USE/“TRY
r+ sen,

1;.

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 c:/CTRC/DateDue.Indd-p.15

RESIDENTIAL CLASS INDICATORS:
Census-Based Techniques

for Modeling Neighborhood Quality
and Social Class in Urban Areas

BY MIKE SOBOCINSKI

AXXXXXXXX

Urban and Regional Planning 889: Master's Research

Advisor: Dr. Roger Hamlin

A Research Paper in Partial Fulfillment of the Requirements
for the Master of Urban and Regional Planning Degree,
Michigan State University, Dec. 2000

THESIS

PREFACE

CONTENTS

ACKNOWLEDGEMENTS

KEY TO ABBREVIATIONS

KEY TO ABBREVIATIONS IN EQUATIONS

l.

2.

7.

APPENDICES

INTRODUCTION

BACKGROUND INFORMATION AND KEY CONCEPTS
THE RC1 MODEL EXPLAINED
APPLICATION OF THE RC1 MODEL
CRITIQUE OF THE RCI MODEL '
AREAS FOR FURTHER RESEARCH

CONCLUSION

REFERENCE LIST

iii

vi

vii

ix

25

44

54

62

85

86

93

PREFACE
The Structure of this Paper

After an introductory section, this research paper will proceed with a lengthy
discussion Of the important concepts being addressed, providing references to
background materials that led up to and enabled this new research. In order to render my
explanations of a complicated subject more approachable and interesting, I have taken the
liberty of writing in the first person and relating numerous ideas and explanations (such
as the identification of weaknesses and areas for further research) where I feel they are
most meaningful, even if in an organizational sense they would be expected to appear in a
separate section of the paper dealing more specifically with such topics.

My explanation of the progress of my research is generally chronological in the
section explaining the RC1 model—a deliberate choice which I felt would assist readers
in becoming gradually acquainted with the concepts, my evolving mathematical model,
and the abbreviations that I use in the equations and discussion of my model's concepts.
Again, the narrative at times is a bit personal, which I felt was acceptable to keep under .
consideration the subjective aspects of the topic. My Objective in occasionally using
some personal anecdotes is to link the mathematical abstractions of my model with a
more colloquial explanation and "common sense" conceptualization of what I have done,
and why. C

When equations are given, later in this paper, I have in most cases immediately
followed them with explanations of the variables that they refer to in abbreviated form.
These variables are all either from the decennial census (1990 STF 3A) or are derivable

from them in relatively straightforward ways (the examples of the RC1 model used

census data downloaded from the web site at http://venus.census.uov/cdrom/lookup). To assist
readers in understanding these equations, and to avoid excessive repetition in the
explanations accompanying them, I have included, following this preface, an alphabetical
guide, called "Key to Abbreviations in Equations," that will help make their meaning and
use more understandable. My abbreviations also follow a format that is fairly
straightforward, once the reader is familiar with all the main concepts involved in the
topic. Preceding the "Key to Abbreviations in Equations" is a smaller "Key to
Abbreviations." All of the longer abbreviations used in equations are made up of
combinations of these smaller abbreviations. I believe that once the subject is
understood, the use of these abbreviations will make discussion and application of the
concepts much easier than if lengthy phrases or are Used to describe each variable and
condition.

The first four equations presented in the paper are mere prototypes, whose simpler
structure allows the reader to better understand the logic of the lengthier equations that
follow. In addition, this renders the evolution of my ideas'more clearly, enabling my
work to be more easily analyzed, critiqued, and amended to correct for flaws that are
found. Various assumptions that are part of the model can be seen more clearly as they
appear during my chronological description of its creation. Using this style of exposition,
I hope to equip my readers with the means to adjust my equations to match different
assumptions that they may make, or different research goals that they may have. The
equation that, at the time Of writing, I found most useful, is Equation 8. Equations 7 and
9 are considered simpler but similarly sound. I follow this set of equations with another

(Equations 10 through 12) that I chose at this time to abandon as less accurate and useful.

I include them because I believe it possible that they may be found to be useful at some
later time as more thinking and research is done on this specific topic, and thus the
equations relate to one of the areas for further research that I have identified.

Following my main explanatory sections is an example of how my research can
be'applied and interpreted. This leads almost directly into an assessment of some of the
weaknesses in my model, and precautions about certain applications of it that I feel at this
point are not yet justifiable. The final, lengthy section takes numerous concepts of the
model and proposes how they may be profitably used in numerous future studies.

Several appendices appear at the end of the document. These are tables that are referred
to by various passages Of the text, but which were not included as figures within the text,

due to their burdensome sizes of at least two pages each.

1."

ACKNOWLEDGEMENTS

I must thank those in MSU's Urban and Regional Planning Department for their
patience during the long exploratory and research stages Of this paper's development,
particularly my advisor, Dr. Roger Hamlin. Thanks also goes to my friends and family
who assisted in the research, either through direct feedback on the quality of the RC1
models as applied to their areas (Buxees Singh in Buffalo, John Runge in Detroit, John
and Lynn Welch in Evanston, and Tim London in Alcona County), through the provision
of transportation in field surveys of those areas (Martha Runge in the Detroit UA, Erick
Williams in the Lansing UA), or through the provision of needed research materials and
equipment (Craig Anderson for maps and computer resources). Thanks to these and the
rest of my family and friends for their support, and tolerance Of the many hours I spent in
seclusion rather than in their company. Thanks to my co-workers and employers (past
and present) for their support, feedback, and provision Of the time and equipment I

needed to complete this research (or related research that contributed to it)—the

Emergency Management Division of the Michigan State Police, the Tri-County Regional

1 Planning Commission, and the City of Lansing Department of Planning and

Neighborhood Development. Thanks also to the various faculty throughout the College
Of Social Science at Michigan State University, who made it such a good learning

environment in which to explore endless topics Of interest and relevance.

vi

%
CR

GQ
GR

HU

Md
MSA
ni

OO

pci
PP
' RCI

R0

SMOC

UA

UDA

KEY TO ABBREVIATIONS
proportion
contract rent
group quarters (also "gq")
gross rent
household
housing unit
institutional
mean
median
MetrOpolitan Statistical Area
noninstitutional

owner-occupied (also "00")

~ persons

per capita income

persons per

Residential Class Indicator
Ranally Metropolitan Area
renter-Occupied (also "ro")
selected monthly owner costs
census tract

urbanized area (also "ua")

user—defined area

vii

V value
wm with a mortgage

wom without a mortgage

viii

KEY TO ABBREVIATIONS IN EQUATIONS

The following is a list of the abbreviations used in equations throughout this paper.

gqiRCIt

gqniRC It

quCIt

%GQn

%Gth

%GQt

MdCRt

MdCRua

Mth

Mqua

MGRt

The term of an RCI equation that estimates the quality of
institutional group quarters in a tract (or other small area of study)

The term of an RCI- equation that estimates the quality of
noninstitutional group quarters in a tract (or other small area of
study)

The term of an RCI equation that estimates the quality of group
quarters in a tract (or other small area of study)

The proportion Of persons in a tract (or other small area of study)
who inhabit institutional group quarters, found by dividing the
number of persons in institutional group quarters in an area by the
total number of persons in that area

The proportion of persons in a tract (or other small area of study)
who inhabit noninstitutional group quarters, found by dividing the
number Of persons in noninstitutional group quarters in an area by
the total number of persons in that area

The proportion of persons in a tract (or other small area of study)
who inhabit group quarters, found by dividing the number of
persons in group quarters in an area by the total number of persons
in that area

The median contract rent of renter-occupied unitsin a tract (or
other small area of study)

The median contract rent of renter-occupied units in the entire
urbanized area (or other large area used as a standard of
comparison)

The median value of owner-occupied housing units in a tract (or
other small area of study)

The median value of owner-occupied housing units in the entire
urbanized area (or other large area used as a standard of
comparison) '

The mean gross rent of renter-occupied units in a tract (or other
small area of study)

MGRua

MVt

MVua

MVwmt

MVwmua

MVwomt

MVwomua

%OOhht

%OOpt

OORCIt

%Omept

%OOwompt

pciGQit

The mean gross rent of renter-occupied units in the entire
urbanized area (or other large area used as a standard of
comparison) -

The mean value of all owner-occupied units in a tract (or other
small area of study)

The mean value of all owner-occupied units in the entire urbanized
area (or other large area used as a standard of comparison)

The mean value of all owner-occupied units with a mortgage in a
tract (or other small area Of study)

The mean value of all owner-occupied units with a mortgage in the
entire urbanized area (or other large area used as a standard of

comparison)

The mean value of all owner-occupied units without a mortgage in
a tract (or other small area of study)

The mean value of all owner-occupied units without a mortgage in
the entire urbanized area (or other large area used as a standard of
comparison)

The prOportion of households in a tract (or other small area of
study) that live in owner-occupied housing units, found by dividing
the number Of owner-occupied households in an area by its total
number of households

The proportionof person in a tract (or other small area of study)
that live in owner-occupied housing units, found by dividing the
number of persons living in such units in that area by the total
number of persons living in that area

The term of an RCI equation that measures the quality of owner-
occupied housing units in a tract (or other small area of study)

The proportion of persons in a tract (or other small area of study)
that live in owner-occupied housing units with a mortgage

The proportion of persons in a tract (or other small area of study)
that live in owner-occupied housing units without a mortgage

The per capita income of persons living in institutional group
quarters in a tract (or other small area of study), found by dividing
the aggregate income of all persons living in institutional group

pciGQnit

pciGQt

pciGQua

pciOmeua

ppOOHUt

ppOOHUua

ppOOHmeua

RCI%

quarters in an area by the number of persons living in institutional
group quarters in that area

The per capita income of persons living in noninstitutional group
quarters in a tract (or other small area of study), found by dividing
the aggregate income of all persons living in noninstitutional group
quarters in an area by the number Of persons living in
noninstitutional group quarters in that area

The per capita income of persons living in group quarters in a tract
(or other small area of study), found by dividing the aggregate
income Of all persons living in group quarters in an area by the
number of persons living in group quarters in that area

The per capita income of persons living in group quarters in the
entire urbanized area (or other large area used as a standard of
comparison), found by dividing the aggregate income of all
persons living In group quarters in that area by its total number of
persons living In group quarters

The per capita income of persons living in owner-occupied
housing units with a mortgage in the entire urbanized area (or other
large area used as a standard of comparison), found by dividing the

' aggregate income of all persons living in owner4occupied units

with a mortgage by the total number of persons living in owner-
occupied units with a mortgage

Persons per owner-occupied housing unit in the tract (or other
small area of study), found by dividing the aggregate number of
persons in owner-occupied housing units in the tract by the number
of owner-occupied housing units in the tract

Persons per owner-occupied housing unit in the entire urbanized
area (or other large area used as a standard of comparison), found
by dividing the aggregate number of persons in owner-occupied

housing units in the area by its total number of owner-occupied

housing units

Persons per owner-occupied housing unit with a mortgage in the
entire urbanized area (or other large area used as a standard of
comparison). I did not succeed in finding or deriving this data for
this study, and had to substitute ppOOHUua for this variable. .

The proportion of a tract RCI over the area RCI (RCIt —=- RCIua),

producing a measure that may allow for the comparison of tracts
between metropolitan areas, or in the same area over time, given

xi

RCIt

RCIua

%ROhht

%ROpt

roRCIt

SMOCwmua

the assumption that class position and relations are relative to local
social conditions rather than fixed standards or objective physical
conditions of housing

Residential Class Indicator for a tract (or other small area of study)

The RCI rating that represents an entire UA (or other large area
used as a standard of comparison), obtained simply by treating
information on that area as if it were for a single tract and applying
the RCIt formula to it

The proportion of households in a tract (or other small area of
study) that inhabit renter-occupied housing units, found by
dividing the number of renter-occupied households in an area by
its total number of households

The proportion of persons in a tract (or other small area of study)
that inhabit renter-occupied hOusing units, found by dividing the
number of persons in such units by the total number of persons
living in that area

The term of an RCI equation that measures the quality of rental-
occupied housing units in a tract (or other small area of study)

The'mean selected monthly owner costs of owner-occdpied units

with a mortgage, for the entire urbanized area (or other large area
used as a standard of comparison)

xii

1. INTRODUCTION
Introduction to the Problem Being Researched

Despite the extensive amount of research on the topic of social class, it is still a
difficult task to identify where a person, family, or household fits into a class or
stratification framework in an American city. Similarly, although most peOple have and
express ideas about areas of a city that they feel are "good" or "bad" to live in, or
somewhere in-between, it is often very difficult to pinpoint just how good or how bad an
area is. Both social class and neighborhood quality seem linked together, bitt there is no
clear measure in general use to express the nature of this relationship. One of the
problems is that there is no simply defined and easily measurable condition that defines
one's "social class." Also, while similar types of residential arrangements may be easy to
compare to each other, a sizable urban area contains quite a diversity of living
arrangements that tend to confound attempts to make comparisons between them.
Although owning a home is generally indicative of greater wealth and status than renting
an apartment, there may be areas within a city where the construction of a new apartment
building is a boon to the neighborhood, and provides housing ‘of a clearly higher quality
than the old and 'shoddy, crumbling houses'that surround it.

While it can sometimes be quite easy to reach agreement on whether some person
is lower or upper-class, or which neighborhood is one of the nicest or worst in a city,
there are many cases between these extremes that are more difficult to classify. Is our
definition of how class is to be measured powerful enough that any randomly selected
person in the country can be assessed and placed within a descriptive framework that is

meaningful to most peOple? If a location within a city were to be selected at random, is

there a way to assess how "desirable" an area it is to live in, or what "class" of person
would be likely or expected to live there? The goal of my research is to allow any
residential area or neighborhood within an urban area (of sufficient size) to be rated as to
its quality, and to use such ratings to help assess the class position, or socioeconomic
status, of those who are living there. My research enables such analysis by providing
equations that, using readily available census data, produce ratings called Residential
Class Indicators (RC1), that assess the residential quality of an area, and thus suggest
something about how its inhabitants fit into a contemporary framework Of social
stratification.
The Need for the Research

There is a need for this kind of research, in the fields of sociology and urban
planning, to better apply the concept of class or social stratification to a description or
model of urban areas in American society, in a fOrm that is usable by those without
extensive mathematical or statistical expertise. This research provides urban and regional
planners with'a relatively simple way of assessing the residential areas they oversee.
Urban geographers should also find my proposed model to be of interest in identifying
noteworthy socioeconomic patterns in residential location decisions and trends in an area,
since it has similarities with the "social area" analysis tools used in their discipline. This
research offers a way to measure concepts of class and neighborhood quality that are
usually too subjective (or controversial) to allow clear and explicit analysis, and possibly
even evaluation Of planning decisions that could affect them. If others agree that my
Residential Class Indicators are sufficiently accurate and useful, their application could

allow certain types of difficult planning decisions to be conceptualized in a clearer and

more objective fashion. Sociologists interested in measuring social class are provided
with an extra indicator that can be used both to classify their subjects of study within a
stratification framework, and to make class distinctions between subjects who, by other
measures, had been assigned to a similar class despite obvious differences in lifestyle,

affiliation, and consumption habits.

2. BACKGROUND INFORMATION AND KEY CONCEPTS
The Concept of Class
Having a background in sociology, I have long been aware of the concept of
social class, and this concept (along with the related one of social stratification) has
guided many aspects of my research. The idea of social class is not completely without
controversy. It is a fact that not everyone in our society enjoys the same quality of work,
income, status and lifestyle (this is called social differentiation—see Kerbo 10). Exactly
how to best measure the ways that people live differently, and what the overall
significance of these differences is, is a subject that is still much debated. It might be
said that the whole concept revolves around attempts to judge other people, by trying to
decide what standards of judgement best match societal/cultural standards (that
themselves are subject to criticism for favoring some at the expense of others—Kerbo
155-158). The endeavor makes sense from a sociological view as a means of reducing
the enormous complexities of socioeconomic relations in modern Society into a model
that is relatively straightforward and comprehensible (if somewhat lacking in predictive
power as a result). The debates about class and measurement of inequality have in recent
decades generally taken the form of using and defending one of two general positions,
which I will call the class and the stratification approaches. I will summarize and
interpret these two predominant approaches to measuring inequality in American society,
then discuss how these have informed and affected my research (based mainly on Kerbo
90-95, 128-153, 155-158).
What I shall call the class View of social inequality emphasizes distinct social

groups ("classes") with differing and often conflicting needs and goals. People in a

society, according to this view, can be generally classified according to a number of
characteristics that approximate how much value or power they have in that society
(Kerbo 12-14, Gilbert and Kahl 16-18). It is believed in this view that certain
characteristics, such as one's wealth, education and occupation, tend to correspond with
each other, and are of very great significance in determining how an individual is treated
by others, the rewards that are given to (or withheld from) him or her, the subculture that
the person operates within, and the sorts of "life-chances" that a person has. "Life-
chances" may be viewed not only as a person's chances for survival with good health and
nutrition, but also in terms of the number and quality of choices available to the person
through his or her life, some of which may be deemed to assist the person in achieving
. "success" as it is defined by the culture, and passing such benefitson to his or her family
and offspring (Gilbert and Kahl 2-3).

I It becomes apparent as these concepts of inequality are explored that there is no
- clear, undisputed means of categorizing every‘person as a part of one or another distinct
class groups (Gilbert and Kahl 46-47, 78-82).' Rather, there are categories of people
whom most would agree belong at either end of the spectrum of inequality, and a vast
majority of persons who have conflicting statuses or interests and therefore must be
classed somewhere in the middle (Kerbo 192, 281). There are many suggested
characteristics that can be used to judge others (see Gilbert and Kahl 12-15, 37-3 8), but
few people in society rate at either extreme on all of these characteristics. General
categories of classes in this classification scheme could be called lower class, working
class, middle class, and upper class, although the exact number of categories can vary

quite a bit (Gilbert and Kahl 28-31). Such groups are presumed to have certain

tendencies in terms of their cultural ideals, spending habits, and other characteristics
generally related to the class concepts mentioned before (Kerbo 290-291, 318-322,
Gilbert and Kahl 112). In popular conceptions of class, each category tends to have
stereotypical features, and persons tend to be classified according to_ the extent to which
they match these class stereotypes (Gilbert and Kahl 306-307). For example, a lower
working-class person would be stereotypically expected to have at most a high school
diploma (possibly less), to work in a blue-collar, heavily supervised job, which promotes
in the worker a culture of gruffness and latent hostility (Ehrenreich 107-121). Whenever
such a person is indeed observed, the stereotype, and therefore the presumed validity of
the class concepts, may be reinforced in those who utilize a class-based model of society
(examples Of this kind of conceptualization can be clearly seen in'Fussell 29-50). The
concept of class is indeed very pertinent when certain patterns are observed that do match
the pre-defined types. As I will discuss later, however, if our goal is to understand,
through classification, the goals, attitudes, and lifestyles of all persons in a country (not
just point to those who neatly fit into categories), there are a number of problems with
class concepts that have yet to be resolved (Kerbo 177-185, 192, Gilbert and Kahl 241-
244)

To illustrate how the concept of class is applied, a small category of people can be
defined who on every reasonable measure are "lower class." A good example of this
would be those who are impoverished, of inferiOr intelligence and health, poorly
educated, unemployed, and a part of some countercultural group that is Opposed by
powerful societal forces (such as illegal immigrants in prison for serious crimes).

Practically any social scientist can be expected to agree that those in such a category,

meeting all the mentioned criteria, are indisputably lower-class. At the other end of the
inequality spectrum, consider someone who is very rich, highly educated, in charge of
many other people (such as through an important corporate or political structure), and
held in high regard by large numbers of people for epitomizing the qualities that are
predominantly deemed by their culture to be virtuous. Such a person can clearly be
expected to be judged as upper-class by a consensus of social scientists. But what of the
large majority of people who are Somewhere in the middle and have some qualities that
will be judged as higher-class, but at the same time, other characteristics that can be
judged as lower-class? (Gilbert and Kahl 16-17, 25-26, 308-310)

What I call the stratification approach to American inequality tends not to attach
discrete labels to categories of people, and instead focus on distinct characteristics such

as income or amount of education, with which persons or households may be placed at an

' approximate point in a continuum. Besides being easy to measure, such variables

‘ generally are of a type that do not Split into a few clear categories, but are instead based

on a person's comparative position in that continuum of values for each variable of
interest (Kerbo 177-183). Any class assessment deriving from this conception of social
stratification would be based on some guess as to the relative importance of each
variable, rather than how well each case fits a stereotypical class image. A fellow who
grew up in a fairly wealthy family,‘but has a temporary job as a phone solicitor while he
attends pre-law classes at a local cOmmunity college, is difficult to place in a distinct-
class, but can be judged on measurements of several distinct variables—an income of
$16,000 per year, 13 years of formal education completed, and a job whose status is

extremely low (according to at least one survey of occupational status—see Kerbo 181 or

Gilbert and Kahl 40). In this sense, then, most Americans are "middle class" because
there are usually some factors on which they can be judged as somewhere other than the
extreme top or bottom (see also Gilbert and Kahl 234-235).

The "class" view of inequality fits a critical, conflict-based view of society
because its distinct class categories are usually applied in a context that emphasizes the
qualitative inferiority of one class and superiority of another, and therefore points to a
struggle between such classes over the distribution of rewards in society (Kerbo 90-95).
The multitude of potentially ambiguous or conflicting indicators of societal position are
subsumed into a few basic class groupings that are designed to be hierarchically ordered.
The "lower middle-class" office worker, for example, is judged in this view to be better-
off than the "upper working-class" skilled blue-collar tradesman. "

The "stratification" view of inequality is less inherently critical of society because
it tends to be unconcerned with judging the motivations underlying the distribution of
benefits that are observed (Kerbo 90-95, 378-3 84). Rather, it seeks to describe, in
quantitative terms, the relative extent of such inequality, or the comparative position of
an individual within a statistical distribution of benefits. This distribution of benefits is
generally attributed to market forces of various kinds, with the effect that the nature and
fairness of the status quo is left unchallenged. The finding of quantitative aspects of
distributed benefits allows a hierarchical comparison of socioeconomic position, but
generally only for that particular aspect that was chosen for measurement and comparison
(for an example, see Kerbo 181-185). If incOme is chosen for comparison, we may find
that a blue-collar tradesman earns significantly more money than a clerical office worker

(Kerbo 32). If occupational injury and mortality rates are measured, the office worker's

lower rate of injury may rate him or her as higher-class (Kerbo 288-290). How to resolve
conflicts between contradictory statuses does not seem to have been worked out at
present—judgements will vary with the goals and biases of the observer (for examples,
see Gilbert and Kahl 28-31, 37, 43-47, 241-244, Kerbo 190-193). An approach from
economics could attempt to quantify the medical costs of the blue-collar injuries and
compare job incomes only after subtracting the respective costs of occupational injuries.
An economist might claim that good and bad aspects of a job will be reflected in its pay,
so that more demanding jobs are matched with higher remuneration. One of the virtues
of the class approach to inequality is that by not being distracted by the politically
suggestive aspects of limited or single indicators, it can demonstrate that the preceding
kind of claim is freq'uently not true. Rather, many of the lowest paid jobs are also
dangerous and in many other ways comparatively undesirable. These necessary but
undesirable jobs are frequently filled, not throngh the raising of pay rates as the simplistic
economic View would suggest, but by having large numbers of people placed in positions
where they are unable to earn needed money in any other legitimate way except by taking
these undesirable jobs (Kerbo 324-327). Thus, it is the rule of a limited supply of (but
high demand for) jobs that described this, rather than the "fair exchange" of pay for
equivalent work, although a minimum wage and government transfers may mitigate the
pain of such job-market conditions in some instances (Kerbo 307-310, 330-340, Gilbert
and Kahl 278-280).

In this research project, while recognizing the scientific and philosophical
strengths and contributions of both perspectives, I focus mainly on the problem of finding

an accurate measurement of one aspect of inequality—the quality of residential areas——

10

rather than the use of such measures in SOCietal criticism. In this, I share the approach of
the stratification view of inequality. However, my chosen variable to be measured has a '
number of characteristics that overlap with the "class grouping" view of inequality, and
my measures may later be founduseful in societal criticism. I will explain some Of the
uses and theoretical implications of my research toward the end of this paper. First, other
important concepts must be introduced, and a history and description of how my
measurements evolved and can be applied will then be given.
The Concept of Neighborhood Quality

In urban planning and geography, the concept of neighborhood quality arises
nearly as much as the concept of class does in sociology. The whole point of economic
and community development, and redevelopment, is to improve in some tangible way the
socioeconomic conditions of an area. Housing should be of decent quality, and located to
allow access to jobs, shopping, and urban services (So and Getzels 363). Measurement
of irnprovement in an area's quality might involve the number of new jobs, new housing
units, or increased housing values. Neighborhood preservation has the goal of slowing or
preventing the decline of an area due to physical aging and weathering of the housing
stock, conflicting land use goals, and economic/demographic changes in the area.
Fortunately, the goals are a bit less controversial on this subject among planners than
those of class analysis are among sociologists. Planning controversies tend to arise about
how best to achieve goals that most believe to be laudable.

The analysis of the subject of neighborhood quality is also a bit more
straightforward. It is cormnonly agreed that there is a competitive market for land, and

that actors in this market are presumed to be acting in their own self-interests, with the

11

goal of urban planners identified as that of anticipating and resolving conflicts that may
arise as a result of individuals' and groups' competing goals for the land and its
environment (SO and Getzels 71-78, 309, 330-332). In terms of measurement procedures,
one of the simplest and most straightforward approaches to the assessment of residential
quality is to consider the market value of housing units in an area. Given the basic
economic assumption that all positive and negative features of each housing unit will be
reflected in its market value, such values act as a kind of weighted average Of the sum of
good and bad elements affecting its use for residential purposes (Anderson and
Funderbunk 137-144). As when a crude economic approach was applied to low-paying
jobs, however (in the preceding section on "class" and sociology), there is a corollary to
be found here in that some housing is inherently undesirable and yet finds inhabitants
willing to pay disproportionately high amounts for it because they are in a position where
alternatives are unavailable (Harvey 548-549). Thus, supply of quality housing tends to
be low when compared to the demand for it, and the existence and enforcement of
construction codes, among other regulations, keeps the price of available housing from
going too low (Harvey 558-559, Jacobs 419-420).

Many conditions may affect the price of housing. Of course there are many
endogenous factors that relate purely to the structure itself—number of bedrooms,
bathrooms, complete plumbing and heating facilities, square footage, construction
quality, and so on. These features may be overridden by other, exogenous ones that have
nothing to do with the property itself—taxation rates, school district, distance from
necessities or amenities, and environmental conditions (So and Getzels 311).

Environmental conditions may be physical, as in the exposure Of the structure to flooding

 

la

12

or pollution hazards. They may also be sociocultural, when the norms and standards of
an area and its inhabitants run counter to those valued by the housing shopper. In this
latter category may be placed concerns about image (including so-called "visual
pollution"), noise, crowding, crime, or ethnic/cultural differences (Anderson and
Funderbunk 139-140).

The point of my research is to attempt to measure, using readily-available census
data, the quality of residential areas within all parts Of an urbanized area, and connect
such measures with broader patterns of inequality and stratification in that area. Just as
my concern with social stratification is to be able to evaluate all residents of an urban
area, my goal in assessing residential quality is to evaluate the status of all inhabitants of
the area—not just homeowners. My model of residential quality therefore includes
measures to evaluate those living in rental units and in grorip quarters. The focus is on
the people as well as the housing. These distinctions, and the merging of residential
quality with sociological class analysis, show that I am addressing the topic somewhat
differently from real estate analySts.

Sociologists and urban geographers frequently employ models that are too
demanding or time-consuming for routine use by planners. (See Ley 75-77 for an
overviewof social area analysis and factorial ecology. While social area analysis has
many similarities to my approach, factorial ecology assumes a comfort with advanced
techniques of multivariate statistical analysis. Both require extensive time to calculate or
set up in a computer.) My model provides an approach that I feel is informative and can

be readily used, with just a bit of study, by planners and researchers who have only a

“O

13

basic knowledge of algebra and statistics, and access to common census data through a
library or computerized source.
The Traditional Class Indicators and Unit of Analysis.

In studies of stratification and inequality, three good indicators of social class, or
"socioeconomic status," are a person's income, occupation, and education (Hess,
Markson, and Stein 182-183). Although I have said "person," it is generally the
household that is the unit Of analysis for stratification studies. I acknowledge that it is far
easier to study household characteristics than individual characteristics, but I do not share
the view that it is intrinsically more correct to examine the stratification of households
rather than individuals in modern society. This could easily lead to a separate paper of its
own (see‘ Kerbo 192 as this point relates to studies of the status of‘women, so I will
merely give a number Of criticisms as a way of pointing out some inadequacies of
household-based conceptions of social class. I feel these flaws are worth pointing out
even though I ultimately had to abandon using an individual view of class in the
development of my model, due to the fact that most available data is patterned to be
useful only for a household analysis.

The traditional view of social stratification basically assmnes that all members of
a household are of the same social class. I believe this assumption is probably fine in
many instances. For example, a traditional agrarian system has minimal measurable
economic links between relatively self-sufficient farming households and the merchant,
activities of the towns (Wallerstein 24). As industrialization proceeded and capitalist
systems spread throughout Europe, America, and elsewhere, the number of measurable

monetary interactions between a household and other parts of society increased

l4

continually. New forms of taxation, compulsory education, and the increasing
specialization of new production processes had the effect of transforming old economic
systems (based on divisions of labor within relatively autonomous households, and light
exchange networks between such households and towns/ governments) into new ones in
which previously unmeasured household production became commodified (that is,
became organized in a way that allowed for market valuation and exchanges of those
goods and services—see Wallerstein 13-43). The direct labor value of children within
the household, for example, was to be challenged by the abstract value of increases in the
marketability of their labor within a tax-supported and compulsory schOOling system.
The labor costs of household tasks performed domestically by household members could
be compared with the prices of modern devices such as dishwashers and drying
machines, and the lessening of time required for household labor allowed for additional
family labors to be sold in broader markets for wage remuneration (Harvey 5 54—5 55).
By the 19505 and 19605, certain econOmic and cultural changes became clear
' which I believe demonstrate the desirability of shifting to an individual-level analysis of
stratification rather than a household-based one. Rising numbers of houSeholds had two
working parents, and so the "family wage" concept started to be reshaped (Gilbert and _
Kahl 106). The whole assumption that one parent (the father) functions as the head of the
household became harder to sustain, as there were too many obvious cases where a
woman's income was much greater than her husbands, and her job more prestigious and
highly-skilled. Although married couples became more likely to include comparable
OCCupational status as one of their marriage criteria, disparities are still common and

complicate class analysis at a household level (Gilbert and Kahl 120-125, 237-239).

’3

15

Also, the breaking down of fixed gender roles and stereotypes freed many women

and men from the traditionally rigid life-cycle concepts. Many married couples pursued

' separate careers and began choosing not to have children. Women did not automatically

have to seek marriage and mothering roles (see Gilbert and Kahl 74-76 on "pink-collar"
jobs). Many couples alsodecided that marriage itself was not vital to their plans.
Female-headed households have increased (Gilbert and Kahl 289-291). The gender-
based restrictions on so many parts of society and the workforce began toidisappear.
Glass ceilings in many professions were pushed up or broken through. Female access to
higher education became equal to (and in some measures now exceeds) that of men
(Hess, Markson, and Stein 211-218).

We have also observed the rise and unprecedented prolifefation of youth
subculture and niche markets. One of the big changes in the 1950s and 19608, this trend
has showed no signs of stopping. Young people under age 18 have more spending
money than ever, and entertainment, recreational, and fashion markets and trends aimed
at youth have expanded to include older persOns as these youths age. The increasing
separation between youth activities and adult ones has enhanced the creation and
maintenance of distinct youth subcultures with each generation, which in many ways
seem to be just as distinct as any cultural differences presumed to separate the traditional
social classes, and which seems most closely affiliated with lower-class values
(Ehrenreich 91-96). On the flip side, we also have seen a significant rise in the
percentage of elderly persons, and in many cases, a lowering of the retirement age,

resulting in increases in recreational spending and political power for this group as well.

16

Many traditional views of social class ignore the strong correlation between age
and the traditional class indicators of education, occupation, and income, assuming that
children are trained in such a way that they can be considered that same class as their 1
parents (or at least, parent of the same gender). Classic studies of status attainment
(Kerbo 369-373) seem to have ignored the effects of age on occupational status (see
Hess, Markson, and Stein 228-229 for an overview of age-based inequalities). While it is
true that occupational inheritance (or at least a correlation between a parent and child's
occupational status, education and income) is a demonstrable feature of the stratification
system, the number of deviations from this expected inheritance is so large that it must
not be ignored (Kerbo 349-354). In addition, even if one's ultimately expected social
class is at a certain high level, in many cases, this is not achieved instantly, but must be
worked toward. College students of lower middle class (or below) origins, for example,
typically have lower-working class jobs or even live in poverty for significant periods of
time while working to attain middle-class or professional status (Ehrenreich 75-78). '
There is no agreement as to what aspects of class positiOn should be emphasized at any
given time and which should be ignored. The poor student may complain that she is
lower class, because of her low~ income and menial job while in college, but her life-
chances are significantly better than the poor person in that job who is not in college.
This shows the significance of the educational component of class—in many cases, the
student would have a much better job, were it not for the commitment to spend so much
time and money in pursuit of educational advancement. Is the student lower class
(income), working class (occupation), or middle class (education), or some average of the

three? If the student is considered middle class in this example because Of her life-

-l7

chances, is this still a true assessment if she dies before achieving true middle class
status—especially if from an occupational injury or lack of money or insurance for health
care? Since college completion is of far greater importance than mere attendance (Kerbo
376) there is no certainty that a currently enrolled student will Obtain and successfully use
a degree, it would seem logically invalid to place too much weight on assessments of
future attainment, when current lifestyle conditions are so different in comparison.

Although I have been presenting a case for an individual-level analysis of
stratification, critical data on income is provided mainly for households. The amount of
income available to individual members of a household generally cannot be determined
by this data source. Income information is typically collected for entire households or
families, and the way that this income is distributed within households—is not detailed. As
part of my model's development, I proposed a measure based on per-capita calculations,
but decided to abandon it as it became clear that it was too sensitive to household sizes
(similar to the income analysis problems reported in Hess, Markson, and Stein 181). It
should also be noted that many of my critiques of household-level analysis were rooted in
a framework presented by a school of thought that insists that a household-level of
analysis is the most appropriate one to use (Wallerstein 23-26, Wallerstein and Smith
234-252). It is possible that my reluCtance to accept household-level analysis is rooted in
biases from my own backgroundand class position (Gilbert and Kahl 120-125).

A Class Indicator Based on Residential Location

I have mentioned three traditional indicators of social class as being income,

education, and occupation, all of which have means of measurement and which to some

degree are present in decennial census data (although occupation is typically measured in

l8

terms of occupational prestige). I will discuss these a bit, and their limits, and argue for
the consideration of my suggested new indicator, which is based on residential location,
in future stratification studies and urban analysis. The Residential Class Indicator (RC1)
provides an assessment of residential quality which otherwise lacks a straightforward
measure from a single census variable. The importance of residential location for class
analysis will also be explained.

Income is an irnportant variable for indicating class or status position because it
can be used to purchase many of the other indicators, whether leisure and cultural
pursuits, education, political influence, or even, through the presence of a surplus that
allows for investments, more income! Part of the lifestyle component of class is
considered to be that of consumption—the choices of how one spends one's money (see
Harvey 553-556, Veblen 68-101, 133-139, and all of Fussell). Education and income are
measured directly by the decennial census questionnaire, and consumption may be
assumed to correlate with income. The correlation may not be ideal, however, since we
do not know from the income data how much of a person's (or family’s or household's)
reported income is available for spending or investment. A College graduate may
technically have a high income, except that most of it in some cases may be unavailable
due to large debts that have accumulated. A more accurate class indicator would
I probably be wealth, which is not asked about by the census bureau (Kerbo 28—3 1, 38-40,
Gilbert and Kahl 101-104). A high-class person may technically have no income at all
during a given year because he is able to live off of accumulated wealth. In such
instances, my proposed indicator of residential quality, based on available measures of

housing values and housing expenditures, would seem to be a better indicator of that

 

l9

person's lifestyle and overall wealth. Housing data is also more visible and subject to
verification, and so is less prone to false answers than a question on income.

Some views of class hold that affiliation is very important, not only from the
status gained or lost by associatiOn with one's acquaintances, friends, and family, but also
in an economic or "life-chances" sense, when someone is assisted in their needs or
endeavors by gifts or favors from wealthier or more influential family or friends.
Education and occupation may be considered correlates of class affiliation, but here also
is an excellent example of where my indicator of residential location would be helpful in
assessing this aspect of a person's class position. The place where a person lives suggests
some affiliation or similarity with others who reside in that area. There are many
instances where choices of residential location are shaped by where one's relatives and
friends live (Gilbert and Kahl 135-137). In addition, class interests may be shaped by
residential location (Harvey 559-560).

In some cases, a person's other census-measured characteristics may not match his
‘ or her true class preferences and affiliations. For example, ’a 'mOderate income person
may spend an unusually largeportion of her money to afford housing in an area that she
considers to be of good quality, based on her background and tastes. Conversely, a
wealthy man who wishes to disdain a rich lifestyle and continue to affiliate with the
working folk he grew Up with may stay in a run-down old house near his old factory.

The choice of where to live can be considered an important indicator of culture,
affiliation, and consumption patterns for a person, and therefore my goal was to measure

patterns of housing inequalities which would reflect this.

IO

20

Other variables that have been considered a part of class standing include political
power and class consciousness (Turner 220-228, Hess, Markson, and Stein 175, 177, 184,
Gilbert and Kahl 13-14), authority over the actions of self and others (Kerbo 112-116),
and socialization and family background (Gilbert and Kahl 13, Roberts 238, Hartigan 8).
These variables in some ways relate to those measured in the census, but are not directly
dealt with in my research because of the difficulty of clearly linking census variables to
them.

There are numerous references to the importance of residential location in one's
class position, in both popular and theoretical literature on class and urban areas. Often,
the subject is mentioned casually as a matter of common knowledge, but is seen as a true
condition from a number of distinct interpretations of its significahce. Those who have
studied poverty note the reality of zones where the poor are concentrated in much higher
proportions than elsewhere (Kerbo 326, Roberts 237). An analysis of the richer areas
reveals areas of exclusionary zoning, gated communities, and "exclusiveness," (So and
Getzels 48, 51, 282-283, Fussell 76-83, Gilbert and Kahl 132). A broader approach
shows aspects of differentiation between residential locations throughout the class strata
(Palen 154-158, 197-199, Ley 55-92, Muller 63-65). One school Of social theory fits
residential segregation into a broader framework of class conflict (Harvey 560).

More empirical and complex analyses have develOped in the ecological school of
class analysis (Kerbo 182). While the classic stratification studies of Robert and Helen
Lynd ("Middletown") and Lloyd Warner ("Yankee City") focused on small communities,
a study by Coleman and Rainwater in the 19705 assessed class positions for Boston and

Kansas City (Kerbo 126-127, Gilbert and Kahl 33-38). In urban studies, the "Chicago

21

school" researchers had begun extensive analysis of residential segregation patterns
focusing on race and ethnicity (Ley 60-61). Following these initial studies were many
others which included numerous indicators of class position or socioeconomic status in
their geographic analyses, so as to identify numerous types of neighborhoods or, at a
larger level, "social areas" (Ley 62-67). Social area analysis included a measure of
"social rank" based on occupation and education (Palen 104-106, Ley 75-76). This
marked a significant methodological step forward from initial crude and very generalized
descriptive modelsof urban land use, such as the concentric zone, sector, and multiple
nuclei models (Hess, Markson, and Stein 538-539, Palen 90-103, Ley 72-75). For
assessing patterns of change, the concepts of invasion and succession were introduced,-
followed by more sophisticated concepts such as filtering and neighborhood development
cycles (Hess, Markson, and Stein 539, Palen 87-90, Ley 248-268).

My research idea started from a considerationof the simple ecological approach
of mapping single census variables for block groups or tracts throughout urban areas. (I
considered factor analysis to be far too demanding in its time requirements, and too
difficult more most Urban planners to use because of the number of variables considered
and the statistical technique used to analyze them.) Urban areas are examined because
census data for them is plentiful and linked with pre-defined geographic areas (tracts and
block groups) which are delineated finely enough to enable useful spatial studies of
residential distributions. Also, a fuller range of stratification is likely to be present in
urban areas than in nonurban ones, since our society's functions and culture are mainly

urban for most of its population G’alen 3-5).

‘43

22

With regard to the boundaries for a city, I consider a Contiguous built-up area of
non-rural density, oriented around centralized areas of greater density and higher land
uses, to be an economic and social whole that I call an urban area. The individual
political and corporate municipalities with fixed geographic boundaries (in Michigan,
these can be either cities, villages, or townships) which compose the urban area as a
whole are certainly of interest for any urban study, but‘in my opinion these boundaries
are overly distracting for most researchers. For my purposes of residential class analysis,
discussion of central city/suburban differences are wholly unsatisfying and much too
imprecise for modeling inequalities. I agree completely with David Rusk's attitude that
"the real city is the total metrOpolitan area" (Rusk 5, 7). My own research on
contemporary urban areas, and a consideration of the locational tr"ends over the last
several decades (for example, Muller 62-82) has led me to conclude that so-called
suburbs are in most cases now functionally, economically, and visually merely part of an
urban area as a whole. Instead of a city annexing adjacent areas as it grows,
contemporary cities now grow across political boundaries, with the areas in outlying
jurisdictions basically serving the same functions as the fiinge of the central city in the
era when it was politically self-contained. The geographic distributiOn of land uses is
merely seen on a larger scale now, and in many cases, the plaCement of cities' corporate
limits seems quite arbitrary in relation to its effect on actual land uses. Arr observer in the
field would frequently find it impossible to tell where city ends and "subur " begins,-

except when there are actual signs posted to mark the "transition." In this research, I find

i the census-defined "urbanized area" to be a fairly good demarcation of the functional

boundaries of an urban area (see Palen 115-116 for definitions of urban area measures).

23

I tend to reject using the (commonly used) Metropolitan Statistical Area (MSA)
values as an area of analysis. This is because I have been dissatisfied with the way that
the MSA is defined in terms of entire counties, which is a particularly crude geographic
delineation when dealing with cities of only medium-size. Its use leads, for example, to
the claim that the Lansing "area" was larger in 1990 than the Flint "area," which any
detailed comparison of the two areas would Show to be untrue. The MSA includes many
relatively unassociated rural areas as being a part of the metrOpolitan Lansing area, even
when their distance has caused most land valuation effects from the city to have
dissipated (following the distance decay principle in economic geography—see Haynes
and Fotheringham 12-13, 15). I did not want a measure whose values were thrown
substantially off by the inclusion of distant small cities, such as Olivet, and the large
agricultural areas throughout Lansing's three associated MSA counties. I also noted the
crudeness of the measure in the case where Battle Creek is assigned Calhoun County as
its MSA, and Kalamazoo is assigned Kalamazoo County as its MSA, even though the
city of Battle Creek is contiguous with the boundary of Kalamazoo County. Rather than
being a part of the same MSA (as in the case of Saginaw-Bay City-Midland), they are
treated as separate; while included in the Battle Creek MSA is a moderate-sized city like
Albion, which is 25 miles away (and therefore has a much smaller " gravitational" force of
interaction with Battle Creek than Kalamazoo's urbanized area, which is 23 miles away
and has more than a dozen times the population).

Fora while I was impressed with the Rand-McNally Corporation's "Ranally
Metro Area" (RMA), which comprises minor civil divisions rather than entire counties.

It defined Lansing's RMA as the two central cities, plus some dozen townships (and the

24

small cities and villages within them) in which half or more of the residents were
assessed as commuting to the central area, and therefore economically and socially
associated with it. When I found that the census bureau's "Urbanized Area" (UA)
measure more compactly defined the Lansing-East Lansing area according to contiguous
and associated urban land uses, regardless of local political boundaries, I have preferred it
ever since. Later I will describe how to apply the model to any urban area, regardless of
how it is defined, or how to use an areas custom-defined by the researcher (called a "user-

defined" area, abbreviated herein as "UDA").

_b
dd

25

3. THE RCI MODEL EXPLAINED
The Origin of this Model of Residential Inequality

My contemplation of residential class indicators pre-dates the formation of this
formal research project regarding them, and in some ways even pre-dates my formal
training as an urban planner. Therefore, some of the evolution and testing of my ideas
did not follow the classical social-scientific tradition, such as performing an initial
literature review. Many who have given a critical analysis of scientific methods have
pointed out that there is a great deal of research that did not follow a classical model, but
is then written up in a way that suggests that it had (Mills 56-5 8, 69-71, Merton 4-7). In
following the recommendations of such analysts for greater candor, I will present in this
section a description of how my creative process for the RCI model actually occurred.
The literature review proper was given in the preceding section. The chronological
descriptions in this section will help to explain the many aspects of the RC1 model in a
more approachable fashion (paralleling how I actually determined them), but will also
allow this paper to be of use to those who study methods of scientific research, and the
creative aspect Of the research process. A possible drawback to this chronological
approach is that there are a number of ideas included in my descriptions which do not
ultimately figure into the RC1 model I am recommending. The inclusion of these extra
ideas may be distracting for some readers, but they have a value for facilitating (1) a
critique of the model, (2) the correction of any flaws found in my reasoning and research
process, and (3) the identification of areas for further research on this topic (described in

a section at the end of this paper).

26

During my readings and classes in sociology, urban planning, and related
disciplines such as geography and economics, I noted that the idea of neighborhood
quality often arose as a useful concept, but was rarely dealt with in an objectively
measurable sense. Although such measures exist (references are in the previous section),
they can be too difficult for most people to calculate, and too time-cOnsuming for most
planners to conveniently research. By contrast, the RC1 model allows the calculation of
residential inequalities for an entire urban area in just a few hours, for those who have
spreadsheet software, computer-formatted census information (or a link to the census
intemet site), and a map of cenSus tracts or block groups (also obtainable over the intemet
from http://factfinder.census.gov/ ). Even if none of this equipment is available, a researcher
can go to census referencematerials in a library, locate a few key ,values for the urban
area, and then look up and calculate values for individual tracts, using only a couple
minutes per tract if a hand calculator is used.

The concepts that were molded into my RCI model date back nearly ten years and
originated in my use of census data to help me select relatively affordable locations to
rent apartments in the Lansing-East Lansing Urbanized Area. Initially, I examined
traditional social indicators such as the percentage of residents in poverty, and per capita
income. However, since I was mainly looking for affordable but decent-quality housing,
my concern with such sOcial indicators became of secondary importance, although I
noted that an apartment could be rented in many different types of neighborhoods in the
Lansing-East Lansing area at prices that were not so different from each other. This is

especially true for those who can split the cost of a housing unit with another person, a

27

Option that enabled me, as a young college student, to live near one of the wealthiest
subdivisions in the whole area (in southern Okemos).

Over the years, I noticed more about the relationship between the social
characteristics of an area and the housing values the census lists for that area. When I
sought to rent an apartment of my own, I mapped out the values for some census
variables for the entire urbanized area. I found a section near downtown Lansing where
the apartments were very affordable, and the housing values were not too low. There was
a sizable percentage of people in poverty in that tract, but Since almost everyone there
had completed high school, I figured this merely meant that they were students like me.
It soon became clear that my naive review of a few basic indicators had some severe
weaknesses. Although the apartment complex itself that I lived in. was spacious and an
excellent deal for the money, the nature of the area accounted for its relative cheapness. I
had to get used to the sound Of gunshots and nearby domestic violence, and my daily
walks to the bus stop often included hostile or undesirable encounters. When I entered
graduate school and gladly moved to a nicer area close to campus, I began again to
consider how census data might be employed to assess neighborhood quality.

My previous mistake was to exaggerate the importance of the educational
component, and to not have noticed that the fairly decent housing values applied only to
the nine percent of units that were owner-occupied. I began to tinker withways to make
a combined measure of neighborhood quality, for there were too many rental units in.
most areas to make housing values alone a reliable indicator of residential quality. I
wanted to make some sort of an index by combining variables that were not measured in

the same terms. .

28

The actual mathematical formulation of the RC1 model was inspired by the
various ratio-based measures I had seen in the Urban Planning profession, such as
location quotients and shift-share analysis (as in Klosterman 113-186). By using and
extensively critiquing such analytic techniques, I felt I had acquired a good understanding
of how "simple division," the use of ratios, could be productively applied to the
evaluation of social data. Needing a standard to which census tract values could be
compared through ratio techniques, I chose the median values for the urbanized area——
contract rents for rental units, and housing values for owner-occupied units. I made little
ratios for each type of unit which compared its median tract values to those of the whole
Lansing-East Lansing area, and added weighting factors that added up to unity so that
when the ratios for each housing type were added tOgether, the result would appear in the
same form that the original ratios had been. Multiplying by 100 put the ratio results into
a familiar, percentage-style range—the tract value is expressed as a percentage of the
overall UA (or UDA) value Users of the RC1 model would therefore only have to
develop a comfort with one rating scale to begin interpreting the results of their
calculations.

It became clear that some sort of adjustment factor was needed so that renter-
occupied (RC) units that matched the median rent for the urbanized area (U A) would not
be counted as equal to owner-occupied (00) units that matched the median value of the
urbanized area, since RC units tend to be socioeconomically much different than 00.
units. Intuitively (based on my general knowledge of the area from over 10 years of

living there, and about 5 years of delivering pizzas), I estimated a rental adjustment factor

29

of 0.6, which I now believe to have been a very good first guess. I put together an initial
RCI formula, which looked like this:

(Equation 1) Prototype 2-Component RCI Model

RCIt = %OOhht[ MdV’ )+%R0hht(—M 0.6 (100)
Md MdCR

 

Vua ua

where "RCIt" means the residential class indicator Of the tract, "%OOhht" refers to the
number of owner-occupied housing units in the tract divided by the total number of
housing units in the tract, and %ROhht is similarly the proportion of renter-occupied
housing units in the tract, the number of rental units divided by the'total number of units.
These proportions serve as weights to balance the ratios of the Mth (median value of
owner-occupied units in the tract) over the Mqua (median value-Of owner-occupied
units in the whole urbanized area) and the MdCRt (median contract rent of rental-
occupied units in the tract) over the MdCRua (median contract rent of rental-occupied
units in the whole urbanized area). Thus, each type of housing is compared with the
urbanized area standard, weighted, and summed, then multiplied by 100 to convert the
rating to a more ordinary-feeling, percentage-style number. I call it a model at this point
because the ratings are intended to be mapped out or plotted by a Geographic Information
System (GIS) for an entire urbanized area to provide a context with which to assess the
position of any given tract or small area. The numbers gain meaning from a comparative
assessment of how a tract or block grotrp rating compares with others, and to the larger
UA or UDA area used as a standard of comparison.

The model seemed to hold up pretty well in describing the appearance and "feel"
of actual areas I saw around Lansing, and so I mapped out large portions of the Detroit

UA and the Grand Rapids UA using it. When I explored those areas (I grew up around

30

Detroit and was already a bit familiar with it) I was very pleased at this simple model's
power to predict the type Of neighborhood I would see, mile after mile—wealthy,
average, run-down, or frightening!

I had some others evaluate what I had done (not an unbiased group, perhaps, but
producing some good initial feedback on the model). My stepfather, who had grown up
in Detroit, perused a tract map of that city based on my prototype RCI equation, and
found the ratings to be accurate—except for the area around Wayne State University. I
then applied my model to a block group analysis of the city of Evanston, Illinois (part of
the Chicago UA), made a map, and it evaluated by a couple of friends who had lived
there for almost a decade. The feedback was generally favorable, but also confurned for
me that the ratings in university areas were problematic. I requested from a friend in the
Buffalo UA a few locations that he knew well that I could map out and have him
evaluate. I mapped out three areas in this fashion, with the favorable feedback that my
numbers were, "in a relative sense, spot on." He noted however, that I had not adequately
identified the fact that one particular area was not just middle class, but had sections with
mansions in it and so should have weighed in as upper middle class. As a mathematician,
he pointed out that this oversight could have arisen because I was using median values in
my assessments. I had already felt a bit uncomfortable about the mathematics of dividing
medians, but since the medians were readily available from my CD-ROM census sources,
and were producing generally good estimates, I had continued to use them. It was also
considered traditional to use medians as a measure of central tendency when dealing with
monetary figures, since the tendency of such values to be greater than zero, yet have no

fixed upper limit, skews their distributions to the right.

31

I should also note that, in following the " class " approach to inequality, I had,
through comparing my RCI ratings with field conditions, come up with estimated class
divisions that could be matched with the ratings this prototype formula produced. The

estimated class divisions at that time were as follows:

 

 

 

 

 

 

 

 

RCI ratinL Estimated "social class" category
Below 40 Lower class

40-59 Lower working class

60-89 Upper working class

90-135? Lower middle class

135? -?? Upper middle class

?? and above Upper class

 

 

I did not feel that I could accurately pinpoint the cutpoints Of the upper classes,
since there did not seem to be any exclusively upper class areas in most of the places I
studied. Areas of the Grosse Pointe cities northeast of Detroit received very high ratings
of as high as 500 (for Grosse Pointe Shores), but although this was therefore clearly
upper class, there were too few of these areas to define where the lower rating boundary
of such a classification should be placed. I had intended such categories to be helpful for
persons unfamiliar with my model to interpret and evaluate the ratings it produced.
However, I chose to keep the ratings in their original raw form rather than standardize
them within a fixed rating scale (such as from 1 to 100).

Initially, I also chose not to adjust them so that they had a clear central point, such
a rating of 100, which could be considered average. Rather, a rating of abOut 85 seemed
to be the "natural" average of the RC1 ratings, and my assignment of class categorieswas
roughly intended to duplicate the kind of class distribution figures found in past research
on American stratification (Kerbo 182, 272, Gilbert and Kahl 26, 34, 235). In an

explanatory sheet, I commented on the subjective nature of one's impressions of an area;

 

32

whereas I might feel uncomfortable living in an area rated below 50, for another person

this point might be lower, such as at a rating of 40, or higher, such as a rating of 80. I did

A not wish to try to interpret the meanings of the ratings for others, but simply provide a

measurement of neigthrhood quality that would allow the more precise expression of
such preferences. Later, as the implications for social stratification became clearer, I
developed means to standardize the ratings around 100 for ease of interpretation (the
RCI% rating, which will be explained later), and to rank areas using percentiles, for
comparison with or placement in a stratification hierarchy.

My initial feedback, and continued field observations, confirmed my impressions
that the majority of RCI ratings were indeed measuring comparative neighborhood
quality. There were two types of areas, however, that the model did not seem to work
well for. Downtown areas seemed to be rated excessively low, and where university
areas were measured, the campus area would rate extremely low, while adjacent student
areas seemed by comparison to be rated as too high. I believe that the low ratings the
model observed in downtown areas were accurate, but that my own sense of such areas
had been distracted by their important commercial and business functions. I had trouble
accepting the ratings until I recalled that they were based on residential characteristics.
My intuitive assessment Of downtowns emphasized their important and often prestigious
business uses, which often attract higher-class persons to them. I had alsofailed to take
into consideration the impact of the "interaction effects" (discussed as an area for further
research) of people travelling into the area from elsewhere and seeming to change its
character. When I took a closer look at the quality of the housing actually available in

these areas, I decided that the RC1 ratings were appropriate.

33

It was the treatment of university areas that made me determined to change the
model, after I assessed (and then moved into) an area of East Lansing known locally as
the "student ghetto" (see Ley 65). While the MSU campus itself had received ratings in
the 305 from my prototype equation, the area just north of it was rated in the 805.
Although this area was one in which "riots" had occurred, many locals dismissed the
significance of such events as merely being large student parties, which sounds innocent
enough. Their dismissive attitude was wrong, making the assumption that college
students can be treated simply as the future middle-class (again, refer to Ehrenreich 75-
78, 91-96). Although the area has some expensive and nice apartments, the majority of it
contains moderate-quality dwellings that are subdivided or shared by numerous students,
and others. (The city of East Lansing has recently passed ordinances to address this
problem by certifying each rental unit for a specific number of occupants.) There are
substantial numbers of people in the area that have no official connection with the
university, including various transient persons and some who are undergoing medical or
psychiatric treatment. Crime is quite high—at least, low-level crimes such as vandalism,
public drunkenness, noise disturbances, littering, and public urination were so frequent
that it seemed no one was concerned with enforcing the laws against them in that area.
There is also a high level of assault and theft crimes. While crime is not necessarily a
good indicator of social class, having much greater correlations with a young male ,
population than with traditional class indicators (Nettler 102-106, 113; see also Jacobs
146-148), some other information on the area is very suggestive. A full 65% of the area's
residents were classified by the census as living in poverty, and 27% lived in group

quarters when the census was taken in 1990.

 

If

34

I decided that my initial crude model was weak at dealing with these areas
because it had ignored the residential conditions of people who do not live in households.
The census records data on such persons under the category of those living in group
quarters, which includes rooming houses of 10 or more units, homeless shelters, medical,
psychiatric, and rehabilitation treatment facilities, fraternities, dormitories, and transient
persons living in street locations. I added to my model an assumption that group quarters
could be considered the lowest quality of residence, and reduced my 2-component RCI
ratings so that the percentage of inhabitants living in group quarters were effectively
counted as having a rating of zero, for weighting purposes, just as the rental-occupied
term in the prototype equation was reduced by being multiplied by 0.6. (This apparently
severe judgement to treat group quarters as a zero rating was made, in part, because the
census provides no information about the housing expenses of people living in group

quarters.) The new, 3-component equation I began using was

 

 

(Equation 2) Prototype 3-Component RCI Model
Mth Mth
RCIt: 1-‘VG t ‘VOOhht + ‘VROhht 0.6 100
( 0 Q)[( 0 )[Mqua) (0 )(MdRua)( )]( )

where %GQt refers to the number of persons in group quarters divided by the total
number of persons in the tract, and the central part of the formula is the same as in
Equation 1. The tacking on of this adjustment factor was helpful, as it reduced the rating

of Tract 41 (part of the "student ghetto") from the 803 to the 505, which I felt was much

'more appropriate (a later model treating GQ residents as higher than zero lifted the rating

into the 603). I later realized that equation 2 mixes a population proportion with

household proportions. A more mathematically consistent way to treat the group quarters

J

35

component (which produces results that are very close to Equation 2) would be to shape

the equation in the form of

 

 

(Equation 3) Variation on Prototype 3-Component RCI Model
MdVr Mth
RCIt: O‘VG t + °/OO t + ‘VRO t 0.6 100
[(0 Q) (0 p)[Mqua) (0 p)(MdRua)( )]( )

in which the %GQt term would always be equal to zero and can therefore be removed
from the equation, as in

(Equation 4) Simplification of Equation 3

 

 

Mth Mth
RCIt: 0/00 t + 0/R0 t 0.6 100
[( ° ‘0 )(Mqua) (° p )[MdRua)( )l( )

so long as %OOpt now refers to the proportion of persons in owner-occupied units in the
tract (rather than the proportion of households) and %ROpt is similarly a proportion of
persons rather than households. Equation 3 allows a value other than zero to be selected
as a reduction factor for the group quarters term, should that be determined to be more
appropriate. After all, it seems harsh to assign all nursing home or dormitory residents an
RCI value of zero, which wOuld be the same rating assigned to the homeless or
imprisoned in this model.
The Evolution of a Refined RCI Model

Once I declared this to be the subject of my master's research paper, I began to
seriously question and reconsider every component and assumption of myifirst RCI
equations. Once I gained access to modern spreadsheet software (previous calculatiOnS
were all done with paper and hand calculators) I knew that there was no longer any

reason to tolerate the simplifying but questionable aspects of the early RCI model. The

first change was to stop using statistical medians and instead use the more

36

mathematically proper statistical mean. Not only had it become clear that the median
was ignoring too much information that could affect residential quality by pulling a tract
average up or down, but another clear weakness of the median emerged as well when I
started working at a block group rather than tract level of analysis. I would find a tract
rating of, for example, 50. Then, when calculating RCI values for all of the block groups
within that tract, I would sometimes find that all of the block groups had ratings higher
than 50. The use of medians was preVenting a proper averaging of component parts
within the larger areas being analyzed. Using means instead of medians, a subdivided
area will have some parts rated above, and some below, the value of the broader area they
compose, or else all ratings will come out about the same (as in a very homogeneous
area).

I also found the surprising fact that some distributions of monetary information
are in fact skewed to the left. In most cases dealing with rents, incomes or housing
values, the median will be lower than the mean, as the mean is easily pulled up by a
relatively few cases of very high values, due to their distance from the typical value in
that area, whereas in this study, there are never any reported negative values that could
pull a mean down as easily. In quite a few instances, however, I found areas whose
median values were higher than their mean values. An example would be where the
bottom 40% of a tract is valued very low, and the top 60% has moderate reported values,
all of which are quite close to each other. The median reflects only the top 60%, whereas
the mean includes all available information on housing values—in this case summarizing
the area as being of lower quality than the median had suggested. I therefore now use the

mean in every instance, although this usually needs to be derived from census data by

37

taking aggregate values or aggregate rents and dividing these by the total number of
analytic units (usually persons or households) in the "universe" for which those
aggregates were tabulated. Thus, a mean value of owner-occupied units is derived by
taking the census figure labeled as the aggregate value Of "selected owner-occupied
units" (called the universe for that variable) and dividing it by the total number of
"selected owner-occupied units." This extra work handled fairly eaSily by the use of
spreadsheet software, once census data is entered or downloaded into it.

Next, I had to convert the crude estimate of 0.6 as an adjustment factor modifying
the renter-occupied ratio to a value that was theoretically justifiable and objectively
derivable. An assumption of my approach is that owner-occupiedunits, and their values,
are the norm against which other residential arrangements must be compared. (In one
school of thought, owner-occupancy is considered a useful norm whose promotion helps
legitimize capitalism in our society, giving a greater vested interest in the system to a
larger portion of the population—Harvey 551'.) The measurement of owner-occupied
units is expressed in terms of value, a measure which has an imperfect correlation with
income (and other class indicators), and which can be more indicative of the sorts of class
charaCteristics not always addressed by income and other census variables. For example,
housing constitutes a major component of most people's wealth, and to this extent may be
more indicative of class position than income (Gilbert and Kahl 102-103)., Rental
housing units are measured in terms of monthly rents, which are not directly comparable
to the norm of housing values observed for owner-occupied units. Similarly, group
quarters had up to this point not really had its quality measured at all, and so it was

certainly not directly comparable with the other varaibles. It was necessary to have some

38

means of fairly comparing one type of measurement with another, so as to assess all
residential types in an area of study. My estimated rental adjustment factor of 0.6 had
allowed such comparisons by treating a unit whose rent was two-thirds higher than the
mean rent in the UA as the residential class-equivalent of the average owner-occupied
housing value in the UA.

My initial assessment of group quarters was that they would never improve the
overall residential quality of an area. I had to question this assumption as well, since not
all group quarters are detrimental to an area's quality of life. There are some nursing
homes, for example, that are not only quite nice, but also very expensive (although only a
fraction of this expense may be considered to relate to housing costs and quality). I had
to review all available decennial census variables and their definitions to determine howl
could best compare measures that were expressed in very different units. I found more
detailed information on the categories of group quarters, and contemplated giving
different weights to these (so that prisons would be lower than nursing homes, for
example), but this technique would have ignored the variations between different-quality
nursing homes, or between local jails and federal prisons. Moreover, the actual values of
the weights assigned to each category of group quarters wOuld still seem quite subjective
and even arbitrary.

Finally, I found that the census provides data on per capita income .for those in
group quarters. I had hoped to find information measuring residential quality, rather than
mere income, which as I pointed out is one of the traditional variables. for class analysis,
but decided that for the group quarters term‘ of my equations, it was really the best

available census, indicator of residential quality of life. Using it could draw distinctions

39

between group quarters of differing quality, on the assumption that higher-income
persons would tend to inhabit group quarters of higher quality. Also, most areas have a
very low percentage of their residents living in group quarters, and so the use of income
as a proxy variable seemed a fair way to factor in the effects of group quarters while the
RC1 indicator as a whole remained distinctly different from a mere analysis of income
distribution. Since owner-occupied units are the standard of comparison, the income of
those in group quarters in a particular tract would be compared with the UA standard for
that type of residential arrangement, and this ratio would be weighted by a factor that
makes it comparable to owner-occupied lifestyles. With this factor needing to be based
on a comparison of income, the standard of comparison would be a proportion based on
dividing the average income for those in grOUp quarters in the UA,, by the average per

capita income for those living in owner-occupied units in the UA:

 

 

 

(Equation 5) Refined Group Quarters RCI Component
gq R C It _ pciGQt pciGQua _ pciGQt
pciGQua pciOmeua pciOmeua

where quCIt is the group quarters component of the area's RCI, pciGQt is the per capita
income of those in group quarters in a tract, pciGQua is the per capita income of those in

group quarters in the urbanized area, and pciOmeua is the per capita income of those in
owner-occupied units with a mortgage in the urbanized area.

The factor to adjust the rental-occupied term of the RC1 was similarly derived, but
rather than using the less-related variable of income, a variety of other variables were
available which had a more direct bearing on residential quality of life. I ultimately
selected " gross rent" (rather than contract rent) as the key variable for rental unit quality.

This would be compared With the "selected monthly owner costs" (SMOC) of selected

, r
Rt.

40

owner-occupied housing units, to produce a ratio from which a more empirical and
theoretically valid factor could be substituted for my initial estimate of 0.6 to adjust rental
units and allow a comparison with the owner-occupied standard. In this procedure, I
faced another choice, however, for the census information on owner-occupied units
provides not only a measure of all owner-occupied units, but also a sub-division of this
category based on whether or not there is a mortgage on the owned property. The
differences between households with a mortgage, without a mortgage, and overall, are
often substantial. I selected a comparison of owner-occupied households with a mortgage
since a household will typically progress from a rental unit to an "owned" unit with a
mortgage, rather than without one (Ley 243). Thus, the comparison of rental and owned
units is based on the ratio between monthly gross rent and selected monthly owner costs
with a mortgage (SMOCwm).

SMOCwm includes mortgage payments, property taxes, and costs of utilities and
upkeep on a home and was intended by the Census Bureau to be a good summary of the
monthly costs of home ownership. The gross rent value sums the contract rent and
average utility and related costs to give a total monthly cost for a rental unit's housing _
expenses. Comparing these two measures seemed an excellent way of relating. RO

expenses to 00 expenses, especially since the inclusion of mortgage payments and

property taxes in the SMOCwm values provides a clear link between monthly housing

costs and the preferred class-indicator variable of housing values.
Selected monthly owner costs without a mortgage (SMOCwom) are much less
than with a mortgage. Overall SMOC values are a weighted average of SMOCwom and

SMOCwm, and so are also significantly less than SMOCwm. In the case of the Lansing-

41

East Lansing UA, the ratio of mean gross rent (MGR) to overall SMOC is roughly 0.685,
which is not too far from my initial estimated rental adjustment factor of 0.6.
(Interestingly, the ratio of per capita income between all renters and owners in the UA is
an even closer 0.589.) The ratio of the MGR to the SMOCwm comes out as a lower
factor of 0.556, and was adopted as the most theoretically defensible adjustment factor.

The equation for the rental component of the RC1 is therefore

 

 

(Equation 6) Refined Rental-Occupied RCI Component
m R C It ___( MGRt ] MGRua J =[ MGRt J
MGRua SMOCwmua SMOCwmua

where roRCIt is the rental component of the RCI, MGRt is the mean gross rent in the
tract, MGRua is the mean gross rent for the urbanized area, and SMOCwmua is the
average selected monthly owner costs for "selected owner-occupied units" with a
mortgage in the urbanized area. (In my downloaded census data, I had to calculate these
average values by dividing aggregate rent and aggregate SMOC by the number of R0
and 00 units that the aggregate values had been tabulated for. Other sources of census
data may provide the values of these averages plainly and not require derivation from
aggregate values.) 3 -- 1 h

Since the owner-occupied term of the RC1 equation is essentially the same, the
new basic equation for the model is expressed as

(Equation 7) The Basic RCI 3-Component Model

RCIt = [(%GQthqRCIt) + (%ROpt)(roRCIt) +(%00pt)(ooRCIt)k100)

0r

pciGQt MGRt ) ( MVt ]
RCIt= ‘VG t + °/R0 t + °/00t 100
[( 0 Q )[pciOmeua] ( o p )(SMOCwmua ( o p) MVua ( )

 

 

 

42

where RCIt is the RCI rating for the tract (or area of analysis), %GQt, %ROt, and %OOt
are all the prOportions of persons living in the three identified kinds of residential
arrangements, MV is the mean value of owner-occupied units, MGR is mean gross rent,
SMOCwm is selected monthly owner costs with a mortgage, pci is per capita income, and
the small letters "ua" and "t" are again designating whether the measure is for the
urbanized area or tract.

I will again note that while I have selected an urbanized area as the standard of
reference, any other area may be chosen to replace it, so long as data is available for that
area and the area is one that it makes sense to compare the smaller areas to. Similarly, I
have beencalling such smaller areas "tracts" but they could also be combinations of
tracts, a block group, or any other area that is significantly smaller than the area of
comparison and for which adequate data is available. More exploration of such
variations on, and applications of, the basic RCI model will appear in a later section of
this paper, addressing areas for further research.

Although I have called it the "Basic RCI Model," I have never actually used
Equation 7, because once I had conceptualized the Basic RCI Model, I realized that I
could split the GO component into two subcomponents, each of which has per capita
income data available for it in the census. These components are institutional group 1
quarters, such as prisons and hOspitals, and noninstitutional group quarters, such as
dormitories and homeless shelters. That way I would be utilizing more of the available
data in my calculations.

The full equation, as I currently use it, is:

43

(Equation 8) The Advanced 4-Component RCI Model

RCIt = 100 x
[(% GQit )(gqiRCIt ) + (%GQnit )(gqniRCIt ) + (% ROpt )(roRCIt) -I- (% OOpt )(ooRCIt )],
or

pciGQit
pciOmeua

 

 

RCIt = (100)[(%GQit)[

]+(%Gan)( paGQm’ J4.

pciOmeua

 

 

(%ROpt )( SMggtfrizua ) + (% 00p! )[ MAI/Ta j]

with the "i" and "mi" in the group quarters components referring. to institutional and
noninstitutional group quarters, respectively. This will be the equation that is primarily
used in the next section of this paper, showing an application of the RCI model.

There is also of course, a basic 2-component version of this model, shown in
Equation 9, below, which follows the same reasoning and form as“ Equation 4 except that
arithmetic means are used in place of medians, and the arbitrary 0.6 rental adjustment
factor has been replaced. In cases where there are no group quarters in an area, the

ratings from this model will come out the same as in Equation 8.

(Equation 9) Basic 2-Component RCI Model

‘ RCIt=[(%00pt( MV’ )+(%R0pt)( MGR’ )](100)
MVua

 

 

SMOCwmua

Please note that when calculating RCI values with a spreadsheet, areas lacking values for
R0 or 00 components will need some'cell contents deleted in order for the formulas
programmed into other cells to work properly and calculate ratings for those areas. The
cells needing deletion will be those where formulas have caused error messages to
appear, such as those indicating an attempted division by zero, or the absence of a value

in some referenced cell.

44

4. APPLICATION OF THE RCI MODEL
A Sample Application in a User-Defined Area

Having explained various ways of measuring neighborhood quality, it will now be
helpful to illustrate an actual application of an RCI model. When I was developing and
testing various RCI equations for correlation with what I knew of the Lansing-East '
Lansing UA, at one point I fed eight of the mean-based equations into a spreadsheet
containing tract data for the entire UA and examined the resulting ratings to see where
and why deviations occurred. (It was this process that enabled the analysis and critique
of my model that appears in other sections of this paper. I A printout of the application of
four of these equations can be found in Appendix A.) The result was an eventual
determination that the four-component model (Equation 8) was the most theoretically
sound. It is that equation I will use in the example applying my model to an analysis of
an urban area. In Appendix B is a two-page table Showing the results of my calculations
for every census tract that is at least partly included in the Census Bureau's definition of
the Lansing-East Lansing Urbanized Area.

The fact that some tracts have areas lying outside the UA means that the
calculations for those tracts will be Slightly off from what they should be if a user-defined
area were chosen that included those tracts in their entirety. For many researchers who
simply wish to research a few tracts, it is much easier to overlook these slight differences.
(Although I consider the amount of inaccuracy introduced by using the UA figures like
this to be less than the inaccuracy of using the MSA figures with all tracts from the three-

county Lansing area, those who prefer using the MSA will enjoy the benefit that its

45

boundaries will match completely with those defined by its collection of component
census tracts.)

Most beginning or casual RCI researchers will plug in the required numbers and
. produce RCIt ratings for every tract in their area of study, then map this out using a GIS
or by writing or color-coding rating numbers onto existing census tract maps of the area.
This in itself is very informative, immediately showing the relative quality of the housing
throughout the metro area. For the most thorough and accurate study using a UA or other
irregular area, serious researchers must take note that all variables in the formulae that
refer to the entire UA should be replaced with values calculated (using a spreadsheet)
from the sums and averages of the entire area chosen for analysis, as I do for this
example. The design of the RC1 model is such thata geographic redefinition of the area
used as a standard of comparison will likely affect all the produced RCI ratings as a
result. The customized UA I created for this example differs from the census-defmed UA
only in that it consists entirely of undivided census tracts. I used a spreadsheet to
recalculate all the values. needed for RCI modeling. My custom-defined area is identified
in Appendix B as a user-defined area (U DA), and for comparison, UA figures are also
shown in an adjacent row of data beneath it.

Appendix B shows the population of each census tract, ratings for each
component of the RC1 model, a total RCI rating for the tract, and an additional rating
following it, called RCI%, which'is the ratio of the RCIt rating over the RCIua rating for
the entire defined area (the UDA). The RCI% and RCIua ratings will be explained
shortly. On the next page is a map of the Lansing-East Lansing UA, with census tracts

and RCI% ratings illustrated on it.

3H3.
nagging. "snag

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

moo. u .8: warrzm no 356 2.
Ewe: .2. 53...?
.1 Is 93.: 2. 33a HIJmT.
n u .. Jo l III $3.,-

1 0 mo 2 _ 3

32:5 xHum 2:3 «Page. msmzmo 2.! a: /
<0: @2524: Pm<w1w2_mz<._ a. N» 3...
NF mus.
/ .2 l A.
.. ”NI
K S. :. .
NNN or r.» .2»
v E
i. if
2. . v
i momz. 2,3 \ilillllll A
will f?- 0—1 h?—
8 . Ax.ola.j\
£111; . no. 52 v
£3... £3...
nu. 259,32 1:.de
_ mm.
:« _
m3
l‘l—fillll'lll‘f‘ll‘ll‘ll l I}

 

 

 

 

 

 

.LZI—l

.. r129

 

 

 

 

 

 

 

 

 

 

 

 

PN-

 

0:
£3...
FF. jma

 

/. o:

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

47

An Application at the Block Group Level

After much field research to compare Obtained ratings with actual area conditions,
I am convinced that an analysis by census blOck group is more revealing and accurate
than an analysis of census tracts alone. Many census tracts may be composed of several
block groups of distinctly different character, and different RCI ratings will usually result
from calculations on these smaller areas. Calculations and mapping by block group is
much more time-consuming but its greater geographic accuracy will reveal much new
detail in many areas as a result. On the following page is a block group map of a portion

of the Lansing UA, which can be compared to the tract map on the previous page.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. . .5 m. 3 E.
a _O_ b¢ If N
2. «E: murrzm no <.HU¢
g 33.7.1 Seoul liliil lull-lust. I. 2. c
«x. . . .lllllJ
_
_W (J Pfi‘NIWZ-mz.<l. 9P
4.; 62 szoogm mameu . Nm rh wlfil
oz“, .5 «62:. .. 1 ma. . . S 3
fl . . I m: m . 2.
3.20.,— = I MN... m .D l.
H. a: .. me 2
.- 3 mm
. GZ _ $63
.3 .3 E.
To .3 3
MN 5 . w
. o
m. 3 am on 3 II 3 .
WDL (u .. mm no .111
1‘" .3s .2 MM _J PM w¢ a?
. m. IIIII I InIti N.» 3
TS ca
3 I S s -
a.
_
2 K. T
t.
2. I. I
rm. K r
02 um~ N: . . Ill Ill ii ,I. ..
# qw Nu N.
«O. rn; Nm
.3 w or
. _ s. a.
w2_m2<4 Hm<m II I
A
.3
3.0 n. e m:
_ a: an. ‘ hm

 

.9
The Interpretation of RCI Data

In addition to producing maps of the relative quality of neighborhoods, some
additional research is recommended. Each UA, or large area with which tracts are
compared, itself has an RCI, the RCIua rating, which can be found simply by treating the
UA data as if it were for a single tract, and applying the RCIt formula to it. The resulting
rating munber represents the relative quality of the UA, when compared with its
component tracts (until now I have described only informal cOmparisons of tracts with
each other). The RCIua therefore provides an overall standard against which to compare
the individual tract ratings. (The RCIua ratings appear in Appendix A under the "UDA
stats" or "UA stats" rows.)

There are at least two good ways to compare the RCIt ratings with the urbanized
area as a whole. One is to use them in a ratio, (RCIt/RCIua)x100, which will produce
numbers showing the RCIt as a percentage of the RCIua. I abbreviate this indicator
RCI%, and it can be seen in the last column of Appendix B. It has the effect of
standardizing the ratings so that 100 designates the norm for that UA, and RCI ratings are
expressed as percentages of that norm. This sort of expression will be easier for people
to interpret than the unstandardized RCIt ratings.

The other way to compare RCIt ratings with the entire urbanized area is to express
the tract ratings as percentiles of the UA's population, as shown in Appendix C. To do
this, tracts are sorted by rating and a cumulative total of the tracts' populations is
tabulated. The percentile of a given tract is found by dividing the cumulative population
of it, and all tracts that received a lower RCI, by the total population of the UA (or UDA).

In Appendix C, tract 42 is listed first because it has the lowest RCIt rating. All persons in

50

the tract (5,442) are treated as being represented by that tract's RCI rating. The 5,442"d
person in the population ranked by RCIt would be placed at the 2nd percentile of the UA'S
population as a whole. The median-quality tract for the Lansing-East Lansing UDA
would be number 202.02, which is at the 50th percentile. Such percentiles could serve as
a means for assigning class categories within certain rating ranges, such that they broadly
match with the class distributions found in other studies of stratification (referenced
earlier in this paper). Other ecological studies have assessed geographic areas using
percentiles in a manner similar to this (Ley 80-82).'

A person living in a tract may rightly feel that its rating does not represent his
own household. Since the RCIt rating measures the quality of an area using weighted
averages (with adjustments), and since not all areas are homogeneous, such a problem is
to be expected. The RCIt rating accounts for all of the residential types in an area. While
an analysis by block group tends to be more precise, in some cases it may also make
sense to examine each component of the RCIt rating, by housing type. Appendix B
shows four columns with the component RCI ratings: gqiRCI, gqniRCI, roRCI, and
ooRCI. (When examining housing components of an area, it is helpful to express each
ratio as in percentage-style values, by multiplying the ratio by 100. This way, all RCI
values are expressed in the same terms and can be directly compared. The standardizing
RCI% technique can also be used, producing measures such as the gqiRCI%, gqniRCI%,
and so on.) To select one area as an example, a tract such as 44.03 has a large difference
between its rORCI rating (42.5) and its ooRCI rating (103.3). When each of these

components is weighted by the population living in each residential type, the overall tract

51

rating is 51.4, which is 64.8% of the rating for the UDA area as a whole (the tract's
RCI%), and located at the 22nd percentile for the urban area.

It may make sense, when disparate housing types are clumped together like this,
to separate the tract or block group rating into its components, so long as the housing
types are known to be Spatially (or otherwise) separated from each other. The predefined
census boundaries may therefore be amended using whatever more detailed land use
information the researcher has confidence in, and such areas may be split into two or
more parts. (For more discussion of this, see the section on "Critique of the RC1
Model.") In cases where extensive land use knowledge is possessed, it may be
meaningful to rate and rank-order, for the entire UA, all component RCIs for areas which
are indeed distinct from those grouped with them in the same tract, thus enabling an even
more precise geographic analysis than the use of census block groups can provide. This
subdivision of areas should only be done, however, when the researcher is certain that
such areas are quite distinct from the others. Areas without clear distinctions, such as
those having a mixture of owner-occupied and renter-occupied houses along a small grid
of streets, should not be separated by type, but must be rated as a whole area, due to the
extensive interaction between the different types of residents.

While the use of increasingly detailed delineation of areas provides more
information about residential patterns, it may seem to some that such information comes
at the expense of measuring the broader "feel" of an area that is affected by extensive
interactions with those surrounding it (see Jacobs 150-152 for a critique of the notion that
urban neighborhoods are discrete or self-contained entities). There is something to be

said for such a criticism, but one of my explicit assumption in RCI modeling is that such

52

"interaction effects" from adjacent areas are already reflected in reported housing values
(Harvey 548).

An alternative technique of attempting to overcome the limitations of the use of a
single rating to describe areas of residential diversity, could be to include an analysis of
ratings based on tract quartile information, in addition to the standard RCI based on the
mean or median. Two areas may have the same RCIt rating, but one may be found to
have a much broader range of residential class within its area (as revealed by quartile
ratings that are spaced farther from the initial central rating) than the other.

Researchers interested in a detailed area analysis would probably want to examine
categorical data on housing values and rental rates (rather than the summary mean or
median values and rents). Certain value or rent categories could be assigned RCI values
based on comparisons with those of the UA, and each tract or block group would then
have certain percentages of its residents measured as part of these RCI categories and
weighting them in an overall average for that area of study. This sort of analysis can be
expected to provide a much more accurate weighting of different components of a
heterogeneous area, and should be explOred further. This detailed technique was
considered a bit too time consuming (and mathematically distinct from the other RCI
equations) to include any assessment or applications of it in this current research paper.

The main point to keep in mind when interpreting RCI ratings is that they are
Statistically measuring central tendency within an area, not dispersion. They are also-
very sensitive to how the areas of analysis have been defined. Still, my application of
them in the study of several urban areas has convinced me that the RCI ratings are

generally successful in modeling the distribution of different-quality residential types

throughout a metro-area, and identifying how the quality of individual areas compares

with others.

53

“i

54

5. CRITIQUE OF THE RCI MODEL
Weaknesses of the Model '5 Concepts, Assumptions, and Applications

I have identified some weaknesses of the RC1 model throughout other sections of
this paper, from the initial concepts of class to its potential applications in fiiture research.
There are a number Of weaknesses that will be mentioned separately in this section, even
though some of these are mentioned elsewhere in the paper. Some weaknesses are fully
addressed only in this section.

Some of the model's shortcomings are rooted in the use of the census as an
information source. Census data is readily available and therefore to need data only from
that one source is a big advantage. of using the RC1 model. However, the types of
information provided were selected in advance by the census bureau, and in many cases
need some manipulation to create the type of data used in RCI calculations.

One of the major weaknesses in the RC1 model is that the geographic areas
(census tracts and block groups) are sometimes defined in ways that do not match the
purposes of the RC1 model. There are numerous areas where a portion Of the housing
and inhabitants do not really match the others observed in that area, but do match up with
an adjacent one. Since RCI ratings are assigned for entire areas, there is a noticeable loss
of representativeness for tracts or block groups that include items better matching some
adjacent tract (others have noticed this also—see Hartigan 295-296). In many instances,
this problem is rooted in the tendency of the census to preserve historic tract boundaries,
to facilitate longitudinal analysis. In other instances, the boundaries seem to have been
deliberately drawn so as to "dilute" the statistical effects of particular areas. For example,

the census-defined place of Haslett comprised two census tracts in the 1990 census—

55

tracts 47 and 48 in Ingham County. An Old cottage area just southwest of Lake Lansing
that was pretty run down at that time was divided between these two tracts, and so a tract
level analysis of the area shows nothing below middle class. A better geographic
delineation of neighborhood types would have taken half a dozen streets from that area
and given them a working-class rating (probably in the 605, if its RCIt were actually
calculable), quite distinct from the middle class areas to its north and southeast. The data
and RCI model provide no way to distinguish the areas from each other—they are treated
as a unit. Even when dealing with block groups, this remains a problem for many areas.

A second weakness of the model is that it expresses a measure of central
tendency, but not dispersion, for the housing in each area. Two areas may receive the
same rating, but feel quite different because a homogeneous area's" rating may accurately
represent its entirety, but a heterogeneous area's rating could be derived from a large
number of possible combinations that happen to average out to the resulting RCI. In
some cases, by examining adjacent tracts, a trend can be seen and an assumption made
that within a tract, a location adjacent to a lower-rated tract is of lower quality than the
tract as a whole, while a location on the other side of the tract, adjacent to a higher-rated
area, is of higher quality than the tract as a whole, with the two balancing out in the
averaging formula which assigns a rating to the entire tract. In many cases, though, it is
only local knowledge of an area that can confirm whether such a pattern actually exists.
For the most part, the accuracy of the model is limited to the validity and homogeneity of
the tract or block group areas defined by the census.

Ideally, I would have used a measure of housing value for every type of housing,

but proxy variables of gross rent and per capita income had to be substituted for this

56

when rental housing and group quarters were assessed. I did not attempt to assess the
residential impact of vacant housing units or seasonal housing units. Second homes are
also unassessed, although such units should be quite indicative of class position
(Meyersohn and Malone 324-325). In some areas, these types of units compose a very
high percentage of the total, and so should be accounted for in some way, but suitable
information suggesting what the impact of these units would be was not readily available,
and to burden the model with additional, yet-untested assumptions about them seemed
unwarranted.

A possible additional problem may arise with the overinterpretation of RCI data.
The model's measurements are ultimately somewhat limited, and after all the normal RCI
calculations have been performed, only additional research (using"other information
sources) will illuminate further the nature of an area (Roberts 236, Jacobs 11-16).
Assumptions, for example, that a $120,000 house is less desirable in an area rated 50 than
in one rated 120, are not only unwarranted and indefensible, but also violate one of the
key assumptions of the model, which is that the value is determined by a market
assessment of all known benefits and drawbacks of the housing unit. Some of these
benefits may turn out to have nothing to do with residential quality (such as a high land
value from profitable location and zoning for commercial uses) but even when they all
do, such specifics are not addressed by the model: Although the $120,000 house in the
area rated 50 may be much larger and in better condition than one of the same price in the
area rated 120, the latter's small size and worse Condition may be offset by a better

location or better school district. The same pricing indicates that these features are

57

equivalent in terms of the regional housing market (but not necessarily equivalent for
consumers with specific housing or locational needs—Ley 242-243).

An important point that should be kept in mind is that there is probably a
tendency, since I have addressed the issue of residential stratification in terms of class, to
confuse the residential aspect of the RC1 with other aspects of social class. A low rated
area in the RC1 model will not necessarily have high crime rates, for example. A high
rated area may still have a significant‘amount of poverty (an income-based variable).
Although correlations might be expected between indicator variables of class, of which
the RC1 is one, they must not be taken for granted—especially when in many cases the
data on many other indicators is readily available in the census and therefore does not
need to be estimated indirectly by the RC1. The RC1 is intended to measure a somewhat
different aspect of social class than other class indicators do, and its distinct nature should
be kept in mind by all of its users.

Where RCI ratings indicate that an area is "lower class," for example, recall that
the RC1 is but one of many class indicators. Specifically labeling an area as "lower class"
or a "slum" can have harmful repercirssions, whereas the reality of social organization in
lower-quality housing areas may be much different from the many assumptions that
might be made about it (Gans 64-68, 99, 119-120, 126-127, Jacobs 354-358). Social

class is a product of social relations in society—"lower classes" are a part.of society, and .

their status can't be changed through zoning, planning, or design techniques which
address only physical aspects of an area (or even attempt to cause certain types of people

to leave the area). It is a part of our American social system that class relations, or

conditions of inequality, are reflected in the layout of our cities (Jacobs 8-9) and that

58

urban forms are based on competition between different groups or interests (Harvey 558-
559)

In terms of assessing overall social class, the RC1 model is merely one indicator
among many others—not a definitive way of accurately assigning all citizens into a
particular class category, although it can and does suggest placement at an approximate
point on a stratification continuum based on class aspects of residential location. If the
RC1 is examined along with other variables such as the education and income levels in an
area, the variables together should generally succeed in categorizing areas in terms of
overall class assessments. Note that education is measured for individuals rather than
households, again suggesting the pertinence of an individual-level analysis of
stratification despite the traditional preference for household-level analysis.

Some might consider it a theoretical weakness that I did not claim that a class
approach or a stratification approach was notably preferable for this type of analysis.
While the continuous nature of the RC1 measure aligns it most closely with the
stratification approach, I have shown that this measure could be used to estimate the
cutpoints demarcating class categories, and the assignment Of particular areas to those
categories. An assumption could even be made, if desired (and supported by an analysis
of other indicator variables), that such areas would have the subcultural characteristics
relating to those classes, possibly even aspects of class consciousness in cases where the
census defined areas actually function as neighborhood communities.

In many cases my selection of particular census variables may deserve criticism.
For example, the choice of using owner-occupied units with a mortgage was one that I

deemed most suitable, and have found satisfactory in its results, but which was selected

7‘

59

mainly due to the limited alternatives provided in the census data. All of my selected
measures for each RCI component are considered to be merely the best I saw available in
readily accessible census data, not necessarily the best possible choices if other data are
available.

Another noteworthy weakness in the model is the difficulty of a user accurately
and objectively defining an urban area within which filil stratification is presumed to
exist, to enable the valid assessment of tracts and block groups. Although I assume that
the census UA is of sufficient size, it probably excludes many wealthy persons who
commute to the city and act as a community member there, but who choose to live in
essentially rural areas, or in exurban "small towns" (Palen 201, 204). Such commuters
have high transportation costs, but enjoy rural taxation rates and residential costs of living
which in many cases allows their residential quality of life to be much higher than could
be obtained for a comparable price in the city (Harvey 549). Knowing of no way to
definitively distinguish, from the data, exurban or rural commuters from those who live a
truly rural lifestyle, I simply excluded such areas from comparison. The cost of living
seemed sufficiently different that I was certain that a different standard of comparison
must be used to assess rural areas—urban wages can be expected to purchase a lot more
there. Still, the variation in RCI ratings from including or excluding fi'inge areas turns
out not to be that large, so this may not be a great problem for the model. .

Probably of greater importance is the model's use of survey-based data from .
census summary file 3A. Not only is error a part of the model because the data was
obtained from a questionnaire instrument where respondents may not answer accurately,

there is also a sampling error that can be fairly significant when assessments are made at

60

the tract or block group level (or even smaller, when housing components are
considered). The use of more complicated, multi-component versions of the model will
likely cause some of these errors to be compounded.

Some questions are based on "selected" households only, such as information on
housing values and gross rent. The model weights responses to these questions according
to the prOportions of a different group of respondents, however, which in one instance
introduces a clear problem into the use of the RCI formulae. There are certain areas
where particular categories of housing compose only a small portion of the total, and
these proportions are sufficiently small that samples of "selected" units turn out to be
zero. In the formula, there is a small weight attached to a component RCI value that is
left undefined. In these instances, the undefined component will be treated as
contributing no additional housing quality to the component summation for that tract,
effectively counting that residential“ segment. of the area as having an RCI of zero.
Fortunately, since these components compose only a small fraction of the entire area, the
reductive effect 'of this anomaly on the RCIt rating is quite small. In cases where an
entire tract has little or no population, a zero rating will be produced by the RC1 formula
as a result. '

One extra issue on the use of census data concerned rental information. The
census identifies a segment of renter-occupied units for which there is no cash rent paid
for the unit. At first I was inclined to include this segment in my calculations of mean
rent, by treating their monthly rental costs as $0, since I intuitively pictured such persons
as merely staying in a friend or family member's otherwise vacant unit. According to

census definitions, however, the "no cash rent" category includes many persons who

1"

61

contribute services in exchange for their housing, such as the management of an
apartment complex , which can have considerable value. I decided not to count these
cases at all when calculating rents, with the result that their presence in an area will not at
all affect its rating. Although my choice was arbitrary, the effect on the ratings for all
areas I have calculated to date has been negligible. If an area is observed which has a
significant proportion of inhabitants with "no cash rent" paid, some alternative solution

should be found to address this category of renters.

ft

62

6. AREAS FOR FURTHER RESEARCH
Equation Modification
I have presented my model in such a way that it is very easy to change its
equations to reflect different assumptions than those I have used. A person may disagree,
for example, with the reasoning behind the rental adjustment factor I have proposed, and
wish to change it. A different researcher may decide to drop the group quarters
components of the equations and weight the two remaining factors according to their
proportion of households, rather than pOpulation. These are examples of perfectly
acceptable adjustments that I intended to be "tried out" in my model at some future point.
Another interesting adjustment that could be made is to represent group quarters
differently in the equation. Following my prototype model, certain kinds of group
quarters, such as prisons, could be assigned a zero (or even negative) rating, so as to
better match the sort of effect such institutions are presumed to have in an area. Other
group quarters, such as nursing homes, might be weighted differently, or assumed to have
no effect on the surrounding area. It was my choice not only to count everyone in an area
as having an effect on it, but to have the ratings assigned to such persons rooted in actual
census data, to reduce subjectivity. However, more use of the model, and comparisons of
its resultant ratings with actual comparative neighborhood conditions, may reveal through
"trial and error" that certain values or weights are more appropriate than the census-based
variables I chose. Where such improvements can be found and agreed to better model
actual residential conditions, their use Should of course be encouraged. There are a large
number of variables that can be considered to potentially affect residential quality, and I

have used only a few, in which I had more confidence, in my model so far.

63

One of my ideas that I eventually decided was not yet developed enough to
improve the model, was to increase its complexity by using the individual person as the
unit of analysis throughout. Toward the beginning of this paper I reported that I believe
such a measure would be preferable (or at least worth investigating) because I mistrust
the validity today of assuming that all members of a household are of the same social
class. I attempted to correct for this by assuming that per capita housing values (or
expenses) could be used as an approximation of the individual position of its members,
by dividing household figures by the average number of persons reported in households
in each tract or block group area. I guessed that the reality that at least one household
member in a family with kids will have a higher class position than some other in the
household, might not matter too much in view of the averaging nature of the model in
general. An individualized measure, like "per 'capita housing expenses," derived either
from the rent divided by the number of persons, or the housing value (or SMOC) divided
by the number of persons, would also help to match these components with the already
individualized analysis done on those in group quarters.

The conditions I mainly wished to assess through individualized measures of
residential class were those relating to the effect that household size (and possibly
crowding) might have in creating inaccuracies in RCI ratings. That is, wouldn't a family
of eight,living in one apartment, be lower in class position than a family of two, living in
the same apartment? The per capita housing expense of the large family is clearly much
lower! Similarly, wouldn't a single person living in a house have a higher class position
than a large family living in that same house? The per capita housing value is much

higher for the single-person household. I

64

My develOpment of per-capita RCI models involved two-component, three-
component, and four-component models, just as I had created for the household-based
models described earlier. These three equations appear below, and share the weakness
that I was unable to find or derive data on ppOOHmeua, and had to substitute
ppOOHUua for it instead when applying these equations to actual census data.
(Equation 10) Rejected 2-Component Per-Capita RCI Model

RCIt = [(%ROpt)(roRCIt)+(%00pt)(ooRCIt)k100)
(Equation 1 1) Rejected 3-Component Per-Capita RCI Model
RCIt = [(%GQthqSRCIt) + (%ROpt)(roRCIt) + (%00pt)(ooRCIt)k100)
(Equation 12) Rejected 4-Component Per-Capita RCI Model

RCIt = 100 x
[(%GQithqiRCIt) + (%GQnit)(gqniRCIt) + (%ROpt)(roRCIt) + (%OOpt)(ooRCIt)]

 

 

 

 

 

where

quCIt = PciGQt gqi R C It = pciGQit gqni R CIt = pciGQnit
pciOmeua ’ pciOmeua ’ pciOmeua ’

r 0 RC It _ MGRt -:- ppROHUt and ooRCIt = MVt + ppOOHUt

. SMOCwmua + ppOOHmeua ’ MVua + ppOOHUua

NOTE: ppOOHmeua (from the denominator of the roRCIt equation) could not be found or derived from the census
data I was using, and so ppOOHUua was used instead as a substitute.

When I actually applied these models to the Lansing-East Lansing UA, I felt that
the quality of match between the ratings and my perception of the areas themselveshad
declined from that of the household-based models. Eventually I decided that this was
because this new, per capita averaging model was too distorted by variations in
household size (just as household-based income studies are—Hess, Markson, and Stein

181). Using these models, single-person households are treated as having a much higher

 

{a

65

class position than was realistic. This is because a single-person household must spend a
certain amount of money to maintain itself—on items whose benefits could, in a multi-
person household, be shared by others, with the result of a much lower per capita
expense. As an example, We can look at changes in the poverty line based on the size of
a household. If a one-person household is in poverty with an annual income of $8,000,
but not in poverty with an annual income of $10,000, this does not mean that a two-
person household is in poverty at a. $16,000 income and above poverty with an income of
$20,000. Rather, the two-person household is considered above the poverty line at about
$13,000, and a three-person household at $16,000. The attempt to derive meaningful per
capita class assessments fails for my housing variables because it is based on simple
averages rather than a curvilinear standard, such as are used in poverty calculations. A
more flexible standard of comparison that doesn't merely use averaging will need to be
incorporated into the equations in order to render the per capita model appropriate for the
kind of analysis it was intended for. If household size could be taken into accoUnt so that
one of its occupants is compared to a standard for a person living alone, and then any
additional occupants are compared at some standard that is a fraction of that for the. first
(similar to the distribution of poverty levels that are based on household size), then the
expanded formula might work. This is why I have included lengthy descriptions relating
to it in this paper, and suggest it as an area for further research.

An extra problem with per capita models is that since I found no way to calculate
the number of persons per owner-occupied unit with a mortgage, I had instead to
substitute the value for all owner-occupied units. This 'can be expected to have a

significant effect, since according to the census, the incomes (and I would guess, ages

66

and family sizes) of those in owner-occupied housing units without a mortgage was quite
different from those with a mortgage. If it is actually possible to calculate the "with a
mortgage" segment properly, and therefore derive a more adequate estimate of per capita
owner-occupied housing values, these rejected equations may ultimately prove to have
much-increased validity.

I also considered splitting the owner-occupied component of the RC1 equations
into two terms, just as I had done with the group quarters component. Initially, I had
rejected the idea because, if OOwom was to be treated as the other components and
adjusted for comparison with the Ome standard, I could not conceptualize a valid way
to link differences between SMOCwom and SMOCwm to their inhabitants' class
positions. Any such linking to me seemed to introduce unverifiedassdmptions about the
characteristics of these two categories of owners. Adj ustrnents to 00 subcomponents
should be based on housing values rather than a SMOC-derived adjustment factor, since I
have based the model on the assumption that housing values are the most revealing
indicator of class position. Therefore, a new component should be created by comparing
the mean values of 00 units with a mortgage in the tract (MVwmt) with the respective
mean value for the UA (MVwmua), and similarly comparing the mean values for 00
units without a mortgage (the ratio of MVwomt over MVwomua).

I didn't start to imagine how this addition to the formulae might improve the
analysis until just before this paper was completed. Its main justification seems to be
mainly to use more data in the model's calculations. It is certainly worth testing,
however, because of the possible class-related differences between residents of OOwom

units and those of Ome units. For example, OOwom residents may be longer-terrn

‘33

67

inhabitants of an area, and Ome residents may be newer inhabitants, and the life-cycle
position of a person or family is certainly important to housing submarkets, as well as
correlating with age, income, lifestyle, and other class indicators (Ley 242-243, Hess,
Markson, and Stein 228, 230, Gilbert and Kahl 87-89). Therefore, I include in this

section for further research the following proposed five-component model:

 

 

(Equation 13) PrOposed 5-Component RCI Model
RCIt = (100) (%GQit) pc'GQ” +(%GQ,,,-,) 10chth +
pciOmeua pciOOWMua

MGRt
SM 0C wmua

 

(%ROpt)[ J + (%OOwompt)(—MM-t—J + (%Omept)[m)]

MVwomua MVwmua

Of course, the most advanced RCI model I currently envision would be one that adjusts
equation 13 for individual-level analysis, as discussed in the preceding section with
equations 10 through 12. This would probably take some time to develop though (if it is
even feasible), and many researchers may not want to take the extra care to use such
expanded models unless the resulting ratings were found to be of significantly improved
accuracy or greater validity.
- Transitional Zone Analysis

If the RC1 model is both valid and accurate, its application will identify areas in
the city where there are significant class differences between adjacent or proximate
areas—areas which I will call transitional zones. In many cases, transitional zones are
quite sizable and are based on geographic distance. A middle class neighborhood (for
example, with an RCI% of 125) may sit adjacent to an upper-working class neighborhood
(let's say with an RCI% of 95), on the other side of which are the seedy areas of town

(with RCI% ratings under 65, let's say), so that the distance between incompatible class

(j

68

areas is substantial. In other cases, the transitional zone may be small in distance, but use
a river or wall or expressway (or the proverbial railroad tacks) to limit interaction
between areas. A study of such areas in the field, including interviews with residents,
and analysis of physical design characteristics, should prove quite informative for urban
and regional planners who are concerned with screening and buffering around otherwise
incompatible areas.

One of the concerns of better-Off residents is the maintenance of certain spatial,
physical, or psychological impediments discouraging extensive interaction with lower
class areas nearby. These concerns are acknowledged in planning hearings, although
planners themselves are understandably reluctant to admit the influence of class criteria
in their decisions, keeping such criteria implicit or Unstated, for the most part. In some
cases, the issue is plain to see, however; planners are often faced with the dilemma of
where to place govemment-subsidized (i.e. lower-class or lower working-class) housing.
The RCI measure could provide a quantifiable aspect to the study of what size of class
differences are generally tolerable between adjacent areas, and what techniques of design
may help increase such tolerances (bushes, lighting, the orientation of buildings,
windows, and access points, the use of gates, alarm systems, security staff, and so on).

Interaction Studies

An interesting area for future research could be to measure the amount of
interaction that occurs between different classes of neighborhoods. Certain types of
transportation modeling and market analysis examine household economic
characteristics. In similar fashion, it may be found that despite the assessed market-

values of particular residential locations, the actual quality of life there may be affected

(“:1

69

by proximity to or access from lower-class areas in ways that have not factored
sufficiently into the market pricing of the unit. The RCIt rating alone may be misleading,
and need adj ustrnent for the effects of interaction between it and lower (or higher) class
areas.

A good example is found in the setting around the campus Of Michigan State
University. The campus itself has housing that consists almost completely of dormitory
units, and receives a gqniRCI rating of about 15, since two or three low income~ persons
frequently share a single room of small Size, and rooms generally lack complete
plumbing, kitchen, and bathroom facilities, which must be shared among groups of
residents (which is why these have been classified as group quarters by the census
bureau). The surrounding census tracts receive much higher RCILratings of 38, 51, 68,
and 71. The notorious block parties/riots that have received much media coverage in the
last couple of decades might, according to the RCIt ratings, be expected to have taken
place in the lowest-rated areas—the campus itself, and possibly a couple of the low-rated
adjacent areas. The reality was that although areas of the campus were involved, the .
majority of the destruction and criminal activity took place in two tracts to the north and
northeast of campus (locally sometimes known as the "student ghetto") which received
RCIt ratings of 68 and 71. A detailed analysis Of areas based on local land use
knowledge suggests why these locations were most affected.

To the west of campus are university-run apartments which have complete
facilities in each unit but are Often overcrowded (about 20% of the units have more than
One person per room, 11% have more than 1.5 persons per room) and which obtained an

roRCI rating of 37 (see Nettler 151 for these definitions of crowding). These apartments

70

are in three groups, Spartan Village (in tract 44.02, which has an RCIt of 38), University
Village, and Cherry Lane Apartments (both of the latter are in tract 44.03, which has an
RCIt of 51). Much of these units are relatively far from the center of campus, and are
inhabited by foreign and older students and staff at the university, many of whom have
families. Tract 44.03 also contains a significant area of owner-occupied middle-class
homes.

To the northeast of campus is tract 43.02, which has an RCIt rating of 71 and in
which is located the infamous Cedar Village Apartment Complex, at which many of the
civil disturbances have begun. To the north of campus is tract 41, whose RCIt rating
calculates as 68, and which contains the main commercial, recreational, and office areas
serving the campus and its student population. Although crimes Show a much greater
positive correlatiOn with the presence of young males, there is alSo a weaker but

significant correlation with variables such as wealth (a negative correlation), rental values

‘ (a negative correlation), neighborhood characteristics ("slum" status), and urban area size

(Nettler 100-106, 111-113, 141, 144-147). Although I have given cautionary statements
about not overinterpreting the RC1 ratings, and the necessity of considering additional
variables before assessing an area's overall social class, the RCIt values in this case seem
not to have identified these areas properly. If we assume that riotous parties are a lower-
class activity, did the RC1 ratings at all suggest the possibility that such damage and
disorderly conduct would occur in tracts 41 and 43.02? I have already given some
indication of the spatial and demographic aspects that would explain the ability of the
lower-rated traCts of 44.02 and 44.03 to avoid riot problems. As it turns out, an analysis

of RCI ratings at the block group level reveals some areas within the two riot-prone tracts

a”:

s‘v

71

that receive very low ratings (compare the two maps in the RCI application example
earlier in this document). In addition, however, I suggest it could be helpful to consider
the effects of interaction between the lowest-rated areas—the dormitories on the campus
itself—and the tracts to their north.

Recall that the RC1 is intended to measure residential quality, rather than the
quality of commercial or industrial areas. In my experience, the quality of interaction in
commercial areas often does not match the character of the adjacent residential areas very
well, because commercial areas tend-to be situated in a way that enables traffic from
around the region to gain access to its shopping, Office, and work amenities. It being
generally profitable to attract as many customers as possible, retail commercial areas
accessible by less prestigious persons tend not to be as exclusiveas residential areas, with
the result that persons in major commercial areas tend to reflect the diversity of the
urbanized area as a whole, rather than merely the surrounding neighborhoods. So
although about 17,000 students live on campus in tracts 42 and 44.01, the vast majority of
these travel into tract 41 for recreational and ShOpping activities. Tract 41 also has direct
bus links with lower quality sections in the city of Lansing, and attractions (such as a
clinic which buys blood plasma) which appeal to a lower class of residents.

A detailed assessment Of land use shows a. general pattern of ascending social
class the farther one lives from the campus boundaries. There are areas adjacent to the
campus that share its low residential quality, but do not benefit from the direct
institutional support and administration of the university. Tract 43.02 has about 10% of
its 3,211 residents living in noninstitutional group quarters, and its gqniRCI component

rating is only 23. Tract 41 has about 27% of its 5,017 inhabitants living in

«J

72

noninstitutional group quarters, which have a gqniRCI rating of 22. SO on the one hand,
the majority of each tract's residents live in fairly decent (roRCI ratings in the 705) rental
units and merit a residential label of working class (fitting with tending to be productively
engaged in school, yet employed in lower-status temporary and lower paying jobs, and
with their youth and not-yet-attained higher education holding them back from a higher
status). On the other hand, there is a sizable minority of the population in this area living
in rooming houses, fraternities, and shared housing arrangements that reflect a lower-
class income, culture and lifestyle. It is also possible that, despite such conditions, many
of the inhabitants of nearby apartments have chosen to reside near such arrangements
because of a cultural or class affiliation with the activities that go on there. In contrast,

many of MSU'S students choose to live in otherwise middle class areas surrounding the

_ university, in tracts 38.01 (20% of persons aged 3 and above enrolled in public college),

38.02 (26% in public college), 39.02 (30% enrolled), 40 (30% enrolled), 43.01 (40%
enrolled), 45 (20% enrolled), 49.01 (17% enrolled), 49.02 (27% enrolled), and 50.01
(11% enrolled). It might be inferred that in these areas, the residential quality better
matches preferences originating in a middle-class or upper-working class childhOod. The
roRCI rating in most of these areas is in the 605 and 703, Showing their relative
affordability (since that RCI component is based on gross monthly rental costs) and
therefore suggesting the importanCe of lifestyle and consumption preferences in the
selection of residential location, despite the notably lower income available to most .
students.

Although procedures for a more general analysis of this type are not yet detailed,

it can be seen how the RCI ratings are useful for descriptive and eXplanatory purposes.

(“I

mt

73

Further research on the nature of interaction between areas could enhance the explanatory
power of the RC1 model beyond merely considering issues of stratification. Perhaps the
incorporation of a gravity model of interaction (Haynes and Fotheringham 23-24) would
provide new analytic breakthroughs, or perhaps more simplistic approaches could be
similarly effective. For example, interaction effects between the campus and its
surrounding tracts could be estimated by combining their data and treating them as a
single tract. As subdivision of tracts was recommended for a detailed residential
assessment, combining of areas could be useful when there seem to be interaction effects
not adequately reflected in the housing values and rents for an area.

Applications in Urban Planning

With much additional research, the RCI model might eventually be used to assist
residential planning decisions, by estimating the impact that new developments will have
on the quality of an area. Some means would have to be found to convert present values
to the equivalent of the census data, but also to account for shifts in values that may result
from the impact of a new development. On the other hand, there may not be such shifis
due to screening, buffering, or interaction-reducing security and design measures
surrounding the new development.

For example, a planner may be calculating the locations of subsidized housing
units in an area, and wishes to distribute such units so that each community has its "fair
share," yet minimize the presumed deleterious effects on surrounding areas. Supposing
that these subsidized units have their roRCI calculated as a 30. The planner resists the
temptation to locate them in a lower working class area with a rating in the 405, even

though residents of this area probably would publicly register few complaints about the

I ,1

I

“IF-I.

 

 

I I

74

location such developments, because of a Concern that the ratings of these areas would be
pulled into the lower class range below 40. The planner instead spreads out the housing
units so that no RCIt rating is lowered more than a couple of points—insufficient to drOp
an area down a class. In areas where there is a great difference between the original RCI
and the project's estimated RCI, the planner investigates techniques of lowering the
impact through location and design measures. The new residents Should be able to enjoy
the benefits of living in nicer-quality areas, having access to jobs, schools, shopping,
parks, or other amenities accounting for the area's higher RCI, and may also enjoy
increased interactions with their wealthier neighbors, rather than feeling trapped and
socially isolated in a "ghetto" or "Slum" setting.

Let's say that one area of project development in this hypothetical example had an
original RCIt rating of 84, and the planner calculated that the new rating would be about
81 as a result of the new development. With the release of new census data 7 years after
the project's development, however, the new rating calculates as a 68! Housing values
had declined significantly between census periods. Although it is not certain that such a
drop was due to the effect of the subsidized housing units, the planner is left to wonder
about this, and attempt to defend the location decision against accusations that the
development started the neighborhood into an unstoppable decline.

Although invented for example purposes, such a scenario illustrates my current
reluctance to employ the RC1 model for residential "impact studies." The RCI model is
much too new and too little tested to bear such burdens at this point. Future verifications
of the model's assumptions and accuracy may eventually enable this sort of use, however,

or even attempts to use the model for projections about future patterns of residential

75

quality throughout an urbanized area. Such applications are, in my opinion, a long ways
off, and I believe that the initial models presented in this paper would be used
irresponsibly if wielded in such a fashion (refer also to Jacobs 417-439).
Use of the RC1 for Historical Trend Studies

Normally, historical trend analysis of economic data requires using inflation
information to adjust dollar values from one year and allow comparison with those from
another year. The usual technique is to examine changes in a cost-of-living index, such
as the consumer price index, and make a ratio of the two values to produce a multiplier
that will allow the values from one year to be expressed in terms of the other year's
monetary values. However, information about housing costs are one of the items used to
determine values of the consumer price index, and the rate of change for housing costs
may be quite different than for other costs, thus making the consumer price index as a
whole less than ideal for accurately measuring changes in housing expenditures over the
long term. Also, housing standards change as society does, and the older, more crowded
living arrangements that were the urban norm 70 years ago are now considered sub-par
by many consumers today. Changes in technology (for example, the number and
capacity of electrical outlets) and code requirements (such as larger setbacks between
structures, for fire safety, and provisions for parking and vehicular access) have rendered
many older housing units to be inferior than newer ones. Since class is a relative thing, in
which comparisons are made within a society, a measure of historical changes in class
should be measured using a flexible-tool that these difficulties won't affect.

The RCI model, by basing its class assessments on the relation between an overall

urban area and its component parts (or neighborhoods), seems to provide this flexibility.

76

Both the RCI% rating and percentile ratings that can be determined for an area (as Shown
in Appendix C) are relative measures I created to allow comparisons to be made between
independent systems, such as between different historical periods. Information on
changes over time in a residential area (neighborhood succession) can be studied using
the RCI model, or variations on it that allow the use of different indicator variables from
earlier census periods.

Researchers of historical trends should keep an eye open for a few quirks that the
data may produce in this sort of analysis. First, area that are completely stable and
practically unchanged over time may appear to grow worse, if wealth and standards of
living in the surrounding areas have increased. Second, if the size of an urban area
increases over time, researchers Should expect the range of stratification observed in that
area to increase, as described in the next section.

Use of the RCI for Comparative Studies between American Cities

Just as one urban area in separate historical periods of time can be considered two
independent stratification systems for-RCI modeling (in that measured values in
economic variables are relatively incompatible, rather than "independent" in the statistical
sense of one condition not affecting another), so also can separate urban areas be
consideredindependent. This is because the costs of living may be much different in the
cities that are chosen for comparison, as well as the ranges of their stratification systems.
My choice of creating the model around a comparison of small tract areas with an overall
urban area, rather than a state or nation as a whole, was due to the large effect that cost-
of-living differences would have had on such comparisons. Thus, an important

assumption of the RC1 model, which is only fully revealed when considering

77

comparisons between UAs, is a localized form of stratification in which the UA serves as
a class reference for all its inhabitants. In this function it is viewed as being, for the most
part, a more important reference point for status assessment than the state or nation of
which it is a part. (For a differing view, see the next section, which assesses stratification
using larger frames of reference.)

For the RCI models presented so far, it is assumed that the biggest impact. of class
distinctions (at least for the majority of people who do not have great wealth) is to be felt
at the local level, and that people's frame of reference is most strongly shaped by the
immediate area that they live in (the UA, for example). If this is a fair assumption, then a
direct comparison Of RCI ratings for tracts in different urbanized areas will not only be
misleading due to the differences in cost-of-living (which, after all, shOuldn't in itself be
that difficult to adjust for) but also due to the extremes of variation around the averages
which are used as a standard. In the Lansing-East Lansing UDA (which in today's
America I consider to be a medium-sized metro area), having a population of 289,509,
the top RCI% rating was only 206.8, for tract 39.01 (the "White Hills" subdivision in East
Lansing). This tract is considered top-notch for the Lansing area, but doesn't seem like
much compared to the rich areas around larger cities such as Chicago or Los Angeles.
For example, whenI applied my prototype RCI (Equation 1) to a tract analysis Of the
Detroit UA, the tract composing the village of Grosse Pointe Shores received a rating
over 500!

On the flip side, there were many tracts in Detroit whose housing values were so 3
low that they were expressed in the census only in terms of being less than some minimal

value, such as "under $10,000" or "under $15,000" (the cut Off point is different

78

depending On whether the data is obtained from the printed books or from the CD-ROM).
Large areas of Detroit received RCIt ratings below 30, without even adjusting for the
effect of group quarters (since Equation 1 did not include this), whereas in the Lansing
UDA, such a low rating was only obtained for the downtown tract (Tract 14), and two
tracts on MSU's campus that are almost exclusively dormitories.

It may be that in smaller urban areas, extremes of wealth and poverty are at too
small a scaleato show up well in their own individual tracts, and so get diluted or
averaged in with surrounding areas that are closer to the norm. An alternative
explanation is that the extremes of wealth and poverty in large cities are due to qualitative
differences in their economic and social functions, and that, along with the wider range of
values in indicator variables, we have a similarly wider range in relative social status.
People without money or homes are in many ways similar no matter where they are—
they are at the bottom of the social ladder. People at the bottom in a large City, though,
tend to be farther from the norms, and farther from the people at the top, than we see in
small or medium-sized areas. This may account for the qualitative differences Observed
for large cities, in addition to the proportional differences one would expect from
increased size (Ley 346-348, 350)

V The description I have given underlying the RCI model is also in accord with the
premises of central place theory (King 31-32) in which the large city is the focus of a
region and has land valuations shaped by its important functional roles, within a
hierarchy of different cities, playing more powerful regional (or national or global) roles

which cause economic referents to be measured on a more competitive scale. The

importance of such hierarchies also figure prominently in the contemporary world-

79

systems approach to urban analysis (Lyons and Salmon 102-103), in which the presence
of global networks and competition finds expression in the organization of larger and
more important cities.

The mathematics Of the RC1 model is also rooted in the sociological concept of
relative poverty, in which poverty is defined in relation to the concepts of social class,
and social relations of inequality, which are subscribed to in a cultural environment at a
particular time and place. Relative poverty is based on how a person's needs,
expectations, or standards of living are defined by the particular social, economic, and
cultural environment in‘which that person is socialized and earns a living (Kerbo 308,
Gilbert and Kahl 276-277). The standard of class comparison issue will be discussed a
bit more in the next section. I .

Thus, the RC1 model would establish the residential component of social class
with reference to standards for a particular urban area, by calculating an RCI% rating. I
hypothesize that the RCI% rating should allow accurate comparison of tracts located in
completely different urban areas, which might even be across the country from each
other. This hypothesis is based not onlyOn the presumedvalidity of the RC1 model, but
also on the assumption that urban conditions throughout the country have enough
qualitative similarity so as to allow comparison. If such a comparison were made
between cities in different nations, with srifffiiently different cultures and societal
structures, the results, while technically comparable (after having adjuSted for differences
in currency and census variables), would not necessarily point to similar lifestyles or

standards of living. Rather, they would probably be providing information on

international lifestyle differences between those of the same (societally relative) class

80

standing. This in itself could make for some very interesting research, comparing, for
example, the living conditions of skilled manual laborers (upper working class) in Brazil
or Indonesia with those in rich countries like the United States. The variables chosen to
measure class, however, were baSed on American society, and probably have numerous
weaknesses when applied to most other regions of the world.

Use of the RC1 for State, Regional, and National Analysis of Stratification

As mentioned before, a researcher who operates from different assumptions than I
have may still be able to use the RC1 model for research and analysis. If a researcher
does not Share my assumption that class is primarily defined with reference to the
urbanized area of residence, a different standard of comparison may be employed. A
broader region may be defined (such as the southern lower peninsula of Michigan), or the
entire state or country may be used as the standard. Such an approach to the model might I
assume that any regional differences in cost of living are due to Significant benefits
derived from living in that area, such asia greater availability of jobs or, in hierarchical
terms mentioned in the last section, the presence of more capital or wealth in the area.

If such an assumption is indeed warranted, then a corollary would be that class is
best defined in relation to the entire country (or other large-scale area of comparison);
that persons actually compare themselves with reference groups that may not even be

visible to them in their lives. An argument Can be made for this because of the extensive

'
'

effect of modern communications, but my current opinion is that such an assumption was
much larger than I wanted to make at this point. I believe that lower income persons, for
example, compare their lot with the nearest example of a stratification system, which may

be their immediate environment, or the nearest urbanized area, and that if such an area is

81

important enough, its role in a national or world economic system will cause areas of it to
reflect conditions elsewhere in the nation or world (for example, that infant mortality
rates in sections of Detroit are comparable to third world rates—Ley 336). A poor person
living in Monroe County (on Michigan's border with Ohio) is likeliest to compare his
residence either to the city of Monroe, or to the adjacent large metro areas of Detroit or
Toledo, rather than the rural poor in Montana, Alabama, Appalachia, or any other distant
area. Clear exceptions to this could be if the person travels to Or communicates with
those areas, such as from media reports or family correspondence, but this would be a
selective and biased comparison with such areas, not a balanced assessment of conditions
throughout the country, which would require systematic analysis. It seems to make sense
to use a state or a researcher-defined regional area as a standard of class comparison,
where a knowledge of that area suggests that it is economically and culturally valid to .
treat it as part of the same stratification system, and where inhabitants of the smaller area
under study are likely to have experience of such conditions with which to make class
comparisons.

A plausible argument can be made that the wealthier classes, due to their
generally more extensive travels, affiliation, and education, are as a result sufficiently
aware of national stratification conditions, and sufficient in their local influence, to

r,-
I
l’/I‘

impose such nationally-based standards on those in their local environment. My
experience in rural areas of Michigan suggests that although the wealthier classes in an
area are aware that they are not so wealthy compared to many other areas, for local

purposes they act as a kind of elite, functioning in their communities at a higher level

than a state or national assessment of their class position would suggest. Persons who in

~13

82

a large icity might be judged as "merely" lower middle class, due to their occupation,
education, and income, fiequently hold influential positions in rural, relatively poor areas.
My observation of this has led me to prefer, for this research, the assumption that class is
relative to the standards of a fairly localized area. If this assumption is untrue, or
unshared by other researchers, the RC1 model can easily be adjusted so that the values
that tracts are compared with are drawn from state or national data.
Use of the RC1 Outside of Urban Areas

It may be possible to apply the RC1 model to judge rural and small-town areas.
Following the same sort of reasoning I just detailed in the preceding subsection, if an area
is chosen that makes sense to act as a standard of comparison for smaller areas within it,
then the usual RCI analysis may proceed. I do not at this point make any claims for the
validity and usefirlness of such an application, for several reasons. First, I am unsure how
the boundaries of the larger area of comparison would best be defined. Such an area
must have sufficient diversity to show the presence of a meaningful stratification system,
but otherwise would seem to have boundaries that were chosen in a rather arbitrary
fashion, unless their selection is informed by research that I am currently unaware of.
Second, areas that are not defined as urban by the census may not have data available in
the spatial detail enabled by the use of census tracts and block groups. Such convenient
areas Of "neighborhood" analysis are defined only within MSAs. The third shortcoming
is related to the second: the large size of the geographic areas the census uses in rural
areas is such that in manycases it is doubtful whether they constitute a neighborhood, Or

meaningful residential grouping. While some patterns and regularities doubtlessly exist

83

in rural areas, the RC1 analysis presumes a level of interaction between inhabitants of an
area that may no longer apply in areas of very low population density.

Nevertheless, attempted rural applications of the RC1 model could still be useful
for sociologists and regional planners. When I was in an earlier stage of developing my
model, I applied the prototype Equation 1 to Alcona County, in Michigan's northern
lower peninsula, adjacent to Lake Huron. I used the whole County as the larger area of
comparison, and then rated all the political subdivisions within it (11 townships and one
very small city). Although the ratings did seem to reflect residential disparities, there was
not a great amount of variation within the county, and the highest-rated area may have
been assessed as higher than the others merely because of the presence of a resort
community (which on investigation turned out to be a private, "gated" one and thus
presumably of less influence on the township as a whole than the RCI rating might
suggest). Thus, the model's weakness that comes from having the ratings affected by the
census bureau's selection of an area's geographic boundaries seems merely to be
amplified as it is applied to larger areas of low population density. one reason that I
prefer to use a UA rather than an MSA for the a standard of comparison is that I observed
that most MSAs have numerous rural tracts at their peripheries which I felt
uncomfortable describing with a single RCI rating.

Combining the RC1 with Other Class-Related Variables ,

As described before, social class is not a distinct or simple concept comprising
only a few clear indicators. My intention is to have residential location (as measured by
the RC1 model) considered as an important indicator Of class position, in addition to

others that have-been identified. The RCI model should not be evaluated purely in class

84

terms, although correlations with other indicators can be expected, as well as with
variables such as crime rates that might also be associated with social class. The RCI is
based on variables related to housing values and expenditures, and so is most directly an
assessment of relative neighborhood quality, under the assumption that increases in
reported values or expenditures reflects either increasing housing quality, or other
valuable benefits of a particular location or environmental setting.

It is intended to be easy to incorporate the RCI measures of class into any
multivariate framework of stratification. The trick with this is of course to assign it the
correct weight in comparison with the other factors. My research has not led me to
explore the specifics of how this has been done, but if the concept of social class is to
have respectable scientific validity, it should be possible to obtainand analyze specific
objective data about persons or households and apply a class category or stratification
measure to them, which others can agree is descriptive and meaningful beyond the
original data that composed it. Researchers who have dealt with the use of data on
education, occupation, income, or other measurable variables are invited to use the RC1
model in expanding their work, and assess how well it fits in with other concepts of class

and stratification.

85

7. CONCLUSION

My own field research to investigate the meaning and accuracy of RCI ratings as
a measure of residential quality have led me to believe that it is a useful modeling
technique. The ratings also provide a good framework in which to discuss otherwise
subjective impressions of how "good" or "bad" an area is. While it is insufficient in itself
to assign people to specific social class categories, it does assign an approximate location
for residential areas in a hierarchical continuum that serves as an important indicator of
social class. It also provides localized assessments of residential inequalities, based on a
widely accessible data source. Its clearly-defined techniques and assumptions give it a
degree of objectivity that allows results to be checked and verified by other researchers,
or adjusted to allow for applications in research based on somewhat different assumptions
and goals. Many examples of such future applications have been suggested in this paper,
and I believe that many of these will be of great use and interest to urban planners,

sociologists, and geographers.

APPENDIX A

APPENDIX B

APPENDIX C

86

APPENDICES
A printout of the RCI ratings resulting from the application of four
different equations. (Referenced on pages 44 and 49)
A table showing the results of my calculations for every census
tract that is at least partly included in the Census Bureau's
definition of the Lansing-East Lansing Urbanized Area. Also
includes RCI% ratings. (Referenced on pages 44, 45, 49, and 50)

A table sorted by the RCIt ratings calculated for tracts in the

Lansing-East Lansing UA, using the 4-component household

model (Equation 8). This sort has allowed the spreadsheet
software to then calculate percentile rankings for each tract, as

show in the rightmost column. (Referenced on pages 49 and 76)

29.01
29.02
31.01
31.02
32
33.01
33.02
34
35
36.01
36.02
37
38.01
38.02
39.01
39.02
40
41
42
43.01
43.02

Persons

2159
1575
2705
3526
2140
2162
3032
3698
2208
2523
2667
1598

207
2525
1331
1105
3878

739
4258
2429
1866
3605
3563
2541
2413
3371
2891
3329
3406
1302
3406
2573
3360
2625
2614
3017
4312
4075
5944
4555
2859
2214
4286
4175
5017
5442
5836
3211

4719
5815
4819
6412
4715
521
4719
4615
579
601
4715
3719
2112
4519
7919

‘15612
9512
3615
4319
501
7219
6215
5919
71.8
5419
62]’

, 6219
7419
5719
6415

. 105.6

46.3
57.4
100.2
84.0
55.5
61.9
51.7
55.5
114.5
85.4
204.7
84.7
112.5
63.0
0.4
96.8
66.0

APPENDIX A

RCls Calculated by household:
2-component 4-component

47.9
58.5
48.2
64.2
50.9
53.2
48.0
46.5
57.9
60.1
47.6
38.2
27.8
45.9
79.8
156.2
95.2
53.1
43.8
50.1
72.9
62.6
‘ 59.0
71.8
55.0
62.7
62.3
75.8
57.3
64.6
105.6
46.6
57.4
100.2
84.0
55.5
61.9
51.7
56.1
114.5
85.4
206.8
91.4
113.0
67.9
14.6
96.8
70.6

4912
6315
4919
6619
4012
7319
5519
4119
6812
661'
4415
5319
4415
6015
7312

18519

10411
5919
4719
5119
821'
7413
641)
8419
631'
6715
7219
7719
8211

1041'

10619
4319
61:7

10015
9712
651
5315
48]’
6111

11615

12112

20519
9912

12412
6011

197
9711
721

Per capita RCI calculations:
2-component 4-component

49.2
63.6
49.3
66.8
43.6
74.8
55.3
41.9

‘ 68.2

661'
4415
5319
511
6015
7312

18519

10411
7511
4719
511)
821'
7419
6419
8419
6319
6715
7219
7819
8211

1041'

10619
4412
61:7

10015
9712
651

.5315
481'
6119

11615

12112

20711

10519

1241'
6519
149
9711
761'

de#

44.01
44.02
44.03
45
46
47
48
49.01
49.02
50.01
50.02
51
52.01
52.02
53.02
53.03
53.04
54
55.01
55.02
57
65

101.02
102.01
102.03
102.04
111.02

201.01

201.02
201.03
201.04

202.01

202.02

203.01

203.02

214

. UDA stats
UA stats

UDA=user-defined area.
UA=Census Bureau's Lansing-East Lansing Urbanized Area

Persons
10450

3039
2930
3536
2686
2718
6924
5723
3889
4240
3500
3502
4771
1804
4742
4878
3627
7701
3870
4268
4562
4059
2190
2287
1793
1273
3064
2463
2747
4178
4014
2588
4505
5178
3951
7586

289509
265151

0.2
38.0
51.4
83.5

166.9
101.2
97.5
126.9
149.6
188.2
174.5
58.0
50.6
93.2
92.2
65.2
60.7
75.6
110.7
93.5
119.8
49.9
100.4
87.0
50.9
86.9
123.1
170.5
85.2
104.6
85.9
118.0
73.2
110.3
117.3
99.3
77.9
76.6

RCls Calculated by household:
2-component 4-component

16.9
38.0
51.4
83.5
166.9
101.5
97.5
126.9
149.6
188.2
176.2
58.1
50.6
93.2
92.3
65.2
61.0
75.6
111.3
93.6
119.8
49.9
100.4
87.3
50.9
86.9
123.1
170.5
85.2
105.3
85.9
118.0
73.2
111.3
117.3
101.5
79.3
78.1

0.5
44.8
58.5
95.7

180.5
113.0
110.1
124.7
155.7
160.8
154.2
52.0
54.2
92.8
99.5
70.3
70.2
77.4
108.2
86.7
109.4
56.8"
126.9
88.8
59.1
74.8
126.1
147.5
113.6
103.6
87.8
121.8
72.9
118.3
107.0
101.3
82.7
80.5

Per capita RCI calculations:
2-component 4-component

17.3
44.8
58.5
95.7
180.5
113.3
110.1
124.7
155.7
160.8
155.9
52.1
54.2
92.8
99.6
70.3
70.5
77.4
108.8
86.7
109.5
56.8
126.9
89.0
59.1
74.8
126.1
147.5
113.6
104.4
87.8
121.8
72.9
119.3
107.0
103.5
84.1
81.9

APPENDIX 8

Component RCI ratings: Summary tract ratings:

Tract # Persons gqiRCl gqniRCl roRCI ooRCI RCIt RCI%

1 2159 0.0 0.0 56.6 44.0 47.9 60.4
2 1575 0.0 0.0 42.4 84.3 58.5 73.8
3 2705 36.2 0.0 52.6 45.1 48.2 60.8
4 3526 0.0 0.0 56.5 67.8 64.2 80.9
5 2140 51.2 0.0 56.7 47.3 50.9 64.2
6 2162 4.5 49.6 48.6 75.3 53.2 _ 67.1
7 3032 0.0 3.9 47.6 49.9 48.0 60.5
8 3698 0.0 0.0 53.1 41.0 46.5 58.6
9 2208 0.0 0.0 62.8 55.0 57.9 73.0
10 2523 0.0 0.0 57.1 61.7 60.1 75.8
12 2667 0.0 0.0 55.3 41.0 47.6 60.1
13 1598 0.0 20.1 39.4 36.1 38.2 48.1
14 207 29.3 25.0 23.8 89.1 27.8 35.1
15 2525 0.0 0.0 43.7 52.2 45.9 57.9
16 1331 0.0 0.0 63.2 83.1 79.8 100.6
17.01 1105 0.0 0.0 72.5 170.1 156.2 197.0
17.02 3878 0.0 0.0 54.9 106.5 95.2 120.0
19 739 0.0 150.0 36.7 , 83.0 53.1 66.9
20 4258 0.0 0.0 47.6 38.0 1 43.8 55.3
21 2429 0.0 0.0 56.0 45.1 50.1 63.1
22 1866 0.0 0.0 51.6 78.3 72.9 91.9
23 3605 0.0 0.0 56.8 63.9 62.6 78.9
24 3563 0.0 0.0 53.7 61.0 59.0 74.4
25 2541 0.0 . 0.0 58.1 74.2 71.8 90.6
26 2413 0.0 22.5 52.1 56.6 55.0 69.3
27 3371 0.0 . 0.0 60.5 63.1 62.7 79.1
28 2891 0.0 0.0 50.5 65.6 62.3 78.5
29.01 3329 27.2 0.0 62.0 88.0 75.8 95.5
29.02 3406 0.0 0.0 57.3 56.6 57.3 72.2
31.01 1302 0.0 0.0 53.0 91.1 64.6 81.5
31.02 3406 0.0 0.0 63.8 112.2 105.6 133.1
32 2573 0.0 97.1 42.6 47.8 46.6 58.8
33.01 3360 0.0 0.0 43.4 69.6 57.4 72.4
33.02 2625 0.0 0.0 62.0 102.4 100.2 126.4
34 2614 0.0 0.0 53.0 93.3 84.0 106.0
35 3017 0.0 0.0 55.3 55.7 55.5 70.0
36.01 4312 0.0 0.0 52.9 64.4 61.9 78.0
36.02 4075 0.0 0.0 41.6 60.3 51.7 65.2
37 5944 24.7 0.0 42.8 65.3 56.1 70.7
38.01 4555 0.0 0.0 62.7 145.9 114.5 144.3
38.02 2859 0.0 0.0 70.9 118.3 85.4 107.7
39.01 2214 47.3 0.0 62.8 285.1 206.8 260.8
39.02 4286 81.1 18.2 60.8 131.3 91.4 115.3
40 ~ 4175 0.0 12.7 59.6 151.3 113.0 142.5
41 5017 148.3 17.2 83.7 103.2 67.9 85.6
42 5442 0.0 14.3 42.0 0.0 14.6 18.4
43.01 5836 0.0 0.0 86.0 107.8 96.8 122.0

43.02

3211

87.0

17.9

75.9

0.0

70.6

89.1

de#

44.01
44.02
44.03
45
46
47
48
49.01
49.02
50.01
50.02
51
52.01
52.02
53.02
53.03
53.04
54
55.01
55.02
57
65

101.02

102.01

102.03
102.04
111.02

201.01

201.02
201.03
201.04

202.01

202.02

203.01

203.02

214

UDA stats
UA stats

UDA=user-defined area,
UA=Census Bureau's Lansing-East Lansing Urbanized Area

Persons
10450

3039
2930
3536
2686
2718
6924
5723
3889
4240
3500
3502
4771
1804
4742
4878
3627
7701
3870
4268
4562
4059
2190
2287
1793
1273
3064
2463
2747
4178
4014
2588
4505
5178
3951
7586

289509
265151

Component RCI ratings:
gqniRCl

gqiRCl

0.0
0.0
0.0
0.0
0.0

, 0.0
28.0
0.0
0.0
0.0
28.8
44.9
0.0
0.0
0.0
0.0
15.0
0.0
24.4
15.4
19.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
30.9
0.0
36.3

43.9 -

46.9

16.8
0.0
0.0
0.0
0.0

27.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

22.5
0.0
0.0
0.0

30.7
0.0
0.0
0.0
0.0

22.8
0.0
0.0
0.0
0.0
0.0

28.1
0.0
0.0
0.0

135.5
4.5

62.5

17.6

17.5

roRCI
40.1
37.0
42.5
67.9
77.9
44.0
60.3
64.9
64.9
72.9
37.0
46.6
46.7
46.0
69.3
43.0
55.1
53.0
71.1
51.9
67.2
48.7
63.6
62.3
41.9
65.2
50.4
67.0
65.5
49.3
66.5
62.3
58.6
69.6
57.0
48.0
56.0
56.5

ooRCI

49.5
148.5
103.3
103.1
179.8
133.1
118.1
153.9
206.2
220.6
192.6

62.2

54.2
101.8
104.8

75.9

80.9

85.4
116.2

100.9

123.7
51.5
106.6
90.3
54.2
90.0
134.6
171.3
127.6
131.4
91.3
128.6
84.8
129.6
120.3
111.7
100.0
100.0

Summary tract ratings:
RCIt RCI%

16.9 21.4
38.0 47.9
51.4 64.8
83.5 105.3
166.9 210.5
101.5 128.0
97.5 123.0
126.9 160.0
149.6 188.7
188.2 237.3
176.2 222.2
58.1 73.3
50.6 63.8
93.2 117.6
92.3 116.3
65.2 82.2
61.0 77.0
75.6 95.4
111.3 140.4
93.6 118.0
,1 19.8 151.1
49.9 62.9
100.4 126.6
87.3 110.1
50.9 64.2
86.9 109.5
123.1 155.3
170.5 215.0
85.2 107.4
105.3 132.8
85.9 108.3
118.0 148.8
73.2 92.3
111.3 140.4
117.3 148.0
101.5 127.9
79.3 100.0
78.1 100.0

de#
42
44m
14
4M02
13

20

15

32
12

65

21
521“
102.03

4403
2K502
19

26
35
37
29112
3301

51

24
10
5304
361”
28
23
27

3101
5303
41
«#302
25

22
202.02
54
291M
16

APPENDIX C

Summary tract ratings:

RCHt

$46
1619
2713
381)
3812
4313
4519
4615
4615
4715
4719
4819
4812
4919
5031
5015
5019
5019
5114
51f7
531
5312
551)
5515
56:1
5713
5711
5719
58:1
5815
591)
6031
611)
6119
6213
6215
62f7
6412
6415
6512
6719
7015
71.8
7219
7312
7515
7513
7913

RCH96

1811
214
351
4719
481
5513
579
586
588
601
604
605
608
629
631
638

642‘

6412
6413
6512
6619
671
6913
7019
707'
7212
7214
731)
7313
7313
7414
7513
7719
781)
7815
7819
791
801)

815 '

8212
8515
891
9015
‘919
9213
'9511
9515
10015

Persons

5442
10450
207
3039
1598
4258
2525
3698
2573
2667
2159
3032
2705
4059
2429
4771
1793

2140.

2930
4075

739
2162
2413
3017
5944
3406
3360
2208
3502
1575
3563
2523
3627
4312
2891
3605
3371
3526
1302
4878
5017
3211
2541
1866
4505
7701
3329
1331

5442
15892
16099
19138
20736
24994
27519
31217
33790
36457
38616
41648
44353
48412
50841
55612
57405
59545
62475
66550
67289
69451
71864
74881
80825
84231
87591
89799
93301
94876
98439

100962
104589
108901
111792
115397
118768
122294
123596
128474
133491
136702
139243
141109
145614
153315
156644
157975

Cumulative percentile

A
OCDNVCDU’IN

11
12
13
13
14
15
17
18
19
20
21
22
23
23
24
25

' 26

28
29
30
31
32
33
34
35
36
38
39
40
41.
42
43

46
47
48
49
50
53
54
55

Tract #
45

34
201.02
381x:
201.04
102.04
102.01
39112
53112
52112
55112
1702
431M
48
3302
101.02
214
47
201.03
3102
203.01
551M
40
381M
203.02
202.01
57
111.02
491M
4K102
1701
46
201.01
£fl102
501M
391M

'UDA stats

Summary tract ratings:

RCHt

8315
841)
8512
8514
8519
861)
8713
9114
9213
9312
9315
9512
9613
975
10012
10014
1CH.5
1CH.5
10513
10515
111.3
111.3
1131)
1145
1173
1181)
11913
1231
12619
14915
15612
16619
17015
17612
18812
20613
7913

UDA=user—defined area

R£H96

10513
1061)
10714
1077'
10813
10915
1101
11513
11613
11726
1181)
1201)
1221)
1231)
12614
12615
1271)
1281)
13213
1331
14014
14014
14215
14413
1481)
14813
1511
15513
1601)
18837
1971)
21015
2151)
22212
23713
26013
1001)

Persons

3536
2614
2747
2859
4014
1273
2287
4286
4742
1804
4268
3878
5836
6924
2625
2190
7586
2718
4178

3406.
5178 .

3870
4175
4555
3951
2588
4562
3064
5723
3889
1105
2686
2463
3500
4240
2214
289509

161511
164125
166872
169731
173745
175018
177305
181591
186333
188137
192405
196283
202119
209043
211668
213858
221444
224162
228340
231746
236924
240794
244969
249524
253475
256063
260625
263689
269412
273301
274406
277092
279555
283055
287295
289509

Cumulative percentile

56
57
58
59
60
60
61
63
64
65
66
68
70
72
73
74
76
77
79
80
82
83
85
86
88
88
90
91
93
94
95
96
97
98
99
100

93

REFERENCES

Anderson, Fred B., and Dale R. Funderbunk. "Real Estate Valuation: A Pedagogical
Approach" in Classic Readings in Real Estate and Development, ed. Jay M. Stein.
Washington, DC: Urban Land Institute, 1996.

Ehrenreich, Barbara. Fear of Falling: The Inner Life of the Middle Class. New York:
Pantheon, 1989; HarperPerennial, 1990.

Fussell, Paul. Class: A Guide Through the American Status System. New York: Simon
& Schuster, 1983; Touchstone, 1992.

Gans, Herbert J. The War Against the Poor: The Underclass and Antipoverty Policy.
New York: Basic Books, 1995.

Gilbert, Dennis, and JosephA. Kahl. The American Class Structure. A New Synthesis,
4th ed. Belmont, CA. Wadsworth, 1993.

Hartigan, John, Jr. Racial Situations: Class Predicaments of Whiteness in Detroit.
Princeton, NJ: Princeton University Press, 1999.

Harvey, David. "Labour, Capital, and Class Struggle Around the Built Environment in
Advanced Capitalist Societies" in Classes, Power, and Conflict: Classical and
Contemporary Debates, eds. Anthony Giddens and David Held. Berkeley, CA:
University of California Press, 1982.

Haynes, Kingsley E., and A. Stewart Fotheringham. Gravity and Spatial Interaction
- Models. Scientific Geography, ed. Grant Ian Thrall, vol. 2. Beverly Hills, CA:
Sage,1984.

Hess, Beth B., Elizabeth W. Markson, and Peter J. Stein. Sociology, 4th ed. New York:
Macmillan, 1991.

Jacobs, Jane. The Death and Life of Great American Cities, Modern Library ed. New
York. Random House, 1993.

Kerbo, Harold R. Social Stratification and Inequality. Class Conflict in Historical and
Comparative Perspective, 2nd ed. New York: McGraw- Hill, 1991.

King, Leslie J. Central Place Theory. Scientific Geography, ed. Grant Ian Thrall, vei. 1.
Beverly Hills, CA: Sage, 1984.

Klosterman, Richard E. Community Analysis and Planning Techniques. Savage, MD:
Rowman & Littlefield, 1990.

Ley, David. A Social Geography of the City. New York: Harper Collins, 1983.

94

Lyons, Donald, and Scott Salmon. "World Cities, Multinational Corporations, and Urban
Hierarchy: The Case of the United States" in World Cities in a World System, eds.
Paul L. Knox and Peter J. Taylor. Cambridge: Cambridge University Press, 1995.

Merton, Robert K. Social Theory and Social Structure, enlarged ed. New York: The
Free Press, 1968.

Meyersohn, Rolf, and Donal Malone. "Social Meanings of Second Homes for Urban
Dwellers" in The Apple Sliced: Sociological Studies of New York City, eds.
Vernon Boggs, Gerald Handel, and Sylvia Fava. Prospect Heights, IL: Waveland
Press, 1984.

Mills, C. Wright. The Sociological Imagination. London: Oxford University Press,
1959.

Muller, Peter O. Contemporary Suburban America. Englewood Cliffs, NJ: Prentice-
Hall, 1981.

Nettler, Gwynn. Explaining Crime, 3rd ed. New York: McGraw-Hill, 1984.
Palen, J. John. The Urban World, 4‘h ed. New York: McGraw-Hill, 1992.

Roberts, RObert. "The Class Structure of the 'Classic Slum'" in Class, ed. Patrick Joyce.
Oxford: Oxford University Press, 1995.

Rusk, David. Cities Without Suburbs, 2“d ed. Washington, D.C.: The Woodrow Wilson
' .Press, 1995.

So, Frank S., and Judith Getzels. The Practice of Local Government Planning, 2nd ed.
Washington, D.C.: International City Management Association, 1988.

Turner, Jonathan H. Leonard Beeghley, and Charles H. Powers. The Emergence of
Sociological Theory, 3rd ed. Belmont, CA: Wadsworth, 1995.

Veblen, Thorstein The Theory of the Leisure Class. London: Viking, 1967; Penguin,
1.994

Wallerstein, Immanuel. Historical Capitalism. LondonzVerso, 1983.

Wallerstein, Immanuel, with Joan Smith. "Households as an Institution of the World-
Economy" in The Essential Wallerstein. New York: The New Press, 2000.

llllllllllllllllllllllllllllllllllllllllJIHIUIHIIIIHIHHII

 

31293 02637 8244