N
Black Alone
1.6%
(±.1)
1.57%
(±0.12)
D 1.17%
1.17%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.45%
(±0.15)
D 2.49%
2.57%
(±0.1)
E
Totals
100% 100%

74.7% 74.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 52 OC Summary Statistics of Cells with Grocery Stores
Cells with Grocery Store Retailers: 316 / 25.3% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
17.87%
(±0.39)
C
17.22%
(±0.33)
C 4.94%
4.53%
(±0.1)
C
Hispanic (any race)
34.2%
(*)
42.26%
(±0.66)
D
46.07%
(±0.58)
D 8.67%
10.71%
(±0.17)
D
White Alone
41.4%
(±.1)
35.33%
(±0.48)
D
32.36%
(±0.39)
D 10.49%
8.95%
(±0.12)
D
Black Alone
1.6%
(±.1)
1.57%
(±0.17)
D
1.52%
(±0.14)
C 0.4%
0.4%
(±0.04)
E
All Other Race(s)
3.3%
(±.1)
2.98%
(±0.21)
E
2.83%
(±0.18)
C 0.84%
0.75%
(±0.05)
E
Totals
100%
100%

100%

25.3% 25.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

120

Overall, the cells with licenses surpassed the disparate distribution threshold. On the
other hand, the zero-retailer scenario manifested a distribution that was slightly more than race
neutral. Although this was to be expected because as the number of cells with zero
licenses/retailers increases towards the county-wide total, the population dynamics will approach
the county-wide profile.
Convenience Stores (N=347) rank third by number of licenses in the list of retailer
categories. Moreover, as Table 53 indicates, the number cells without Convenience Stores
(N=1,888) was greater than scenarios with Liquor Stores (N=1,835) and Grocery Stores
(N=1,856). However, even though the sample size continued to approach the county-wide value
(N=2,172), this scenario surpassed the disparate distribution threshold. This suggested a race-
biased function at least in part operated for the absence of convenience stores in the built
environment between Hispanic and White populations. One potential factor related to this
distribution could be that the majority of Convenience Stores had Type 20 licenses (N=296)
while the rest had Type 21 (N=51), unlike Liquor Stores and Grocery Stores which primarily had
Type 21 licenses.

121

Table 53 OC Summary Statistics of Cells with Zero Convenience Stores
Cells with Zero Convenience Store Retailers: 1,888 / 76.9% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.04%
(±0.28)
N 15.01%
15.42%
(±0.19)
N
Hispanic (any race)
34.2%
(*)
29.86%
(±0.39)
D 26.32%
22.97%
(±0.25)
D
White Alone
41.4%
(±.1)
45.03%
(±0.36)
C 31.86%
34.65%
(±0.23)
C
Black Alone
1.6%
(±.1)
1.54%
(±0.12)
D 1.21%
1.18%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.53%
(±0.16)
D 2.56%
2.72%
(±0.1)
E
Totals
100% 100%

77.0% 76.9%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 54 covers the cells with Convenience Stores scenario. Here again, both
overrepresentation of Hispanic populations and underrepresentation of White and Asian
populations were observed. Moreover, seven out of ten evaluation points exceeded the disparate
distribution threshold.

122

Table 54 OC Summary Statistics of Cells with Convenience Stores
Cells with Convenience Store Retailers: 284 / 23.1% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
17.73%
(±0.39)
C
17.11%
(±0.34)
D 4.5%
4.09%
(±0.09)
C
Hispanic (any race)
34.2%
(*)
48.68%
(±0.72)
D
50.08%
(±0.65)
D 7.88%
11.22%
(±0.17)
D
White Alone
41.4%
(±.1)
29.26%
(±0.46)
D
28.6%
(±0.41)
D 9.54%
6.75%
(±0.11)
D
Black Alone
1.6%
(±.1)
1.68%
(±0.18)
D
1.64%
(±0.16)
D 0.36%
0.39%
(±0.04)
E
All Other Race(s)
3.3%
(±.1)
2.65%
(±0.2)
D
2.58%
(±0.18)
D 0.77%
0.61%
(±0.05)
D
Totals
100%
100%

100%

23.1% 23.1%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Gas Stations (N=270) followed Convenience Stores in the list of retailer categories by
license count. Like Convenience Stores, Gas Stations primarily held Type 20 licenses (N=255),
but there were a small number with Type 21 licenses (N=15). Table 55 confirmed that as the
number of cells with absence of retailers (Gas Stations) approaches the full county-wide cell
count, bias attenuates into the margin of errors and becomes race neutral.

123

Table 55 OC Summary Statistics of Cells with Zero Gas Stations
Cells with Zero Gas Station Retailers: 1,954 / 84.7% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.52%
(±0.29)
N 16.53%
16.53%
(±0.19)
E
Hispanic (any race)
34.2%
(*)
33.17%
(±0.42)
N 28.97%
28.1%
(±0.27)
N
White Alone
41.4%
(±.1)
42.41%
(±0.36)
N 35.07%
35.93%
(±0.24)
N
Black Alone
1.6%
(±.1)
1.54%
(±0.13)
D 1.33%
1.3%
(±0.08)
E
All Other Race(s)
3.3%
(±.1)
3.37%
(±0.16)
C 2.82%
2.85%
(±0.11)
E
Totals
100% 100%

84.7% 84.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 56, on the other hand, manifested a nearly symmetrical 5% difference of
over/under representation between Hispanics and White Populations. Moreover, the other
minority populations are nearly all within the margins of error for the expected and observed
percentages. Overall, the disparate distribution threshold was exceeded between Hispanics and
Whites, but the Asian population impact was effectively race neutral.

124

Table 56 OC Summary Statistics of Cells with Gas Stations
Cells with Gas Stations Retailers: 218 / 15.3% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.47%
(±0.53)
N
19.7%
(±0.48)
E 2.98%
2.98%
(±0.08)
E
Hispanic (any race)
34.2%
(*)
39.89%
(±0.86)
D
39.69%
(±0.78)
D 5.23%
6.1%
(±0.13)
D
White Alone
41.4%
(±.1)
35.77%
(±0.62)
D
35.69%
(±0.56)
D 6.33%
5.47%
(±0.09)
D
Black Alone
1.6%
(±.1)
1.75%
(±0.22)
E
1.74%
(±0.2)
E 0.24%
0.27%
(±0.03)
E
All Other Race(s)
3.3%
(±.1)
3.12%
(±0.29)
E
3.17%
(±0.25)
E 0.51%
0.48%
(±0.04)
E
Totals
100%
100%

100%

15.3% 15.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Orange County had 159 Pharmacies with either a Type 21 (N=135) or Type 20 (N=24)
alcohol license. As Table 57 indicates, the absence of retailer sample size (N=2,031) for this
category was even closer to the county-wide cell count (N=2,172) than previous categories. As
expected, the population dynamics with such a large sample was approaching the county-wide
percentages. As such, the absence of pharmacies appeared to be a race neutral function.

125

Table 57 OC Summary Statistics of Cells with Zero Pharmacies
Cells with Zero Pharmacy Retailers: 2,031 / 89.2% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.68%
(±0.3)
N 17.4%
17.54%
(±0.2)
E
Hispanic (any race)
34.2%
(*)
34.06%
(±0.44)
N 30.49%
30.37%
(±0.29)
E
White Alone
41.4%
(±.1)
41.37%
(±0.37)
N 36.91%
36.88%
(±0.24)
E
Black Alone
1.6%
(±.1)
1.58%
(±0.13)
D 1.4%
1.41%
(±0.09)
E
All Other Race(s)
3.3%
(±.1)
3.32%
(±0.16)
C 2.97%
2.96%
(±0.11)
E
Totals
100% 100%

89.2% 89.2%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 58, on the other hand, presented a new population distribution dynamic. Although
there still appeared to be an overrepresentation bias with the Hispanic population, the White
population was effectively race neutral, being neither over nor under represented. This was a
departure from the census tract version where the Hispanic population manifested an
underrepresentation and the White population indicated a slight overrepresentation. These
differences were most likely related to MAUP issues in the census tract analysis. Overall, this
scenario narrowly manifested some absence of race neutral distribution, but only where multiple
pharmacies in a cell were a factor.

126

Table 58 OC Summary Statistics of Cells with Pharmacies
Cells with Pharmacy Retailers: 141 / 10.8% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
18.14%
(±0.59)
N
16.96%
(±0.54)
C 2.11%
1.97%
(±0.06)
N
Hispanic (any race)
34.2%
(*)
35.3%
(±0.95)
E
37.27%
(±0.9)
C 3.71%
3.83%
(±0.1)
E
White Alone
41.4%
(±.1)
41.64%
(±0.74)
E
41.04%
(±0.69)
E 4.49%
4.51%
(±0.08)
E
Black Alone
1.6%
(±.1)
1.49%
(±0.26)
E
1.43%
(±0.24)
E 0.17%
0.16%
(±0.03)
E
All Other Race(s)
3.3%
(±.1)
3.43%
(±0.35)
E
3.3%
(±0.32)
D 0.36%
0.37%
(±0.04)
E
Totals
100%
100%

100%

10.8% 10.8%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

There were fifty alcohol retailers identified as Department Stores (i.e. Targets and
K-Marts) in Orange County. Table 59 reinforced the previous observations that as the absence of
retailers in the built environment increases, the absence distribution approaches a race neutral
function.

127

Table 59 OC Summary Statistics of Cells with Zero Department Stores
Cells with Zero Dept Store Retailers: 2,122 / 96.3% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.46%
(±0.31)
N 18.79%
18.74%
(±0.2)
E
Hispanic (any race)
34.2%
(*)
33.96%
(±0.45)
N 32.93%
32.7%
(±0.3)
E
White Alone
41.4%
(±.1)
41.68%
(±0.39)
N 39.87%
40.13%
(±0.25)
E
Black Alone
1.6%
(±.1)
1.57%
(±0.14)
D 1.51%
1.51%
(±0.09)
E
All Other Race(s)
3.3%
(±.1)
3.33%
(±0.17)
C 3.21%
3.21%
(±0.11)
E
Totals
100% 100%

96.3% 96.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

However, unlike previous retailer categories, there are no instances where two or more
Department Stores occur in one cell in Orange County. This is shown in Table 60 where the
Pop % Cells and (Pop x L) % Cells columns have the same values. While this scenario exhibited
overrepresentation of Hispanic populations and neutral representation in the presence of
Department Stores, the census tract summary statistics indicated un underrepresentation of
Hispanic populations and overrepresentation of Asian populations. As with the discrepancies
between cell and census tracts Pharmacy statistics, these differences were likely due to MAUP-
related census tract boundary issues. Overall, this scenario exceeded the disparate distribution
threshold for Hispanics and Whites.

128

Table 60 OC Summary Statistics of Cells with Department Stores
Cells with Dept Store Retailers: 50 / 3.7% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
20.77%
(±1.12)
E
20.77%
(±1.12)
E 0.72%
0.77%
(±0.04)
E
Hispanic (any race)
34.2%
(*)
40.28%
(±1.7)
D
40.28%
(±1.7)
D 1.27%
1.49%
(±0.06)
D
White Alone
41.4%
(±.1)
34.1%
(±1.25)
D
34.1%
(±1.25)
D 1.53%
1.26%
(±0.05)
D
Black Alone
1.6%
(±.1)
1.63%
(±0.44)
D
1.63%
(±0.44)
D 0.06%
0.06%
(±0.02)
E
All Other Race(s)
3.3%
(±.1)
3.23%
(±0.56)
D
3.23%
(±0.56)
D 0.12%
0.12%
(±0.02)
E
Totals
100%
100%

100%

3.7% 3.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Wholesale Clubs was the final category for analysis. As Figure 38 shows, there were only
sixteen Wholesale Clubs with alcohol licenses in OC. As a result, the population sample size for
the absence of Wholesale clubs was 2,156 cells with 99.3% of the county population (see Table
61), while the population size for the presence was only 0.7% (see Table 62). As expected with
the sample size only 16 cells short of the county-wide cell count, the absence of Wholesale Clubs
was effectively the county-wide populations dynamics within the margins of error and race-
neutral. On the other hand, the summary statistics for the 16 retailers was swamped by the
margins of error and was excluded as unreliable.

129

Table 61 OC Summary Statistics of Cells with Zero Wholesale Clubs
Cells with Zero Wholesale Club Retailers: 2,156 / 99.3% of OC Population

Pop %
County
Pop %
Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
19.46%
(±0.31)
N 19.37%
19.32%
(±0.21)
E
Hispanic (any race)
34.2%
(*)
34.21%
(±0.46)
N 33.95%
33.96%
(±0.3)
E
White Alone
41.4%
(±.1)
41.43%
(±0.39)
N 41.1%
41.13%
(±0.26)
E
Black Alone
1.6%
(±.1)
1.57%
(±0.14)
D 1.56%
1.56%
(±0.09)
E
All Other Race(s)
3.3%
(±.1)
3.33%
(±0.17)
C 3.31%
3.31%
(±0.11)
E
Totals
100% 100%

99.3% 99.3%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Figure 38 OC Cells with Wholesale Clubs

130

Table 62 OC Summary Statistics of Cells with Wholesale Clubs
Cells with Wholesale Club Retailers: 16 / 0.7% of OC Population

Pop %
County
Pop %
Cells
Dist
(Pop x L)
% Cells
Dist
%
Expected
%
Actual
Dist
Asian Alone
19.5%
(±.1)
25.56%
(±3.11)
E
25.56%
(±3.11)
E 0.14%
0.19%
(±0.02)
E
Hispanic (any race)
34.2%
(*)
32.54%
(±4.13)
E
32.54%
(±4.13)
E 0.25%
0.24%
(±0.03)
E
White Alone
41.4%
(±.1)
36.78%
(±3.32)
E
36.78%
(±3.32)
E 0.3%
0.27%
(±0.02)
E
Black Alone
1.6%
(±.1)
1.98%
(±1.3)
E
1.98%
(±1.3)
E 0.01%
0.01%
(±0.01)
E
All Other Race(s)
3.3%
(±.1)
3.14%
(±1.1)
E
3.14%
(±1.1)
E 0.02%
0.02%
(±0.01)
E
Totals
100%
100%

100%

0.7% 0.7%

*Estimate is controlled, margin of error treated as zero.

C: Race Correlated | D: Disparate Distribution | E: Margin of Error | N: Race Neutral | X: Exclude

Table 63 summarizes the (Pop x L) % Cells column values and overall distribution
assessments of the license/retailers in scaled population grid cells scenarios. As the table
indicates, the Hispanic population had the most overrepresentation scenarios. Likewise, the
majority of scenarios exceeded the threshold for disparate distribution.

131

Table 63 OC Cells with Alcohol Licenses Population Summary
Asian Hispanic White Black Other Dist
Orange County
19.5%
(±0.1)
34.2%
*
41.4%
(±0.1)
1.6%
(±0.1)
3.3%
(±0.1)
N/A
All Licenses
17.86%
(±0.17)
44.37%
(±0.29)
33.26%
(±0.2)
1.62%
(±0.07)
2.89%
(±0.09)
D
Type 21
18.42%
(±0.22)
41.26%
(±0.38)
35.6%
(±0.28)
1.64%
(±0.1)
3.08%
(±0.12)
D
Type 20
17.08%
(±0.25)
48.63%
(±0.47)
30.03%
(±0.3)
1.58%
(±0.11)
2.67%
(±0.13)
D
Liquor Stores
17.97%
(±0.33)
43.64%
(±0.59)
33.78%
(±0.41)
1.69%
(±0.15)
2.92%
(±0.17)
D
Grocery Stores
17.22%
(±0.33)
46.07%
(±0.58)
32.36%
(±0.39)
1.52%
(±0.14)
2.83%
(±0.18)
D
Convenience Stores
17.11%
(±0.34)
50.08%
(±0.65)
28.6%
(±0.41)
1.64%
(±0.16)
2.58%
(±0.18)
D
Gas Stations
19.7%
(±0.48)
39.69%
(±0.78)
35.69%
(±0.56)
1.74%
(±0.2)
3.17%
(±0.25)
D
Pharmacies
16.96%
(±0.54)
37.27%
(±0.9)
41.04%
(±0.69)
1.43%
(±0.24)
3.3%
(±0.32)
C
Department Stores
20.77%
(±1.12)
40.28%
(±1.7)
34.1%
(±1.25)
1.63%
(±0.44)
3.23%
(±0.56)
D
Wholesale Clubs
25.56%
(±3.11)
32.54%
(±4.13)
36.78%
(±3.32)
1.98%
(±1.3)
3.14%
(±1.1)
X

*Estimate is controlled, margin of error treated as zero

C: Race Correlated | D: Disparate Distribution | N/A: Not Applicable | X: Exclude

Table 64 summarizes the Pop % Cells values and overall distribution assessments of the
zero license/retailers in scaled population grid scenarios. As the table indicates, the Hispanic
population manifested reduced numbers in the census tracts without licenses, while the White
population was overrepresented in most census tract without licenses. The distribution patterns

132

of the absence of alcohol licenses at the cell level also provided two interesting patterns. First,
somewhere between a sample size of 1,888 cells (Convenience Stores) and 1,945 cells (Gas
Stations), the absence of license/retailer population dynamics started to closely track the county-
wide dynamics. Second, sample sizes 1,888 and lower showed a consistent overrepresentation
bias for White populations and underrepresentation of Hispanic populations. The Asian, Black,
and Other populations across all samples tracked closely to their county-wide dynamics, with
values either slightly under or over the margins of error.
Table 64 OC Cells with Zero Licenses Population Summary
Asian Hispanic White Black Other Dist
Orange County
19.5%
(±0.1)
34.2%
*
41.4%
(±0.1)
1.6%
(±0.1)
3.3%
(±0.1)
N/A
All Licenses
20.47%
(±0.31)
25.81%
(±0.39)
48.49%
(±0.4)
1.52%
(±0.14)
3.7%
(±0.18)
D
Type 21
19.96%
(±0.24)
29.39%
(±0.33)
45.6%
(±0.31)
1.52%
(±0.11)
3.52%
(±0.14)
D
Type 20
20.41%
(±0.26)
28.44%
(±0.35)
46.01%
(±0.33)
1.53%
(±0.11)
3.6%
(±0.15)
D
Liquor Stores
19.7%
(±0.27)
31.3%
(±0.38)
44.01%
(±0.34)
1.53%
(±0.12)
3.46%
(±0.15)
C
Grocery Stores
20.06%
(±0.28)
31.46%
(±0.39)
43.46%
(±0.35)
1.57%
(±0.12)
3.45%
(±0.15)
C
Convenience Stores
20.04%
(±0.28)
29.86%
(±0.39)
45.03%
(±0.36)
1.54%
(±0.12)
3.53%
(±0.16)
D
Gas Stations
19.52%
(±0.29)
33.17%
(±0.42)
42.41%
(±0.36)
1.54%
(±0.13)
3.37%
(±0.16)
N
Pharmacies
19.68%
(±0.3)
34.06%
(±0.44)
41.37%
(±0.37)
1.58%
(±0.13)
3.32%
(±0.16)
N
Department Stores
19.46%
(±0.31)
33.96%
(±0.45)
41.68%
(±0.39)
1.57%
(±0.14)
3.33%
(±0.17)
N
Wholesale Clubs
19.46%
(±0.31)
34.21%
(±0.46)
41.43%
(±0.39)
1.57%
(±0.14)
3.33%
(±0.17)
X

*Estimate is controlled, margin of error treated as zero

C: Race Correlated | D: Disparate Distribution | N: Race Neutral | N/A: Not Applicable | X: Exclude

133

Overall, these results suggested a compelling argument for two race/ethnicity and alcohol
licenses correlations. First, the Hispanic population is overrepresented in areas within ~0.5 miles
(the approximate length of a cell side) of most alcohol retailers. Second, as the number of
licenses/retailers in the built environment increases, the licenses are less likely to occur in White
dominant cells. Whether similar correlations exist with other race/ethnicities is not clear. While
there were some observations inconsistent with the county-wide population dynamics, they
occurred with small sample sizes or were within the margins of error to be more than anecdotal
observations.
4.2.2. Scaled Population Grid Alcohol License Density
The next step in analyzing the distribution of alcohol licenses at the cell level was to
determine the density of alcohol licenses per cell. Figure 39 shows the license density per cell for
all licenses in Orange County as a single group, the Type 21 licenses, the Type 20 licenses, and
for licenses by each category of retailer (except Wholesale Clubs). First, scatter plots were
generated for each race/ethnicity cell percentage versus each cell’s licenses and retailers counts.
Next, linear regressions were performed with the licenses/retailers per cell as the dependent
variable for each scatter plot and the resulting trend lines were coded red if the race/ethnicity
indicated a positive correlation with increasing population percentage and green if the
race/ethnicity indicated a negative correlation with increasing population percentage. Finally, the
results were rejected if the p-value was above 0.05. Because each cell has a whole number of
licenses with a very small range of potential values, the individual regressions become less
meaningful as their slope approaches zero.

134

For the regressions with p-values less than 0.05, if the slope polarities between two or
more populations were inconsistent, then the likelihood of a race-neutral distribution function
was rejected and assumed to be a disparate distribution. To be clear, only the differences in the
slope polarities between the populations were assessed. This methodology allowed for a quick
visual inspection to determine the potential existence of race-neutral versus race-correlated

Figure 39 OC Alcohol Licenses per Cell

135

distributions in the distributions per cell of the licenses/retailers. Out of ten regression scenarios,
two (Department Stores and Wholesale Clubs) produced slopes with a value of zero (effectively
a null condition since the slopes did not have a polarity) and two (Type 21 licenses and Gas
Stations) were completely rejected because two or more results had p-values greater than 0.05.
Figure 40 shows that the trend line polarities between the White and Hispanic
populations for All Licenses were reversed and their associated p-values were less than 0.05,
while the Asian regression was rejected because the p-value was greater than 0.05.
Notwithstanding the rejected Asian regression, the White and Hispanic results met this study’s
criteria for a race-correlated distribution.

Figure 41 shows the regression results for the Type 21 licenses per cell scenario. Here,
the high p-values for White and Asian regressions rendered the results inconclusive for
comparisons and the entire scenario was rejected. Figure 42 depicts the Type 20 regressions
which had two valid results with opposite slopes. Therefore, that scenario met the criteria for
race-correlated distribution.

Figure 40 OC Linear Regressions on All Licenses per Cell

136

While the expectation was to compare the minority regression slope polarities to the
White slope polarity, the Liquor Store regressions presented a scenario where the White
regression was rejected (see Figure 43). Moreover, the Liquor Store category was composed
entirely of Type 21 licenses, where the White and Asian regressions were rejected. This result
supported the merit of analyzing both license type and retailer type and that race-correlated
distributions can occur between minority populations as well as the majority population.

Figure 41 OC Linear Regressions on Type 21 Licenses per Cell

Figure 42 OC Linear Regressions on Type 20 Licenses per Cell

137

For the remaining regression scenarios—Grocery Stores (see Figure 44), Convenience
Stores (see Figure 45), and Pharmacies (see Figure 46)—the Hispanic regressions produced
results with a p-values below 0.05. While the Asian regression was rejected for Grocery Stores
and the White regressions were rejected for Convenience Stores and Pharmacies the non-rejected
slopes had opposite polarities to the Hispanic slopes. Thus, these three categories met the criteria
for race-correlated distribution.

Figure 43 OC Linear Regressions on Liquor Stores per Cell

Figure 44 OC Linear Regressions on Grocery Stores per Cell

138

Table 65 summarizes the results for the percent population versus licenses per cell
regressions. For all the non-rejected results, the trend line slope polarities for the Hispanic
population were always opposite to the White and Asian populations. These results met the
study’s threshold for the presence of race-correlated distribution.

Figure 45 OC Linear Regressions on Convenience Stores per Cell

Figure 46 OC Linear Regressions on Pharmacies per Cell

139

Table 65 Licenses per Cell Linear Regressions Trend Line Polarity Summary
Asian Hispanic White Black Other Dist
All X
+ ─
X
─
C
Type 21 X
+
X X X X
Type 20 X
+ ─
X
─
C
Liquor Stores
─ +
X X X C
Grocery Stores X
+ ─
X
─
C
Convenience
Stores
─ +
X X X C
Gas Stations X X X X X X
Pharmacies
─ +
X X
─
C
Department
Stores
X X X X X X
Wholesale Clubs X X X X X X
─ Negative Slope +Positive Slope X: Excluded C: Race Correlated

Another useful metric is the distribution of alcohol licenses based on their density per
1,000 population. Figure 47 provides distribution maps of the metric for All Licenses, Type 21,
Type 20, and all the retailers except Wholesale Clubs. However, there were 100 cells with a total
population of less than 1,000 that also contained one or more licenses; as a result, some of those
cells appeared in a higher license band than the number of licenses in the cell. This was
particularly evident for Department Stores which only had one retailer per cell, but the cells with
populations under 1,000 appeared to have more (see Figure 47 bottom right).

140

Scatter plots were generated for each race/ethnicity cell percentage versus each license
type and for each retailer category per 1,000 population in each cell. Linear regressions were
performed with the licenses/retailers per 1,000 population as the dependent variable and
race/ethnicity percent as the independent variable for each scatter plot. The resulting trend lines
were coded red if the race/ethnicity indicated a positive correlation with increasing population

Figure 47 OC Alcohol Licenses per 1,000 Population per Cell

141

percentage and green if the race/ethnicity indicated a negative correlation with increasing
population percentage. Finally, the results were rejected if the p-value was greater than 0.05. For
example, the results for Department Stores and Wholesale Clubs were rejected as all
race/ethnicities had p-values greater than 0.05. For the regressions with p-values less than 0.05, if
the slope polarities between two or more populations were inconsistent, then the likelihood of a
race-neutral distribution function was rejected and assumed to be a race-correlated distribution.
Starting with All Licenses, all the regressions had p-values below 0.05 (see Figure 48).
Moreover, the White trend line had the opposite slope polarity to both the Asian and Hispanic
trend lines. Likewise, the Type 21 regression p-values were all below 0.05 and the White trend
line had the opposite slope polarity to the Asian and Hispanic lines (see Figure 49) Finally, for
the Type 20 regressions the Asian result was rejected, but the White and Hispanic trend lines had
opposite polarities. All the scenarios satisfied the criteria for race-correlated distributions since
they all had one or more races with opposite slope polarities.

Figure 48 OC Linear Regressions on All Licenses per 1,000 People per Cell

142

The regressions for the retailer categories followed similar patterns, with all satisfying the
criteria for race-correlated distributions. For Liquor Stores, both Asian and Hispanic trend lines
were the opposite polarity of White (see Figure 51). The Asian regression result was rejected for
Grocery Stores, but the Hispanic and White trend lines had opposite polarities (see Figure 52).
The same polarity occurred for Asian and Hispanic regressions with Convenience Stores, but
both were opposite to White (Figure 53). The Asian results were rejected for both Gas Stations
and Pharmacies, but Hispanic and White regressions had opposite polarities for both (see Figure
54 and Figure 55).

Figure 49 OC Linear Regressions on Type 21 Licenses per 1,000 People per Cell

Figure 50 OC Linear Regressions on Type 20 Licenses per 1,000 People per Cell

143

Figure 51 OC Linear Regressions on Liquor Stores per 1,000 People per Cell

Figure 52 OC Linear Regressions on Grocery Stores per 1,000 People per Cell

Figure 53 OC Linear Regressions on Convenience Stores per 1,000 People per Cell

144

Table 66 summarizes the results for the percent population versus licenses per 1,000
people regressions. Overall, there were only two scenarios where the results were rejected:
Department Stores and Wholesale Clubs. The non-rejected results for the other scenarios, on the
other hand, satisfied the criteria for race-correlated distributions.

Figure 54 OC Linear Regressions on Gas Stations per 1,000 People per Cell

Figure 55 OC Linear Regressions on Pharmacies per 1,000 People per Cell

145

Table 66 Cell Linear Regressions per 1,000 People Trend Line Summary
Asian Hispanic White Black Other Dist
All
─ ─ +
X
+
C
Type 21
─ ─ +
X
+
C
Type 20 X
─ +
X X C
Liquor Stores
─ ─ +
X
+
C
Grocery Stores X
─ +
X X C
Convenience
Stores
─ ─ +
X X C
Gas Stations X
─ +
X X C
Pharmacies X
─ +
X X C
Department
Stores
X X X X X X
Wholesale Clubs X X X X X X
─ Negative Slope +Positive Slope X: Excluded C: Race Correlated

146

4.2.3. Scaled Population Alcohol License Hot Spots: Getis-Ord Gi* Statistic
A Getis-Ord Gi* statistic—specifically the ArcGIS Pro Optimized Hot Spot tool—was
run for All Licenses to determine the presence of statistically significant clustering at the cell
level. The tool output was reviewed to identify the neighborhood distance (~1.8 miles), and then
the tool was run again for the Type 21 licenses, Type 20 license, and for each retailer category
(see Figure 56). The distance band was set to ~1.8 mile for each run.

Figure 56 OC Optimized Alcohol License Hot Spots Based Upon Cell Boundaries

147

As depicted in Figure 56, the Optimized Hot Spot analysis detected areas of hot spot
clustering in the northern portion of the county and some cold spot clustering in the southern
portion for most license and retailer combinations. However, the Pharmacy category showed no
clustering, while the Department Store category showed negligible hot spot clustering.
The next step was to perform race/ethnicity summary statistics on the confidence level
bins for All Licenses, Type21, Type 20, Liquor Stores, Grocery Stores, Convenience Stores, and
Gas Stations. The remaining retailer categories were not evaluated due to insufficient sample
sizes. Table 67 has the summary statistics for All Licenses. The Hispanic Population was
consistently overrepresented in the Hot Spot bins and underrepresented in the Cold Spot bins
while the converse was true for the White population. The Asian population, however, was
overrepresented only in the 95% bin of the Hot Spots and underrepresented in all other bins.
While the results in the Cold 99% bin were unreliable due to the sample size (N=3), the other
bins had sufficient samples to evaluate the difference between observed and county-wide values.
Overall, the majority of values exceeded the disparate distribution threshold of 10% difference
from the county-wide proportions.

148

Table 67 OC Cell Based All Licenses Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
340
20.91%
(±0.11)
C
55.0%
(±0.2)
D
20.45%
(±0.11)
D
1.43%
(±0.04)
C
2.22%
(±0.05)
D
Hot 95%
Confidence
93
16.0%
(±0.05)
D
42.64%
(±0.09)
D
37.11%
(±0.06)
D
1.47%
(±0.02)
E
2.77%
(±0.02)
D
Hot 90%
Confidence
59
17.13%
(±0.04)
D
36.76%
(±0.06)
C
40.96%
(±0.04)
N
1.81%
(±0.02)
D
3.35%
(±0.02)
E
Cold 90%
Confidence
177
18.45%
(±0.04)
C
12.82%
(±0.03)
D
63.11%
(±0.05)
D
1.35%
(±0.01)
D
4.27%
(±0.02)
D
Cold 95%
Confidence
158
16.35%
(±0.03)
D
10.8%
(±0.03)
D
67.52%
(±0.06)
D
0.82%
(±0.01)
D
4.51%
(±0.02)
D
Cold 99%
Confidence
3
14.44%
(±0.0)
X
13.37%
(±0.0)
X
67.77%
(±0.01)
X
0.31%
(±0.0)
X
4.11%
(±0.0)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

The summary statistics of the Type 21 licenses, while varying in degree, followed the
same pattern as All Licenses (see Table 68). Moreover, the Type 20 summary statistics indicated
that the population living in the 235 Hot 99% Confidence cells was almost 65% Hispanic, nearly
twice their representation across OC (Table 69). This was a significant degree of
overrepresentation considering that there were only 711 cells with alcohol licenses. Overall, the
majority of values for the Type 21 and Type 20 licenses surpassed the disparate distribution
threshold of 10% difference from the county-wide proportions.

149

Table 68 OC Cell Based Type 21 Licenses Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
235
23.94%
(±0.1)
D
47.79%
(±0.16)
D
24.15%
(±0.1)
D
1.53%
(±0.04)
E
2.59%
(±0.04)
D
Hot 95%
Confidence
105
15.86%
(±0.05)
D
54.38%
(±0.1)
D
26.26%
(±0.06)
D
1.19%
(±0.02)
D
2.31%
(±0.03)
D
Hot 90%
Confidence
51
16.41%
(±0.03)
D
46.79%
(±0.05)
D
32.98%
(±0.04)
D
1.34%
(±0.01)
D
2.48%
(±0.01)
D
Cold 90%
Confidence
90
16.49%
(±0.02)
D
9.95%
(±0.02)
D
68.36%
(±0.04)
D
0.88%
(±0.01)
D
4.32%
(±0.01)
D
Cold 95%
Confidence
21
15.8%
(±0.01)
D
11.63%
(±0.01)
D
67.39%
(±0.02)
D
0.78%
(±0.0)
D
4.4%
(±0.01)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Table 69 OC Cell Based Type 20 Licenses Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
235
18.04%
(±0.09)
C
63.59%
(±0.18)
D
15.12%
(±0.08)
D
1.47%
(±0.04)
E
1.78%
(±0.04)
D
Hot 95%
Confidence
124
21.44%
(±0.07)
C
39.71%
(±0.11)
D
34.06%
(±0.07)
D
1.82%
(±0.03)
D
2.97%
(±0.03)
D
Hot 90%
Confidence
75
17.53%
(±0.04)
D
37.06%
(±0.07)
C
40.39%
(±0.05)
N
1.57%
(±0.02)
E
3.45%
(±0.02)
E
Cold 90%
Confidence
133
20.2%
(±0.04)
N
11.23%
(±0.03)
D
62.92%
(±0.06)
D
0.89%
(±0.01)
D
4.76%
(±0.03)
D
Cold 95%
Confidence
3
26.0%
(±0.01)
X
14.66%
(±0.01)
X
55.02%
(±0.01)
X
0.84%
(±0.0)
X
3.48%
(±0.0)
X

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

150

Turning to the retailer categories, the Liquor Store category had the greatest number of
retailers (N=418), all of which had Type 21 licenses. Table 70 shows that the Liquor Store
summary statistics generally followed the results of the Type 21 licenses, although there was
only a single cold bin. These statistics indicated that the heaviest clustering of Liquor Stores
occurred in Hispanic dominant cells.
Table 70 OC Cell Based Liquor Stores Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
281
24.2%
(±0.1)
D
46.39%
(±0.17)
D
25.35%
(±0.1)
D
1.47%
(±0.04)
E
2.58%
(±0.05)
D
Hot 95%
Confidence
120
15.47%
(±0.05)
D
50.14%
(±0.11)
D
30.16%
(±0.06)
D
1.56%
(±0.02)
E
2.66%
(±0.03)
D
Hot 90%
Confidence
65
14.15%
(±0.03)
D
57.27%
(±0.08)
D
25.34%
(±0.04)
D
1.07%
(±0.01)
D
2.18%
(±0.02)
D
Cold 90%
Confidence
136
27.39%
(±0.06)
D
11.16%
(±0.05)
D
55.54%
(±0.07)
D
1.53%
(±0.03)
E
4.37%
(±0.03)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Grocery Stores, Table 71, had second highest retailer count (N=412), and those retailers
held either a Type 20 (N=112) or a Type 21 (N=300) license. The Optimized Hot Spot analysis
also indicated less clustering overall for Grocery Stores compared to Liquor Stores. But like
Liquor Stores, the summary statistics showed the greatest clustering in Hispanic dominant cells.

151

Table 71 OC Cell Based Grocery Stores Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
66
9.14%
(±0.04)
D
81.42%
(±0.11)
D
7.51%
(±0.03)
D
0.9%
(±0.02)
D
1.02%
(±0.02)
D
Hot 95%
Confidence
46
21.39%
(±0.04)
C
59.54%
(±0.07)
D
15.63%
(±0.03)
D
1.38%
(±0.01)
D
2.07%
(±0.02)
D
Hot 90%
Confidence
30
34.61%
(±0.04)
D
37.22%
(±0.06)
C
24.42%
(±0.03)
D
1.25%
(±0.01)
D
2.5%
(±0.01)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Moving to the two remaining two categories, Convenience Stores (N=347) had 296 Type
20 licenses, while Gas Stations (N=270) had 255. Table 72 has the summary statistics for
Convenience Stores and Table 73 the statistics for Gas Stations. Both these categories also
showed that Hispanic population was overrepresented in all bins while the White population was
underrepresented.
Table 72 OC Cell Based Convenience Stores Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
250
17.49%
(±0.09)
D
61.23%
(±0.19)
D
17.91%
(±0.09)
D
1.4%
(±0.04)
D
1.97%
(±0.04)
D
Hot 95%
Confidence
48
23.32%
(±0.03)
D
44.34%
(±0.06)
D
27.61%
(±0.03)
D
1.68%
(±0.01)
E
3.05%
(±0.02)
C
Hot 90%
Confidence
56
25.69%
(±0.04)
D
39.59%
(±0.07)
D
30.21%
(±0.04)
D
1.69%
(±0.02)
C
2.81%
(±0.02)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

152

Table 73 OC Cell Based Gas Stations Optimized Hot Spot Summary Statistics
Cells Asian Dist Hispanic Dist White Dist Black Dist Other Dist
County
Statistics
2172
19.5%
(±0.1)
N/A
34.2%
*
N/A
41.4%
(±0.1)
N/A
1.6%
(±0.1)
N/A
3.3%
(±0.1)
N/A
Hot 99%
Confidence
54
20.5%
(±0.04)
C
48.63%
(±0.06)
D
26.89%
(±0.04)
D
1.65%
(±0.01)
E
2.33%
(±0.01)
D
Hot 95%
Confidence
106
20.42%
(±0.06)
N
45.17%
(±0.1)
D
29.96%
(±0.06)
D
1.85%
(±0.02)
D
2.6%
(±0.03)
D
Hot 90%
Confidence
78
27.93%
(±0.05)
D
37.42%
(±0.07)
A
30.78%
(±0.05)
D
1.4%
(±0.02)
D
2.47%
(±0.02)
D

*Estimate is controlled, margin of error treated as zero

C: Correlated | D: Disparate | E: Margin of Error | N: Race Neutral | N/A: Not Applicable | X: Exclude

Overall, the Cell Based Optimized Hot Spot scenarios indicated high clustering in
Hispanic dominant cells and White population underrepresentation. The majority of Dist column
values also surpassed the disparate distribution threshold.
In order to observe the potential occurrence of more macro-scale clustering, the Hot Spot
Analysis was performed using a 3-mile distance band as a proxy for a reasonable distance an OC
resident would drive to a convenient retailer (see Figure 57). Although this was less than double
the optimized distance band value used above, and the same distance band value used to create
the census tract optimized hot spot analyses, the observational results were rejected as too hyper-
clustered for useful analysis. Moreover, the Pharmacy and Department Store categories still
indicated insignificant clustering at this distance band value.

153

While setting the distance band value to arbitrarily longer distances did produce evidence
of macro-clustering with the Pharmacy category, it further increased the hyper-clustering of the
other scenarios. For example, setting the distance band value to 6.4 miles (the value used to
create the optimized hot spot results with the census tracts) resulted in the northern part of the
county becoming a solid hot spot while the southern part of the county was a solid cold spot.

Figure 57 OC Alcohol License Observational Hot Spots Based Upon Cell Boundaries

154

Ultimately, the attempt to generate a Cell-based Observational Hot Spot Analysis was abandoned
as the results did not appear to provide any meaningful insight.
4.3 Quantification of Race Neutral and Disparate Distributions
This study utilized six different analytical methods applied to census tracts and five
different methods applied to scaled population grid cells—110 different evaluation points—in an
attempt to quantify race-based distribution patterns of alcohol licenses (see Table 74). Out of the
110 evaluation points, 29 results were excluded from analysis because they had less than 20
census tracts or cells represented in the results or their linear regression p-values were greater
than 0.05. The remaining 81 points were suitable for race-neutral versus disparate distribution
analysis. Of those 81 results, 74 deviated from what would be expected if race-neutral functions
were operating. Finally, 34 results manifested a race-correlated distribution, while the remaining
40 results exceeded thresholds defined to identify occurrence of disparate distributions.
Looking specifically at the licenses results, All Licenses and Type 20 Licenses
manifested 7 disparate distribution and 3 race-correlated results out 11 tests. Type 21 Licenses,
on the other hand, had 5 disparate distribution results and 4 race-correlated results. These
outcomes suggest that California’s race-neutral licensing regulations alone cannot address the
various factors that together result in some communities of color having higher distributions of
alcohol retailers than their county-wide presence would predict, while majority white
communities have fewer alcohol retailers then their county-wide presence would predict.
As for the retailer results; first, Wholesale Clubs was effectively excluded from analysis
because of the small sample size (N=16). Next, Department Stores, Pharmacies, and Gas Stations
exhibited some race-neutral distributions and had the least disparate distribution results. Liquor

155

and Convenience Stores, on the other hand, exhibited disparate distribution in the majority of
their results. Grocery Stores appeared to be race-correlated with some disparate distribution.

156
Table 74 Summary of Distribution Evaluation Points

All
Licenses
Type
21
Type
20
Liquor
Stores
Grocery
Stores
Convenience
Stores
Gas
Stations
Pharmacies
Dept
Stores
Wholesale
Clubs
Definitions:
C: Race-Correlated Distribution
D: Disparate Distribution
N: Race Neutral
X: Excluded from analysis
Census Tracts
Presence D C D D D D C C D X
Absence D D D D C D C N N N
Density per
Square Mile
C C C C C C C C X X
Density per
1,000 People
X C X C C X X C X X
Optimized
Hot Spots
D D D D X D X X X X
Observational
Hot Spots
D X D D X D D X X X
Scaled Cells
Presence D D D D D D D C D X
Absence D D D C C D N N N N
Density per
Cell
C X C C C C X C X X
Density per
1,000 People
C C C C C C C C X X
Optimized
Hot Spots
D D D D D D D X X X

157

4.4 Census Tracts Versus Scaled Population Grid: Outcome Variations
As a general observation, there are significant variations between census tracts and scaled
population grid cells at the atomic level—individual cells, tracts, and results of computations.
This should come as no surprise as the two approaches are based on significantly different spatial
units. Take for example Type 20 licenses (see Table 75). First, the total area covered by census
tracts with Type 20 licenses is 446.8 square miles compared to 123.8 square miles covered by the
cells; census tracts cover more than triple the cell area. Second, the tract-based population
(N=2,050,182) is nearly double the cell-based population (N=1,085,899). However, these
variations at the atomic level between scale provide further insight as to populations and licenses
distribution patterns.
Table 75 Comparison of Type 20 Licenses between Tracts and Cells

Tracts: 364 / 65.0% of OC Population
446.8 mi
2
/ N=2,050,182 pop
Cells: 450 / 34.4% of OC Population
123.8 mi
2
/ N=1,085,899 pop
%
Expected
%
Actual
Pop %
Tracts
(Pop x L)
% Tracts
%
Expected
%
Actual
Pop %
Cells
(Pop x L)
% Cells
Asian Alone 12.67%
12.78%
(±0.17)
19.67%
(±0.26)
18.99%
(±0.19)
6.71%
6.12%
(±0.11)
17.79%
(±0.33)
17.08%
(±0.25)
Hispanic (any race) 22.22%
25.9%
(±0.26)
39.87%
(±0.41)
43.04%
(±0.31)
11.77%
15.54%
(±0.2)
45.16%
(±0.58)
48.63%
(±0.47)
White Alone 26.9%
23.25%
(±0.2)
35.79%
(±0.31)
33.38%
(±0.22)
14.25%
11.22%
(±0.14)
32.6%
(±0.39)
30.03%
(±0.3)
Black Alone 1.02%
1.04%
(±0.07)
1.61%
(±0.11)
1.66%
(±0.08)
0.54%
0.57%
(±0.05)
1.65%
(±0.15)
1.58%
(±0.11)
All Other Race(s) 2.16%
1.99%
(±0.09)
3.06%
(±0.14)
2.93%
(±0.1)
1.15%
0.97%
(±0.06)
2.81%
(±0.17)
2.67%
(±0.13)

For instance, while the Asian, Hispanic, and White values in the tract-based
(Pop x L) % Tract column and the cell-based (Pop x L) % Cells column both show that the
Hispanic population has more exposure to Type 20 licenses, the tract-based values have less

158

variance than the cell-based values. This indicates that in the aggregate more of the Hispanic
population lives within ~0.5 miles of a Type 20 retailer than the other two races. Thus, the
variations between the methods provides an insight that would not be apparent from either
method alone. Overall, the two methods are complimentary and when compared together provide
insight into patterns and distributions that operate at different scales.

159

Chapter 5 Conclusion
Systemic racism in the built environment is present when minority groups experience greater
detriments or fewer benefits than nearby majority populations. Some detriments are easily
identified: pollution, crime, dumps, and toxic water. Benefits, on the other hand, are not always
as easily identified, while their absence from the built environment is just as impactful. Take for
example access to clean water—a benefit often taken for granted—which promotes healthy
communities, while its absence—a benefit denied in Flint, MI—precipitates a public health
crisis. Or the unexpected benefit of open green spaces, which decreases environmental heat
retention, while the lack of open green spaces, in the form of continuous concrete and asphalt
surfaces, results in increased urban temperatures. These forms of systemic racism are often
referred to as disparate impacts, which is a way of describing the effects of these detriments on
the community. This study introduced the more nuanced term disparate distribution to describe
the uneven distribution of a benefit or burden across the built environment.
5.1 Finding Disparate Distributions of Alcohol Licenses in Orange County
This study focused on whether race and ethnicity correlated with the distribution of
alcohol licenses in the Orange County built environment because one of the major factors in the
distribution of those licenses is a race-neutral licensing regulation. Alcohol licenses were chosen
because the density of alcohol retailers and the sale of alcohol has been correlated with negative
impacts on the communities where alcohol retailers are concentrated and the population lives
within close proximity. While there has been less research on the impacts to the community
based on type of retailer, it is reasonable to assume that the dissociative impact of easy access to
alcohol in a community is not completely diminished simply because the access is through a
pharmacy, grocery store, or some other type of retailer with associative qualities. Therefore, the

160

type of retailer became another variable of interest in analyzing distribution patterns of alcohol
licenses with reference to the racial/ethnic composition of the local population.
To analyze whether there were disparate distributions of Type 20 and Type 21 alcohol
licenses, this study leveraged multiple straightforward statistical methods, commonly available
datasets, and two forms of areal sampling to address and mitigate issues related to MAUP,
aggregation, and spillover effects that can exaggerate or attenuate distribution analysis.
Moreover, multiple reliable observations increased the objective quality of the results. Overall,
this study found that both Type 20 and Type 21 licenses exceeded thresholds for disparate
distributions across almost all evaluation points. For example, the Hispanic population was
consistently overrepresented—exceeding their county-wide population representation proportion
by more than 10%—in the licensing Hot Spot analyses. Likewise, the Liquor Stores,
Convenience Stores, and Grocery Stores, also exceeded the thresholds for disparate distributions
across the majority of evaluation points. On the other hand, Gas Stations, Pharmacies, and
Department Stores also exhibited disparate distributions, but with significantly less evaluation
points in agreement. In other words, these categories produced mixed summary statistics results
of disparate distributions and no or few hot spot hots with statistical significance.
This study also identified four principal forms of disparate distribution. The first is
overrepresentation of a minority race/ethnicity in the presence of licenses/retailers (presence).
The second is underrepresentation of a minority race/ethnicity in the locations where
licenses/retailers are not distributed (absence). The third is underrepresentation of the majority
race in the presence of licenses/retailers. The fourth is overrepresentation of the majority race in
the locations where licenses/retailers are not distributed.

161

These patterns were often found together, but there were instances where some of the
evaluation points were race neutral and only one or two of the disparate distribution forms
manifested in the other evaluation points. For example, the Hispanic population was almost
always overrepresented in the presence of licenses/retailers, but the Asian population was
frequently neutral or underrepresented. While this study focused on the Asian, Hispanic, and
White populations to access disparate distributions, the Black and Other populations results
manifested unique representation profiles different than the Asian, Hispanic, and White
populations. For example, the Black population was more frequently overrepresented in the
presence of licenses/retailers than the Asian population, but not as often as the Hispanic
population, whereas the Other population was frequently underrepresented in the scenarios.
Thus, as to the forms of disparate distributions, this study found that the Hispanic
population was consistently overrepresented in the presence of licenses/retailers and
underrepresented in locations where licenses/retailers did not occur. Likewise, the White
population was consistently underrepresented in the presence of licenses/retailers and
overrepresented in locations without licenses/retailers. In other words, all four principal types of
disparate distribution were found in Orange County. Moreover, Orange County is not unique or
exotic in a way that would explain the disparate distributions.
These results indicate that requiring laws and regulations to avoid recognition of race is
likely insufficient to ensure race-neutral distributions of benefits and detriments in the built
environment. As a matter of public policy, laws and regulations should focus on race-neutral
distributions of benefits and detriments across society, which can only occur by recognition that
race matters. At least since the late 20
th
century, America has undertaken, as a matter of public
policy, the project of addressing and correcting racism in all forms. This study, with its

162

development of the concept of disparate distributions, extends that initiative with a framework of
methodologies and datasets that can be replicated, tailored, and deployed to identify systemic
racism in the built environment.
5.2 Limitations of the Study
One of the main limitations of the study is the accuracy of the race/ethnicity estimates at
both the census tract and scaled population grid cell levels. Specifically, the scaled population
grid data was created by areally weighted spatial interpolation of census tracts and LandScan
2018 data; any errors in the census tract estimates or LandScan 2018 population data would be
multiplied as a part of the grid creation process.
Another limitation is this study relied on a single snapshot of the active licenses on the
date of license data acquisition. While the entire dataset of licenses is likely fairly stable over
time, there are constant minor changes as businesses move, licenses are transferred, new
businesses start, and other businesses fold.
Also, the study only examined retailers with alcohol licenses, however, grocery stores,
convenience stores, gas stations, etc. may exist in the built environment that do not have alcohol
licenses. Examining the distribution of the alcohol-licensed and non-licensed retailers would
provide greater insight as to how race correlates with retailer type distributions.
5.3 Areas for Future Study
This study provided a proof of concept framework for a multiscale disparate distribution
analysis in a single California county. However, the datasets exist to analyze all counties in
California as the licensing regulations operate at the state level. Moreover, this study could be
used as a template for a temporal analysis of the changing license distributions over time.

163

Furthermore, while this study aggregated licenses to census tracts and scaled population
grid cells, the spatial interpolation could be carried one step further to create zones around each
license point to aggregate the local population to the license points at other scales for analysis.
This study would also benefit from inclusion of additional socio-economic factors such as
income, education level, and housing statistics. Finally, examining retailers with and without
alcohol licenses would provide greater insight into race-correlated retailer distribution patterns.

164

References
Alsan, Marcella, Owen Garrick, and Grant Graziani. 2019. "Does Diversity Matter for Health?
Experimental Evidence from Oakland.” American Economic Review 109, no. 12: 4071-
4111. doi:10.1257/aer.20181446.
Archer, Deborah N. 2019. “The New Housing Segregation: The Jim Crow Effects of Crime-Free
Housing Ordinances.” Michigan Law Review 118, no 1 (February).
https://dx.doi.org/10.2139/ssrn.3331751.
Barcelona Field Studies Centre. 2020. “Simpson’s Diversity Index.” Last Modified January 19,
2020. Accessed February 2, 2020.
https://geographyfieldwork.com/Simpson'sDiversityIndex.htm.
Barocas, Solon and Andrew D. Selbst. 2016. “Big Data's Disparate Impact.” California Law
Review 104: 671. Available at SSRN: https://ssrn.com/abstract=2477899 or
http://dx.doi.org/10.2139/ssrn.2477899.
Berke, Ethan M., Susanne E. Tanski, Eugene Demidenko, Jennifer Alford-Teaster, Xun Shi, and
James D. Sargent. 2010. “Alcohol Retail Density and Demographic Predictors of Health
Disparities: A Geographic Analysis.” American Journal of Public Health 100, no. 10:
1967-1971. doi:10.2105/AJPH.2009.170464.
California Business and Professions Code (BPC), Division 9 Alcoholic Beverages, Chapter 6
Issuance and Transfer of Licenses, Article 1 Applications for Licenses (2019).
California Department of Alcoholic Beverage Control (ABC). 2020. “Licensing Reports.”
Retrieved February 18, 2020. https://www.abc.ca.gov/licensing/licensing-
reports/#weekly-data-export.
California Legislature. Assembly (AB). 1994. Hearing before the Assembly Committee on
Governmental Organization. AB 2897, Undue Concentration of Liquor Licenses, April
12.
Cooksey-Stowers, Kristen, Marlene B. Schwarzt, and Kelly D. Brownell. 2017. “Food Swamps
Predict Obesity Rates Better Than Food Deserts in the United States.” International
Journal of Environmental Research and Public Health 14, no. 11: 1366-1386.
doi:10.3390/ijerph14111366.
Dark, Shawna J. and Danielle Bram. 2007. “The Modifiable Areal Unit Problem (MAUP) in
Physical Geography.” Progress in Physical Geography 31, no. 5: 471-479.
doi:10.1177/0309133307083294.
Drewnowski,A., D. Arterburn, J. Zane, A. Aggarwal, S. Gupta, P.M. Hurvitz, A.V. Moudon, J.
Bobb, A. Cook, P. Lozano, D. Rosenberg. 2019. “The Moving to Health (M2H)
Approach to Natural Experiment Research: A Paradigm Shift for Studies on Built

165

Environment and Health. SSM - Population Health 7: 100345.
https://doi.org/10.1016/j.ssmph.2018.100345.
Drummond, Jim. 2014. “City Formalizes Process for Liquor Licenses.” Orange County Register,
June 26. https://www.ocregister.com/2014/06/26/city-formalizes-process-for-liquor-
licenses.
Dwivedi, Pallavi, Dina Huang, Weijun Yu, and Quynh Nguyen. 2019. “Predicting Geographical
Variation in Health-Related Quality of Life.” Preventive Medicine 126: 105742.
doi:10.1016/j.ypmed.2019.05.030.
Fisher, Joshua B. Maggi Kelly, and Jeff Romm. 2006. “Scales of environmental justice:
Combining GIS and spatial analysis for air toxics in West Oakland, California.” Health &
Place 12, no. 4: 701-14. doi:10.1016/j.healthplace.2005.09.005.
Google. 2020. “Overview. Google Maps Platform, Web Services, Geocoding API, Overview.
Accessed February 17, 2020.
https://developers.google.com/maps/documentation/geocoding/overview.
Groos, Maya, Maeve Wallace, Rachel Hardeman, and Katherine Theall. 2018. “Measuring
Inequity: A Systematic Review of Methods Used to Quantify Structural Racism.” Journal
of Health Disparities Research and Practices 11, no. 2: 190-206.
https://digitalscholarship.unlv.edu/jhdrp/vol11/iss2/13.
Halonen, Jaana I., Mika Kivimäki, Marianna Virtanena, Jaana Pentti, S.V. Subramanianc, Ichiro
Kawachi, and Jussi Vahtera. 2013. “Proximity of Off-Premise Alcohol Outlets and Heavy
Alcohol Consumption: A Cohort Study.” Drug and Alcohol Dependence 132, no. 1: 295-
300. doi:10.1016/j.drugalcdep.2013.02.022.
He, Li, Antonio Páez, and Desheng Liu. 2017. “Built environment and Violent Crime: An
Environmental Audit Approach Using Google Street View.” Computers, Environment
and Urban Systems 66: 83-95. https://doi.org/10.1016/j.compenvurbsys.2017.08.001.
Hippensteel, Christopher L, Richard C Sadler, Adam J Milam, Victoria Nelson and C Debra
Furr-Holden. 2019. “Using Zoning as a Public Health Tool to Reduce Oversaturation of
Alcohol Outlets: an Examination of the Effects of the New ‘300 Foot Rule’ on Packaged
Goods Stores in a Mid-Atlantic City.” Prevention Science 20: 833-843.
doi:10.1007/s11121-018-0947-9.
Hoffman, Jeremy S., Vivek Shandas, and Nicholas Pendelton. 2020. “The Effects of Historical
Housing Policies on Resident Exposure to Intra-Urban Heat: A Study of 108 US Urban
Areas.” Climate 8, no. 1: 12. https://doi.org/10.3390/cli8010012.
Hurvitz, Philip M., Anne V. Moudon, Colin D. Rehm, Laura C. Streichert, and Adam
Drewnowski. 2009. “Arterial Roads and Area Socioeconomic Status Are Predictors of
Fast Food Restaurant Density in King County, WA.” International Journal of Behavioral
Nutrition and Physical Activity 6:46. http://www.ijbnpa.org/content/6/1/46.
Doi:10.1186/1479-5868-6-46.

166

Infogroup, Inc. 2020. List of Alcohol Retailing Businesses in Orange County, CA. Retrieved
February 17, 2020, from ReferenceUSA database.
Jelinski, Dennis E. and Jianguo Wu. 1996. “The Modifiable Areal Unit Problem and
Implications for Landscape Ecology.” Landscape Ecology 11, no. 3: 129-140.
https://doi.org/10.1007/BF02447512.
Johnson, Christopher and Jackson Hoops. 2019. “The War on Drugs and the Case for
Rehabilitation.” Brigham Young University Prelaw Review 33, no. 17.
https://scholarsarchive.byu.edu/byuplr/vol33/iss1/17.
Kau, James, Lu, Fang, Munneke, Henry. 2018. An Unintended Consequence of Mortgage
Financing Regulation – a Racial Disparity. Journal of Real Estate Finance and
Economics (November):1-40. doi:10.1007/s11146-018-9683-y.
Kim, Yushim and Yongwan Chun. 2019. “Revisiting Environmental Inequity in Southern
California: Does Environmental Risk Increase in Ethnically Homogeneous or Mixed
Communities?” Urban Studies 56, no. 9: 1748-1767. doi: 10.1177/0042098018803227.
Kogure, Katsuo and Yoshito Takasaki. 2019. “GIS for Empirical Research Design: An
Illustration with Georeferenced Point Data.” PLoS ONE 14, no. 3: e0212316.
https://doi.org/10.1371/journal.pone.0212316.
Kramar, David E, Aaron Anderson, Hayley Hilfer, Karen Branden, and John J. Gutrich. 2018.
“A Spatially Informed Analysis of Environmental Justice: Analyzing the Effects of
Gerrymandering and the Proximity of Minority Populations to U.S. Superfund Sites.”
Environmental Justice 11, no. 1: 29–39.
Krivo, Lauren J., Ruth D. Peterson, and Danielle C. Kuhl. 2009. “Segregation, Racial Structure,
and Neighborhood Violent Crime.” American Journal of Sociology 114, no. 6: 1765-
1802. doi:10.1086/597285.
Kubrin, Charis E. and John R. Hipp. 2016. “Do Fringe Banks Create Fringe Neighborhoods?
Examining the Spatial Relationship between Fringe Banking and Neighborhood Crime
Rates.” Justice Quarterly 33, no. 5: 755-784.doi: 10.1080/07418825.2014.959036.
Leyk, Stefan, Andrea E. Gaughan, Susana B. Adamo, Alex de Sherbinin, Deborah Balk, Sergio
Freire, Amy Rose, Forrest R. Stevens, Brian Blankespoor, Charlie Frye, Joshua
Comenetz, Alessandro Sorichetta, Kytt MacManus, Linda Pistolesi, Marc Levy, Andrew
J. Tatem, and Martino Pesaresi. 2019. “The Spatial Allocation of Population: A Review
of Large-scale Gridded Population Data Products and Their Fitness for Use.” Earth
System Science Data 11: 1385-1409. https://doi.org/10.5194/essd-11-1385-2019, 2019.
Liu, Jodi L., Bing Han, Deborah A. Cohen. 2015. “Beyond Neighborhood Food Environments:
Distance Traveled to Food Establishments in 5 US Cities, 2009-2011.” Preventing
Chronic Disease 12(E126): 150065. DOI:http://dx.doi.org/10.5888/pcd12.150065.

167

Montello, Daniel, R. and Paul C. Sutton. 2013. An Introduction to Scientific Research Methods
in Geography and Environmental Studies. 2nd ed. London:SAGE Publications.
National Association for the Advancement of Colored People (NAACP). 2015. “Criminal Justice
Fact Sheet.” Accessed November 10, 2019.
http://naacp.org/criminal-justice-fact-sheet.
Obermeyer, Ziad, Brian Powers, Christine Vogeli, and Sendhil Mullainathan. 2019. “Dissecting
Racial Bias in an Algorithm Used to Manage the Health of Populations.” Science 366
(October): 447-453.
Oka, Masayoshi and David W. S. Wong. 2016. “Spatializing Area-Based Measures of
Neighborhood Characteristics for Multilevel Regression Analyses: An Areal Median
Filtering Approach.” Journal of Urban Health 93: 551-571.
doi:10.1007/s11524-016-0051-z.
Orange County Historical Society (OCHS). 2020. “Orange County Chronology.” Accessed
February 4, 2020. http://www.orangecountyhistory.org/wp/?page_id=40.
Popkin, Barry M., Kiyah Duffey, and Penny Gordon-Larsen. 2005. “Environmental Influences
on Food Choice, Physical Activity and Energy Balance.” Physiology & Behavior 86,
no. 5: 603-613. doi:10.1016/j.physbeh.2005.08.051.
Pulido, Laura, Steve Sidawi, and Robert O. Vos. 1996. “An Archaeology of Environmental
Racism in Los Angeles.” Urban Geography 17, no. 5: 419-439.
Rose, Amy N., Jacob J. McKee, Marie L. Urban, Edward A. Bright, and Kelly M. Sims. 2019.
“lspop2018.” LandScan Global Population Database 2018 Oak Ridge, TN: Oak Ridge
National Laboratory. https://landscan.ornl.gov/landscan-datasets.
Ruther, Matt, Stefan Leyk, and Barbara P. Buttenfield. 2015. “Comparing the Effects of an
NLCD-derived Dasymetric Refinement on Estimation Accuracies for Multiple Areal
Interpolation Methods.” GIScience & Remote Sensing 52 no. 2: 158-178.
http://dx.doi.org/10.1080/15481603.2015.1018856.
Smith, Tara A. and J. S. Onésimo Sandoval. 2019. “A Spatial Analysis of Homicides in Saint
Louis: The Importance of Scale.” Spatial Demography 7: 57–82.
https://doi-org.libproxy1.usc.edu/10.1007/s40980-018-00046-8.
Strand, Palma Joy. 2019. “The Invisible Hands of Structural Racism in Housing: Our Hands, Our
Responsibility.” University of Detroit Mercy Law Review 96, no. 2 (January): 155–185.
Sweeney, Latanya. 2013. “Discrimination in Online Ad Delivery.” Queue 11, no. 3: 10-29.
doi:https://doi.org/10.1145/2460276.2460278.
Texas Dept. of Housing and Community Affairs et al. v. Inclusive Communities Project, Inc., et
al. 135 S. Ct. 2507 (2015).

168

Tobler, Waldo R. 1970. “A Computer Movie Simulating Urban Growth in the Detroit Region.”
Economic Geography 46: 234-40. doi:10.2307/143141.
Unger, Jennifer, B., Robert O. Vos, Jasmine Siyu Wu, Kimberly Hardaway, Ada Y. Li Sarain,
Daniel W. Soto, Christopher Rogers, and Jane Steinberg. 2020. “Locations of Licensed
and Unlicensed Cannabis Retailers in California: A Threat to Health Equity.” Preventive
Medicine Reports 19: 101165. https://doi.org/10.1016/j.pmedr.2020.101165.
U.S. Census Bureau. 1994. Geographic Areas Reference Manual. [Washington, D.C.] U.S. Dept.
of Commerce, Economics and Statistics Administration, Bureau of the Census.
https://www.census.gov/programs-surveys/geography/guidance/geographic-areas-
reference-manual.html.
U.S. Census Bureau. 2017. “Table DP05: ACS Demographic and Housing Estimates.” American
Community Survey. Retrieved from
https://data.census.gov/cedsci/all?text=dp05&t=Race%20and%20Ethnicity&tid=ACSDP
1Y2017.DP05&g=0500000US06059&hidePreview=false&vintage=2017&cid=DP05_00
01E.
U.S. Census Bureau. 2018. “Understanding and using American Community Survey data: What
all data users need to know.” Retrieved from https://www.census.gov/programs-
surveys/acs/guidance/handbooks/general.html.
U.S. Census Bureau. 2020. “QuickFacts Orange County, California.” QuickFacts. Retrieved
from https://www.census.gov/quickfacts/orangecountycalifornia.
U.S. Environmental Protection Agency (USEPA). 2018. “Environmental Justice.” Accessed
November 10, 2019. https://www.epa.gov/environmentaljustice.
Wolch, Jennifer, R., Jason Byrne, and Joshua P. Newell. 2014. “Urban Green Space, Public
Health, and Environmental Justice: The Challenge of Making Cities ‘Just Green
Enough.’” Landscape and Urban Planning 125: 234-244.
https://doi.org/10.1016/j.landurbplan.2014.01.017.
Yoo, Youngmin and Andrew P. Wheeler. 2019. “Using Risk Terrain Modeling to Predict
Homeless Related Crime in Los Angeles, California.” Applied Geography. 109: 102039.
https://doi.org/10.1016/j.apgeog.2019.102039.
Young, Robert, Laura Macdonald, and Anne Ellaway. 2013. “Associations Between Proximity
and Density of Local Alcohol Outlets and Alcohol Use Among Scottish Adolescents.”
Health and Place 19, no. 100: 124-130. doi:10.1016/j.healthplace.2012.10.004.

169

Appendix A: ABC License Types
Type Description
01 Beer Manufacturer
02 Winegrower
03 Brandy Manufacturer
04 Distilled Spirits Manufacturer
05 Distilled Spirits Manufacturer's Agent
06 Still
07 Rectifier
08 Wine Rectifier
09 Beer and Wine Importer
10 Beer and Wine Importer's General
11 Brandy Importer
12 Distilled Spirits Importer
13 Distilled Spirits Importer's General
14 Public Warehouse
15 Customs Broker
16 Wine Broker
17 Beer and Wine Wholesaler
18 Distilled Spirits Wholesaler
19 Industrial Alcohol Dealer
20 Off-Sale Beer & Wine
21 Off-Sale General
22 Wine Blender
23 Small Beer Manufacturer
24 Distilled Spirits Rectifier's General
26 Out-of-State Beer Manufacturer's
Certificate
27 California Winegrower's Agent
28 Out-of-State Distilled Spirits Shipper's
Certificate
29 Wine Grape Grower's Storage
Type Description
40 On Sale Beer
41 On Sale Beer & Wine – Eating Place
42 On Sale Beer & Wine – Public Premises
47 On Sale General – Eating Place
48 On Sale General – Public Premises
49 On Sale General – Seasonal
51 Club
52 Veteran’s Club
57 Special On Sale General
59 On Sale Beer And Wine – Seasonal
60 On Sale Beer – Seasonal
61 On Sale Beer – Public Premises
62 On-Sale General Bona Fide Public Eating
Place Intermittent Dockside Vessel
64 Special On-Sale General Theater
67 Bed and Breakfast Inn
70 On Sale General – Restrictive Service
71 Special On-Sale General License
72 Special On-Sale General For-Profit
Theater, Napa
75 On Sale General – Brewpub
78 On Sale General Wine, Food and Art
Cultural Museum
80 Bed and Breakfast Inn – General
82 Direct Shippers Permit
83 On-Sale General Caterer's License
86 Instructional Tasting License
87 Neighborhood-Restricted Special On-Sale
General License
88 Special On Sale General for Historic
Cemetery

Source: ABC (2019)

Abstract (if available)

Abstract Systemic racism, institutional racism, structural racism: these are the terms used to describe unequal minority participation in job markets, over representation in the criminal justice system, and lack of access to and enjoyment of clean and safe neighborhoods. Studies in social justice and environmental justice are now starting to quantify structural racism by utilizing Geographic Information Systems and applying analytic methods of Geographic Information Science. One area ripe for study of structural racism is whether race-neutral laws and regulations promote race-neutral distributions in the built environment or perpetuate existing structural racism. ❧ The distribution of alcohol retailers in Orange County, California, provided an opportunity to explore how a race neutral regulation—in operation for over two decades—has impacted the built environment. Exploring the distribution of alcohol retailers informs our understanding of structural racism because a higher density of retailers has been correlated with negative impacts on neighborhoods such as increased crime, negative health outcomes, and poverty. Moreover, California’s alcohol licensing regulations are race-neutral and as such do not consider race as a factor in determining the approval or rejection of a license application. ❧ This study analyzed the February 18, 2020 inventory of active off-site retail sales alcohol licenses in Orange County and compared the distribution of licenses with race/ethnicity across the county. The comparison was repeated at two spatial scales: census tract and a scaled population grid based on the Oak Ridge National Laboratory’s LandScan 2018 dataset with 30 arc-second cells (∼0.5 miles). This study found that Hispanic populations were consistently overrepresented in census tracts and cells where alcohol licenses were found. This result suggested that requiring laws and regulations to avoid recognition of race is insufficient to ensure race-neutral distributions of benefits and detriments in the built environment.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A spatiotemporal analysis of racial disparity in the distribution of superfund sites within Santa Clara County, California

PDF

The impact of severe coastal flooding on economic recovery disparities: a study of New Jersey communities following Hurricane Sandy

PDF

Investigating bus route walkability: comparative case study in Orange County, California

PDF

An accessibility analysis of the homeless populations' potential access to healthcare facilities in the Los Angeles Continuum of Care

PDF

Mapping future population impacts caused by sea level rise in Huntington Beach and Newport Beach: comparing the cadastral-based dasymetric system to past dasymetric mapping methods

PDF

Spatial analysis of veteran access to healthcare in Los Angeles County

PDF

Identification and analysis of future land-use conflict in Mecklenburg County, North Carolina

PDF

Distribution and correlates of feral cat trapping permits in Los Angeles, California

PDF

Exploring the impact of the natural gas leak on home values in Porter Ranch, California

PDF

Closed landfills to solar energy power plants: estimating the solar potential of closed landfills in California

PDF

Risk analysis and assessment of non‐ductile concrete buildings in Los Angeles County using HAZUS‐MH

PDF

Preparing for immigration reform: a spatial analysis of unauthorized immigrants

PDF

The role of GIS in asset management: County of Kauai Department of Parks and Recreation a need for an asset management program

PDF

Estimating at-risk population for lead service lines induced lead exposure and their correlation to socioeconomically disadvantaged neighborhoods in Milwaukee, Wisconsin

PDF

An evaluation of Esri’s tapestry segmentation product in three Southern California communities: Manhattan Beach, Santa Monica, and Venice Beach

PDF

Development and evaluation of a USV based mapping system for remote sensing of eelgrass extent in southern California

PDF

Geospatial analysis of the Round Fire: a replication of burn severity analyses in the Sierra Nevada

PDF

The geographic connotations of reincarceration: a spatial analysis of recidivism in Washington State

PDF

Wrong way driving in San Antonio, Texas: a transportation route study using Network Analyst

PDF

Urban green space accessibility and environmental justice: a GIS-based analysis in the city of Phoenix, Arizona

Asset Metadata

Creator Gulledge, Kelly Woody (author)

Core Title An analysis of racial disparity in the distribution of alcohol licenses and retailers in Orange County, California

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 09/04/2020

Defense Date 08/10/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Alcohol,alcohol retailer,disparate distribution,disparate impact,Liquor Store,OAI-PMH Harvest,race-neutral legislation,racial disparity,structural racism

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ruddell, Darren (committee chair), Vos, Robert (committee member), Wu, An-Min (committee member)

Creator Email gulledge@usc.edu,kwgull@workinglaw.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-367836

Unique identifier UC11666382

Identifier etd-GulledgeKe-8942.pdf (filename),usctheses-c89-367836 (legacy record id)

Legacy Identifier etd-GulledgeKe-8942.pdf

Dmrecord 367836

Document Type Thesis

Rights Gulledge, Kelly Woody

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA