University of Southern California Dissertations and Theses

Investigating the association of historical preservation and neighborhood status in Detroit, 1970-2015

(USC Thesis Other)

Investigating the association of historical preservation and neighborhood status in Detroit, 1970-2015

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

Investigating the Association of Historical Preservation and Neighborhood Status
in Detroit, 1970-2015

by

David Rosas Flores

A Thesis Presented to the
Faculty of the USC Graduate School
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
Master of Science
Geographic Information Science and Technology

December 2018

Copyright © 2018 by David Rosas Flores

To my mother

iv
Table of Contents
List of Figures ................................................................................................................................ vi
List of Tables ................................................................................................................................ vii
Acknowledgements ...................................................................................................................... viii
List of Abbreviations ..................................................................................................................... ix
Abstract ........................................................................................................................................... x
Chapter 1 Introduction .................................................................................................................... 1
1.1 Detroit Historical Districts ...................................................................................................1
1.1.1. Implementation ..........................................................................................................2
1.2 Motivation ............................................................................................................................4
1.3 Research Goals .....................................................................................................................5
1.4 Study Organization ..............................................................................................................6
Chapter 2 Background .................................................................................................................... 7
2.1 History ..................................................................................................................................7
2.1.1. Historical Background ...............................................................................................7
2.1.2. Black Migration .........................................................................................................9
2.1.3. Rise of the Industrial City ........................................................................................10
2.1.4. Decline .....................................................................................................................11
2.2 Historic Preservation Research ..........................................................................................14
2.2.1. Research about neighborhood quality ......................................................................15
Chapter 3 Methodology ................................................................................................................ 18
3.1 Data Sources ......................................................................................................................18
3.1.1. Indicators of Neighborhood Status ..........................................................................20
3.1.2. Interpolation .............................................................................................................23
3.1.3. Extent of Historical Protections ...............................................................................27
3.1.4. Difference-in-Differences Analysis .........................................................................31
3.1.5. Regression Analysis (Temporal Progression) ..........................................................33
Chapter 4 Results .......................................................................................................................... 35
4.1 Hot Spot Analysis ..............................................................................................................35
v
4.2 Difference-in-differences ...................................................................................................36
4.2.1. Existing Designations ..............................................................................................37
4.2.2. Area Percentage in a Census Tract ..........................................................................39
4.2.3. Building Structure Count .........................................................................................41
4.3 OLS Regression Results.....................................................................................................44
4.4 Spatial Autocorrelation ......................................................................................................47
Chapter 5 Conclusion .................................................................................................................... 49
References ..................................................................................................................................... 56

vi

List of Figures
Figure 1 City of Detroit, Michigan ................................................................................................. 3

Figure 2 Interpolation diagram (NHGIS datasets) ........................................................................ 25

Figure 3 Example of census tracts that experienced change between 1980 and 2015 ................. 26

Figure 4 Building structures hot spot analysis .............................................................................. 36

Figure 5 DiD results: experimental group recovery trend for existing designations .................... 38

Figure 6 DiD results: experimental group stabilizing tend for existing designations .................. 39

Figure 7 DiD results: experimental group recovery trend for historic designation coverage of
<1% ................................................................................................................................ 40

Figure 8 DiD results: experimental group recovery trend for historic designation building count
of <100 ........................................................................................................................... 42

Figure 9 DiD results: experimental group recovery trend for historic designation building count
<10. ................................................................................................................................ 43

Figure 10 Natural logarithm (LN) historic district area coverage percentage results ................... 47

vii
List of Tables
Table 1 Historic districts and protected census tracts over the study period ................................ 19

Table 2 Census tracts from 1970-2015 ......................................................................................... 23

Table 3 Number of census tracts that experienced change ........................................................... 27

Table 4 Census tracts total ............................................................................................................ 31

Table 5 Summary of difference-in-differences trends by varying spatial thresholds ................... 37

Table 6 OLS regression output ..................................................................................................... 45

Table 7 Spatial autocorrelation results .......................................................................................... 48

viii
Acknowledgements
I am grateful to my thesis advisor, Professor Vos for providing me with guidance and
challenging me throughout this research, my committee members Professor Su Jin Lee and
Professor Katsuhiko Oda, and to Professor Darren Ruddell and the rest of the faculty at the
Spatial Science Institute for their teachings. I would like to thank my longtime friend Bertram
Riddick for giving me access to learning spaces and Professor Alan Green for providing a critical
opinion on my work. I am grateful for the data provided to me by the National Historical
Geographic Information System, Director Dave Van Riper, my friend Joeleo Smith and to the
City of Detroit, a resurgent city.

ix
List of Abbreviations
ACS American Community Survey
DiD Difference-in-differences Analysis
GIS Geographic Information System
MAUP Modifiable Areal Unit Problem
NHGIS National Historical Geographic Information System
OLS Ordinary Least Squares
SEMCOG Southeast Michigan Council of Governments
SHPO Michigan State Historic Preservation Office
UAW United Automobile Workers

x
Abstract
Cities throughout the United States have adopted historical designations in order to
protect the historic architectural resources and promote economic development of areas that
carry a cultural significance to their communities. Detroit, a city in steep economic decline from
1970 until 2015, has also attempted to use historical preservation to promote economic
development in particular neighborhoods. The role of historic preservation has rarely been
considered in a city in steep, ongoing economic decline. The study presents spatial analysis
techniques that can help determine the association, if any, of historical district designations with
neighborhood rise or fall. By using various approaches to count structures and measure preserved
space within census tracts, a difference-in-differences (DiD) analysis and an ordinary least
squares regression model was developed to test the association of preservation and neighborhood
status change from 1970 to 2015.
The results indicate that census tracts with a historic designation showed less decline and
quicker improvement in neighborhood status when compared to census tracts with no
protections. To further corroborate DiD results, ordinary least squares analysis indicated
statistically-significant relationships between the percentage of historical district coverage and
historically protected building counts and changes in all but two indicator values, but these
results should not be accepted as evidence of causality because there is significant spatial
autocorrelation. Also, Detroit’s socio-economic conditions differ from other metropolitan
statistical areas in the U.S. Further research is needed in other cities during periods of economic
decline and with additional control variables.

1

Chapter 1 Introduction
The adoption of Detroit’s historical district ordinance 161-H in 1976, allowed the City of
Detroit to protect historical resources from a variety of perspectives. In the words of the statute,
the law preserves areas of the city with historical resources of the following types: “cultural,
social, spiritual, economic, political, engineering, or architectural history or its archeology” (City
of Detroit 1976, Sec. 25-2-1). This study focuses on the ways in which the law preserves the
built environment, particularly historically significant buildings, and the affect this may have on
economic and social status of neighborhoods over time.
This study presents spatial analysis techniques that can help determine the effects, if any,
of historical district designations on neighborhood rise or fall. By using difference-in-differences
analysis and an ordinary least squares (OLS) regression model, the research investigates the
relationship between census tracts that contain some type and degree of historical protections as
a treatment group, pairing them against census tracts that contain little or no historical
protections as a control group. Analyzing decennial census and American Community Survey
(ACS) data, the research utilizes four indicators of neighborhood status: median household
income, vacancy percentage, owner occupancy percentage, and percentage of white population.
If census tracts experiencing various forms of historical protections rise in status or decline in
status less than census tracts that did not have protections, this provides evidence of the impact
of the historical designations in the law.
1.1 Detroit Historical Districts
Throughout cities in the United States, a historical district designation is generally
assigned to a particular area as a subset of land use regulation in order to preserve a unique

2

identity to the area (Costonis 1989). In most cases, it is guided by historic preservation of
landmarked buildings. In New York City, for example, a commission is in place to review the
appropriateness of any work performed on the exterior of buildings within a historic district
(Allison 2005). This maintains the historic character of the neighborhood in the physical sense.
Detroit’s historical districts originate from the outcome of a failed appeal to the zoning
board, in order to avoid the demolition of two houses across the street from the house of Beulah
Groehn Croxford in 1965. Her activism, and that of a few neighbors, initiated the process of
exploring what it would take to obtain historical protections for their own neighborhood of West
Canfield, Detroit. To many, she deserves to be named the mother of historic preservation in
Detroit (Bragg 2015).
As this was something that had never been done before in Detroit, with no legal
precedents at the local or state level, it was necessary to validate such designation by creating a
city ordinance. Detroit adopted historic district ordinance 161-H in 1976, allowing the city to
regulate the construction, reconstruction, addition, alteration, repair, moving, excavation, and
demolition of resources in historic districts within city limits. The ordinance allows for historic
preservation to be declared a public purpose, protecting buildings of significant value to a
community and safeguarding the cultural heritage of the city (City of Detroit, ordinance 161-H).

1.1.1. Implementation
A panel of seven residents of local historic districts appointed by the mayor are given the
task to implement the purpose of ordinance 161-H. The Historic District Commission’s purpose
(HDC) is to ensure the preservation of historically and culturally significant areas in the city
(Planning and Development Department 2016). Upon receiving the request by a person or

3

business of a local area, the city council may adopt a resolution directing the advisory board to
perform a study to determine whether that area meets the criteria to become a historic district.
Upon receipt of substantial evidence demonstrating definite significance for designation under
the provisions described in ordinance 161-H, the city council, at its discretion, may first adopt a
resolution for interim historic designation, and later on upon the recommendation of the advisory
board assign permanent historical district status (Detroit Section 25-2-4)
Furthermore, the commission promotes the use of historic districts for the education,
pleasure, and welfare of city residents and aims for the stabilization and improvement of
property values in historic districts and its surrounding neighbors. The commission attempts to
foment community pride and strengthen the local economy.

Figure 1 City of Detroit, Michigan

4

While the historic district advisory board surveys and researches the proposed areas, the
city council by ordinance may establish additional designations, reject proposed areas, eliminate
historic designation area, and modify boundaries of existing historic districts. As of 2015, based
on data provided by the Planning and Development Department, Detroit has 140 historic district
designations. As seen in Figure 1 above, historic districts are conglomerated primarily around
downtown Detroit, and each historic district has unique spatial characteristics, varying in the area
covered from several city blocks to as small as one building. This study in Chapter 3 explores
different measures for establishing the extent of historic protections to define experimental
versus control neighborhoods. The measures were created by assigning historic designations
based on spatial conditions instead of being based solely on the provided boundary delimitations.
1.2 Motivation
Over the past 60 years, hundreds of communities across the Rust Belt have lost
population (Bertron 2013). Left in a state of peril, as problems compounded upon each other,
Detroit went from a booming city to a city in obvious decline. In 1955, at the height of its
success, almost two million people lived and worked in Detroit, yet by 2015 the population had
shrunk to approximately 675,000 (Farr 2011).
Detroit, like many other cities in the region, is searching for strategies to manage vacant
land and abandoned properties. Historic preservation is touted as a way of revitalizing
neighborhoods that are deemed eligible due to the location of historically significant building
structures within its boundaries (Bertron 2013). Designation and preservation of historic districts
is believed to be an important tool in efforts to promote economic development in blighted urban
areas (Coulson et al. 2003).

5

Detroit’s deep economic and social divide makes the study area a unique case study in
that many indicators are so low that a minimal rise or slowed decline related to historical
protections would be considered a positive outcome towards recovery. Coulson et al. (2003)
utilize Fort Worth, Texas as a study area due to the extent to which historical designations had
been implemented. The study area closely resembled the demographical statistics of other
metropolitan areas in the U.S., so it can be argued that the results can be used as an indicator of
effects of historical protections. The Metro Detroit statistical area in contrast is so far skewed
negatively when compared to the U.S. average that results have to be analyzed only for that
region.
Previous studies have compiled a strong theoretical articulation utilizing census
indicators as means of measuring neighborhood quality, nonetheless existing literature makes
relatively limited use of GIS technologies. To contribute to existing literature on the topic of
historic district designations and the effects on neighborhood rise, fall or stasis, this study will
explore various spatial analysis techniques to uncover, significant relationships, if any, between
neighborhood quality and preservation of the built environment.
1.3 Research Goals
The research aims to investigate viable methods to spatially associate preservation of
historical structures with neighborhood rise, fall, or stasis in the city of Detroit. The study tests
the hypothesis that census tracts that contain within their boundaries some degree or type of
historical protection experienced greater improvement, remained unaltered, or declined less in
the face of broad declines in neighborhood status throughout Detroit. The hypothesis is tested
using a variety of spatial methods for gauging the impact of the historical designation within

6

each census tract. The spatial analysis is used to explore different ways of defining the extent to
which neighborhoods are associated with preservation of historical structures.
1.4 Study Organization
This study contains four additional chapters. Chapter 2 begins with a brief overview of
the history of the City of Detroit. In order to clearly comprehend the study area, it is important to
understand what events occurred that set current living conditions. Also, it explores related
literature to identify methods and neighborhood quality indicators from similar or comparable
studies. Chapter 2 sets the foundation for the methodology that will be presented in Chapter 3, a
difference-in-differences analysis (DiD) and ordinary least squares (OLS) regression to
investigate the historical association between the concentration of historical structure
preservation in certain census tracts and neighborhood trajectories over a period from 1970 to
2015. Chapter 4 presents results obtained from the DiD analysis, as well as explores for
statistically-significant associations between historic preservation and neighborhood status.
Chapter 5 discusses data limitations, as well as drawing conclusions about the results,
understanding that such results should be interpreted as a case study of effects of historic
preservation. It also provides suggestions for future research.

7

Chapter 2 Background
In order to understand the circumstances and conditions that surround our study area, this
chapter undertakes a brief look at the history of Detroit and summarizes existing literature related
to historical preservation as it relates to neighborhood economic and social status.
The historical background section explores important events in the history of Detroit that
have influenced and shaped the study area. Ultimately presenting how historical events relate to
each of the chosen neighborhood status indicators.
The literature review exhibits previous works that have undertaken to study how historic
preservation relates to the condition of neighborhood status. It also discusses existing studies that
utilize rigorous data analysis methods to explore historic preservation and suggests how such
research might be expanded by incorporating the use of GIS technologies.
2.1 History
Detroit took the nickname “Motor City” due to its thriving automobile industry. At its
height around the 1950’s, the city boasted some of the highest wages in the U.S., highest single-
family home ownership percentage, and a population of over 1,800,000. As of 2015, its
population had plunged 63% since 1950, thousands of structures and lots had been abandoned,
and it had become the most racially segregated city in the U.S. (Boyle 2001). What caused such a
pronounced and rapid decline?
2.1.1. Historical Background
On July 24, 1701, an expedition of over 100 men at the command of Antoine Loumet de
Lamothe Cadillac, upon surveying the terrain on a narrow point of the Detroit River, set to

8

constructing a small fort: “Fort Pontchartrain du Detroit,” and thus the city of Detroit was born
(Woodford 2001).
Cadillac’s original commission was to build an outpost that would prevent British
encroachment on the region and allow the French to gain control of the rich fur trade in what is
now Michigan and then was part of the Northwest territory. Situated in the heart of the Great
Lakes region, near the eastern edge of the Great Central Plain of North America, over the years
the region prospered due to trading between French, British, and American Indian merchants
(Parkins 1918). While the population in the area grew slowly, it was not until 1805 upon the
establishment of the Territory of Michigan with Detroit as its capital by President Thomas
Jefferson, that the population began to flourish. The creation of a city plan and ordinances
defined the layout of how the city was to grow and gave the traditional spatial characteristics of
Detroit: spacious, large lots, with single-family homes. It helped Detroit to lead the nation among
major cities in the number of homes occupied by single owners. Ultimately this characteristic
also produced problems from which Detroit still suffers to this day, such as the lack of effective
transportation (Woodford 2001).
In 1835, white settlers in Michigan adopted a constitution, and declared themselves a
state. However, it was not until 1837 that Michigan having settled a territorial dispute with Ohio,
was finally admitted into the Union (Woodford and Woodford 1969). The two decades following
statehood were a time of continued growth. The population in Detroit reached 6,927 in 1836, by
1840 it hit 9,124, and by 1850 it had doubled. (Woodford 2001). There also was the beginning of
Detroit’s first real heavy industry, the manufacture of railroad cars. The industry employed a
number of skilled workmen and established the foundation of manufacturing plants from which
inventions and improvements of mechanical nature came out (Woodford and Woodford 1969).

9

Offering an abundance of machine-shop facilities and skilled labor, the thesis that Detroit was
well-equipped to offer the automobile industry a home is widely supported (Lewis and Goldstein
1983). Nonetheless, other contributing factors such as the availability of raw materials, access to
transportation routes (land or water), development of new technology, and a large stock of
immigrant labor played an important role. This was particularly important during the U.S. efforts
in the two World Wars, as Detroit’s industry was called upon to supply armaments, tanks, boats,
and chemicals (Booza and Metzer 2004).
2.1.2. Black Migration
A haven from slavery and Negrophobia that existed in other parts of the country, an
increase on black population began around 1840. From 193 Black residents in Detroit, the
number grew to 587, and the city provided a public sphere where Black communities and
families could thrive and prosper. The first permanent Black institutions in Detroit, were the
churches, and by 1846 the small community had three. The basements were used as colored
public schools, and their chapels as political halls. It is important to mention that the contiguous
geographical area in which the small Black population lived in 1860 permitted some
differentiation in residential patterns. In all wards, wealthy Black owners of real estate clustered
together, apart from the renters. Housing units for the rest of the Black population were in poor
condition, and many of them lived in old dwellings, and reflecting the working-class level of
many workers (Katzman 1973).
The “Great Migration” in the two decades between 1910 and 1930 brought a large
number of black southerners into Northern cities. One of the most significant demographic
events to occur in the U.S. during the twentieth century, it produced a dramatic redistribution of
the black population (Tolnay 2003). In 1910 the black population in Detroit stood at 1.2% of the

10

total population, by 1930 that number had increased to 9.1% (Gavrilovich and McGraw 2000).
Migration from southern states allowed workers steady work, but also favorable worker laws
(Farr 2011). With the adoption of more restrictive immigration policies by the U.S. during World
War I, northern employers in order to fulfill the shortage in the labor force had to consider
southern blacks as a source of inexpensive labor (Tolnay 2003). The “Great Migration”
contributed in large part to the increase of Detroit’s black population during the first part of the
20
th
century, while the total numbers and percentages continued to increase at a steady pace in
later decades, another contributing factor for the increase was the exodus of white population
starting in the 1950’s. By 1970, the percentage of black population had increased to 44.5% of the
total population (Gavrilovich and McGraw 2000).
2.1.3. Rise of the Industrial City
The shift in population from the core of the city to the outer rings occurred at the time of
Detroit’s greatest population expansion in the 1890’s. A conglomeration of ethnic and racial
neighborhoods, with common community names, arise, such as: Dutchtown, Polacktown, and
Corktown (Irish immigrants from County Cork). The ties to birth, culture, and language bound
together ethnic colonies within a common geographical area, developing in most cases an
autonomous social system (Katzman 1973). Immigrants from all over the world began
descending upon the city, clustering around their own ethnic groups. Census records from 1880
show a breakdown of twenty-six different nationalities. Industrial development required a
substantial labor force, and while most factory jobs were taken by native-born Americans, as
demand grew so did the need for additional labor. The first automobile appeared on the streets of
Detroit in 1896, small mechanic shops throughout the city gave birth to an industry that would
later be revolutionized by Ford’s Model-T mass production on an assembly line. The automobile

11

industry would transform the city into the sixth largest city in the United States with a bustling
population and a skilled labor force with a steady income stream that in turn translated into rising
social status of Detroit’s neighborhoods (Woodford 2001).
The car industry led to quick urban growth as workers moved to the city. An example of
this took place in the neighborhood of Highland Park, which was later designated a historic
district, the neighborhood grew from 400 to 40,000 residents between 1900-1920. With 139
square miles, Detroit by the 1920’s had grown to the size (i.e., physical extent) it remains today
(Farr 2011).
A significant factor that takes place with all migration is the introduction of new cultural
elements that are assimilated by the population that currently reside in the area. While interaction
between new immigrants and local Detroiters took place in the workplace, for the most part the
city remained highly segregated. Originally at the end of the 1800’s, lines divided distinct ethnic
populations and there was little consensus understanding of race. As an example, by the 1920’s,
the U.S. census counted foreign-born of European descent and their U.S.-born children as white,
but in separate categories from those whose parents were U.S.-born. Third-generation
immigrants gained full membership to the white American race after the 1924 immigration
restrictions which ensured no interaction between them and new generation immigrants
(Roediger 2005).
2.1.4. Decline
At the height of Detroit’s economic golden age in the 1940’s and 1950’s, the automobile
industry boomed. The United Automobile Workers (UAW) union, at the height of its powers,
assured that all of its members, overwhelmingly white males, shared the prosperity. This made
its members amongst the highest paid industrial workers in the nation. To a lesser degree,

12

working class black Detroiters enjoyed the benefits of steady employment and higher wages
when compared to black workers in other parts of the country. Some among these black workers
had the opportunity to boosted into a prodigious middle-class.
In its pursuit of higher wages and benefits for its majority white members, in the 1940’s
the UAW abandoned the progressive agenda that once challenged corporate power in the 1930’s.
Dubbed the “Treaty of Detroit,” the UAW and the auto companies reached an accord that ruled
labor relations for the following three decades, giving control of investment decisions and plant
conditions to the automakers. Effectively splitting the working class in two, by the 1960’s, union
members in its majority white males enjoyed more security, while non-members, in its majority
blacks or females, received low wages and few benefits (Boyle 2001).
As homeownership increased after World War II throughout major metropolitan areas in
the US, local variations in the housing situation of blue-collar workers depended in part upon the
local industrial and wage structure. In comparison to other metropolitan areas, wages were
considerably higher in Detroit during the 1940’s as the steel and automobile industry provided
many secured and well-paid jobs (Harris 1990). The city flourished with new buildings in
downtown, and new housing tracts that allowed more families access to homeownership. Detroit
was the flagship of American industrialization (Boyle 2001).
Although incomes were roughly comparable between white and black auto workers, the
standard of living diverged the most at housing. There was simply insufficient decent housing for
black workers at any price. White workers had more control over housing patterns than they did
over employment practices. The single most important measure of success in America is
homeownership. By the 1940’s, first-and-second generation immigrants, who by then had

13

already assimilated as white, were more often homeowners than native Detroiters (Petterson
1979).
Median household income in Wayne County, where Detroit is located was 13% higher in
1950 when compared to the national median income (Social Explorer 2013). A strong middle
class, in its majority white, once made Detroit America’s capitalist dream town. Having fallen
the longest and the furthest, Detroit is often considered to be the greatest urban failure in the U.S.
(Binelli 2013).
Once the symbol of modernity’s great tool room in the first half of the twentieth century,
since the 1960’s Detroit has become the symbol of urban decay. Many historians and scholars
have struggled with the question of how its decline from prominence occurred so quickly. A
series of events drove to its boisterous decline, including a riot in 1967 that exacerbated the
existing racial divide, the subsequent collapse of the domestic automobile industry that destroyed
the city’s economic base, and the exodus of one third of its population (Boyle 2001).
While fear for crime and violence following the riots of 1967 contributed to the exodus to
the suburbs by Detroit’s white residents, the migration had actually started as early as the 1950’s.
However, the pace of the migration sped up later, so that between 1970 and 1980, more than
310,000 white residents moved to the suburbs, taking with them a significant portion of the city’s
economic base and also depriving the city of important professional service and leadership
(Woodford, 2001). It is important to emphasize that white flight alone was not the only cause for
the devastation of inner-city Detroit. However, it is important to note its significance since this
study uses the percentage of white population as an indicator of neighborhood rise, fall or stasis
within the city of Detroit.

14

Across the Rust Belt region, places that used to pride themselves on the affordability of
single-family houses now have thousands of empty buildings and vacant lots (Bertron 2013). The
value of property fell sharply, which fueled the exodus, and a population loss of about a million
people from 1950 to 1980. People left so quickly that many buildings were left abandoned (Farr
2011). A direct consequence of this phenomenon was the steady rise of the vacancy percentage
in our study area which reached slightly over 30% in 2015, in comparison to the 13% U.S.
vacancy rate for the first quarter of 2015 (Census Bureau 2015). The vacancy rate of particular
neighborhoods is another measure used in this study.
2.2 Historic Preservation Research
Preserving the built environment through the enforcement of local ordinances or historic
landmark designations has often been viewed to be in direct opposition to economic
development. Debunking this idea is a primary focus of existing historic sites preservation
literature which focuses on preserving existing building structures through reuse or
reclassification. Smaller-scale level grassroots efforts such as historical neighborhoods or
districts present a noticeable impact on the community that can later translate to improvements at
a city level (Farr 2011).
Historic Preservation can contribute an essential perspective to reshaping cities by
helping articulate their unique identity (Bertron 2011). According to the Michigan State Historic
Preservation Office (SHPO), historic preservation as a planning and economic development tool
enables communities to manage how they will grow and change.
An idea not to the liking of many communities presents the displacement of poor and
ethnic minorities as something necessary and worth undertaking. The increase in the number of
affluent and well-educated residents to historic district or protected areas enhances city revenue

15

by means of higher property taxes. Instead of placing the blame on government failure to
produce affordable housing, this argument presents the scenario that a city that attracts more
affluent residents can aggressively finance affordable housing (Byrne 2002).
In many instances historical sites preservation is believed to be directly linked to changes
to neighborhood’s economic and social status. However, through analysis of demographic and
economic data in a study that examined nine historic districts in New York City, Allison (2005)
showed that there is no indication of a direct relationship between neighborhood rise and either
historic district designation or the drive to create one. The research is composed of nine case
studies, each comparing two census tracts that encompass a historic district (treatment group)
with two neighboring census tracts (control group). While comparing census data indicators
changes over a period from 1950 to 2000, the study draws conclusions based on a small
percentage of historic districts. New York City has over 80 historic designations. It also only
utilizes a simple on or off measure for whether a census tract has a historic designation that does
not involve the use of GIS.
2.2.1. Research about neighborhood quality
The complexities involved in obtaining and geocoding historical addresses and measuring the
impact of preservation on the neighborhood social environment have rarely been undertaken
(Rose et al. 2004). Determining the effects of revitalization efforts through historical districts
over a period of time requires a comparison analysis of indicators to determine how each one has
affected neighborhood status. In a study that took place in Salt Lake City, Utah from 1992 to
2000, a natural experiment stemming from a neighborhood revitalization project allowed
researchers to produce neighborhood maps with a gradation scale for each neighborhood status
variable at the property level and aggregate block level. The intervention focused on an area that

16

was declining, and its new direction to alleviate further deterioration of urban fabric (Perkins et
al. 2009). The study compared 60 blocks from two adjacent and similar neighborhoods with
propensity score matching and utilized three sources of data: a resident survey, an environmental
assessment, and archived issued building permits. A strength of this study was the use of GIS to
analyze large cluster samples and geocoding of all data sources to select through spatial analysis
neighborhoods with similar spatial characteristics.
Urban expansion is driven by demographic and economic factors (Liu et al 2005) and is a
traditionally a transformation process from a predominantly agricultural society to a modern
metropolitan society. Studies on urban sprawl have primarily focused on the use of Landsat data
comparing satellite imagery between different periods of time. While studying spatio-temporal
patterns in China, Liu et al., identified through visual interpretation of Landsat data, modes of
urban land expansion. During the period between 1990-1995, most of the urban expansion came
from cultivated land. This particular phenomenon is inverted from the process that is currently
taking place in Detroit. As the metropolitan area experienced a rapid decline in quality of life due
to loss of jobs, and population migration to other more prosperous or solvent cities, many areas
in the city became abandoned. Long-term population loss is recognized as a major challenge in
older industrial cities throughout the rust belt (Bertron 2011). Attempting to bring a higher
quality of life in the face of sustained loss of population, some cities have policies and programs
to ease restrictions and speed up building permits processes. While the attempt is to rapidly
transform the affected areas, the policies only take into consideration an economic stand point,
neglecting historic preservation elements that are important for the current local population
(Bertron 2011).

17

Researchers often use income and housing value taken from census data as an indicator
for neighborhood quality. Unfortunately, it is also limited because it only offers nationwide and
metropolitan level summaries not providing more granular geographic data (Holzer 2017). Using
census tracts as the unit of observation, five demographic and housing indicators were examined
in order to measure historic preservation and neighborhood change in Fort Worth, TX between
1990 and 2000 (Coulson et al. 2004).
Another study by Leichenko et al. (1999), in which Coulson is involved, expands on the
idea that historic designations have a positive impact on property values and analyzes the effect
across several cities in the state of Texas. Using multivariate regression models to assess the
impact of historic designations on residential property values, the authors are able to establish
basic explanatory variables and create a matrix that assigns a value to additional variables that
compensate for any differences in square footage and room composition.
There is a limited number of studies that explore effects of historic designations
utilizing GIS. Research for the most part focuses on statistical computations that do not take into
consideration spatial characteristic other than the designated boundaries. The study reported
below in contrast explores different spatial analysis techniques that will consider additional
variables, such as the areal extent of coverage of historically-designated space within census
tracts and the counts of historical building structures to which historical protection applies to
differentiate treatment and control groups.

18

Chapter 3 Methodology
One purpose of this study is to present various spatial analysis techniques that can help
determine the effects, if any, of historical district designations on neighborhood rise or fall. The
following sections offer an overview of the research design and particular spatial analysis
methods employed, as well as present potential issues that may arise when comparing spatial
data over a period of time.
3.1 Data Sources
The study analyzes four census indicators as measures of neighborhood status over time:
median household income, owner occupied percentage, vacancy percentage, and percentage of
white population. A common phenomenon that occurs in comparing census tracts from one
historical period with another is that they can be split or consolidated, and their boundaries
altered in complex ways based on population totals (Logan et al. 2014). To overcome the
obstacle of presenting historical data rendered on different spatial boundaries, interpolation by
area weights method was utilized to overlay and match to 2010 census tracts. Further details are
given below.
Overall, the study first employs a difference-in-differences analysis to compare census
tracts that contain historical protections and aims to examine the impact on neighborhood rise or
fall against census tracts that do not have such protections. Second, the study uses regression
analysis to determine whether any associations that are observed result from spatial
autocorrelation.
The National Historical Geographical Information System (NHGIS) provided the main
source of data for this study for the four variables that represent neighborhood status. The

19

NHGIS draws from the decennial census data for the years 1970, 1980, 1990, and 2000. The
study also uses American Community Survey (ACS) data for the years 2010 and 2015. The
NHGIS was selected as a source due to its vast historical archive of decennial data, and the
interpolation methods used to extract historical census data.
Decennial data provided by the NHGIS does not offer data summarized at a smaller scale
than census tracts for years prior to 1980. While it was explored if data summarized at the census
block group level better aligned with the boundaries of local historic districts, due to the
unavailability of data at that scale, it was necessary for this study to utilize census tracts as a
study unit. As local historic district designations in the city of Detroit first started with the
approval of ordinance 161-H in 1976, the study explores the period from 1970-2015. The time
frame provides a solid base that takes place prior to the origination of historic district
designations, providing a starting point as well as a long interval of time that historically covers
the city’s decline.
Table 1 Historic districts and protected census tracts over the study period
Year Historic Districts Census Tracts
1980 22 19
1985 44 27
1990 54 32
1995 62 40
2000 74 50
2005 101 56
2010 123 64
2015 139 73

The number of historic districts periodically increased as new designations were
approved by the City council. Table 1 presents the cumulative total number of historic districts
established and the total number of associated census tracts with 2010 boundaries every five

20

years from 1980 to 2015. As new areas continued to gain protections throughout the study
period, it was necessary to establish a threshold year for the DiD analysis that would allow
sufficient time had passed in order to evaluate any effects of historic designations. The historical
protections included in the DiD analysis for this study came into place by the end of the year
1995. For the purposes of the DiD analysis, this means that census tracts have a minimum 20
years to perceive whether historical protections influence changes in neighborhood status. They
represent 44% of the current historic districts in the City of Detroit. Note below that for the OLS
analysis, the entire set of historical protection districts was used.
NHGIS also provided geometry files for GIS as shapefiles that can be rendered utilizing
GIS software. During this study, ArcMap 10.5 was the software of choice. All boundary files are
derived primarily from U.S. Census Bureau’s TIGER/Line files with numerous additions to
represent historical boundaries that do not appear in TIGER/Line files. NHGIS also erases
coastal water areas in all years to produce polygons that terminate at the coasts and Great Lakes
shores (NHGIS 2018).
The extent of historical protections in each census tract were derived from a GIS analysis of
Detroit’s historic districts. Further details on this data source and how it was analyzed to measure
the extent of historical protections are given below.
3.1.1. Indicators of Neighborhood Status
Previous literature has used a variety of indicators to determine neighborhood status. In
each case, the researcher, based on the study question, chose the variables best fit to explain
effects over a period of time. While Ceballo et al. (2005) chose median household income as the
only census indicator to determine neighborhood status, Coulson et al. (2004) used a
combination of five different indicators, among them vacancy percentage and the Simpson

21

diversity index. For this study the following indicators are used: median household income,
owner occupied percentage, vacancy percentage, and percentage of White population.
Median household income, is the most common census indicator used to measure an
area’s actual economic status. A higher median household income is associated with wealth and
purchasing power. Households with a higher than average median household income are able to
select the most attractive neighborhoods, as they have more mobility (Holzer 2017). Sometimes
they are able to revert and transform neighborhoods from lower to higher status. An increase of
income on local historic districts in Detroit would indicate an increase in neighborhood status.
All monetary values displayed in this study are in 2015 dollars and were adjusted for
inflation using the Consumer Price Index (CPI). The CPI is a statistical estimate constructed
from the average change over time in the prices paid by urban consumers for a market basket of
consumer goods and services (Bureau of Labor Statistics 2018).
At the height of Detroit’s opulence in the 1950’s, the city was at the forefront of owner-
occupied, single-family residences in the U.S. Owner occupancy in the U.S. is an indicator that is
associated with stability and wealth. Owning a home is a path to wealth, and neighborhoods with
a high percentage of owner-occupied dwellings are generally viewed as higher status
neighborhoods than those with mostly renters. The owner-occupied percentage was calculated by
diving the number of owner-occupied housing units by the total number of housing units. An
increase on the percentage of owner occupancy represents stability, and a rise in neighborhood
status.
Losing a third of its population left the city of Detroit with numerous abandoned building
structures. During Detroit’s decline, many areas suffered from abandoned dwellings and lots,
which caused a decline in property tax revenue and led to bankruptcy proceedings in 2013.

22

Vacancy percentages increased as the demand for housing decreased due to the loss of
population throughout the city. A constant flow of tax revenue is an important issue for any
municipality in order to be able to provide public services. Neighborhood status has been in
nearly constant decline across the city when compared to the study benchmark of 1970, and a
rise of vacancies and abandoned properties with no owners has been a determining factor on the
shortage of property-tax revenue. The vacancy-percentage rate is calculated by dividing the
number of vacant dwellings by the total number of housing units. A low vacancy suggests a
neighborhood is popular and in demand (Holzer 2017). A vacancy percentage decrease of census
tracts containing local historic district designation would indicate a rise in neighborhood status.
Detroit has become the most segregated city in the U.S., as higher income households
(predominately white) migrated to the suburbs, they left economically disadvantaged households
(mostly blacks) to live in the urban core (Freeman and Braconi 2007). The phenomenon of white
population flight from the city of Detroit started to occur around the 1960’s. Several historians
argue that the riots of 1967 were the catalyst to the phenomenon, but evidence points towards the
exodus already taking place as early as the 1950’s. It is important to mention that there were
many other contributing factors aside from the 1967 riots, such as whites gaining higher incomes
than other races that increased their mobility and access to newly constructed housing tracts that
led them to relocate to suburbs outside the city.
Nonetheless, the exodus deprived the city of Detroit of a significant portion of its
economic base. An increasing middle-class in central-city neighborhoods that includes white
households, could help desegregate the metropolitan region (Lee et al. 1985). Therefore, for our
study purposes, a return of white population to live in or around local historic districts would
indicate a rise in neighborhood status. Contributing factors such as higher incomes and

23

educational attainment would benefit the neighborhoods so long as the existing residents were
not displaced.
3.1.2. Interpolation
A common situation faced by researchers using areal data is discrepancies in the
boundaries of reporting units (Logan et al. 2014). Census tract boundaries change over time due
to increase or decrease of population. The city of Detroit, like many other cities in the Midwest
region of the US, has experienced a large decrease in population, similarly Detroit’s census tracts
boundaries over time have changed. During the study period research, Detroit in 1970 had 420
census tracts, by 1990 the number had decreased to 321, and during the last reporting boundaries
of 2010 (which still conforms to 2015 reporting data), the number decreased yet further to a total
of 297 census tracts.
Table 2 Census tracts from 1970-2015
Year Census Tracts
1970 420
1980 344
1990 321
2000 314
2010 (2015) 297

Prior to every U.S. census, for redistricting purposes and other planning and policy
functions, state and local officials identify small areas for which they wish to receive population
totals. As a result, units that were defined on previous census could be split or consolidated, and
their boundaries altered in complex ways (Logan et al. 2014). In order for this study to research

24

changes in units over time, the interpolation with area weights method is necessary to create a
bridge between years.
The first step is to establish a benchmark boundary layer to which all other boundary
layers will be overlaid. For our study, estimates for 1970-2015 are based on tract boundaries
provided by the NHGIS, and the most recent gathered decennial data, 2010, was used as the
benchmark census tract layer. All census tract boundary layers from the years 1970, 1980, and
1990 were overlaid and merged to a 2000 Census tract polygon layer. The results provided a new
single layer for each of the paired years which included both records that experienced change,
which were displayed with multiple records, and the records that remained the same, displayed
with a single polygon and data record. All newly created layers were matched and overlaid to a
2010 census tract polygon layer.
In 2010, the census bureau created a more detailed census tract polygon layer for the U.S.
The NHGIS modified the provided layer by removing coastal lines and projecting land-based
areas. They also created a 2000 census tracts layer that aligns to the 2010 census tracts layer but
creates discrepancies with previous census tracts years of 1970-1990. To better analyze changes
over the study period, both 2000 census tracts were used during the interpolation with area
weights method.

25

Figure 2 Interpolation diagram (NHGIS datasets)

Once the 2000 census tracts polygon layer was selected as the base layer, the
interpolation by area weights process was initiated. The next step overlaid the 1970 tracts
boundary file on to the 2000 tracts boundary file utilizing the union tool on ArcMap. The
resulting polygon layer included shape length (perimeter) and area in square feet. Once the shape
area was obtained, it was transformed to obtain the percentage of overlay area that changed over
time. The field calculator tool supplied each census tract with a percentage overlay number by
dividing the overlay area by the original area then multiplying the result by one hundred.
At this point the file layer was ready for census indicators data to be attached, a table
containing all indicators for each individual year was joined to the newly created census tract
layer which contained the percentage overlay utilizing the join tool in ArcGIS. The join made it
possible to obtain only the percentage of each indicator that was contained within the overlay
boundaries of the newly created polygon layer. As an example, the median household income for
1970 pertaining to each area in the overlay layer was obtained by multiplying the original
number by the overlay percentage divided by 100, the result represents the interpolated amount
utilizing the area weights method.
1990 Census Tracts

1980 Census Tracts
1970 Census Tracts
2000 (2010) Census
Tracts
2000 (1970-1990)
Census Tracts
2010 (2015) Census Tracts

26

It should also be noted that there are potential errors from an interpolation that is based
only on area weights. As an example, areas within a census tract may be uninhabitable or
unpopulated, creating the results to be skewed to one portion of the tract (Logan et al. 2014).
While the city of Detroit experienced population loss, many census tracts were consolidated. The
exodus created a large inventory of abandoned dwellings, and many became focal points of
criminal activity. In order to bring a solution, the city demolished over 13,000 building
structures from January 2014 to May 2018. As a consequence, large areas of vacant land are now
prevalent in Detroit. It is important to note that in some cases the existence of large amounts of
vacant land on a census tract may have skewed interpolation results. Nonetheless, as better
options would require extensive heads-up digitizing to create new datasets, the study uses the
aereal weights method that is standard in the literature. It is suggested for future studies to use
interpolations methods that would take into consideration the balance between habitable and
vacant land.

Figure 3 Example of census tracts that experienced change between 1980 and 2015
As Detroit experienced an extreme loss in population, census tracts boundaries changed
accordingly. Over the study period it was common for two census tracts to become one. As

27

shown above on figure 3, the example presents how the boundary of four census tracts from
1980 changed over time to become two census tracts in 2015.
The distribution of tract changes is reviewed in detail for 1970-2015 in Table 3, which
shows the number of tracts that did not change, two to one, three to one, and complex types of
changes that involved multiple tracts.
Table 3 Number of census tracts that experienced change
Type of Change From 1970-2000 From 1980-2000 From 1990-2000 From 2000-2010
No change 249 275 303 214
Two to one 125 56 15 70
Three to one 37 12 3 24
Many to many 9 1 0 6

The final step on the interpolation by area weights process was to summarize the overlay
percentage results. Utilizing the dissolve tool on ArcMap, each indicator statistical field was
summarized utilizing the census tract number or identification code. The NHGIS utilizes the
code GISJOIN to identify each individual census tract number. The dissolve process allows for
all four indicator statistical fields to be processed simultaneously.
3.1.3. Extent of Historical Protections
The city of Detroit hosts an online database called the Detroit Open Data Portal. It is a
platform that allows the general public to download datasets which are provided by various city
departments. For this study, the Detroit Planning and Development Department provided the
local historic districts polygon shapefile which contained the official boundaries of Detroit’s 140
historic districts. The provided dataset was joined to a 2010 census tract dataset already
containing the interpolated data from 1970 to 2015 to enable further analysis. The resulting
layering was used to establish treatment groups (tracts with historic protections) and control

28

(tracts without historic protections) groups. When overlaid, those 2010 census tracts that
coincided with the boundaries of local historic districts were giving the value of 1, and those that
did not, were giving the value of 0.
Upon an initial visual inspection of the local historic districts, it was determined that most
of the districts converged around Detroit’s downtown section, and each individual district varied
in area and shape. Once, overlaid with the 2010 census tract layer, it was also evident that many
historic districts extend over multiple census tracts.
Upon approval by the city council, historic designations are assigned. The city council is
the only entity that can revoke or modified district boundaries. The intent of safeguarding and
preserving historic building structures and/or sites through historic designations is first supported
by local residents of the area. They then bring to the attention of the city council the areas and
structures that would benefit from the designation, whether it is a collection of buildings or a
single building structure. Once approved, each district has a unique boundary, and depending on
the structures inside the boundary contains a specific amount of building footprint. Thus, it is
important to note that in this study the treatment group is not selected at random, but rather is
provided through a process involving community activists and City of Detroit, which likely takes
into account historical values in existing building as well as other factors.
In addition to creating a unified census geography with which to analyze the effects of
historical preservation over a long time period, this study also explored different spatial
thresholds to evaluate which census tracts would be considered as part of the treatment group,
and which would be part of the control group. The measures used relied on current historic
designation polygons, but went beyond these to measure the effect of these designations given
the size and number of buildings in each census tract.

29

The first measurement of historic preservation is simply whether or not there is any
existing designation (any polygon or part of polygon) that intersects with a given census tract. A
second measure indicates the intensity of historic preservation by measuring the area percentage
of a census tract covered by polygons with historic designations. To obtain the overlay
percentage of each census tract, it was necessary to utilize the union tool from the overlay toolset
(Spatial Analysis) on ArcMap 10.5, with both 2010 census tracts and local historic districts
shapefiles as input datasets. The output layer included all data from both datasets and provided
the overlay area. The overlay area percentage was obtained by diving the overlay area by the
census tract area multiplied by 100. Upon selecting by attributes only the areas with a historic
designation, it was then possible to summarize the overlay percentage within each census tract
by utilizing the dissolve tool in the generalization toolset (Data Management toolbox) in ArcMap
10.5. The overlay percentage was summarized using the provided NHGIS census tract number or
GISJOIN identification code.
The next step was to explore changing the threshold of the area percentage covered by a
historic designation that would consider a census tract as being part of the treatment or control
groups. Although, historical district boundaries closely match those at the census block group,
the unavailability of census indicator data at the census block group level prior to 1980, made it
necessary for this study to utilize census tracts as the study unit. Census tracts are much larger in
area compared to historic districts or census block groups. That difference in size means that the
overlay percentage of most historic districts is quite small. Taking this into consideration, the
two threshold values were established when determining control groups for two different DiD
analyses: any census tract with an overlay percentage of 1% or below would be considered as

30

part of the control group and on a different DiD analysis, any census tract with an overlay
percentage of 10% or below would become part of the control group.
A third way of measuring the intensity of historical preservation in each census tract is to
count the number of protected structures. Therefore, additional exploration was performed
utilizing building structure footprint in the city of Detroit. The Southeast Michigan Council of
Governments (SEMCOG) data portal provides the digital footprint of each building in Southeast
Michigan (http://semcog.org). The dataset named building Detroit, contained building type,
building ID, city ID, year built, median height, parcel ID, residential square feet, and non-
residential square feet. While this study only focused on the building structure count contained
within each census tract as part of the local historic districts, further studies can explore
additional spatial parameters utilizing building structure footprint such as square feet and
building type.
Using the select by location tool for building structures contained within a local historic
district, it was then possible to create a new layer named historic building structures. The newly
created layer was then joined utilizing the spatial join tool, to the 2010 census tract layer. The
resulting output layer provided the building count for each census tract. As the 2010 census tract
had previously determined which census tracts contained historic designations, it was then
possible to explore changing the parameters on the treatment group. By manipulating the
building count threshold, the study explored if DiD results would change when census tracts part
of the original treatment group would become part of the control group.
The following threshold counts were used to determine control groups on two additional
DiD analysis reports. Any census tract with a building structure count of 1 to 10 became part of

31

the control group for one report, and similarly on a second report, any census tract with a
building structure count 1 to 100 was considered as part of the control group.
Table 4 Census tracts totals
Control Group Treatment Group
Existing Historic Designation 221 76
Existing Historic Designation by 1995 257 40
Area 1% or below by 1995 280 17
Area 10% or below by 1995 285 12
Count 1 to 10 by 1995 281 16
Count 1-100 by 1995 286 11

3.1.4. Difference-in-Differences Analysis
Difference-in-differences (DiD) is a popular technique used to assess the effect of
treatment over time in comparison to the control group (Holzer 2017). A DiD analysis uses
longitudinal data from treatment and control groups to obtain an appropriate counterfactual to
estimate a causal effect (Columbia 2016).
When studying the effects of home prices before and after the creation of historic districts
in the Boston-Cambridge-Quincy MSA between 2000-2007, Heintzelmand & Altieri (2011) used
DiD to control for higher starting value homes in historic districts by carefully selecting control
and treatment groups with matched starting home values. By using a DiD analysis, this study
was able to isolate the effects of historic district designations on each of the selected census
indicator between 1970-2015. Assuming that in the absence of intervention both matched
treatment and control neighborhoods would have similar outcomes, it is then possible to measure
any effects that historic districts designations would have on census tracts that enclose one within
its boundaries.

32

A positive DiD value when using the indicators of median household income, owner
occupied percentage, and percentage of white population, and a negative value for vacancy
percentage imply that neighborhood quality in census tracts that contain a local historic district
increased when compared to non-local historic district designation census tracts.
For the study, the DiD required data that measures historical protections to be cumulative
up to a cut-off year. As noted above, protections established in 1995 or earlier were included.
Therefore, as we examine changes in decennial census data, the baseline for the study was
established utilizing reported data from 1970, 1980, and 1990. Additional comparable measuring
points in the years 2000, and 2010 provided assessment references, and the final reported data
used for the DiD analysis was ACS data for 2015. A total of 62 historic districts and 40 census
tracts associated with them were formed by 1995 are part of the treatment group for this study.
Combined they represent 44% of the current historic districts in the City of Detroit.
DiD requires a parallel trend assumption, although both our treatment and control groups
do not start at the same neighborhood status at the time of first measurement, it is assumed that
the treatment group if not treated would follow a parallel movement similar to that of the control
group. The DiD provides a visual form of analysis that allows for rapid comparison of trend lines
that may show linear or non-linear changes in either improving or declining neighborhood status
for each variable explored in this study. This is quite useful in this study where Detroit over the
study period may show results below the original measuring point in the opening decades with
possible upward trends at the end of the study period. The DiD allows for visualization of
whether declines in neighborhood status are more or less pronounced and recoveries more or less
steep in treatment versus control groups. The DiD analysis for this study served as an exploratory
analysis technique, and further exploration is possible through the use of a regression analysis

33

model, which allows for continuous rather than only discrete measurement of the intensity of
historic preservation.
3.1.5. Regression Analysis (Temporal Progression)
A series of ordinary least squares (OLS) analyses tested whether the change was related
to local historic district designations based on two different continuous independent variables.
For the OLS analysis, there is no need to use a cut-off year. Instead, the spatial extent of the
historical protections is multiplied by the number of years’ protections have been in place in a
given census tract. The two independent variables for the OLS were the census tract percentage
covered by a local historic district and historical building structure count multiplied by the
number of years each historic designation had been in effect.
The OLS regressions examined the association between the independent variables listed
above and the change in neighborhood status indicator values from 1970 to 2015. If the
coefficient had a statistically significant p-value <0.05, then it indicates there was a relationship
between the test variable and the change in the indicator value. When the dependent variable
represented the change from 1970 to 2015 for median household income, owner occupancy
percentage and percentage of white population, a positive coefficient indicates an association of
historical preservation with a rise in neighborhood status, and when the dependent variable
represented a change in vacancy percentage, a negative coefficient indicates an association of
historical preservation rise in neighborhood status.
As an example, the first test evaluated the relationship between the census tract
percentage of area covered by a historic district designation times the amount of years each
designation had been in effect and the percentage change of each indicator over the period of
time from 1970 to 2015. This design evaluated the relationship of each census indicator

34

individually, so four different OLS models were used in order to isolate test results. The
residuals of tests that presented statistically significant p-values were further analyzed utilizing
spatial autocorrelation.
The next series of tests evaluated the relationship between the count of historical
building structures within each census tract times the amount of years each designation had been
in effect and the percentage change of each of the four indicators over the period of time from
1970 to 2015. In all tests, OLS models that showed statistically significant coefficients, and their
residuals were further analyzed utilizing spatial autocorrelation testing.

35

Chapter 4 Results
This chapter presents the results of the various spatial analysis techniques presented in
chapter 3. DiD results over the study period from 1970 to 2015 indicate that census tracts that
contained a local historic district designation experienced a slight increase in neighborhood
quality in comparison to census tracts that did not contain one within its boundaries. Due to the
drastic socioeconomic changes that took place in the city of Detroit during that period, it is
important to note that each census indicator shows an upward trend in 2015, yet most all census
tracts values are still below the original starting point in 1970. OLS Regression results found a
positive relationship of local historic district area percentage in a census tract and the indicators
of median household income and owner-occupied percentage, but the OLS results all show
significant spatial autocorrelation. In order to understand the clustering that results in this
finding, this chapter starts with a brief report on a hot spots analysis of the count of building
structures and then details the DiD and the OLS regression results.
4.1 Hot Spot Analysis
To identify statistically significant spatial clusters, a Getis-Ord hot spot analysis was used
to provide visual identification of statistically significant hot spots in the downtown Detroit area.
With a Gi_Bin 95% confidence level or higher, Detroit’s historic designation areas cluster
around each other near the downtown district (see Figure 4 below). This is an important context
to note before reviewing the results of the DiD and OLS models. The areas with historic
protections generally share a locational characteristic and given their common location also may
share other characteristics that influence their neighborhood status over time.

36

Figure 4 Building structures hot spot analysis

4.2 Difference-in-differences
The local historic district designations were approved and assigned by the Detroit city
council upon review of an extensive report provided by the Historic District Commission. As of
2015, 138 historic districts are designated protected areas in the City of Detroit. The results
presented on this section evaluate the treatment group that encompasses census tracts that were
covered under the designation prior to the end of 1995. The results calculate the average totals of
each census indicators across all the tracts starting in 1970 and ending in 2015. Throughout the

37

study, the experimental group was assigned the value=1 and the control group was assigned the
value=0. In the following section, DiD results are detailed below if there is difference apparent in
the trajectory of the treatment group compared with the control group in the years 2000, 2010,
and 2015 (i.e., when the result is not “null”).
Table 5 Summary of difference-in-differences trends by varying spatial thresholds

Median Household
Income
Owner Occupied
Percentage
Vacancy
Percentage
Percentage of
white
population
Existing
Designation null stable stable recover
>1% of area null recover null recover
>10% of area null recover null recover
>10 buildings null recover null recover
>100 buildings null recover null recover

Table 5 summarizes the results of the multiple difference-in-differences analyses that
show current existing designations as well as trends that occur when various spatial thresholds
are given. When both experimental and control groups are virtually the same, a “null” value is
given, if the status holds up in census tracts with historical protections even as it declines in
tracts without historical protections, a “stable” value is assigned; a “recover” value is given when
both control and experimental tracts decline at the outset of the study but protected tracts recover
faster and/or finish on a better trend.
4.2.1. Existing Designations
When exploring DiD results for existing designations for owner-occupied percentage
change, census tracts with historical protections started with a lower percentage of owner-
occupied housing than census tracts in the control group but managed to hold steady their share

38

of owner-occupied housing while non-protected census tracts on average dropped from around
68% to about 50% owner-occupied housing.

Figure 5 DiD results: experimental group recovery trend for existing designations

By contrast, when exploring percentage of white population change (Figure 5 above),
DiD results indicate that while census tracts with historical protections started with a lower
percentage of white population in comparison to census tracts in the control group they managed
to end with a higher percentage. Both groups experienced a decline at the outset, and non-
protected census tracts on average dropped from around 65% to 10% while census tracts with
historic designation protections presented a recovery from that pattern ending approximately
14% higher than the control group.

0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
1970 1980 1990 2000 2010 2015
Percentage of White Population
1 0

39

Figure 6 DiD results: experimental group stabilizing tend for existing designations
Census tracts that were acquiring any level of historical protections from 1970-1995
initially experienced a decline in vacancy percentage from 1970 to 1980 (Figure 6), after which a
sharp increase occurred as historical protections were coming into place. This followed more or
less a parallel pattern to census tracts without historical protections. It is important to note that
while non-protected census tracts experienced a steady increase from 2000 onwards, protected
census tracts ended the study period with a lower vacancy-percentage in a sort of stabilizing
trend, which creates an approximately 5% difference in vacancy rates at the end of the study
period.
4.2.2. Area Percentage in a Census Tract
Further analysis compared results derived from utilizing the area percentage of a census
tract covered by a historical district designation. The results display changes by calculating the
difference of the average totals of each census indicator when assigning the value=1 to census
tracts with historic designation coverage of <1% and assigning the value of 0 to census tracts
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
1970 1980 1990 2000 2010 2015
Vacancy Percentage
1 0

40

with > 1% of historic district coverage. Additional exploration was performed utilizing a more
stringent threshold. Census tracts assigned the value of 1 with a historic district area coverage
ranging from 10% to 100%, and census tracts assigned the value of 0 with historic district area
coverage under 10%.
For the owner-occupied percentage census indicator for both <1% and <10%, the
treatment group goes up steadily from 30% owner-occupied in 1970 to 40% owner-occupied in
2015. During this same period of time, the control group steadily declines from 65% to 48%. The
treatment census tracts achieved higher levels of homeownership in spite of declines across the
city (i.e., a recovery trend).

Figure 7 DiD results: experimental group recovery trend for historic designation coverage of
<1%
At the onset of the study period in 1970, Figure 7 shows that there was a 20% lower
white population for the census tracts that acquired historic designations from 1970-1995.
Although initially both groups present a parallel downward trend, towards the 2000’s tracts with
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
1970 1980 1990 2000 2010 2015
Percentage of White Population <1%
1 0

41

at least 1% protected area initiate a recovery period. When the control group threshold is set at
<1%, by 2015 the treatment group outperforms the control group by 12.78%. Tracts with 10% or
greater areal coverage of historical protections show almost an identical trend for percentage
white population to those with only 1%.
4.2.3. Building Structure Count
This study also explored the effects of the total building structure count as a parameter to
determine control and treatment groups. By making changes to the threshold count of building
structures, this study explored the effects of the built environment as it relates to the selection of
historic designations. When selecting census tracts that would become part of the treatment
groups the parameters used were: >10 buildings and >100 buildings.
Results presented indicate that census tracts with protected building structures
prior to the end of 1995 increased in status while those without protected structures declined in
status on owner-occupation and percentage white population indicators. The result is almost
identical for both parameters.

42

Figure 8 DiD results: experimental group recovery trend for historic designation building count
of <100
Figure 8 shows the DiD analysis results for owner-occupied percentage for historic
designation building structure count >100. It shows a pronounced recovery for the census tracts
with 100 or more protected structures in spite of declines in the rest of the city. As shown on
Figure 8 above, control and treatment groups at the onset of the study period in 1970 start with
an approximately 30% difference margin between them. By 2015 that difference is reduced to
under 10%.

0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
1970 1980 1990 2000 2010 2015
Owner Occupied <100 Building Count
1 0

43

Figure 9 DiD results: experimental group recovery trend for historic designation building count
<10.

When establishing the threshold building count at <10 as seen on Figure 9 above, census
tracts with historical protections recovered from a steep decline of the 1970’s and present a
greater increase in percentage of white population. It is important to note that census tracts part
of the treatment group initiated almost 20% below the control group and as the study period
progressed both groups directly invert positions with census tracts part of the treatment group
outperforming the control group by 13.92%.
A similar result with smaller margins at the end of the study period is seen when the
threshold is set to <100 building structure count to percentage of white population change.
Census tracts with historical protections started with a lower percentage of white population than
census tracts with no protections but managed to increase their share of percentage of white
population by the end of the study period. Census tracts part of the control group presented a
steep decline of 50% finishing 10% below the treatment group. It is important to note that
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
1970 1980 1990 2000 2010 2015
Percentage of White Population <10 Building
Count
1 0

44

although census tracts part of the treatment group performed better, at the end of the study period
finish 30% lower than the original baseline. In general, the results for building structure counts
and area percentage measures parallel one another for owner-occupied and percent white
population.
4.3 OLS Regression Results
A series of ordinary least squares models individually tested the significance of two
independent variables and four dependent variables. Each independent variable was multiplied
by the amount of years each historic designation was in place. Each dependent variable obtained
represents the amount of change (median household income) and the percentage change (owner-
occupied, vacancy percentage change and percentage of white population change) that has taken
place on census tracts with historic designations from 1970 to 2015.

45

Table 6 OLS regression output
Indicator (Dependent Variable) Coefficient p-value Adjusted R-
squared
Probability Robust
Probability
Independent Variable: Coverage Perecentage
Median Household Income 1324.6524 p<0.01 0.001349 0.001361* 0.001151*
Owner-occupied 0.013690 p< 0.01 0.098626 0.000000* 0.000000*
Vacancy-percentage -0.003478 p< 0.01 0.004833 0.119557 0.049580*
Percentage of white population 0.017323 p< 0.01 0.030018 .001601* 0.000000*
Independent Variable: Building Count
Median Household Income 0.572594 p< 0.01 0.005641 0.102739 0.218171
Owner-occupied 0.000010 p< 0.01 0.073304 0.000002* 0.000000*
Vacancy-percentage -0.000002 p< 0.01 0.001445 0.233019 0.183071
Percentage of white population 0.000011 p< 0.01 0.014858 0.020064* 0.000962*

The results of the 297 census tracts, when exploring median household income change as
a dependent variable, indicate statistical significance when the independent variable was
percentage coverage of a census tract by a historic designation and historically protected building
count. Both are significant at p< 0.01 level with the expected positive slope. The Adjusted R-
Square value for median household income is below 1%, indicating that very little of variability
across the decades is explained by the degree of historical protection.
With only two exceptions, as noted below, the OLS results for all the other independent
variable and dependent variable pairs are similar to median household income. In the case of
owner-occupied housing and to a lesser degree percentage white population, the Adjusted R-
Square values are somewhat higher. This is consistent with the strong findings on these two
census indicators in the DiD analysis.

46

Two exceptions where the relationships are not significant should be noted. One
exception is when the independent variable is building count, in which case there is no
association with vacancy percentage. Here it is interesting to note that when measuring by
coverage percentage, there is a statistically significant association. The second exception is when
the independent variable is building count and the dependent variable is for median household
income. Here again, while the relationship is in the expected direction, the association is not
statistically significant. The different OLS results for percent coverage and protected building
counts highlight the importance of how the spatial measure of historical protection is
constructed.
It is important to note that only building structures that are part of a historic
designation were taken as part of the study. To corroborate that the model was not biased, a
preliminary histogram was visually inspected, and a natural log was created (Figure 10, below).
The distribution is somewhat skewed, but mostly because of the large number of zero values in
the data. As many tracts in the data contain a 0, the natural log would alter the data, and therefore
it was not used. Nonetheless, the results show that the way this variable lays out in the real-
world, it will tend to suppress findings of an OLS analysis. Highlighting the need for this study
to explore different spatial parameters to define control and experimental groups.

47

Figure 10 Natural logarithm (LN) Historic District area coverage percentage results
It is also important to note that a negative coefficient for vacancy percentage change is
the expected direction, as it indicates that a census tract is experiencing neighborhood status rise.
During the OLS, results for vacancy percentage a robust probability of 0.049580 indicated
statistical significance with a negative coefficient of -0.003478.
Following each individual OLS analysis, to corroborate the results and to remove
the possibility of other external factors from contributing to the results, a spatial autocorrelation
analysis was performed on the remaining coefficient residuals.
4.4 Spatial Autocorrelation
A spatial autocorrelation check was performed in order to verify if the residuals were
independently distributed with respect to the location of the census tract. The results indicate that
in all cases, based on positive z-score values, there is likely an underlying spatial process of
clustering in residuals in the data. This is not surprising; given that we know from the hot spots
analysis, which shows that the historically-protected structures cluster near downtown Detroit.

48

Nonetheless, it raises the possibility that some additional factor or factors related to downtown
location explains the associations seen here for median household income, owner occupied
percentage, and vacancy percentage, perhaps over and above the historical protections. This
finding means that it is not possible to infer causality in the associations found between historical
protections and neighborhood status.
Table 7 Spatial autocorrelation results
Indicator (Dependent Variable) Moran's Index Expected Index Variance z-score p-value
Independent Variable: Census Tract Percentage
Median Household Income 0.141928 -0.003378 0.001268 4.1 0.000045
Owner-occupied 0.434961 -0.003378 0.001286 12.2 0.000000
Vacancy 0.535444 -0.003378 0.001266 15.1 0.000000
Percentage of white population 0.859574 -0.003378 0.001293 24.0 0.000000
Independent Variable: Building Count
Median Household Income 0.150446 -0.003378 0.001269 4.3 0.000016
Owner Occupied 0.463269 -0.003378 0.001286 13.0 0.000000
Vacancy 0.546422 -0.003378 0.001266 15.5 0.000000
Percentage of white population 0.877820 -0.003378 0.001293 24.5 0.000000

49

Chapter 5 Conclusion
This study investigated spatial analysis techniques that would help determine the effects
of historical district designations and examined its impact on neighborhood rise or fall. This
section will discuss results, make conclusions, discuss study limitations, and provide suggestions
for future research.
Given the historical background of the city of Detroit, it is important to notice the events
that lead to the socioeconomic conditions of our study period. Based on the DiD and OLS results
in the previous chapter, it is apparent that in general census tracts that have historical protections
are experiencing a stabilization of decline or rise in neighborhood status as of 2015. Nonetheless,
while the upward trend is evident, in most cases, neighborhood status does not yet meet the
original 1970’s baseline conditions established for this study. Future studies should allow
sufficient time to pass to investigate if the results that are present in 2015 are consistent and
strong enough for upward changes to be considered robust reflections of any of associations
explored here.
While there are many similarities in the DiD and OLS results, there are also a few
differences several of which were noted in Chapter 4. It is important to note that in general, a
conservative approach was taken in identifying stabilizing and recovery trends in the DiD
analysis. Because the OLS is based on continuous variables and includes all data on historical
designations up to 2015, including measurement of the length of time of the designation, it is a
more sensitive measure of potential associations.
Owner-occupied percentage on census tracts that contain historical protections has
increased at a steady pace and is the only census indicator that regardless of changes made to
spatial thresholds finished the study period above the initial values. On the other hand, owner-

50

occupied percentage on census tracts with no historical protections remain on a constant
downward trend. A steep decline is present on both control and experimental groups at the outset
of the study period when observing percentage of white population. The percentage of white
population on census tracts with historical protections have recovered quicker than those without
any designations experiencing an increase starting in the 2000’s.
A negative change in vacancy percentage is an indicator of a positive change in
neighborhood status, but DiD results show that from 1970 to 2015 vacancies in the city of
Detroit have increased across the city. This was an expected phenomenon due to the massive
exodus of population that has occurred since the 1950’s. The original assumption that historic
district designations would contribute to lower vacancy percentages than non-historic district
areas in Detroit is inconclusive as DiD results indicate a parallel rise in vacancy percentages
from 1970 to 2010, and a redirection in its initial stages in 2015. Further studies can investigate
the change utilizing future decennial data, or ACS estimates, or by concentrating on areas where
the redirection is clearly evident, as in downtown Detroit.
The use of areal interpolation to determine census indicator values for census tracts in
Detroit introduces uncertainty. During Holzer’s neighborhood comparison study of Minneapolis
and St. Paul (2017), which explores the impact of economic development programs, uncertainty
arises as different neighborhoods belong to different census block group and selecting census
block groups as a geographic unit makes it impossible to determine the impacts of economic
development funding. As census tract boundaries shifted from 1970 to 2015, areal interpolation
is necessary in order to define and create a consistent study unit. Standardizing census tract
boundaries from 1970-2000 to census tract boundaries from 2010 by areal interpolation
determines consistent indicator values for all census tracts in the city of Detroit throughout the

51

study period from 1970 to 2015. While there is still some degree of error in the data, the
compromise is necessary to obtain consistency when comparing census tracts that have shifted
boundaries multiple times over the study period.
As data availability in the city of Detroit increases, future studies could compare results
with a different study unit, exploring perhaps census block group data that better aligns with
local historic district boundaries and evaluating the degree of uncertainty in this test. The NHGIS
portal is only able to provide median household income prior to 1980 at the census tract level.
Through areal interpolation, exploration of decennial data from 1950 and 1960 may expand on
this study’s assessment of Detroit’s socioeconomic conditions that led to its decline.
The lack of a precision match between the historic district polygons and census tracts
means that in the vast majority of cases the historic designation area coverage percentage
coverage is well below 50 percent of a census tract. The large area of a census tract that is not
covered by a historic designation protection may skew results. This is the modifiable areal unit
problem (MAUP), which may be solved in the future by obtaining census indicator data at the
block level or even smaller spatial units. A smaller study unit may be able to better isolate results
and compare them to the overall Detroit metropolitan region.
To overcome the MAUP, new data is needed. Future work could explore non-census data
to build measures of neighborhood status. For example, the use of tax assessor data could depict
property values at the parcel level. The total census population flows over time at the block level
could indicate neighborhoods with sharply diminishing or growing populations. Such measures
might well produce different results than the ones obtained at the census tract level. Increases in
property values and population growth at the census block level may be a more sensitive
measure of potential associations between historic preservation and neighborhood status.

52

It is important to mention the unique socioeconomic conditions that are prevalent in the
city of Detroit. While historic preservation efforts in other cities in the U.S. may show an
association with an increase in neighborhood status, Detroit’s extreme decline has created a
different environment with conditions well below the original starting point of 1970. For that
reason, when analyzing results, it is important to acknowledge that while findings in the
treatment groups present an upward rise towards the end of our study period, when the results are
compared to the initial baseline, those conditions are merely starting to move towards the
original starting point. Historic districts in Detroit have performed better in preserving or
increasing neighborhood status when compared to other neighborhoods without the designation.
Efforts by other cities in the U.S. may perform at a higher rate in part because the initial
conditions did not deteriorate to the extent to which the city of Detroit was exposed. In Detroit,
many census tracts with historical protections at the onset of the study period start at greater
disadvantage far behind the control groups, an indication that perhaps major deteriorations to
theses tracts had already started prior to 1970. These reasons may explain the major disparity in
results between major metropolitan areas in the Midwest region and other areas in the U.S.
The use of decennial and ACS census data to determine the value of each census
indicator also creates data uncertainty. As decennial data provided actual values from 1970 to
2000, ACS data provided estimates for 2010 and 2015. Using the most recent and available data
allowed this study to assess change over a long period of time by setting up a starting point prior
to the assignment of local historic district designations in Detroit in 1978, and ACS data for
2015. It was essential to study such a wide period of time in order to assess the impact of historic
district designations in the city of Detroit since its original inception. Future studies should
analyze other census indicators to determine neighborhood status, as well as specific regional

53

economic data that can help describe smaller spatial units that better align with historical
districts.
While it is difficult to isolate the effects of historical district designations, this study
suggests that over time historical designations are perhaps one of the contributing factors for
median household income, owner-occupied percentage and percentage of white population to
increase, and vacancy percentage to decrease, therefore becoming a factor of neighborhood
status to rise. However, there are likely other factors related to the location of the census tracts
not yet uncovered.
Temporal regression results validate that a positive relationship is present between
historic designated census tracts and the selected census indicators. Nonetheless, based on the
spatial autocorrelation results that were obtained from the OLS residuals, other factors may have
contributed to the increase in neighborhood status seen in Detroit census tracts that host a
historical district designation. What other factors may have contributed? A strong investment on
Detroit’s downtown and financial neighborhoods, which happen to be historical protected areas,
has been spearheaded by the conversion of commercial structures into residential dwellings as
well as the creation of the most compact sports district in the U.S. which has led to more affluent
population to relocate and live in the area.
Although historic districts have been in effect since the 1970’s, new areas are added to
existing conditions as residents petition to the City council for an area to be considered for the
designation. Spatial conditions will change in the future as new areas become part of Detroit’s
historic district designations. Therefore, this study can serve as a starting point on how these new
parameters can be later analyzed.

54

Corroborated by hot spot analysis results, most historical protected building structures are
located in the downtown area. Further studies can explore whether redevelopment efforts in
downtown Detroit will have any spillover effects to surrounding neighborhoods.
Understanding that historical protections may contribute to neighborhood status rise by
attracting more affluent residents, potential displacement due to neighborhood improvements
should be researched in future studies. A major criticism of historical preservation efforts is that
it leads to the displacement of disadvantaged population that can no longer afford to live in their
communities. It will be important to research whether neighborhood status rise benefit the
existing residents or if the increase is the result of more affluent people moving in (Holzer 2017).
There is a limited amount of historical preservation literature that utilizes GIS. To offset
that void, related literature on the fields of historical preservation, sociology, history, and GIS
analysis was necessary to establish a foundation to build upon a methodology. In order to
understand the extent of historical protections and obtain stronger results no single analysis
method with simple measures will give good results. Instead this study has shown the importance
of using multiple spatial analysis methods. Using OLS analysis as an exploratory method, the
research shows the importance of tracking spatial autocorrelation which has not been done in
previous studies.
A surge of revitalization efforts on the urban core of metropolitan areas throughout the
U.S. has sparked the debate whether historic preservation designations are good or bad for
economic development. One purpose of this study was to present various spatial analysis
techniques that can help determine the effects, if any, of historical district designations on
neighborhood status rise or fall. Furthermore, it exhibits the limitations that were encountered

55

and suggests that similar studies expand on the topic to measure the effects of historic
preservation.

56

References
Allison, Eric W. 2005. “Gentrification and Historic Districts: Public Policy Considerations in the
Designation of Historic Districts in New York City.” PhD diss., Columbia University.
Accessed April 23, 2018. http://www.allisongroup.com/ericwallison/wp-
content/uploads/2012/08/EWA_Dissertation_Full.pdf.
Altieri, Jason A. and Martin D. Heintzelman. 2011. “Historic Preservation: Preserving Value?”
The Journal of Real Estate Finance and Economics 46, no. 3 (2013): 543-563.
Bertron, Cara. 2011. “Between a Rock and a Historic Place: Preservation in Postindustrial Urban
Planning.” Master’s Thesis, University of Pennsylvania.
Bertron, Cara. 2013. “Rightsizing Right.” Forum Journal 27, no. 4: 23-33.
Booza, Jason and Kurt Metzger. 2004. “On Some Socio-economic Aspects of Detroit.” In
Shrinking Cities: Complete Works - Detroit, edited by Phillip Oswalt, 44-49. Berlin:
ARCH+Verlag GmbH.
Boscarino, Timothy. 2017. “Local Historic Districts Shapefile.” City Planning Commission.
Detroit Open Data Portal. Accessed March 2, 2018.
http://data.detroitmi.gov/Government/Local-Historic-Districts/hj4t-4mea
Boyle, Kevin. 2001. “The Ruins of Detroit: Exploring the Urban Crisis in the Motor City.”
Michigan Historical Review 27, no. 1, Detroit 300: 109-127.
Ceballo, Rosario, Vonnie C. McLoyd and Teru Toyokawa. 2004. “The Influence of
Neighborhood Quality on Adolescents’ Educational Values and School Effort.” Journal
of Adolescent Research 19, no. 6: 716-739.
U.S. Census Bureau. Census of Population and Housing, 1970, Series PHC(1), Census Tracts.
Accessed May 8, 2018. http://www.census.gov/prod/www/decennial.html
Coulson, N. Edward, Robin M. Leichenko, and David Listokin. 2001. “Historic Preservation and
Residential Property Values: An Analysis of Texas Cities.” Urban Studies 38, no. 11:
1973-1987.
Farley, Reynolds, Charlotte Steeh, Maria Krysan, Tara Jackson and Keith Reeves. 1994.
“Stereotypes and Segregation: Neighborhoods in the Detroit Area.” American Journal of
Sociology 100, no.3 (Nov): 750-780.
Farr, Sara. 2009. “Preservation and the Postindustrial City: Preservation as an Economic
Revitalization and Policy Tool in Pittsburgh and Detroit.” Master’s Thesis, University of
Georgia.

57

Freeman, Lance and Frank Braconi. 2004. “Gentrification and Displacement New York City in
the 1990’s.” Journal of the American Planning Association 70, no. 139-52.
Harris, Richard. (1990). “Working-class Home Ownership in the American Metropolis.” Journal
of Urban History 17, no. 1: 46-69.
Holzer, Richard. 2017. “Evaluating the Minneapolis Neighborhood Revitalization Program’s
Effect on Neighborhoods.” Master’s Thesis, University of Southern California.
Hyde, Charles K., 2001. “Detroit the Dynamic: The Industrial History of Detroit from Cigars to
Cars.” Michigan Historical Review 27, no. 1, Detroit 300: 57-73.
Kappler, Cara and Lindsey Miller. 2009. “Re-imagining the Urban Greenway: An Alternative
Transportation Strategy and Vacant Land Use Plan for the Woodbridge Neighborhood of
Detroit.” Master’s Thesis, University of Michigan.
Katzman, David M., 1973. “Before the Ghetto.” Chicago: University of Illinois Press.
Klepper, Steven. 2001. “The Evolution of the U.S. Automobile Industry and Detroit as its
Capital.” Carnegie Mellon University.
Lewis, David Lanier and Laurence Goldstein, 1983. “The Automobile and American Culture.”
Ann Harbor: The University of Michigan Press.
Logan, John R., Brian J. Stults and Zengwang Xu. 2015. “Interpolating U.S. Decennial Census
Tract Data from as Early as 1970 to 2010: A Longitudinal Tract Database.” The
Professional Geographer 66, no. 3: 412-420.
Manson, Steven, Jonathan Schroeder, David Van Riper and Steven Ruggles. 2017. “Census
Tracts in the U.S. from 1970-2010 Shapefiles.” IPUMS National Historical Geographic
Information System: Version 12.0 [Database]. University of Minnesota. Accessed March
5, 2018. https://data2.nhgis.org/main
Manson, Steven, Jonathan Schroeder, David Van Riper and Steven Ruggles. 2017. “Median
Household Income in the U.S. from 1980-2000.” IPUMS National Historical Geographic
Information System: Version 12.0 [Database]. University of Minnesota. Accessed March
5, 2018. https://data2.nhgis.org/main
Manson, Steven, Jonathan Schroeder, David Van Riper and Steven Ruggles. 2017. “Occupied
Units in the U.S. from 1970-2000.” IPUMS National Historical Geographic Information
System: Version 12.0 [Database]. University of Minnesota. Accessed March 5, 2018.
https://data2.nhgis.org/main
Manson, Steven, Jonathan Schroeder, David Van Riper and Steven Ruggles. 2017. “Owner
Occupied Units in the U.S. from 1970-2000.” IPUMS National Historical Geographic
Information System: Version 12.0 [Database]. University of Minnesota. Accessed March
5, 2018. https://data2.nhgis.org/main

58

Manson, Steven, Jonathan Schroeder, David Van Riper and Steven Ruggles. 2017. “Population
Totals in the U.S. from 1970-2000.” IPUMS National Historical Geographic Information
System: Version 12.0 [Database]. University of Minnesota. Accessed March 5, 2018.
https://data2.nhgis.org/main
Michigan GIS Open Data. 2018. “Buildings Detroit Shapefile.” Southeast Michigan Council of
Governments. Accessed March 22, 2018.
http://arcgisserver.semcog.org/arcgis/rest/services/OpenData/Buildings_Detroit/MapServ
er/0
Perkins, Douglas D., Courtney Larsen and Barbara B. Brown. 2009. “Mapping Urban
Revitalization: Using GIS Spatial Analysis to Evaluate a New Housing Policy.” Journal
of Prevention and Intervention in the Community 37, no.1: 48-65.
Peterson, Joyce Shaw. 1979. “Black Automobile Workers in Detroit, 1910-1930.” The Journal of
Negro History 64, no. 3: 177-190.
Roediger, David R., 2005. “Working Toward Whiteness: How Amrica’s Immigrants Became
White, The Strange Journey from Ellis Island to the Suburbs.” New York: Basic Books.
Schoberth, Leigh. 2014. “Detroit’s Historic Theaters: A Study of Significance and Reuse.”
Master’s Thesis, Clemson University.
Thompson Heather A. 1999. “Rethinking the Politics of White Flight in the Postwar City Detroit,
1945-1980.” Journal of Urban History 25, no. 2: 163-198.
Tolnay, Stewart E., 2003. “The African American “Great Migration” and Beyond.” Annual
Review of Sociology 29: 209-232.
Warren, Donal I., 1975. “Black Neighborhoods: An Assessment of Community Power.” Ann
Arbor: The University of Michigan Press.
Woodford, Arthur M., 2001. “This is Detroit 1701-2001.” Detroit: Wayne State University Press.
Woodford, Frank B. and Arthur M. Woodford, 1969. “All Our Yesterdays: A Brief History of
Detroit.” Detroit: Wayne State University Press.

Abstract (if available)

Abstract Cities throughout the United States have adopted historical designations in order to protect the historic architectural resources and promote economic development of areas that carry a cultural significance to their communities. Detroit, a city in steep economic decline from 1970 until 2015, has also attempted to use historical preservation to promote economic development in particular neighborhoods. The role of historic preservation has rarely been considered in a city in steep, ongoing economic decline. The study presents spatial analysis techniques that can help determine the association, if any, of historical district designations with neighborhood rise or fall. By using various approaches to count structures and measure preserved space within census tracts, a difference-in-differences (DiD) analysis and an ordinary least squares regression model was developed to test the association of preservation and neighborhood status change from 1970 to 2015. ❧ The results indicate that census tracts with a historic designation showed less decline and quicker improvement in neighborhood status when compared to census tracts with no protections. To further corroborate DiD results, ordinary least squares analysis indicated statistically-significant relationships between the percentage of historical district coverage and historically protected building counts and changes in all but two indicator values, but these results should not be accepted as evidence of causality because there is significant spatial autocorrelation. Also, Detroit’s socio-economic conditions differ from other metropolitan statistical areas in the U.S. Further research is needed in other cities during periods of economic decline and with additional control variables.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

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

PDF

Out-of-school suspensions by home neighborhood: a spatial analysis of student suspensions in the San Bernardino City Unified School District

PDF

Investigation and analysis of land use/land cover to determine the impact of policy developments in cities

PDF

Conditions of belonging: life, historical preservation and tourism development in the making of Pelourinho-Maciel, Salvador da Bahia, 1965-1985

PDF

Network accessibility and population change: historical analysis of transportation in Tennessee, 1830-2010

PDF

Effects of building modifications and municipal policies on green cover in Los Angeles County

Asset Metadata

Creator Rosas Flores, David (author)

Core Title Investigating the association of historical preservation and neighborhood status in Detroit, 1970-2015

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 09/20/2018

Defense Date 08/23/2018

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

Tag areal interpolation,census indicators,census tracts,Detroit,difference-in-differences analysis,GIS,historic districts,Historic Preservation,neighborhood change,neighborhood status,OAI-PMH Harvest,spatial analysis,spatial autocorrelation,temporal progression

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Vos, Robert (committee chair), Lee, Su Jin (committee member), Oda, Katsuhiko (committee member)

Creator Email rosasflo@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-73074

Unique identifier UC11670924

Identifier etd-RosasFlore-6775.pdf (filename),usctheses-c89-73074 (legacy record id)

Legacy Identifier etd-RosasFlore-6775.pdf

Dmrecord 73074

Document Type Thesis

Format application/pdf (imt)

Rights Rosas Flores, David

Type texts

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

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

Repository Name University of Southern California Digital Library

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