ESSAYS ON URBAN DEPOPULATION AND MUNICIPAL POLICY
By
Christina Plerhoples

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Agricultural, Food and Resource Economics – Doctor of Philosophy
2013

ABSTRACT
ESSAYS ON URBAN DEPOPULATION AND MUNICIPAL POLICY
By
Christina Plerhoples
Much of the academic literature on urban economics and public finance focuses on
growth – both population growth and budgetary growth. However, most cities do not grow
continuously, and many cities face long periods of decline rather than expansion. The city of
Detroit, for example, dropped in population from 1.8 million in 1950 to 713,777 in 2010. The
city of Saginaw, which is examined in Chapter 1 of this dissertation, went from a peak
population of 98,265 in the 1960s to 51,230 in 2013. It is difficult for city governments to
maintain infrastructure and provide adequate public services to a city that has lost such a massive
amount of population and with it, revenue streams. Little research exists on policies that can be
used by such cities. In this dissertation, I analyze two such policies.
In the first essay, I examine the effect of vacant building demolitions on crime using
block face level monthly panel data from the city of Saginaw, Michigan. Although the United
States Government spends millions of dollars a year on vacant building demolitions, no clear
causal link has been established between demolitions and crime. Results from my analysis
indicate that both demolitions and permits for demolitions actually increase crime rather than
reduce it. These results imply that by the time a building reaches the point of demolition, it may
be too late to reverse the crime trends incurred by neighborhood decline.
In the second essay, I examine the motivations for why cities choose to implement public
employee residency requirement laws, or laws that require employees to live within the city
boundaries. Justifications for these laws range from public safety to public coffer arguments.

However, little is known about the true motivations for why cities choose to implement
residency laws. In this paper, I estimate a parametric duration model with a proportional hazard
function to examine these motivations. In order to perform this analysis, I construct a unique
data set on municipal residency laws for the largest 100 cities in the United States from 1970
through 2007. I also create a set of fiscal stress indicators for each of these cities over time.
Results indicate that cities implement residency laws for economic and racial equity reasons and
remove them when the city improves in comparison to the metropolitan area as a whole. Fiscal
stress and service provision quality do not appear to be major contributing factors.
In the third essay, I look once again municipal residency requirement laws and examine
the impact that these laws have on cities. Despite the prevalence and extensive history of
residency laws, little is known about their impact on city outcomes. In this paper, I provide the
first causal estimates of the impact of municipal residency requirements on the quality of service
provision, municipal fiscal health, and the size and composition of cities. My identification
strategy exploits the timing of municipal law changes as well as the implementation of state bans
on residency laws. Results provide little evidence that residency laws have an effect on fiscal
health or city demographics when they are implemented, but they are correlated with a
temporary increase then long term decrease in crime. The removal of a residency law is not
associated with any significant change in fiscal health, city demographics, or crime.
These three essays provide information that can be used to help declining cities avoid
spending time and money on policies that do not work and remove laws that are harmful. More
research is needed to determine what policies do work in declining cities and how they can best
be implemented.

Copyright by
CHRISTINA PLERHOPLES
2013

DEDICATION
I dedicate this dissertation to my family and friends, whose never ending support made this
process not only possible, but enjoyable.
To my mom and dad, the two smartest, kindest, and most generous people I know. You have
always put yourselves second and have always been supportive, whether it was talking to me at
midnight from the other side of the world, or traveling last minute to help me in a crisis. I love
you!
To my brother Bill, who sometimes knew more about me than I did. You helped me choose the
right path at many forks in the road.
To my brother Tim, who I have always looked up to and who has never let me down.
To my sister-in-law Ellie, whose grace and composure are something I strive to achieve.
To my sister-in-law Alicia, who is the epitome of strong female role model.
To Lyra, Cora, Matthew, and Naomi, who will hold my heart forever.
To Anna and Kelly and the power of three.
To Brian, who makes it all worth it. Sont des mots qui vont très bien ensemble, Très bien
ensemble.
To Tim, Drea, Dale, Sung, Nicky, Laura, Rie, Susan, Xuan, Adam, and too many more to list.
In memory of Asa.

v

ACKNOWLEDGEMENTS

I would like to thank my committee chair, Eric Scorsone, for teaching me firsthand about
local government in Michigan. I would also like to thank my other committee members, Mark
Skidmore, Robert Myers, Leslie Papke, and Ronald Fisher for their help and guidance
throughout this process.
Thank you also to others who provided feedback on these essays: Quentin Brummet,
Timothy Hodge, Sung Hoon Kang, Nicky Mason, Todd Pugatch, Gary Solon, Brian Stacy, and
seminar participants at Michigan State University and the University of Michigan. Thank you to
those at the Federal Reserve Bank of Cleveland for their feedback during my fellowship there,
particularly Dionissi Aliprantis, Timothy Dunne, Daniel Hartley, and Francisca Richter.
Special thanks to Scott Crofoot, Jeff Klopcic, Dan Sherman and others from the City of
Saginaw for providing the data for the first essay. Thanks also to city government
representatives throughout the United States who provided information about the history of their
residency laws through searches of their archived charters and ordinances.
Thanks to previous professors who sparked my interest in economics and public service:
Father Turner, Siddharth Chandra, Shanti Gamper-Rabindran, and many more.

vi

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………...……………………...ix
LIST OF FIGURES………………………………………………………...……………………xii
CHAPTER 1
THE EFFECT OF VACANT BUILDING DEMOLITIONS ON CRIME UNDER
DEPOPULATION………………………………………………………………………………...1
Abstract……………………………………………………………………………………1
Introduction………………………………………………………………………………..2
Conceptual Framework……………………………………………………………………5
Data………………………………………………………………………………………..8
A Brief Background on Saginaw……………………………………………........10
Demolition Process………………………………………………………………11
Empirical Methodology………………………………………………………………….14
Displacements and Spillovers……………………………………………………19
Results……………………………………………………………………………………20
Displacement and Spillover Results………………………………………….…..22
Robustness Checks……………………………………………………………………….23
Strict Exogeneity…………………………………………………………………23
Number of Lags and Leads………………………………………………………24
Arson……………………………………………………………………………..25
Multiple Demolitions…………………………………………………………….25
Conclusion……………………………………………………………………………….26
APPENDIX………………………………………………………………………………………28
BIBLIOGRAPHY………………………………………………………………………………..50
CHAPTER 2
LATENT MOTIVATIONS FOR THE IMPLEMENTATION OF MUNICIPAL RESIDENCY
REQUIREMENT LAWS……………………………………………………………………...…55
Abstract…………………………………………………………………………………..55
Introduction………………………………………………………………………………56
Background………………………………………………………………………………58
Types of Residency Requirements………………………………………………..58
Rationales for Residency Requirement Laws………………………………….…61
Data………………………………………………………………………………………64
Econometric Model………………………………………………………………………68
Results……………………………………………………………………………………71
Residency Law Implementation………………………………………….………71
Residency Law Removal…………………………………………….…………...72
Conclusion……………………………………………………………………………….72
vii

APPENDIX………………………………………………………………………………………74
BIBLIOGRAPHY………………………………………………………………………………..80
CHAPTER 3
MUNICIPAL RESIDENCY REQUIREMENT LAWS AND THEIR IMPACT ON
CITIES……………………………….…………………………………………………………..83
Abstract…………………………………………………………………………………..83
Introduction………………………………………………………………………………84
Background………………………………………………………………………………86
Types of Residency Requirements………………………………………….…….86
Pros and Cons of Residency Requirement Laws………………………….……...89
Empirical Methodology………………………………………………………………….93
Data………………………………………………………………………………………94
Results……………………………………………………………………………………95
The Effect of Residency Requirement Laws on Crime……………………..…….95
The Effect of Residency Requirement Laws on Municipal Fiscal Health……….97
The Effect of Residency Requirement Laws on Demographics………………...100
Conclusion……………………………………………………………………………...102
APPENDIX……………………………………………………………………………………..103
BIBLIOGRAPHY………………………………………………………………………………139

viii

LIST OF TABLES

Table 1.1: Summary statistics for crimes, permits, and demolitions per block face /month…….40
Table 1.2: Frequency of multiple demolitions in one month…………………………………….40
Table 1.3: Frequency of multiple cumulative demolitions over entire sample period…………..41
Table 1.4: Crimes that occurred before a permit was issued on parcels that later underwent a
demolition………………………………………………………………………………………..41
Table 1.5: Crimes that occurred during the permit period on parcels that later underwent a
demolition………………………………………………………………………………………..42
Table 1.6: Crimes that occurred on parcels after a demolition ………………………………….42
Table 1.7: Selected demographic characteristics and mean difference tests for blocks with and
without demolitions …………………………………………………………………………….43
Table 1.8: Crime statistics and mean difference tests for blocks with and without demolitions...44
Table 1.9: Average number of crimes on each block face in each month pre-permit, during
permit, and post-demolition……………………………………………………………………...44
Table 1.10: Variable Specification………………………………………………………………44
Table 1.11: Effect of demolitions and permits for demolitions on crime in Saginaw, Michigan..45
Table 1.12: Dynamic effect of demolitions and permits for demolitions on crime in Saginaw,
Michigan…………………………………………………………………………………………46
Table 1.13: Effect of demolitions on crime with permit months removed………………………47
Table 1.14: Displacement and spillover effects of demolitions and permits for demolitions on
crime in Saginaw, Michigan……………………………………………………………………..47
Table 1.15: Dynamic displacement and spillover effects of demolitions and permits for
demolitions on crime in Saginaw, Michigan…………………………………………………….48
Table 2.1: Number of Cities with Residency Requirements from 1970 through 2010………….75
Table 2.2: Residency Requirement Law Data……………………………………………….......76
Table 2.3: Fiscal Stress Ratios…………………………………………………………………...76

ix

Table 2.4: Descriptive Statistics…………………………………………………………………77
Table 2.5: Variables that Influence Residency Requirement Implementation using Repeated
Demographics……………………………………………………………………………………78
Table 2.6: Variables that Influence Residency Requirement Removal using Repeated
Demographics……………………………………………………………………………………79
Table 3.1: Number of Cities in dataset with Residency Requirements from 1970 to 2010……104
Table 3.2: Descriptive Statistics………………………………………………………………..105
Table 3.3: City Residency Laws………………………………………………………………..105
Table 3.4: Mean Effect of Residency Requirement Laws on Crime…………………………...106
Table 3.5: Mean Effect of Residency Requirement Laws on Crime with Time Trend………...107
Table 3.6: Dynamic Effects of Residency Requirement Laws on Crime………………………108
Table 3.7: Mean Effect of the Removal of Residency Requirement Laws on Crime………….109
Table 3.8: Mean Effect of the Removal of Residency Requirement Laws on Crime with Time
Trend……………………………………………………………………………………………110
Table 3.9: Dynamic Effects of the Removal of Residency Requirement Laws on Crime……..111
Table 3.10: Fiscal Stress Ratios………………………………………………………………...112
Table 3.11: Mean Effect of Residency Requirement Laws on Municipal Fiscal Health………113
Table 3.12: Mean Effect of Residency Requirement Laws on Municipal Fiscal Health with Time
Trend……………………………………………………………………………………………114
Table 3.13: Dynamic Effects of Residency Requirement Laws on Municipal Fiscal Health….115
Table 3.14: Mean Effect of the Removal of Residency Requirement Laws on Municipal Fiscal
Health…………………………………………………………………………………………...118
Table 3.15: Mean Effect of the Removal of Residency Requirement Laws on Municipal Fiscal
Health with Time Trend………………………………………………………………………...119
Table 3.16: Dynamic Effect of the Removal of Residency Requirement Laws on Municipal
Fiscal Health……………………………………………………………………………………120

x

Table 3.17: The Effect of Residency Requirement Laws on Municipal Demographics Collapsed
to Decennial Years……………………………………………………………………………...123
Table 3.18: The Effect of Residency Requirement Laws on Municipal Demographics Collapsed
to Decennial Years with Time Trend…………………………………………………………...125
Table 3.19: The Effect of the Removal of Residency Requirement Laws on Municipal
Demographics Collapsed to Decennial Years…………………………………………………..127
Table 3.20: The Effect of the Removal of Residency Requirement Laws on Municipal
Demographics Collapsed to Decennial Years with Time Trend……………………………….129
Table 3.21: Mean Effect of Residency Requirement Laws on Population Using Yearly Census of
Governments Estimates………………………………………………………………………...131
Table 3.22: Mean Effect of Residency Requirement Laws on Population Using Yearly Census of
Governments Estimates with Time Trend……………………………………………………...132
Table 3.23: Dynamic Effect of Residency Requirement Laws on Population Using Yearly
Census of Governments Estimates……………………………………………………………...133
Table 3.24: Mean Effect of the Removal of Residency Requirement Laws on Population Using
Yearly Census of Governments Estimates……………………………………………………...135
Table 3.25: Mean Effect of the Removal of Residency Requirement Laws on Population Using
Yearly Census of Governments Estimates with Time Trend………………………………...…136
Table 3.26: Dynamic Effect of the Removal of Residency Requirement Laws on Population
Using Yearly Census of Governments Estimates………………………………………………137

xi

LIST OF FIGURES.

Figure 1.1: Density of residential vacancies in Saginaw, Michigan in 2008 and 2009 with
demolitions and crimes…………………………………………………………………………..29
Figure 1.2: Demolitions and permits by month………………………………………………….31
Figure 1.3: Length of time of permits……………………………………………………………31
Figure 1.4: Traditional city block versus block face……………………………………………..32
Figure 1.5: Total residential vacancies in Saginaw, Michigan, 2002-2009…………………..….34
Figure 1.6: Average residential Taxable Value (TV) and State Equalized Values (SEV) in
Saginaw, Michigan, 2002-2009………………………………………………………………….34
Figure 1.7: Average number of crimes on a block face before and after a demolition………….35
Figure 1.8: Example of a block face and its neighbors…………………………………………..36
Figure 1.9: Dynamic effects of permits on crime………………………………………………..37
Figure 1.10: Dynamic effects of demolitions on crime………………………………………….38
Figure 1.11: Average number of crimes on a block face before and after a permit……………..40
Figure 2.1: Changes to residency requirements, 1970 to 2006…………………………………..75

xii

CHAPTER 1: THE EFFECT OF VACANT BUILDING DEMOLITIONS ON CRIME
UNDER DEPOPULATION
Abstract
The United States Government spends millions of dollars a year on vacant building
demolitions. However, no clear causal link has been established between vacant building
demolitions and crime, nor has it been determined whether demolitions affect crime globally or
merely displace it into other areas within a city. I examine these questions using block face level
monthly panel data from the city of Saginaw, Michigan. Because the permitting process for a
demolition induces a change in crime, I do not compare crime after the demolition to crime
before the demolition, but rather crime after the demolition to crime before the permit is issued.
I also analyze the displacement and spillover effects of demolitions and permits on neighboring
block faces through a spatially lagged independent variable model. Interestingly, results indicate
that both demolitions and permits actually increase crime rather than reduce it; demolitions
increase both overall crime and property crime on the block face of the demolition and permits
increase property crime on the block face of the permit. Some negative spillovers do occur onto
contiguous blocks, but these effects are not persistent over time. These results imply that by the
time a building reaches the point of demolition, it may be too late to reverse the crime trends
incurred by neighborhood decline.

1

Introduction
The durability of housing has been found to be the primary reason why urban decline is
more persistent than growth.

1

Because houses are not removed as quickly as they are built, a

negative shock decreases housing prices more than it decreases population which leads to a
slower change than does a positive shock. Because of this, decline is much longer and slower
than growth. Demolitions may be one way to counteract this durability and persistence.
However, little research exists on the effects of demolitions or on vacancies in general.

2

Policy makers and academics suggest that crime is related to both building vacancies and
3

4

depopulation. Not only does crime cause depopulation and vacant structures, but vacant
structures cause crime through increased incidences of arson, the sheltering of criminals, and the
5

creation of general disorder. With this as one justification, the United States Government
6

spends millions of dollars a year demolishing vacant buildings. Between 2008 and 2011, the
US Department of Housing and Urban Development spent almost $200 million on vacant
building demolitions under the Neighborhood Stabilization Program which is only one of several

1
2

Glaeser and Gyorko, 2005
Schilling and Logan, 2008

3

In this paper, I use the term “vacant” to refer to blighted or dangerous buildings. Not all of the
buildings that are demolished are technically vacant. Some still have residents living in them but
are dangerous enough for the city to have the power to demolish.
4
Cullen and Levitt, 1999
5

Winthrop and Herr, 2009; Spelman, 1993; Eastern Pennsylvania Organizing Project and the
Temple University Center for Public Policy, 2001; The Community Research Partners and
Rebuilt Ohio, 2008
6
Demolitions are also justified as a means of increasing surrounding property values, removing
safety hazards, and attracting residents and businesses to the neighborhood.
2

7

funding sources for demolitions. The city of Flint, Michigan alone was awarded over $3
million in 2010 -- the same year in which the number of murders in the city reached an all-time
high. The city budget was so constrained that year that the jail was shut down and as a result,
8

police officers had to issue tickets rather than arrest warrants for many offenses. Are
demolitions the best use of funding in such economically and fiscally stressed cities?
In this paper, I seek to answer part of this question by examining the relationship between
demolitions and crime. Although previous research has shown that high-rise public housing
9

demolitions reduce crime, no research has examined the link between vacant building
demolitions and crime. Vacant building demolitions differ from public housing demolitions in
that they do not redistribute concentrations of low income residents as do public housing
demolitions. Therefore, it is unclear whether crime is similarly affected by vacant building
demolitions.
I fill this gap in the literature by estimating the temporal and spatial effects of vacant
building demolitions on crime. To do this, I construct a unique data set of demolitions and crime
in Saginaw, Michigan and convert it into block face observations, or houses that share a street
face rather than a traditional city block. Because the permitting process for a demolition induces
a change in crime, I do not compare crime after the demolition to crime before the demolition,
but rather crime after the demolition to crime before the permit is issued. In order to check for
spikes in crime before the permit issuance as well as changes in crime induced by the permit
process, I estimate both a contemporaneous model that provides the mean effect of permits and

7
8
9

United States Department of Housing and Urban Development, 2011
For more serious offenses, Flint police officers use the county jail if there is space.
Hartley, 2010
3

demolitions on crime as well as a flexible model that includes lags and leads of both permits and
demolitions. I also analyze the displacement and spillover effects of demolitions and permits
through a spatially lagged independent variable model.
Results indicate that both demolitions and permits actually increase crime rather than
reduce it. One demolition increases all crime by 15% per month and property crimes in
particular by 25% per month from the time of the demolition to the end of the sample period.
Demolition permits increase property crimes by 20.4% per month on average from the time the
permit begins to the time the demolition occurs. Demolitions on contiguous block faces do not
induce a statistically significant mean change in crime on surrounding blocks, indicating that no
mean contiguous spillover effects are occurring. Some spillovers and displacements can be seen
in the dynamic model, but these effects are no persistent over time. Results are robust to a
number of different specifications including varying numbers of leads and lags and the exclusion
of multiple demolitions.
These results show that, in the short run, vacant building demolitions do not reverse the
crime levels and trends of declining neighborhoods. This is consistent with social
disorganization theory which suggests that social capital and cohesion are disrupted when a
neighborhood loses population and the social controls that put limits on criminal activity
deteriorate.

10

Demolitions are not likely to reverse this effect since they do not replace these

social controls. Results may also be consistent with an induced change in the perceived and/or
actual probability of apprehension when a house is demolished. The actual probability of
apprehension increases when there is no shelter to house a crime, and the perceived probability

10

Park and Burgess, 1925; Shaw and McCay, 1932, 1942, 1969
4

decreases due to the appearance of fewer eyes on the street. Both of these effects may cause an
increase in crime.
In their report on the cost of vacant and abandoned properties, The Community Research
Partners and Rebuild Ohio argue that vacancies create a downward spiral for neighborhood
housing markets that is difficult to correct, even with large infusions of public dollars.

11

By the

time a house gets to the point of demolition, it may be too late. Earlier interventions such as
foreclosure prevention may be more successful at reducing crime and neighborhood decline.

Conceptual Framework
There are three theories that lend insight into how demolitions may impact criminal
behavior.

12

First, the rational choice theory of crime hypothesizes that criminals maximize their

economic well-being by comparing the benefits and costs of crime such as fines, imprisonment,
and social stigmatism.

13

If the potential gain from committing a crime is sufficiently greater

than the combined risk of being caught and the size of the punishment, then the criminal chooses
to commit the crime. A demolition may induce a change in the actual or perceived probability of
being caught committing a crime due to the removal of the structure. The actual probability of
being caught committing a crime may increase after a demolition due to the lack of a shelter to
hide the crime. This may increase reported crime on or near the vacant parcel. The perceived
probability of being caught committing a crime may decrease because it may appear that the
number of eyes on the street has been reduced. This may also lead to increased levels of crime

11
12
13

The Community Research Partners and Rebuilt Ohio, 2008
For an excellent review of the crime literature, see Deller, Amiel, and Deller, 2011
Fleisher, 1963, 1966a, 1966b; Becker, 1968a, 1994; and Ehrlich, 1973, 1975
5

on a block face that has undergone a demolition. However, once a house has been demolished,
there is less property to steal or burn, reducing the benefits of crime, so property crime might
also decline. This theory does not provide a clear prediction of the sign of the effect of
demolitions on crime.
These changes in the perceived costs and benefits may cause two types of crime
displacement. First, a criminal may choose to commit a different type of crime because he can
no longer undergo the activity that he had previously undertaken.

14

For example, if a drug

dealer can no longer produce a drug in an abandoned building, he may switch to robbery as a
source of income. I examine this displacement by categorizing crimes into different groups and
examining their changes before and after a demolition. Crime displacement may also occur after
a demolition if a criminal moves to another location in order to commit the same crime. For
instance, a drug producer may simply move to another vacant building to use as a laboratory. I
examine these displacement effects through a spatially lagged independent variable model.
The second theory that relates demolitions to criminal behavior is what has been termed
the ‘broken windows’ theory.

15

This theory states that if a window in a building is broken and

left unrepaired, the rest of the windows in the building will soon be broken as well. Window
breaking does occur on a large scale because some areas are inhabited by determined windowbreakers whereas others are populated by window- lovers. Rather, one unrepaired window is a
signal that breaking more windows costs nothing. In the case of vacant buildings, the theory
implies that one vacant building lying decrepit leads to further crime solely based on the signal
that the probability of being punished is low. This suggests that demolitions cause positive

14
15

Repetto, 1974
Kelling and Wilson, 1982
6

spillover effects in which crime is reduced not only in the immediate area but in surrounding
areas as well. However, this theory is highly disputed. Some authors have found that property
disorder does increase crime and that targeting this disorder is a viable crime reduction
strategy,

16

while others find no support for the broken windows theory nor for the proposition

that broken windows policing is the optimal use of scares law enforcement resources.

17

The third theory that lends insight into the relationship between demolitions and crime is
the social disorganization theory of crime. This theory suggests that social capital and cohesion
are disrupted when a neighborhood loses population and the social controls that put limits on
criminal activity deteriorate.

18

As houses become vacant, crime increases. Previous research

has found this to be true with estimates ranging from $1,472 of public safety money spent per
19

vacant property

to a doubling of crime rates on block faces with open abandoned buildings.

20

A number of studies look at foreclosures rather than at vacant buildings with most finding a
21

positive relationship between foreclosures and crime

and some finding no relationship between

22

the two . However, demolitions are not likely to reverse this effect since they do not replace
these social controls.

16

Braga and Bond, 2008; Braga et. al., 1999; Corman and Mocan, 2005; Sousa and Kelling,
2002
17
Harcourt and Ludwig, 2006; Fagan and Davies, 2000
18
19
20

Park and Burgess, 1925; Shaw and McCay, 1932, 1942, 1969
Winthrop and Herr, 2009
Spelman, 1993

21

Teasdale, Clark, and Hinkle, 2012; Ellen, Lacoe, and Sharygin, 2011; Clark and Teasdale
2005; Stucky, Ottensmann, and Payton, 2012; Arnio, Baumer and Wolf, 2011; Goodstein and
Lee, 2010; Acevedo, 2009; Pandit, 2011; Harris, 2011; Immergluck and Smith, 2006
22
Cui, 2010; Madensen, Hart, and Miethe, 2011; Jones and Pridemore, 2012; Kirk and Hyra,
2011
7

In addition to the demolition, the permitting process for a demolition affects crime as
well. Anecdotal evidence suggests that permits increase property crime because criminals patrol
permitted structures based on public records and permit stickers on the houses.

23

These

individuals then strip vacant and nearby houses of copper pipes, fixtures, and aluminum siding.

24

This change in crime during the permit process may cause persistence in crime in that may
extend into the post demolition period; once a criminal targets a neighborhood, he may be more
likely to return later. However, it is also possible that the permitting process decreases violent
and other types of crime because of the frequent entrance and exit of city officials into the house
and neighborhood.
These theories do not provide a clear hypothesis as to how demolitions will affect crime.
It is possible that both the permit process and demolitions either increase crime or decrease crime
not only on the block face of the permit or demolition but on surrounding block faces as well.

Data
The data for this study were collected in collaboration with the city of Saginaw’s
geographical information systems (GIS) and inspections departments. The GIS department
provided block level demographics and parcel level crime data and the Inspections department
provided detailed demolition data, including permit dates, demolition costs, and sources of
funding. I then converted these data into block face level observations to better capture
neighborhood interactions as discussed in more detail later in this section. These data provide a
unique opportunity to examine the impact of permits and demolitions at a detailed level.

23
24

Personal correspondence with Thomas Fitzpatrick, Federal Reserve Bank of Cleveland
The Community Research Partners and Rebuilt Ohio, 2008
8

The data set consist of 2,443 block faces spanning 24 months: January 2008 through
December 2009. Table 1.1 displays basic summary statistics for crimes, demolitions, and
permits on these block faces. Figure 1.1 illustrates the data geographically. Both maps show the
density of vacant parcels in Saginaw in 2009 in the underlying shades of the chloropleth (the
intensity of the color indicates the degree of vacancy). The first map shows the demolitions that
took place in 2008 and 2009 and the second shows all crimes that took place in those same years.
I break incidents of crime into three categories: all crime, violent crime, and property
crime. Violent crime includes murder and non-negligent manslaughter, forcible rape, robbery,
simple assault, and aggravated assault. Property crime includes burglary and breaking and
entering, purse snatching, theft from a building, theft from a machine, theft from a vehicle, other
larcenies, motor vehicle theft, stolen property, destruction of property, shoplifting, pocket
picking, and arson. Violent crime and property crime do not sum to all crime; there are some
crimes that are not counted as either type such as fraud, intimidation, weapon law violations, and
drug/narcotic violations.
Figure 1.2 displays the number of demolitions and permits that occur each month in my
data set. Only 13 block faces have more than one demolition in a month during the time period
of my data set. This breakdown can be seen in Table 1.2. Thirty eight block faces have more
than one cumulative demolition over the entire time period which can be seen in Table 1.3. The
length of time that a house is permitted for a demolition before it is demolished varies from 0 to
23 months with a mean of 11.33 months. The frequency of permit times can be seen in Figure
1.3.
In order to more accurately model the impact of demolitions on crime, I convert the data
into block face observations rather than block level observations. A block face is all parcels that

9

face each other on a street segment rather than all parcels on a traditional city block as defined by
the census. A traditional city block includes all houses on either side of a block rather than
houses that face each other on a street. An example of a traditional city block and a block face
can be seen in Figure 1.4. The mean number of parcels on a block face is 18.74. A block face
more accurately captures the likely effect of a demolition on crime since it is logical that an
event on one side of the street has a greater impact on a house across the street than it does on a
house on the other side of the block.

Brief Background on Saginaw
Saginaw has undergone rapid depopulation since the decline of manufacturing in the
latter half of the twentieth century. Foreclosures and land vacancies that had already been
increasing were exacerbated by the financial and housing crises of 2007 which is illustrated in
Figure 1.5. Crime is also a serious problem. As mentioned previously, Saginaw ranked as the
most violent city in America from 2003 through September of 2008.

25

Property values have decreased substantially from an already depreciated level since the
financial crisis of 2007, as illustrated in Figure 1.6. Saginaw home prices devalued by 10.3% in
2009 and the unemployment rate reached 19.7%.

26

Because of the lag between the time when a

home loses value and when its official assessed value decreases, assessed property values will
continue to feel the effect of both the recession and depopulation long after they have both

25
26

Burns, 2010
Sperling’s, 2009
10

subsided.

27

Because of these issues, the municipal consulting firm of Plante & Moran projects

that the city of Saginaw could face a $19.9 million deficit by 2014 if leaders do not adjust to
declining revenues and a shrinking population.

28

Demolition Process
Demolitions are a key part of Saginaw’s policy relating to vacant and abandoned
buildings. To implement demolitions, Saginaw uses both Community Development Block Grant
(CDBG) and Neighborhood Stabilization Program (NSP) funds provided by the US Department
of Housing and Urban Development (HUD). The buildings that Saginaw demolishes need not be
vacant; they only need to be qualified as structurally dangerous. They could be owned by private
individuals, corporations, limited liability companies, mortgage companies, the county treasurer,
or the county land bank. The city does not take ownership of the property; it only enforces its
right and obligation to keep the city safe. Once demolished, the property value drops to the
current local value for a vacant lot of equal size.

27

Since the approval of the General Property Tax Act in 1893, property taxes have been the
main source of revenue for local governments in Michigan. However, the property tax structure
was altered in 1994 by Proposal A which placed a constitutional cap on the growth of taxable
value (TV). Since Proposal A was instated, the TV of a property has been allowed to increase
only by the lesser of the rate of inflation or five percent until the property is transferred (not
including additions or new construction). Historically, this value has been below the state
equalized value (SEV) which has led to a general decline in property tax revenues as a
proportion of property values. It was not anticipated that the SEV would begin to decrease and
eventually fall below TV as has already occurred for some properties throughout the state. When
this happens, the property tax paid by the owner follows the fall in SEV. In the short-run,
Proposal A may help to insulate local revenues from the declining home values. However, when
house prices do stabilize and begin to increase, TV will be ratcheted downward and local unit
fiscal capacity may not recover for years.
28
Engel, 2010
11

Once a building is put on the dangerous building list, it is demolished in numerical order.
There are three ways in which a property can be placed on the dangerous building list. First, a
building can be added to the list through a resident complaint filed with the city's complaint
department or clerk's office. Second, a building can be added due to an internal complaint from
a city worker who observed the property firsthand while in the field or received complaints at a
neighborhood meeting. Third, a building can be added through citywide sweeps undertaken by
the inspections department each spring and sometimes in the fall.

29

The first two methods of adding a house to the dangerous building list may systematically
correlated with crime since it is likely that residents are more likely to complain about a property
if there is crime occurring in or near it and city workers may be in the neighborhood because of
crime. I inspect for spikes in crime before permit issuance by including leads of the permit
variable in my dynamic specification and find no evidence of this effect.
If a building undergoes arson, it is immediately demolished. I use this information to
examine the endogeneity of demolitions and arson within Saginaw. I also exclude the 25
emergency demolitions and the arsons that caused them from my dataset because they provide a
direct source of endogeneity. Although this is a concern for the analysis in that it may be
causing me to miss an important crime reduction caused by demolitions, there is additional
concern that the structure of the demolition policy is incentivizing residents to commit arson
against neighboring vacant homes because if a vacant building undergoes arson, it is moved to
the top of the demolition list. If I include these emergency demolitions and arsons in my analysis
I am not able to determine whether the demolition policy causes the arsons, whether the arsons
cause the demolitions, or what I would truly like to know: whether demolitions reduce the

29

Personal correspondence with Scott Crofoot, Dangerous Buildings Inspector, City of Saginaw
12

incidence of arson. I do include all other arson that does not immediately lead to a demolition
which allows me to determine whether demolitions reduce the incidence of arson in the
neighborhood of the demolition. In addition, I perform robustness checks that include and
exclude all arson to determine whether they cause bias in my results.
The demolitions process follows a set procedure. When the house becomes permitted for
a demolition, a notice to the owner is placed at the front entry of the building as part of the
notification requirements. This occurs once at the beginning of the process and again 75-80 days
later when the notice of findings is posted. Fifteen to thirty days after the last posting, the house
is measured to provide cubic volume and determine demolition costs as well as inspected for
asbestos.

30

This activity near the end of the permit period is believed to cause an increase in

theft of anything that has a metal content.

31

Table 1.4 lists crimes that occurred before permit issuance on parcels that underwent a
demolition and Table 1.5 lists crimes that occurred during permit issuance on parcels that
underwent a demolition. The most common crime in both lists is arson, followed by burglary,
larceny, and damage to property. Surprisingly, there are no incidences of drug crimes on these
vacant properties. Table 1.6 lists crimes that occurred on parcels after a demolition took place.
Surprisingly, there were still 4 arsons even after the vacant building was removed. These are
likely arsons in which brush or garbage was lit on fire.

30

The City of Saginaw does not board up houses. Any boarding that occurs is either completed
by the owner or by volunteers during an event that takes place each year before Halloween. If
the city did systematically board up vacant houses it would likely affect the incidence of crime
on that parcel.
31
Personal correspondence with Scott Crofoot, Dangerous Buildings Inspector, City of Saginaw
13

Empirical Methodology
To estimate the causal effect of a demolition on crime, I would like to compare a block
face that underwent a demolition to itself had it not undergone the demolition. This,
unfortunately, is not possible. Alternatively, comparing a block face that underwent a demolition
to one that did not will be misguided since there may be unobserved factors that make block
faces with demolitions different from block faces without demolitions, which can be seen in
Tables 1.7 and 1.8.

32

This leads to inconsistent estimates if using a cross section for analysis.

Estimates that do not control for month fixed effects also lead to misleading results since
a large portion of the demolitions occurred in the winter when crime levels are the lowest. Table
1.9 displays the average number of crimes before a permit, during the permit period, and after a
demolition on the block faces in the city that had a demolition during the data period. The
permit period is associated with a slightly higher average level of total crime and property crime
than the pre-permit period, and no mean difference in violent crime. The post demolition period
is associated with a slightly lower level of total crime than the pre-permit period and a slightly
higher level of property crime, with no difference in violent crime. However, these results likely
reflect the monthly seasonality of crime. Figure 1.7 plots the average monthly number of crimes
before and after a demolition. It appears from this graph that average crime levels are declining
prior to a demolition and then remaining fairly steady after a demolition. However, this also is
most likely a reflection of seasonality.
I therefore use a within estimator to compare block faces that underwent a demolition to
themselves over time which removes the time invariant unobserved heterogeneity of each block
face. However, I do not compare crime after a demolition to crime immediately before a
32

Note that demographic data are only available at the block level (not at the block face level)
so these means are for the block rather than block face.
14

demolition because the permit process for a demolition induces a change in crime that biases
results. If the permit process increases property crime which appears to be the case both in the
data and anecdotally, using the permit period as the base period will bias coefficients
downwards. Alternatively, violent and other crime may actually decrease during the permit
period due to city officials entering and leaving the premises. This will cause these types of
crime to appear to increase after a demolition when in fact they are reverting to the mean.
In order to examine and account for these trends, I estimate both a contemporaneous
model with the variables specified as stock variables and a dynamic model with the variables
specified as flow variables. The stock of demolitions is the cumulative number of demolitions
that have occurred on the block face and the stock of permits is the cumulative number of houses
with permits that have not yet undergone a demolition. The flow variable for each is defined as
the first difference of the stock variable, i.e.:

DemosFlowi ,t = DemosStocki ,t − DemosStocki ,t −1

PermitsFlo wi ,t = PermitsSto ck i ,t − PermitsSto ck i ,t −1
An example of both stock and flow variables for each variable can be found in Table 1.10.
Because my dependent variable, the number of crimes on a block face in a given month,
takes on non-negative integer values (is a count variable), a linear model for E(y│x) is not ideal
because it can lead to negative predicted values.

33

33

Also, because y can take on the value zero

Wooldridge, 2002 p 388
I use levels of crime as my dependent variable rather than the crime rate because no good timevarying estimate of population per block exists in my data set. The only number that I could use
15

with positive probability, a log transformation is inappropriate. Therefore, I assume that

given

xi

yi

takes on a Poisson distribution. The Poisson model is as follows:

f ( yi ) =
λi , t = e
where

x i ,t β

e

− λi , t

λi ,t y

i ,t

y i ,t !

for

y i =0,1,2…

(1)

= E ( yi ,t xi ,t ) = Var ( yi ,t xi ,t ) .34 Provided that

E ( y i ,t xi ,t ) = exp( xβ ) , estimates of β

are consistent even if the mean does not

35

equal the variance -- when there is over dispersion.

I then estimate the above model using multinomial quasi-conditional maximum likelihood

estimation as described in Wooldridge, 1999 using two main equations for

xi ,t β . In the first,

is the number of occupied parcels on each block. However, this number is defined as the number
of parcels with a building on it, not the number of parcels with a person living on it. Therefore,
this number changes when a demolition occurs and would cause simultaneity in my estimation.
In addition, an occupied parcel could have any number of residents on it from a single person
home to a multi-family rental. Therefore a crime rate based on this number would not only be
inaccurate but would also cause bias in my results.
34
Winkelmann, 2008
35

Fixed effects estimations in nonlinear models such as this one generally lead to inconsistent
estimates. However, the Poisson distribution can be arbitrarily misspecified and any kind of
serial correlation can be present and the fixed effects Poisson estimator is consistent under mild
regularity conditions (Wooldridge, 2002 p 648). Provided that E ( y | x ) = exp( xβ ) , estimates of
β are consistent even if the mean does not equal the variance. Therefore, a fixed effects model
is appropriate and over dispersion can be ignored.
16

I specify both permits and demolitions as stock variables and estimate the following equation:

x i ,t β = α 0 + δ 1 Permits i ,t + β 1 Demos i ,t + θ i + γ t + u it

Each variable is measured at the block face and month level.

number of permits issued for block face i in month t.

Demos i ,t

demolitions on block face i in month t. The coefficient

Permitsi ,t

(2)

equals the

equals the number of

δ1 on Permits i ,t

gives us the mean

impact of permits on crime from the month of issuance to the month of the demolition. The
coefficient

β1

on

Demosi ,t

tells us the mean impact of demolitions on crime from the

month of the demolition through the end of the sample period. Block face and month fixed
effects are represented by

θi

and

γt

respectively. Time invariant variables are captured by

the block face level fixed effects and are therefore not included in this equation. I use standard
errors clustered at the block face level that are robust to heteroskedasticity and arbitrary forms of
error correlation within each block face.

I then estimate a second equation for

xi ,t β

which allows the effect of permits and

demolitions to vary over time through use of a dynamic treatment specification. I specify the
permit and demolitions variables as flow variables and include two leads and five lags of the
permit variable and five lags of the demolition variable. I also include a variable that equals the
sum of all demolitions that took place six or more month previously and all permits that took
place six or more months previously. This is specified as follows:

17

xi , t β = α 0 +

5

∑δ

j =−2
5

∑β
j =0

24

j

Permitsi,t − j + δ 6 ∑ Permitsi,t − j +
j =6

(3)

24

j

Demosi,t − j +β 6 ∑ Demosi ,t − j + θ i + γ t + uit
j =6

Variables are once again measured at the block face and month level.

Permitsi ,t − j

equals the number of permits issued for block face i in month t-j. This variable is set to zero if
the block face has undergone the demolition for which this permit was issued in month t-j. This
ensures that the permit and demolition variables are not both greater than zero at the same time,
allowing me to separately identify the dynamic effects of the permit from the dynamic effects of
the demolition.
The

δj

Demosi ,t − j

equals the number of demolitions on block face i in month t-j.

coefficients indicates the cumulative effect of a permit j months ago on crime

now, compared to 3 or more months before the permit and the

βj

coefficients indicates the

cumulative effect of a demolition j months ago on crime now, compared to 3 or more months
before the permit. The second and fourth summations in equation (3) represent all permits and
demolitions that occurred six or more months before month t. The second summation is also set
to zero when the demolition for which the permit was issued occurred in month t-j.
To further check that I am separately identifying the effect of permits and demolitions
and to find the mean effect of a demolition on crime ignoring the permit period, I estimate the

18

following equation with permit months dropped from the sample and demolitions specified as
stock variables:

xi ,t β = α 0 + β 1 Demos i ,t + θ i + γ t + u it

Coefficient

(4)

β1 in equation (4) provides the mean effect of a demolition on crime from

the time of the demolition to the end of the sample period as compared to the pre-permit period.

Displacements and Spillovers
Knowing whether crime changes on one block face due to a demolition does not tell us
whether overall crime changes or is merely displaced into surrounding neighborhoods. I
therefore estimate a model in which I include a spatial lag of demolitions. In other words, I add
into the models above the sum of demolitions and permits that occurred in block faces that touch
block face i, or block face i’s neighbors. An example of a block face and its neighbors is
illustrated in Figure 1.8.
First, I estimate the mean effects model with a spatial lag as follows:

xi ,t β = α 0 + δ 1 Permitsi ,t + β1 Demosi ,t +

δ 2Wi Permitst + β 2Wi Demost + θ i + γ t + u it
I then estimate both temporal and spatial lags together as follows:

19

(5)

xi ,t β = α 0 +

5

∑δ Permits

i ,t − j

j

j = −2
5

∑ β Demos

i ,t − j

j

j =0

24

+ δ 6 ∑ Permitsi ,t − j +
j =6

24

+β6 ∑ Demosi ,t − j +
j =6

24

5

∑δ

Wi Permitst − j

2j

j = −2

5

j =6

(6)

j =0

+ δ 26Wi ∑ Permitst − j + ∑ β2 jWi Demost − j +
24

β 26 ∑Wi Demost − j + γ t + ψ t + uit
j =6

where

Wi

is an 1 x I spatial weights vector that assigns a weight of 1 to block faces that are

contiguous to block face i and a weight of 0 to block faces that are noncontiguous to block face
i.

36

Permits t

is an I x 1 vector of the permits on every block face in the city.

Demost

is

an I x 1 vector of the demolitions on every block face in the city.

Results
Results indicate that, compared to the pre-permit period, one demolition actually
increases all crime at a mean rate of 15% per month and property crime at a mean rate of 25%
per month from the time of the demolition through the end of the sample period as can be seen in
Table 1.11.

37

Permits also increase property crime at a mean rate of 20.4% per month during the

permit period.
36

Drukker, Peng, Prucha, and Raciborski, 2011

37

These results are robust to the use of lags of demolitions and leads of permits as independent
variables in the place of the contemporaneous variables.
20

Results from the flexible model that includes lags and leads of permits and demolitions
can be found in Table 1.12 and graphically in Figures 1.9 and 1.10. The dark line graphs the

δj

and

βj

coefficients and the lighter lines above and below are two standard deviations

above and below these coefficients. Permits increase violent crime by 54.8% the month that the
permit occurs, but this effect dies out after that first month. Permits increase property crime by
43.7% one month after issuance which then increases to a cumulative effect of and 52.2% two
months after issuance. The cumulative effect of permits on property crime six or more months
after permit issuance is 35.2%.
Demolitions increase all crime by 49.5% the month of the demolition which decreases to
45.2% two months after the demolition, ending at a cumulative increase of 30.9% six or more
months after the demolition occurs. Demolitions also increase property crime; by 72.2% two
months after a demolition and a cumulative effect of 63.8% six or more months after the
demolition.
There may be concern that there is a spike in crime before permit issuance because of the
ways in which houses are put on the permit list as discussed in section 3b. If this were true,
using the month immediately prior to permit issuance as a baseline for analysis will lead to
downward bias due to mean reversion after this spike, similar to the effects of an Ashenfelter Dip
found in labor studies.

38

Figure 1.11 plots the average number of crimes before and after permit

issuance. From this graph, there appears to be a slight spike in crime before permit issuance.
38

Ashenfelter and Card, 1985
An Ashenfelter dip refers to the decline in mean earnings of participants in employment and
training programs during the period just prior to participation. This dip causes bias if this period
is used as a baseline because of mean reversion after the dip. This is similar to the increase in
crime immediately prior to a demolition permit. If used as a baseline, the effect of permits and
demolitions on crime may appear to be smaller than they actually are because they are being
compared to an elevated level of crime before the permit.
21

However, when examining this more formally through leads of permits, the leads are not
statistically significant indicating that there is no mean reversion bias occurring in my analysis.
Results from equation (4) in Table 1.13 with permit months dropped from the sample
show similar results to the full sample regression: demolitions increase all crime by 17.2% and
property crime by 29.8% per month. This is the mean effect of a demolition on crime compared
to the pre-permit period from the time of the demolition to the end of the sample period.

Displacement and Spillover Results
Demolitions and permits on contiguous block faces do not induce a statistically
significant average change in crime on the block in question as can be seen in Table 1.14. This
indicates that the cumulative increase in crime that is found in the results above is not caused by
crime displacement from surrounding block faces. This is not to say that crime displacement
from other areas within the city is not occurring, but it is not occurring on average from
geographically contiguous block faces.
When examined dynamically, however, some displacement and negative spillover effects
can be seen as in Table 1.15. There is a statistically significant increase in property crime on a
block face three months after a permit occurs on a contiguous block of 25.9% and 32.3% two
months after a demolition occurs. This could be caused by negative spillover effects of vacant
property on the nearby block, and the attraction of criminals to the area during the permitting
process. Violent crime actually decreases on a block one month after a demolition occurs on a
surrounding block by 39.8%. This could be caused by displacement of the violent crime from
that block face to the block face with the vacant parcel.

22

In addition, the dynamic effect of demolitions and permits on crime on their own block
changes slightly when contiguous permits and demolitions are held constant. Permits now cause
an increase in violent crime of 57.6% during the month of the permit and an increase in property
crime of 41.6% then up to 48.7% two months after the permit, summing to a cumulative effect of
33.2% six months or more after a permit. Holding demolitions and permits on contiguous blocks
constant, a demolition on a block in question increases all crime cumulatively by 30.2% six or
more months after a demolition, violent crime by 61.1% five months after a demolition, and
58.2% six or more months after a demolition.

Robustness Checks
Strict Exogeneity
If demolitions and permits are not randomly chosen but instead are partially determined
by crime, β 2 will be inconsistent due to endogeneity caused by selection bias. This appears not

ˆ

to be the case since the leads on the permit variable in the dynamic specification are not
statistically significant. Hence, strict exogeneity is not violated.
There may still be concern that demolitions are partially determined by crime trends in a
neighborhood; i.e. that demolitions are implemented in neighborhoods where crime is increasing.
In order to examine whether this is true, I perform a robustness check of the model that includes
time trends as follows:

23

CrimeIncidentsi ,t = α 0 +

5

∑δ

j

Permitsi ,t + j × NoDemoi ,t + j +

j = −2

(4)

24

5

j =6

j = −2

δ 6 ∑ Permitsi ,t + j × NoDemoi ,t + j + ∑ β j Demosi ,t − j +
24

∑ β Demos
6

i ,t + j

+ θ i + γ t + ψ t×i + uit

j =6

where

ψ t ×i

are block face by month fixed effects, or block face/month time trends. The

coefficients are once again very similar to the model run without the time trend fixed effects and
of the same sign and significance, which indicates that results are not driven by block face trends
in crime.

Number of Lags and Leads
The specifications above are robust to different baseline months (leads on permits) as
long as the baseline is at least two months before the permit is issued due to the spike in crime
immediately before a permit. Results using a baseline of two, three, and four months before the
permit are very similar to those that use two months before the permit as the baseline.

39

When using one month before a permit as a baseline, results are biased downwards as
expected due to a spike in crime immediately before a house is permitted for a demolition as
discussed above. This makes it appear that crime is decreasing due to the permit when in fact it
is merely reverting to the mean.

39

Results of the robustness checks are available upon request of the author. For the sake of
brevity, they are not included in this paper.
24

Arson
Once a house undergoes any type of fire it is immediately demolished. Because of the
direct endogeneity caused by this policy, I cannot include demolitions due to arson in my
analysis. There is anecdotal evidence to suggest that this demolition policy is at least in part
causing arson in vacant buildings. Homeowners near vacant buildings have learned that if a
vacant building undergoes arson it will be demolished more quickly, which neighbors perceive
as a benefit. Therefore, even without the estimation issues caused by the endogeneity between
demolitions and arson, it would be difficult to determine which incidences of arson were caused
by vacant buildings themselves and which were caused by the structure of the demolition policy.
In order to deal with this issue, I first remove all demolitions that were caused by arson,
and all arsons that led to demolitions from my data set. Because there may still be a concern that
some arsons were undertaken in order to induce a demolition or that some demolitions were
caused by arsons that were not coded properly, I do a robustness check for endogeneity bias of
all regressions by removing all arsons. The results are very similar and of the same sign and
significance, suggesting that this concern is not a major issue.

Multiple Demolitions
Another estimation concern may be that blocks that have more than one demolition in the
same month may be biasing results because the specification assumes that two demolitions have
twice the impact as one. In my sample of 2,510 block faces, 226 block faces have one
cumulative demolition, only 11 blocks have two demolitions in one month, one has three
demolitions in one month, and one has four demolitions in one month.

25

Because there are so few block faces that underwent more than one demolition in a
month, there is not enough variation to separately identify separate variables for the number of
demolitions, or to identify a separate squared term on demolitions. Therefore, to check this
assumption, I estimate a subsample of block faces by dropping those that have more than one
demolition in a month. Results of this robustness check are once again not significantly different
from the baseline results, leading me to believe that this assumption is valid.

Conclusion
Millions of dollars are spent each year demolishing vacant buildings with an often cited
justification that demolitions reduce crime. However, my results show that both demolitions and
permits actually increase crime rather than reduce it. Compared to the pre-permit period, one
demolition increases all crime at a mean rate of 15% and property crime at a mean rate of 25%
from the time of the demolition through the end of the sample period. The permitting process for
a demolition also increases property crime at a mean rate of 20.4%. Although these percentages
seem large, these effects equate to an increase of 0.79 total crimes and 0.51 property crimes per
year caused a demolition and 0.42 property crimes per year caused by a permit because the
percentages are for such a small geographic area and time period.
Although there appears to be some dynamic displacement and spillover effects of crime
from a demolition and permit onto surrounding block faces, the mean displacement and spillover
effects are null. This suggests that crime displacement is not occurring and overall crime in a
neighborhood increases when a demolition or permit issuance occurs.
These results call into to question the use of demolitions as a crime reduction tool and the
methods of permitting a house for demolition. The average de-construction cost of demolishing

26

a house in Saginaw is $5,020.58. This does not include the overhead costs of the city demolition
department or the overhead of administering the grants by US Department of Housing and Urban
Development and other agencies. This money may be better spent on public safety personnel if
crime reduction is the goal.
In the short run it is clear that demolitions do not reverse the crime levels and trends of
declining neighborhoods. Earlier interventions such as foreclosure prevention may be more
successful at impeding neighborhood deterioration and crime trends. Additional research is
needed to determine whether demolitions can reverse long run neighborhood declines.

27

APPENDIX

28

APPENDIX
Tables and Figures
Figure 1.1: Density of residential vacancies in Saginaw, Michigan in 2008 and 2009 with
demolitions and crimes

29

Figure 1.1 (cont’d)

30

Figure 1.2: Demolitions and permits by month

70
60
50
40
30
Demolitions

20

Permits

10
0

Month

Figure 1.3: Length of time of permits

40
35
Frequency

30
25
20
15
10
5
0
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
Permit Time (Months)

31

Figure 1.4: Traditional city block versus block face

Traditional City Block

Map by Christina Plerhoples

32

Figure 1.4 (cont’d)

Block Face

Map by Christina Plerhoples

33

Figure 1.5: Total residential vacancies in Saginaw, Michigan, 2002
5:
2002-2009
3,300
3,200
3,100
3,000
2,900
2,800
2,700
2,600
2,500
2002

2003

2004

2005

2006

2007

2008

2009

Figure 1.6: Average residential Taxable Value (TV) and State Equalized Values (SEV) in
6:
Saginaw, Michigan, 2002-2009

$18,000
$17,000
$16,000
$15,000
SEV

$14,000

TV

$13,000
$12,000
$11,000
$10,000
2002 2003 2004 2005 2006 2007 2008 2009

34

Figure 1.7: Average number of crimes on a block face before and after a demolition
0.7

Mean
Number of
Crimes

0.6
0.5
0.4
0.3
0.2
0.1
0
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Months Since Demolition

35

Mean
Number of
Violent
Crimes
Mean
Numer of
Property
Crimes

Figure 1.8: Example of a block face and its neighbors

Map by Christina Plerhoples

36

Figure 1.9: Dynamic effects of permits on crime

All Crime
1.0
0.5
0.0
-2

-1

0

1

2

3

4

5

6+

4

5

6+

-0.5
-1.0

Months Since Permit

Violent Crime
1.0
0.5
0.0
-2

-1

0

1

2

3

-0.5
-1.0
Months Since Permit

37

Figure 1.9 (cont’d)

Property Crime
1.0
0.5
0.0
-2

-1

0

1

2

3

4

5

6+

-0.5
-1.0
Months Since Permit

Figure 1.10: Dynamic effects of demolitions on crime

All Crime
1.2
0.2
0

1

2

3

4

-0.8
-1.8
Months Since Demolition

38

5

6+

Figure 1.10 (cont’d)

Violent Crime
1.2
0.2
0

1

2

3

4

5

6+

-0.8
-1.8
Months Since Demolition

Property Crime
1.3
0.8
0.3
-0.2

0

1

2

3

4

-0.7
-1.2
-1.7

Months Since Demolition

39

5

6+

Figure 1.11: Average number of crimes on a block face before and after a permit
0.7

Mean
Number of
Crimes

0.6
0.5

Mean
Number of
Violent
Crimes
Mean Numer
of Property
Crimes

0.4
0.3
0.2
0.1
0
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Months Since Permit

Table 1.1: Summary statistics for crimes, permits, and demolitions per block face /month

Variable (per block face /month)
All Crimes
Violent Crimes
Property Crimes
Demolitions
Permits
Permit Time among Permitted Block Faces

No. of
Obs.
58,632
58,632
58,632
58,632
58,632
251

Total Mean
18,197 0.31
4,812
0.08
6,973
0.12
255
0.00
251
0.00
2,789 11.33

Table 1.2: Frequency of multiple demolitions in one month
Demolitions
0
1
2
3
4
Total

Freq.
2,204
226
11
1
1
2,443

40

Std.
Dev.
0.85
0.45
0.48
0.07
0.10
4.43

Min Max
0
24
0
11
0
15
0
4
0
4
1
23

Table 1.3: Frequency of multiple cumulative demolitions over entire sample period
Cumulative Demolitions
0
1
2
3
4
Total

Freq.
2,235
170
31
5
2
2,443

Table 1.4: Crimes that occurred before a permit was issued on parcels that later underwent
a demolition
Offense Description
Frequency
Arson – Residence
26
Burglary – Forced Entry – Residence
14
Larceny
6
Damage to Private Property
6
Assault and Battery/Simple Assault
3
Disorderly Conduct
3
Aggravated Assault – Non-Family -- Gun
2
Burglary – Forced Entry – Non-Residence
2
Larceny – Personal Property from Vehicle
2
Robbery – Business Strong Arm
1
Robbery – Street Gun
1
Arson (other)
1
Arson – Burning of Real Property
1
Burglary – No Forced Entry – Residence
1
Larceny – From Yards/Grounds
1
Fraud (Other)
1
Embezzlement – Business Property
1
rd
Retail Fraud, Theft 3 Degree
1
Disorderly Conduct (Other)
1
Traffic - Furnish False Info to Officer
1
Health and Safety Violations
1
Skipped Number
1
Inspections/Investigations - Lost and Found Property 1
Miscellaneous - General Assistance
1

41

Table 1.5: Crimes that occurred during the permit period on parcels that later underwent
a demolition
Frequency
Offense Description
Arson – Residence
23
Larceny (Other)
5
Burglary – Forced Entry – Residence
3
Aggravated Assault\Non-Family – Gun 2
Robbery - Business Strong Arm
1
Robbery – Street Gun
1
Burglary – Forced Entry – Non-Residence 1
Burglary - No Forced Entry – Residence 1
Damage to Property – Private Property
1

Table 1.6: Crimes that occurred on parcels after a demolition
Frequency
4
1

Offense Description
Arson – Residence
Dog Law Violations

42

Table 1.7: Selected demographic characteristics and mean difference tests for blocks with
and without demolitions
Block faces with no Block faces with >0
Demolitions
Demolitions
Difference P-Value
Population
32.54
39.68
-7.14
0.04
(0.88)
(2.68)
White
16.02
8.04
7.99
0.00
(0.58)
(0.89)
Black
13.37
27.50
-14.13
0.00
(0.61)
(2.41)
Males
15.20
17.90
-2.70
0.12
(0.43)
(1.17)
Females
17.34
21.78
-4.44
0.03
(0.49)
(1.57)
Median Age
25.77
27.14
-1.37
0.39
(0.40)
(0.93)
Number of Households
12.29
13.71
-1.42
0.31
(0.35)
(0.98)
Average Household Size
2.11
2.77
-0.66
0.00
(0.03)
(0.08)
Owner Occupied Housing
7.89
7.47
0.42
0.66
(0.24)
(0.45)
Renter Occupied Housing
4.40
6.23
-1.84
0.04
(0.22)
(0.79)
Vacant
1.25
2.21
-0.95
0.00
(0.05)
(0.20)
Size
273,795
214,886
58,909
0.39
(17202)
(20798)
Standard errors are listed in parentheses below the means

43

Table 1.8: Crime statistics and mean difference tests for blocks with and without
demolitions

-0.037

PValue
0.007

-0.028

0.035

-0.013

0.02

-0.014

0.119

-0.014

Block faces with no
Block faces with > 0
Demolitions
Demolitions
All Crimes
0.237
0.274
(0.003)
(0.013)
0.23
0.258
All Crimes Sans
Arson
(0.003)
(0.012)
Violent Crimes
0.048
0.061
(0.001)
(0.005)
Property Crimes
0.117
0.131
(0.002)
(0.009)
0.117
0.131
Property Crimes
Sans Arson
(0.002)
(0.009)
Standard errors are listed in parentheses below the means

0.119

Difference

Table 1.9: Average number of crimes on each block face in each month pre-permit, during
permit, and post-demolition

All Crime
Violent Crime
Property Crime

Pre-Permit Permit
0.43
0.45
0.12
0.12
0.15
0.18

Block faces with a
Demo, All Months
0.44
0.13
0.17

Post-Demo
0.40
0.12
0.16

All Block faces,
All Months
0.31
0.08
0.12

Table 1.10: Variable Specification
Month
Demos flow
Demo stock
Permits flow
Permits stock

12345
00000
00000
01000
01111

6
1
1
-1
0

789
000
111
000
000

10
0
1
0
0

11
0
1
1
1

12
0
1
0
1

44

13
0
1
0
1

14
0
1
0
1

15
0
1
0
1

16
0
1
0
1

17
1
2
-1
0

18
0
2
0
0

19
0
2
0
0

20
0
2
0
0

21
0
2
0
0

22
0
2
0
0

23
0
2
0
0

24
0
2
0
0

Table 1.11: Effect of demolitions and permits for demolitions on crime in Saginaw,
Michigan
(1)
(2)
(3)
All Crime
Violent Crime
Property Crime
Permits
0.085
-0.048
0.204**
(0.070)
(0.114)
(0.093)
Demolitions
0.150**
0.157
0.250**
(0.065)
(0.135)
(0.104)
Observations
48,504
33,120
41,808
Number of block faces
2,021
1,380
1,742
Robust standard errors in parentheses, clustered at the block face level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of all block faces in Saginaw, MI
from 2008-2009. Crime offenses refer to the number of incidents on each block face in each
month.

45

Table 1.12: Dynamic effect of demolitions and permits for demolitions on crime in
Saginaw, Michigan
(1)
(2)
(3)
All Crime
Violent Crime
Property Crime
Two months before permit
0.038
0.137
0.152
(0.101)
(0.225)
(0.163)
One month before permit
0.027
0.198
0.197
(0.120)
(0.244)
(0.173)
Month of permit
0.202
0.548**
0.097
(0.142)
(0.235)
(0.226)
One month after permit
0.192
0.172
0.437**
(0.171)
(0.266)
(0.210)
Two months after permit
0.210
0.010
0.522***
(0.153)
(0.277)
(0.197)
Three months after permit
0.133
0.368
0.131
(0.175)
(0.236)
(0.238)
Four months after permit
0.169
0.288
0.358
(0.157)
(0.295)
(0.241)
Five months after permit
0.004
0.290
-0.055
(0.163)
(0.261)
(0.236)
Six or more months after permit
0.140
0.136
0.352*
(0.132)
(0.218)
(0.195)
Month of demolition
0.495**
0.635
0.583
(0.235)
(0.522)
(0.392)
One month after demolition
0.168
0.486
-0.131
(0.227)
(0.324)
(0.426)
Two months after demolition
0.452**
0.422
0.722***
(0.203)
(0.296)
(0.265)
Three months after demolition
0.151
-0.112
0.295
(0.211)
(0.410)
(0.346)
Four months after demolition
0.141
0.431
0.396
(0.216)
(0.332)
(0.325)
Five months after demolition
0.187
0.616*
0.072
(0.223)
(0.359)
(0.294)
Six or more months after demolition
0.309**
0.323
0.638***
(0.146)
(0.231)
(0.197)
Observations
30,400
19,072
24,560
Number of block faces
1,900
1,192
1,535
Robust standard errors in parentheses, clustered at the block face level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of all block faces in Saginaw, MI
from 2008-2009. Crime offenses refer to the number of incidents on each block face in each
month.

46

Table 1.13: Effect of demolitions on crime with permit months removed
(1)
(2)
(3)
All Crime
Violent Crime
Property Crime
Demolitions
0.172**
0.161
0.298**
(0.086)
(0.156)
(0.131)
Observations
45,988
31,135
39,354
Number of block faces
2,001
1,356
1,701
Robust standard errors in parentheses, clustered at the block face level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of all block faces in Saginaw, MI
from 2008-2009 with permit months dropped from the sample. Crime offenses refer to the
number of incidents on each block face in each month.

Table 1.14: Displacement and spillover effects of demolitions and permits for demolitions
on crime in Saginaw, Michigan
(1)
(2)
(3)
All Crime
Violent Crime
Property Crime
Permits
0.092
-0.042
0.201**
(0.070)
(0.113)
(0.094)
Permits on contiguous block faces
-0.031
-0.032
0.022
(0.032)
(0.050)
(0.047)
Demolitions
0.167***
0.172
0.256**
(0.064)
(0.138)
(0.105)
Demolitions on contiguous block faces
-0.047
-0.052
-0.005
(0.033)
(0.053)
(0.054)
Observations
48,504
33,120
41,808
Number of block faces
2,021
1,380
1,742
Robust standard errors in parentheses, clustered at the block face level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of all block faces in Saginaw, MI
from 2008-2009. Crime offenses refer to the number of incidents on each block face in each
month.

47

Table 1.15: Dynamic displacement and spillover effects of demolitions and permits for
demolitions on crime in Saginaw, Michigan
(1)
(2)
(3)
All
Violent
Property
Crime
Crime
Crime
Two months before permit
0.045
0.144
0.155
(0.102)
(0.228)
(0.164)
One month before permit
0.036
0.201
0.202
(0.118)
(0.237)
(0.177)
Month of permit
0.204
0.576**
0.087
(0.143)
(0.233)
(0.228)
One month after permit
0.182
0.172
0.416**
(0.172)
(0.266)
(0.210)
Two months after permit
0.197
0.003
0.487**
(0.153)
(0.276)
(0.198)
Three months after permit
0.122
0.366
0.121
(0.177)
(0.237)
(0.240)
Four months after permit
0.164
0.287
0.338
(0.158)
(0.293)
(0.241)
Five months after permit
0.009
0.299
-0.059
(0.162)
(0.264)
(0.234)
Six or more months after permit
0.142
0.155
0.332*
(0.131)
(0.217)
(0.195)
Two months before permit on contiguous block faces
-0.019
-0.081
-0.044
(0.051)
(0.083)
(0.091)
One month before permit on contiguous block faces
-0.036
-0.172*
0.079
(0.054)
(0.100)
(0.094)
Month of permit on contiguous block faces
-0.036
-0.035
-0.040
(0.092)
(0.137)
(0.163)
One month after permit on contiguous block faces
-0.012
-0.200
0.027
(0.095)
(0.138)
(0.138)
Two months after permit on contiguous block faces
0.071
-0.030
0.146
(0.086)
(0.142)
(0.139)
Three months after permit on contiguous block faces
0.076
0.007
0.259*
(0.092)
(0.140)
(0.151)
Four months after permit on contiguous block faces
-0.013
-0.118
0.028
(0.094)
(0.142)
(0.157)
Five months after permit on contiguous block faces
-0.015
-0.109
-0.015
(0.095)
(0.130)
(0.155)
Six or more months after permit on contiguous block
faces
-0.012
-0.188
0.126
(0.078)
(0.125)
(0.133)

48

Table 1.15: (cont'd)

Month of demolition
One month after demolition
Two months after demolition
Three months after demolition
Four months after demolition
Five months after demolition
Six or more months after demolition
Month of demolition on contiguous block faces
One month after demolition on contiguous block faces
Two months after demolition on contiguous block faces
Three months after demolition on contiguous block faces
Four months after demolition on contiguous block faces
Five months after demolition on contiguous block faces
Six or more months after demolition on contiguous block
faces

(1)
All
Crime
0.489**
(0.234)
0.166
(0.230)
0.416**
(0.200)
0.114
(0.215)
0.116
(0.220)
0.188
(0.224)
0.302**
(0.147)
-0.071
(0.162)
-0.051
(0.105)
0.130
(0.101)
0.077
(0.102)
0.043
(0.102)
-0.045
(0.101)

(2)
Violent
Crime
0.672
(0.513)
0.552*
(0.317)
0.398
(0.300)
-0.173
(0.415)
0.401
(0.338)
0.611*
(0.363)
0.345
(0.226)
-0.239
(0.236)
-0.398**
(0.171)
0.084
(0.142)
0.005
(0.176)
-0.052
(0.147)
-0.112
(0.153)

(3)
Property
Crime
0.559
(0.393)
-0.137
(0.427)
0.683***
(0.264)
0.256
(0.348)
0.363
(0.328)
0.053
(0.291)
0.582***
(0.203)
-0.084
(0.265)
0.038
(0.170)
0.323*
(0.194)
0.156
(0.196)
0.029
(0.171)
-0.047
(0.174)

0.007
-0.144
0.162
(0.078)
(0.126)
(0.127)
Observations
30,400
19,072
24,560
Number of block faces
1,900
1,192
1,535
Robust standard errors in parentheses, clustered at the block face level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of all block faces in Saginaw, MI from
2008-2009. Crime offenses refer to the number of incidents on each block face in each month.

49

BIBLIOGRAPHY

50

BIBLIOGRAPHY

Acevedo, Luis G. 2009. “An Empirical Analysis of the impact of home foreclosure on the crime
rate: evidence in Atlanta, GA.” Bryant Economic Research Paper 2 (9).
Arnio, Ashley N., Eric P. Baumer, and Kevin T. Wolff. 2010. “The Contemporary Foreclosure
Crisis and U.S. Crime Rates." Social Science Research 41 (6):1598-1614.
Ashenfelter, Orley and David Card. 1985. “Using the Longitudinal Structure of Earnings to
Estimate the Effect of Training Programs." Review of Economics and Statistics 67 (4).
Becker, Gary S. 1968a. “Crime and Punishment: an Economic Approach.” Journal of Political
Economy 76 (2):169-217.
Becker, Gary S. 1968b. “Nobel Lecture: the Economic Way of Looking at Behavior.” The
Journal of Political Economy 101 (3):385-409.
Becker, Gary S. 1964, 1994, 3rd ed. Human Capital: A Theoretical and Empirical Analysis, with
Special Reference to Education Chicago, University of Chicago Press.
Braga, Anthony A., David L. Weisburd, Elin J. Waring, Lorraine Green Mazerolle, Willima
Spelman, and Francis Gajewski. 1999. “Problem Oriented Policing in Violent Crime Places: A
Randomized Controlled Experiment.” Criminology 37 (3):541-79.
Braga, Anthony and Brenda J. Bond. 2008. “Policing Crime and Disorder Hot Spots: A
Randomized Controlled Trial.” Criminology 46 (3): 577-607.
Burns, Gus. 2010. "FBI stats designate Saginaw the most violent city per capita in America."
The Saginaw News. MLive.com.
<http://www.mlive.com/news/saginaw/index.ssf/2010/01/fbi_stats_designate_saginaw_th.html>
Clark, Lynn M., and Brent Teasdale. 2005. “Subprime Lending, Crime, and Neighborhood
Disorganization: Beyond Internal Dynamics.” University of Akron.
The Community Research Partners and Rebuild Ohio 2008. “$60 Million and Counting: The
Cost of Vacant and Abandoned Properties to Eight Ohio Cities.”
Corman, H. and Mocan, N. 2005. “Carrots, Sticks, and Broken Windows.” NBER Working
Papers 9062 National Bureau of Economic Research.
Cui, Lin. 2010. “Foreclosure,Vacancy and Crime.” Manuscript, Department of Economics,
University of Pittsburgh.
Cullen, Julie Berry, and Steven D. Levitt. 1999. “Crime, Urban Flight, and the Consequences for
Cities.” Review of Economics and Statistics 81 (2):159-169
51

Deller, Steven, Lindsay Amiel, and Melissa Deller. 2011. “Model Uncertainty in Ecological
Criminology: An Application of Bayesian Model Averaging with Rural Crime Data.”
International Journal of Criminology and Sociological Theory 4 (2):683-717.
Drukker, David M., Hua Peng, Ingmar Prucha, and Rafal Raciborski. 2011. “Creating and
Managing Spatial-Weighting Matrices using the spmat Command.”
<http://econweb.umd.edu/~prucha/Papers/WP_spmat_2011.pdf>
Eastern Pennsylvania Organizing Project and the Temple University Center for Public Policy.
2001. “Blight Free Philadelphia.”
Ehrlich, I. 1973. “Participation in illegitimate activities: a theoretical and empirical
investigation.” Journal of Political Economy 81 (3):521-565.
Ehrlich, I. 1975. “On the relation between education and crime.” In F.T. Juster (Ed.) Education,
Income, and Human Behavior. McGraw-Hill Book Co., New York. 313-338.
Ellen, Ingrid Gould; Johanna Lacoe; and Claudia Sharygin. 2011. “Do Foreclosures Cause
Crime?” National Institute of Justice Crime Mapping Conference, Miami, FL.
Engel, Justin. 2010. “Saginaw on Cost-Cutting Restructure Plan: ‘It Can’t be Bloodless’.”
Mlive.com <http://www.mlive.com/news/saginaw/index.ssf/2010/01/saginaw_on_costcutting_restru.html>
Fagan, Jeffrey and Garth Davies. 2000. “Street Stops and Broken Windows: Terry, Race and
Disorder in New York City.” Fordham Urban Law Journal Vol. 28:457.
Fleisher, Belton M. 1963. “The effect of unemployment on juvenile delinquency.” Journal of
Political Economy 71 (6):543-555.
Fleisher, Belton M. 1966a. “The effect of income on delinquency.” American Economic Review
56:118-137.
Fleisher, Belton M. 1966b. The economics of delinquency Quadrangle: Chicago.
Glaeser, Edward, and Joseph Gyourko. 2005. “Urban Decline and Durable Housing.” Journal of
Political Economy 113 (2):345-375.
Goodstein, Ryan, and Yan L. Lee. 2010. “Do foreclosures increase crime?” FDIC Center for
Financial Research Working Paper No. 2010-05.
Griswold, Nigel. 2006. "The Impacts of Tax-Foreclosed Properties and Land Bank Programs on
Residential Housing Values in Flint, Michigan." Michigan State University Master’s Thesis.

52

Harcourt, B. and Ludwig, J. 2006. “Broken Windows: New Evidence from New York City and a
Five-City Social Experiment.” The University of Chicago Law Review 73:271-318.
Harris, Paul E. Lish. 2011. “A Closer Look at Home Foreclosures and Crime: Examining the
Criminal Consequences of Home Foreclosures on Houston Neighborhoods.” The College of
Criminology and Criminal Justice of Florida State University Doctoral Thesis.
Hartley, Daniel A. 2010. “Blowing it up and Knocking it Down: The Effect of Demolishing
High Concentration Public Housing on Crime.” Federal Reserve Bank of Cleveland Working
Paper 10-22.
Immergluck, Dan, and Geoff Smith. 2006. "The External Costs of Foreclosure: The Impact of
Single-Family Mortgage Foreclosures on Property Values." Housing Policy Debate Fannie Mae
Foundation. 17 (1):57-79.
Jacob, Brian A. 2004. “Public Housing, Housing Vouchers, and Student Achievement: Evidence
from Public Housing Demolitions in Chicago.” American Economic Review 94 (1):233-258.
Jones, Roderick W. and William Alex Pridemore. 2012. “The Foreclosure Crisis and Crime: Is
Housing-Mortgage Stress Associated with Violent and Property Crime in U.S. Metropolitan
Areas?” Social Science Quarterly 93 (3):671.
Kelling, George L and James Q. Wilson. 1982. “Broken Windows.” The Atlantic Online
http://www.theatlantic.com/magazine/archive/1982/03/broken-windows/304465/
Kirk, David S. and Derek S. Hyra. 2011. “Home Foreclosures and Community Crime:
Causal or Spurious Association?” Social Science Quarterly. 93 (3):648-670.
Madensen, Tamara D., Timothy C. Hart, and Terance D. Miethe. 2011. “The Impact of
Foreclosures on Neighborhood Crime in Nevada, 2006-09.” University of Las Vegas Center for
the Analysis of Crime Statistics.
Pandit, Caaminee. 2011. “The Impact of Foreclosure on Crime in American Cities”. Sociological
Insight 3:82-109.
Park, Robert E. and Ernest W. Burgess. 1925. The city Chicago, IL: University of Chicago Press.
Repetto, Thomas. 1974. “Crime prevention and the displacement phenomenon.” Crime and
Delinquency 22 (2):166-177.
Schilling, Joseph, and Jonathan Logan. 2008. “Greening the Rust Belt: A Green Infrastructure
Model for Right Sizing America’s Shrinking Cities.” Journal of the American Planning
Association. 74 (4):451-466.
Shaw, C.R. and H.D. McCay. 1932. “Social factors in juvenile delinquency.” In Report of the
Causes of Crime, Vol. 1. National Commission on Law Observance and Enforcement.
Washington, DC: Government Printing Office.

53

Shaw, C.R. and H.D. McCay. 1942, 1969 revised. “Juvenile Delinquency and Urban Areas: a
Study of Rates of Delinquency in Relation to Differential Characteristics of Local Communities
in American Cities.” Chicago, IL: University of Chicago Press.
Sousa, William H. and George L. Kelling. 2002. “Policing Does Matter.” The Manhattan
Institute City Journal Winter.
Spelman, William. 1993. “Abandoned Buildings: Magnets for Crime?” Journal of Criminal
Justice 21 (5):481-495.
Sperling’s Best Places. 2009. “Best Places to Live in Saginaw, Michigan.” Web.
<http://www.bestplaces.net/city/Saginaw-Michigan.aspx>
Stucky, Thomas D., John R. Ottensmann, and Seth B. Payton. 2012. “The Effect of Foreclosures
on Crime in Indianapolis, 2003-2008”. Social Science Quarterly 93 (3):602-624.
Teasdale, Brent, Lynn M Clark, and Joshua C. Hinkle. 2012. “Subprime lending, crime, and
neighborhood disorganization: Beyond internal dynamics.” American Journal of Criminal
Justice 37 (2):163-178.
United States Department of Housing and Urban Development. 2011. “Neighborhood
Stabilization Program Snapshot Reports.”
http://hudnsphelp.info/index.cfm?do=viewSnapshotHome#pagetop
Winkelmann, Rainer. 2008. Econometric Analysis of Count Data 5th ed. Berlin; Springer.
Winthrop, Bob and Rebecca Herr. 2009. “Determining the Co$t of Vacancies in Baltimore.”
Government Finance Review 38 (5).
Wooldridge, J. 1999. “Distribution-Free Estimation of Some Nonlinear Panel Data Models.”
Journal of Econometrics 90 (1):77-97.
Wooldridge, J. 2002. Econometric Analysis of Cross Section and Panel Data 2nd ed. Cambridge,
MA: Massachusetts Institute of Technology Press.

54

CHAPTER 2: LATENT MOTIVATIONS FOR THE IMPLEMENTATION OF
MUNICIPAL RESIDENCY REQUIREMENT LAWS
Abstract
Public employee residency requirement laws have been a common feature of the
th

American urban policy toolkit since the early 20 century. Many cities still require city
residency for their employees, and others are attempting to reinstate such requirements.
Justifications for these laws range from public safety to public coffer arguments. However, little
is known about the underlying factors that motivate cities to implement residency laws. In this
paper, I estimate a parametric duration model with a proportional hazard function to examine
these latent motivations. In order to perform this analysis, I construct a unique data set on
municipal residency laws for the largest 100 cities in the United States from 1970 through 2007.
I also create a set of fiscal stress indicators for each of these cities over time. Results indicate
that cities implement residency laws for economic and racial equity reasons and remove them
when the city improves in comparison to the metropolitan area as a whole. Fiscal stress and
service provision quality do not appear to be major contributing factors.

55

Introduction

Cities throughout the United States cite different reasons for employing residency
requirement laws, or laws that require public employees to reside in the city or county of their
employment. Some cities argue that public safety will be of higher quality if police and firemen
live in the community in which they patrol so that they are more sensitive to their constituents
and are quicker to respond to emergencies. Some argue that residency requirements ensure that
the tax dollars spent by cities to pay for public employee salaries are recirculated in that city
rather than exported to a nearby suburb. Still others argue that residency requirements increase
minority representation and make hiring more representative of the demographics of a city.
In this paper I explore the underlying factors that motivate cities to implement and
remove residency requirement laws by estimating a hazard model that examines the factors that
lead up to the implementation of a residency law. I construct a unique data set on municipal
residency laws for 73 of the largest 100 cities from 1970 through 2006. To the best of my
knowledge, this is the first panel data set of these laws.
To account for city and census tract boundary changes over time, I use Brown
University’s Longitudinal Tract Database which standardizes census data from 1970 through
2000 to 2010 census tract boundaries.

40

I then take the census tracts that are within the city

boundaries in 2010 and compare these over time. Finally, I merge these decennial city
demographics with annual fiscal data from the Census of Governments and create fiscal stress
indicators, as well as annual crime data from the Federal Bureau of Investigation’s Uniform
Crime Reports. Although many justifications are suggested for the implementation of these
laws, I hypothesize that fiscal stress is the true motivation.
40

Logan, Xu, and Stults, 2012
56

Only one previous paper examines this question directly. Eisinger (1983) creates a nonrandom cross section of 74 cities in 1980 to determine the characteristics of cities that have
residency requirements versus those that do not. He finds that the adoption of residency laws is
associated with unemployment, fiscal hardship, population loss, Frost Belt location, and mayorcouncil government. He also finds that, on average, cities which rescinded residency laws
generally displayed stable or high population growth and low unemployment. Eisinger argues
that the true motivation for all residency requirement laws is to improve the overall population
and demographics of a city and that they are not actually a tool of public personnel management
but an attempt to combat local unemployment and encourage the spending of city salaries in the
local economy.
In this paper, I provide the first panel estimates of why cities choose to implement
residency requirement laws. Results indicate that residency requirement laws are more likely to
be implemented by cities with higher rates of unemployment, lower income per capita, and a
larger number of black people living in the city versus the metropolitan area as a whole.
Residency laws are more likely to be removed by cities with a lower income per capita but a
higher income per capita in comparison to the metropolitan area as a whole. This implies that
cities implement residency laws for economic and racial equity reasons and remove them when
the city improves in comparison to the metropolitan area as a whole. Fiscal stress and service
provision quality do not appear to be major contributing factors.

57

Background
Types of Residency Requirements
There are several types of municipal residency requirement laws. The first type, and
the one of primary focus in this paper, is a law that requires municipal employees to live within
the city boundary itself. This requirement can come in the form of a durational residency
requirement in which prospective employees must prove that they have lived in the community
for a prescribed period of time in order to be eligible for employment. The requirement can also
be a continuing residency requirement which demands only that governmental employees live in
the community during the term of their employment. The former has been found to impinge on
the fundamental right to travel, forcing a governmental unit to show a compelling interest to
justify a durational requirement.

41

Some residency requirements have been negotiated by unions

for specific departments. These union negotiated requirements are not recorded in municipal
archives and are therefore not included in my data set. However, this lack of data will only bias
my results towards zero because if union negotiated contracts are in effect as residency
requirements are implemented and removed, the results of the change in policy will be reduced.
Residency laws can also require that municipal employees live within the county
boundary, within the state boundary, or within a certain radius of a municipality. Some states
have used private local hire laws in which preference is given to city residents, but these have
been deemed unconstitutional in at least one case.

42

Cities have also used incentives and disincentives to encourage municipal employees to
live within city boundaries. New York City’s charter requires nonresident city employees to

41
42

Myers, 1986
“Local hire laws: Alaska’s futile attempts at preferential treatment”, 1987
58

make payments in the form of contractually agreed-upon salary deductions similar to an income
43

tax.

Detroit, Michigan has recently instituted a residency incentive program called Project 14

which is meant to encourage police officers who live outside of Detroit to move into the city.
Under Project 14, 200 tax-foreclosed houses are being offered for as little as $1,000 down, plus
up to $150,000 in grants for renovations along with a monthly housing payment of $500 to
44

$1,000.

Residency requirements may be targeted at specific departments or they may be
comprehensive amongst all employees. When targeted at specific departments, the rationale is
usually that they are meant to improve the quality of the service provided by this department.
For instance, most targeted residency requirements are focused on public safety departments
with the belief being that if the public safety personnel live within the city not only will they be
able to respond to emergencies more quickly, but they will also know the neighborhoods better
and will therefore be able to provide better safety to the citizens.

45

Comprehensive residency requirements that demand that all public employees live in the
city or county are most often instituted for economic or demographic reasons. It is difficult to
argue that an accountant’s services will be improved by living in the city. More often, the
accountant is forced to live in the city in order to increase the city’s tax base and to improve the
demographics within the city.

43
44
45
46

46

Boies, 2005/2006
Nichols, 2011
Duncan, 2005
Duncan, 2005
59

Each of these types of residency requirements and incentive programs will have an
impact on the makeup of the city in which it is instituted. In this paper I focus mainly on
comprehensive city boundary residency requirements but also examine some aspects of other
types of residency requirement laws.
Residency requirement laws were common in the early 1900s. However, these laws were
gradually removed from most cities between 1920 and 1960 mainly based on the argument that
such laws were a barrier to hiring the most qualified candidates.

47

There was a brief revival of

the laws during the Great Depression, but by the late 1960s only Philadelphia, Buffalo,
Milwaukee, and a few other big cities retained their residency requirements for police officers.

48

They became fashionable again in the 1970s, and by 1980 nearly two thirds of all cities over
250,000 in population had residency requirements.

49

In 1976, the constitutionality of residency requirement laws was questioned in McCarthy
v. Philadelphia. The Supreme Court upheld the Philadelphia residency requirement, and many
cities soon followed with residency requirements of their own. Many cities that had dormant
residency requirement laws began to enforce them after this court decision. Some states decided
to ban residency requirements at the state level. California did so in 1974, for example,
Michigan in 1999, and Ohio in 2009. Still, many cities throughout this time period reinstituted
residency requirements and they remain a topic of debate to this day. In Michigan legislation
will be introduced in September to remove the ban on residency requirements so that Detroit can

47
48
49

Reed, 1941; International City Management Association, 1976
Fogelson, 1977
Eisinger, 1983
60

re-instate its requirement.

50

Despite the lack of evidence that these laws work, they are still a

topic of discussion and discontent in many cities throughout the United States.

Rationales for Residency Requirement Laws
Justifications for municipal residency requirement laws fall into four main categories.
The first is what is termed the “public coffer theory.” This theory posits that residency
requirement laws reduce local unemployment, benefit the local economy and increase the tax
base.

51

The theory is based on the notion that salaries paid to civil servants should recirculate

within the public and private economy of the city that pays those salaries.

52

Studies

investigating the economic effects of residency laws on municipal labor markets have been
contradictory. Hirsch and Rufalo (1986) find no effect of residency requirements on labor
supply or demand, but do find effects on compensation because they alter the bargaining
environment with unions. Gonzales, Mehay, and Duffy-Deno (1991) find that police
employment is approximately 10 percent higher in cities with residency requirements which they
suspect to be either due to the increase in the number of resident voters who are city employees
and who are more likely to support spending increases, or due to an increase in productivity
caused by the residency requirement. O’Brien (1997) finds that residency laws do not affect
compensation and employment for either police or firefighters. Mehay and Seiden (1986) find
that the gain in efficiency caused by residency requirements appears to have been lost due to the
public employees’ high demand for public services. They also argue that even if such laws do

50
51
52

Angel, 2011
Myers, 1986
Eisinger, 1983
61

reduce per unit cost, the main benefit appears to accrue to the local bureaucracy in the form of
greater output, expenditures, and agency size, and not to employees in the form of higher wages
or to city residents in the form of tax relief.
The second type of rationale for residency requirements relates to the efficiency and
quality of service provision. Many policy makers have argued that requiring municipal
employees to live within the city boundaries or some other radius will improve their ability to
provide services through increased knowledge of the community, 24 hour protection by public
53

54

safety personnel , the fostering of employee concerns about the affairs of their city , the
reduction of absenteeism and tardiness, and greater social symmetry between social servants and
55

their constituents . Smith (1980) provides empirical evidence of a positive relationship
between the efficiency of a police department and the percentage of officers living in the
community they serve. Gonzales, Mehay, and Duffy-Deno (1991) find that the actual crime rate
tends to be below the predicted rate in cities that have residency requirements. However, many
policy makers argue that residency requirements restrict the pool of the workforce which
decreases the quality of applicants.

56

Residency requirements have also been opposed based on

the freedom of employees to live where they please. This is often at the center of court cases
against the laws.
Third, some cities have justified residency requirement laws based on minority
representation issues. In Milwaukee, for example, the firefighters union is arguing for the
preservation of their residency requirements partly in order to encourage the hiring of more black
53
54
55
56

Smith, 1980
City of Cincinnati, 1977
Ector Versus City of Torrance, 1973
Eisinger, 1983
62

and female firefighters.

57

Residency requirements are argued here to make municipal hiring

representative of the demographics of a city. Stein (1986) studies the representativeness of local
governments and finds that the presence of a residency requirement can offset the negative
impact on representative hiring caused by civil service commissions, namely a 3% change.
Finally, the main argument which encompasses all of the previous arguments is that
residency requirements improve the overall population and demographics of a city. Eisinger
(1983) argues that this is the true motivation for all residency requirement laws and that they are
not actually a tool of public personnel management but an attempt to combat local
unemployment and encourage the spending of city salaries in the local economy. He finds that
the adoption of residency laws is associated with unemployment, fiscal hardship, population loss,
Frost Belt location, and mayor-council government. He also finds that, on average, cities which
rescinded residency laws generally displayed stable or high population growth and low
unemployment. The population size and demographics of a city may also benefit as a secondary
effect from the intermediary effects discussed in the previous arguments. For instance, the
intermediary effects from the public coffer theory (in which the city has a larger tax base which
may imply an increased quantity and quality of services provided) may lead to a secondary effect
of an increased number of people moving to the city. The efficiency argument, similarly, implies
that there will be an intermediary effect of an increased quality of service provision such as
decreased crime rates, which may also lead to more people moving to the city. Finally, the
minority representation argument may lead to increased equality and decreased social unrest
which may also make the city a more attractive place to live.

57

Laasby, 2011
63

Data
To the best of my knowledge, no comprehensive historical data on residency
requirements exists. To collect this data, I first searched historical city databases for the 100
largest cities in the United States to find city codes and charters that identify the years in which
comprehensive residency requirements were implemented and removed as well as the details of
each of these laws. However, most cities do not have electronic historical data available so I
then combined this search with a search of law articles and historical newspapers. This also only
produced data on the largest cities. To complete the data set, I then requested information from
the missing cities individually through a series of phone calls to the city archivists. I attempted
to collect data on residency requirements dating back to 1900, but many cities did not begin
recording this data until the 1970’s. Some cities were able to provide documentation to back up
their facts; others simply provided the information based on memory. Out of the 100 cities from
which I attempted to collect data, I was able to collect the history of residency requirement laws
from 1970 to 2012 for 73 cities.
Table 2.1 lists the number of cities in my data set that had a residency requirement law on
the decennial years. Figure 2.1 shows the number of residency laws that were implemented and
removed each year. The negative bars represent removals of the law and the positive bars
represent implementations of the law. Overall, 42 cities implemented various types of residency
laws and 23 cities removed such laws. Table 2.2 summarizes the residency requirement data.
To account for city and census tract boundary changes over time, I use Brown
University’s Longitudinal Tract Database (LTDB) which standardizes census data from 1970
through 2000 to 2010 census tract boundaries. The LTDB provides estimates using 2010
boundaries for a standard set of variables from 1970 through 2000 for both full count variables

64

such as population, age, and race, as well as sample count variables based on the one-in-six
samples from the decennial census and the American Community survey such as income and
employment. I take the census tracts that are within the city boundaries in 2010 and compare
these over time.
I only observe the demographic variables in my data set for the denial years. There is no
perfect way to deal with the missing data for the years in between the census years. Interpolation
and change variables both introduce endogeneity since these methods use future values to predict
current values. I therefore choose to estimate the model using decennial demographic
characteristics repeated each year until the next decennial year.
I then merge the residency law data and the decennial city demographics with annual
fiscal data from the Census of Governments from 1970 through 2006. These fiscal data allow
me to create fiscal stress indicators that measure how stressed the city is in each year as shown in
Table 2.2. I also use fiscal stress measures from the decennial census such as per capita income
and the rate of unemployment.
I define fiscal health using the definition proposed in Groves and Valente (2003). I then
define fiscal stress as a deficiency in any of these measures. Fiscal health is defined as a
government’s ability to maintain solvency in four measures: cash, budgetary, long run, and
service level. Cash solvency refers to a local government’s liquidity and effective cash
management, as well as its ability to pay current liabilities. Budgetary solvency refers to the
ability of the government to generate sufficient revenues to fund its current or desired service
levels. Long run solvency refers to the impact of existing long term obligations on future
resources. Service level solvency refers to the ability of the government to provide and sustain a

65

service level that citizens require and desire. Fiscal stress is therefore defined as a deficiency in
any of these measures.
Because no cash measure is included in the Census of Governments data, I can only
measure cash solvency using a modified debt service ratio. The typical debt service ratio
includes total interest on debt and principal payments. However, I only have information on
total interest on debt from the Census of Governments data. I therefore measure the debt service
as the interest on debt compared to total revenues.
I measure budgetary solvency using an operating deficit ratio which measures the
operating surplus or deficit in comparison to total revenues. A deficit in one year does not
necessarily indicate fiscal stress. A government may have unusually large expenditures in the
current year but have planned for such events by saving money in previous years. Credit rating
agencies are generally concerned when there is a budget deficit in two consecutive years.
Because of this, I include an indicator for a budget deficit in two consecutive years or an
abnormally large deficit in one year.
I also measure budgetary solvency using an intergovernmental revenues ratio, which
measures how dependent the city is on intergovernmental transfers, and a property tax ratio,
which measures the tax base of the city. Property tax is the main revenue source for most cities
and therefore is a good indication of the health of a city’s budget.
I measure service level solvency using crime rates from the Federal Bureau of
Investigation’s Uniform Crime Reports. The only crimes that are available for all years are what
are known as the crime index offenses: murder, forcible rape, robbery, aggravated assault,
burglary, larceny-theft, and motor vehicle theft. I sum all of these crimes into one overall crime
level for each city/year

66

Unfortunately, the Census of Governments data does not provide any measures of long
term solvency. However, these should be captured in the demographic variables that are
included as controls.
Finally, I merge these data with crime data from the FBI’s Uniform Crime Reports on
offenses known and clearances by arrest for each year from 1970 through 2006. The only crimes
that are available for all years are what are known as the crime index offenses: murder, forcible
rape, robbery, aggravated assault, burglary, larceny-theft, and motor vehicle theft. I sum all of
these crimes into one overall crime level for each city/year, and then also examine murder and
vehicle theft separately. I examine murder on its own because of its seriousness, and I examine
vehicle theft on its own because it is thought to be very closely correlated with the actual number
of offenses because victims are likely to report vehicle thefts for insurance purposes. Murders
are also likely to be reported.
There is a large literature pointing to inaccuracies in this FBI data. Namely, many
city/years are endogenously missing because the data is self-reported. Some cities choose not to
report their crime rates in years where they had a particularly high level of crime. Maltz (1999)
advocates for the use of longitudinal rather than cross-section imputation techniques to correct
for this missing data rather than the often used technique of using contemporaneous data from
similar jurisdictions to impute the missing values. I, therefor, use longitudinal imputation
techniques to interpolate the missing values in my data set. Descriptive statistics for each of the
variables used in the regressions can be found in Table 2.4.

67

Econometric Model
To estimate the impact of different covariates on the decision to implement residency
requirements, I estimate a parametric survival model of the time elapsed until implementation of
a residency requirement. I then estimate a second model of the time elapsed from
implementation until removal of a residency requirement in order to examine the motivations for
removal of the laws.
Following Wooldridge, 2010, I model the probability of implementing or removing a
residency requirement as a function of fiscal stress within the city, population, state bans on
residency requirements, income differentials between the city and the MSA, race differentials
between the city and the suburbs. To measure fiscal stress, I include per capita income, the rate
of unemployment, the operating deficit ratio, the intergovernmental transfer ratio, the debt ratio,
and the property tax revenue ratio as defined in the previous section.
I model the probability that a city will implement a residency requirement law using a
proportional hazard model of the following form

λ (t ; x ) = κ ( x ) λ o (t )
I then specify the baseline hazard,

(1)

λo (t ) , using a Weibull distribution because it allows

for the duration to have positive or negative dependence as long as it is monotonically increasing
or decreasing. The hazard then takes the form

λ (t; x) = exp(xβ )αt α −1

68

(2)

To account for time varying covariates, I let x(t ) denote the vector of regressors at time t.
For t ≥ 0 , let X(t ) , t ≥ 0 , denote the covariate path up through time t. Following Lancaster
(1990, Chapter 2), I define the conditional hazard function at time t by

λ[t ; X(t )] = lim
h↓0

P[t ≤ T < t + h | T ≥ t , X(t − h)]
h

(3)

This hazard function allows me to approximate the probability of implementing a
residency requirement law within a short interval, conditional on having not implemented one up
to the starting time of the interval.
I then let T denote the time until implementation of a residency requirement law.
Because I do not observe the entire duration for all cities in my data set (i.e., some cities already
had or still had residency requirements in 1970 and 2010, the beginning and end of my data),
some data points are censored. To account for censoring, I let

cm

be a binary censoring

indicator equal to unity if the duration is censored in interval m, and zero otherwise. Because my
sample includes all 100 largest US cities regardless of their initial or end state, this censoring is
not systematic and therefore does not cause selection bias. This duration is therefore independent
of censoring conditional on the covariates, an assumption of the model.
From this assumption, I can compute the probability that a city will implement (or
remove) a residency requirement law as follows:

69

P( y m = 1 | y m −1 = 0, x, cm = 0) =
P(am −1 ≤ T < am | T ≥ am −1 , x)
 am

= 1 − exp − ∫ λ ( s; x, θ )ds  ≡ 1 − α m (x,θ )
 am −1




(4)

for m=1,2,…,M, where

 am

α m (x,θ ) = exp − ∫ λ ( s; x, θ )ds 
 am −1



where the data is split into M + 1 intervals,

[0, a1 ),[a1 , a 2 ),...,[a M −1 , a M ), [a M , ∞) , and ym

is a binary indicator equal to

unity if the duration ends in the mth interval and zero otherwise.
From this, I construct a log likelihood for a single spell duration model with time-varying
covariates with censoring is as follows:
mi −1

∑ log[α
h =1

where

58

di

h

58

( xim , θ )] + d i log[1 − α mi ( xim , θ )]

= 1 if duration i is uncensored.

Wooldridge, 2002 p 711
70

(5)

I choose not to include unobserved heterogeneity in my model because it is difficult to
relax the strict exogeneity assumption with time-varying covariates and unobserved
heterogeneity included. Also, with single-spell data, I cannot allow general correlation between
the unobserved heterogeneity and the covariates.

59

Results
Residency Law Implementation
Table 2.5 reports the results of what variables affect the implementation of residency
requirement laws. Results show that, conditional on the other covariates, cities with higher rates
of unemployment, lower income per capita, and a larger number of black people living in the city
versus the metropolitan area as a whole are more likely to implement residency requirement
laws. Specifically, a city with a 1 percentage point higher unemployment rate, all else held
equal, is 21.09% more likely to implement a residency requirement law. A city with 1% higher
per capita income is 1.40% less likely to implement a residency requirement law. A city with a 1
percentage point larger number of black people living in the city versus the metropolitan area as
a whole is 1.66% more likely to implement a residency requirement law. This follows the
reasoning that some cities implement residency laws for reasons relating to racial equity in
hiring; if more black people live in the city, the city may implement a residency law to ensure
that hiring is representative of the racial distribution within the city.
Unemployment and income per capita are indicators of fiscal health within the city.
However, the variables that more specifically relate to city finances are not statistically
significant, such as the operating deficit ratio and the debt ratio. This may indicate that

59

Wooldridge, 2002 p 713
71

economic indicators for the city are more correlated with residency law implementation than are
the finances of the city government. This follows the theory that residency laws are
implemented as a tool for helping the overall economy rather than simply the financial
statements.

Residency Law Removal
Cities that remove residency laws are hypothesized to do so when their economy and
fiscal health are improving. I therefore also examine the variables that cause cities to remove
their residency laws. However, many cities in my data set removed their residency laws by force
when a state ban was implemented. Therefore, identification of the removal of residency laws
rests on the few cities who removed their laws by choice. Out of the 42 cities that had a
residency requirement during my time period, 13 removed them because of state imposed bans.
I removed these cities from this analysis because they are perfectly predicted by the ban.
Results, as presented in Table 2.6, show that, conditional on the other covariates, cities
with a 1% higher income per capita are 3.094% less likely to remove their residency law. In
addition, cities with a 1% higher income per capita in the city compared to the MSA as a whole
are 4.647% more likely to remove their residency law. This implies that cities do not necessarily
remove their residency laws when the demographics of the city improve, but when the
demographics of the city compared to the MSA improve.

Conclusion
Residency requirement laws are more likely to be implemented by cities that have higher
rates of unemployment, lower per capita income, and a larger proportion of black people living

72

in the city compared to the metropolitan area as a whole. These results imply that residency laws
are implemented to better the economic health of a city – the “public coffer theory” -- and to
make hiring representative of the racial mix of the city – the minority representation theory. It
does not appear that population levels, public service provision, or municipal fiscal health are
primary reasons for implementing residency requirement laws.
Residency laws are removed by cities with a lower income per capita but a higher income
per capita in comparison to the metropolitan area as a whole. This implies that cities remove
residency laws when the city improves in comparison to the metropolitan area as a whole.
Residency requirement laws have been a common feature in the American urban
th

landscape since the early 20 century. Many cities still enforce residency for their employees,
and some cities are in discussion to reinstate such laws. This paper indicates the underlying
reasons why cities choose to implement such laws and lay the framework for the analysis of
residency laws.

73

APPENDIX

74

APPENDIX
Tables and Figures
Figure 2.1: Changes to Residency Requirements, 1970 to 2006

1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006

-6

-4

-2

0

2

4

Changes to Residency Requirements

Table 2.1: Number of Cities with Residency Requirements from 1970 through 2010

Year
1970
1980
1990
2000
2010

Number of Cities with a Residency
Requirement Law
15
23
22
20
17

75

Table 2.2: Residency Requirement Law Data
Residency Requirement Law
Number of Cities Surveyed
Number of Responses
Number of Cities that had a Residency Law Between 1970 and
2006
Number of Cities that Removed a Residency Law Between 1970
and 2006
Number of Cities that Removed a Residency Law Between 1970
and 2006 because of a State Ban

Table 2.3: Fiscal Stress Ratios

42
23
13

60

Large Operating Deficit Ratio
Operating Deficit Ratio for Two
Consecutive Years
Intergovernmental Revenue Ratio
Debt Ratio
Property Tax Per Capita

60

100
73

= ((Expenditures - Revenues)/ Revenues)>.05
= (Expenditures - Revenues)/ Revenues>0
= Intergovernmental Revenues/ Revenues
= Debt Interest/ Revenues
= Property Tax Revenue/ Population

Each of these variables is measured in 2010 dollars
76

Table 2.4: Descriptive Statistics
Obs

Mean

Std. Dev.

Min

Max

Residency Requirement

2,701

0.29

0.45

0.00

1.00

Population

2,701

594,982

987,746

16,478

8,008,278

Crime

2,701

45,640

68,666

49

682,063

% Unemployed

2,701

6.5%

2.7%

2.1%

19.7%

Income Per Capita

2,701 $8,669.12 $6,369.27

$235.91

$32,163.32

Operating Deficit > 5%
Operating Deficit Two
Consecutive Years
Intergovernmental Revenue
Ratio

2,701

0.54

0.50

0.00

1.00

2,650

0.32

0.47

0.00

1.00

2,701

0.24

0.14

0.00

0.91

Debt Ratio

2,701

0.07

0.04

0.00

0.28

Property Tax Per Capita
# Black in City Divided by #
Black in MSA
Income Per Capita in City
Divided by Income Per
Capita in MSA
Number of People Above
Age 60
State Ban on Residency
Requirements

2,701

$397.54

$362.26

$1.28

$2,514.23

2,701

0.61

0.34

0.00

1.00

2,701

1.05

0.27

0.15

2.23

2,701

90,577

165,035

887

1,373,093

2,701

0.22

0.42

0.00

1.00

Variable

77

Table 2.5: Variables that Influence Residency Requirement Implementation using
Repeated Demographics
Residency Requirement
Law
Log of Population
-0.910
(1.125)
Log of Crime
0.055
(0.187)
% Unemployed
21.090*
(12.490)
Log of Income Per Capita
-1.391***
(0.300)
Operating Deficit Ratio > 5%
-0.268
(0.493)
Operating Deficit Two Consecutive Years
-0.886
(0.573)
Intergovernmental Revenue Ratio
1.158
(2.014)
Debt Ratio
5.230
(5.182)
Log of Property Tax Per Capita
0.471
(0.387)
# Black in City divide by # Black MSA
1.663*
(0.980)
Income per Capita in City divided by Income per Capita in MSA
0.596
(0.688)
Log of Number of People above Age 60
1.283
(1.075)
State Ban on Residency Requirements
-1.551
(1.364)
1,383
Observations
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of the 100 largest cities in the United
States from 1970-2007.

78

Table 2.6: Variables that Influence Residency Requirement Removal using Repeated
Demographics

Log of Population Repeated
Log of Crime
% Unemployed Repeated
Log of Income Per Capita Repeated
Operating Deficit Ratio > 5%
Operating Deficit Two Consecutive Years
Intergovernmental Revenue Ratio
Debt Ratio
Log of Property Tax Per Capita
# Black in City divide by # Black MSA Repeated
Income per Capita in City divided by Income per Capita in MSA
Repeated

Residency Requirement
Law
-1.183
(0.854)
0.285
(0.347)
-25.580
(15.970)
-3.094***
(0.416)
0.262
(0.175)
-0.074
(0.262)
1.645
(1.240)
4.149
(4.079)
0.040
(0.437)
-0.720
(1.522)

4.647**
(2.021)
Log of Number of People above Age 60 Repeated
0.858
(0.805)
Observations
792
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of the 100 largest cities in the United
States from 1970-2007.

79

BIBLIOGRAPHY

80

BIBLIOGRAPHY

Angel, Cecil. “Detroiters tell legislators how ban on residency laws is hurting city.” Detroit
Free Press (Aug. 18, 2011).
http://www.freep.com/article/20110819/NEWS01/108190413/Detroiters-tell-legislators-howban-residency-laws-hurting-city
Boies, Christopher G. “Fleming v. Giuliani.” New York Law School Law Review. (2005/2006)
50:4 pp. 1003-1014.
City of Cincinnati. (1977). Ordinance of 29 June 1977.
Duncan, Brian. “Using Municipal Residency Requirements to Disguise Public Policy.” Public
Finance Review. 33: 84. (2005)
“Ector v. City of Torrance.” (1973) 514 P 2nd 433.
Eisinger, Peter K. “Municipal Residency Requirements and the Local Economy.” Social Science
Quarterly. 64:1 (March 1983) p. 85
Fogelson, Robert. “Big City Police.” Cambridge, Ma. Harvard University Press. (1977)
Gonzalez, Rodolfo A., Stephen L. Mehay, Kevin Duffy-Deno. “Municipal Residency Laws:
Effects on Police Employment, Compensation, and Productivity.” Journal of Labor Research
(Fall 1991) 13:4
Groves, S. M., and M. G. Valente. (2003) “Evaluating financial condition: A handbook for local
government.” Washington, DC: International City/County Management Association.
Hirsch, Werner Z., Anthony M. Rufolo. “Economic Effects of Residency Laws on Municipal
Police.” Discussion paper #279. (January, 1983).
Hirsch, Werner Z., Anthony M. Rufolo. “Residence Laws and Unionization in Municipal Labor
Markets: The Case of Firefighters.” Journal of Labor Research (Winter, 1986) 7:1
International City Management Association. “Personnel Practices in the Municipal Police
Service: 1976.” Urban Data Service Report (Dec., 1976) pp. 2-14
Laasby, Gitte. “Firefighter group backs keeping residency rule.” Journal Sentinel Online.
(2011) http://www.jsonline.com/news/milwaukee/119739199.html
“Local hire laws: Alaska’s futile attempts at preferential treatment.” Alaska Law Review (Dec.,
1987) pp. 359-386.

81

Logan, John R., Zengwang Xu, and Brian Stults. 2012. “Interpolating US Decennial Census
Tract Data from as Early as 1970 to 2010: A Longitudinal Tract Database” Professional
Geographer, forthcoming.
Maltz, M. (1999). Bridging gaps in police crime data. Washington, DC: U.S. Bureau of Justice
Statistics.
Mehay, Stephen L., Kenneth P. Seiden. “Municipal Residency Laws and Local Public Budgets.”
Public Choice (1986) 48:1 pp.27-35
Myers, Ross S. “The constitutionality of continuing residency requirements for local
government employees: a second look.” California Western Law Review (Fall 1986) 23 p 24-41
Nichols, Darren A. “Bank of America donating houses to entice Detroit cops to live in city.”
Detroit News (2011) http://detnews.com/article/20110323/METRO01/103230415/Bank-ofAmerica-donating-houses-to-entice-Detroit-cops-to-live-in-city#ixzz1Jz84pw1v
O’Brien, Kevin M. “Do Municipal Residency Laws Affect Labour Market Outcomes?” Urban
Studies (1997) 34:11 pp. 1759-1769
Reed, Thomas. “Municipal Management.” New York: McGraw Hill. (1941)
Smith, Dennis C. “Police Attitudes and Performance: The Impact of Residency.” Urban Affairs
Review (1980) 15:317
Stein, Lana. “Representative Local Government: Minorities in the Municipal Work Force.” The
Journal of Politics (Aug., 1986) 48:3 pp. 694-713
Wooldridge, Jeffrey. 2002. Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA: MIT Press.

82

CHAPTER 3: MUNICIPAL RESIDENCY REQUIREMENT LAWS AND THEIR
IMPACT ON CITIES
Abstract
Over the past century, many U.S. cities have experimented with residency laws that
require municipal employees to reside in the city or county of their employment. Despite the
prevalence and extensive history of these laws, little is known about their impact on city
outcomes. In this paper, I provide the first estimates of the impact of municipal residency
requirements on the quality of service provision, municipal fiscal health, and the size and
composition of cities. My identification strategy exploits the timing of municipal law changes in
static and dynamic models. In order to perform the analysis, I construct an original data set on
municipal residency laws for 73 of the largest 100 cities from 1970 through 2006. I then
combine these data with information on city demographics from the Decennial Census, fiscal
data from the Census of Governments, and crime data from the Federal Bureau of Investigation.
Results provide little evidence that residency laws have an effect on fiscal health or city
demographics when they are implemented, but they are correlated with a temporary increase then
long term decrease in crime. The removal of a residency law is not associated with any
significant change in fiscal health, city demographics, or crime.

83

Introduction

Over the past century, many U.S. cities have experimented with residency requirement
laws that require municipal employees to reside in the city or county of their employment. New
York City, for example, implemented such a law in 1937, removed it in 1962, and then reimplemented in 1986. An event in 2011 in which a police officer made a racially charged
remark on the internet has rekindled debate over the law with proponents arguing that forcing
police officers to live in the city would increase their awareness and sensitivity toward their
constituents.

61

Detroit, Michigan, implemented a residency requirement in 1913 and was forced

to remove it in 1999 when the state banned such laws. Policy makers in Detroit are calling for
the removal of the ban so that the city can once again enforce the law.

62

The governor of

Wisconsin recently implemented a state ban on residency requirements has incited criticism from
large cities and small villages throughout the state.

63

Despite the prevalence and extensive history of these requirements, little is known about
their impact on municipal service provision, fiscal health, or city demographics. In this paper I
provide the first estimates of the impact of comprehensive municipal residency requirements on
city outcomes. My identification strategy exploits the timing of municipal law changes. I also
examine the static and dynamic effects of the laws. In order to perform the analysis, I construct
an original data set on municipal residency laws for 73 of the largest 100 cities from 1970
through 2006. To the best of my knowledge, this is the first panel data set of these laws that
includes more than two years.
61
62
63

New York News, 2011
Angel, 2011
Stein and Walker, 2013
84

To account for city and census tract boundary changes over time, I use Brown
University’s Longitudinal Tract Database which standardizes census data from 1970 through
2000 to 2010 census tract boundaries. I then take the census tracts that are within the city
boundaries in 2010 and compare these back over time. Finally, I merge these decennial city
demographics with annual fiscal data from the Census of Governments and annual crime data
from the Federal Bureau of Investigation’s Uniform Crime Reports.
Previous research has shown, using a difference in difference of two years, that a
residency requirement can increase a city’s middle class population but only if it is accompanied
64

by a municipal wage premium.

This suggests that in cities where the government is not able to

offer a wage premium over its more wealthy suburbs like in Detroit, Michigan, a residency
requirement will not have its desired demographic effect. In addition, it is likely that when
implemented, a residency requirement will encourage the most qualified employees (the ones
who can easily find a different job) to switch jobs rather than to move. The less qualified
employees will be more likely to move to keep their current job because their outside prospects
are fewer.
However, even if these requirements do encourage only the less qualified employees to
move into the city, these employees may improve in quality because of their increased
connection with the community and quicker response times for emergencies. They also may
have a greater incentive to improve the quality of their work since it will improve the
neighborhood in which they and their families live. In addition, these new residents will increase

64

Duncan, 2005
85

the property tax base for the city at least in the short run, even if expenditures also increase due
to their new incentive to vote for spending increases.

65

My results show that the implementation of a residency law may be correlated with a
slight increase in crime two years after the law is implemented, and then a reduction in crime six
or more years after implementation. The removal of a law is correlated with an increase in
crime, but these results disappear once a time trend is included, indicating that this is likely
caused by city trends rather than by the law itself. Residency laws may be correlated with a
reduction in intergovernmental revenues four and six or more years after it is implemented. This
may imply a reduction in fiscal stress but it is unclear using that indicator alone. Finally,
residency laws and their removal do not appear to be correlated with any changes in
demographics or population.

Background
Types of Residency Requirements
There are several types of municipal residency requirement laws. The first type, and
the one of primary focus in this paper, is a law that requires municipal employees to live within
the city boundary itself. This requirement can come in the form of a durational residency
requirement in which prospective employees must prove that they have lived in the community
for a prescribed period of time in order to be eligible for employment. The requirement can also
be a continuing residency requirement which demands only that governmental employees live in
the community during the term of their employment. The former has been found to impinge on
the fundamental right to travel, forcing a governmental unit to show a compelling interest to

65

Gonzales, Mehay, and Duffy-Deno, 1991
86

justify a durational requirement.

66

Some residency requirements have been negotiated by unions

for specific departments. These union negotiated requirements are not recorded in municipal
archives and are therefore not included in my data set. However, this lack of data will only bias
my results towards zero because if union negotiated contract are in effect as residency
requirements are implemented and removed, the results of the change in policy will be reduced.
Some of my city-year observations that I assign to the control group are actually in the treated
group because of the union negotiated contracts. This makes the difference in coutcomes for the
treatment and control groups less large because some of the control group is actually being
treated.
Residency laws can also require that municipal employees live within the county
boundary, within the state boundary, or within a certain radius of a municipality. Some states
have used private local hire laws in which preference is given to city residents, but these have
been deemed unconstitutional in at least one case.

67

Cities have also used incentives and disincentives to encourage municipal employees to
live within city boundaries. New York City’s charter requires nonresident city employees to
make payments in the form of contractually agreed-upon salary deductions similar to an income
68

tax.

Detroit, Michigan has recently instituted a residency incentive program called Project 14

which is meant to encourage police officers who live outside of Detroit to move into the city.
Under Project 14, 200 tax-foreclosed houses are being offered for as little as $1,000 down, plus

66
67
68

Myers, 1986
“Local hire laws: Alaska’s futile attempts at preferential treatment”, 1987
Boies, 2005/2006
87

up to $150,000 in grants for renovations along with a monthly housing payment of $500 to
$1,000.

69

Residency requirements may be targeted at specific departments or they may be
comprehensive amongst all employees. When targeted at specific departments, the rationale is
usually that they are meant to improve the quality of the service provided by this department.
For instance, most targeted residency requirements are focused on public safety departments,
with the belief that if the public safety personnel live within the city, they be able to respond to
emergencies more quickly and they will know the neighborhoods better and will therefore be
able to provide better safety to the citizens.

70

Comprehensive residency requirements that demand that all public employees live in the
city or county are most often instituted for economic or demographic reasons. It is difficult to
argue that an accountant’s services will be improved by living in the city. More often, the
accountant is forced to live in the city in order to increase the city tax base and improve the
demographics in the city.

71

Each of these types of residency requirements and incentive programs will have an
impact on the makeup of the city in which it is instituted. In this paper I focus on comprehensive
city boundary residency requirements.
Residency requirement laws were common in the early 1900s. However, these laws were
gradually removed from most cities between 1920 and 1960 mainly based on the argument that

69
70
71

Nichols, 2011
Duncan, 2005
Duncan, 2005
88

such laws were a barrier to hiring the most qualified candidates.

72

There was a brief revival of

the laws during the Great Depression, but by the late 1960s only Philadelphia, Buffalo,
Milwaukee, and a few other big cities retained their residency requirements for police officers.

73

They became fashionable again in the 1970s and by 1980 nearly two thirds of all cities with
populations over 250,000 had residency requirements.

74

In 1976, the constitutionality of residency requirement laws was questioned in McCarthy
v. Philadelphia. The Supreme Court upheld the Philadelphia residency requirement, and many
cities soon followed with residency requirements of their own. Many cities that had dormant
residency requirement laws began to enforce them after this court decision. Some states chose to
ban residency requirements at the state level. California did so in 1974, Michigan in 1999, and
Ohio later in 2009. Still, many cities throughout this time period reinstituted residency
requirements and they remain a topic of debate to this day. In Michigan, legislation has been
introduced to remove the ban on residency requirements so that Detroit can re-instate its city
residency requirement.

75

Despite the lack of evidence that these laws work, they are still a topic

of discussion and discontent in many cities throughout the United States.

Pros and Cons of Residency Requirement Laws
Advocates of residency requirements claim four main benefits. The first relates to what
has been labeled the “public coffer theory.” This theory posits that residency requirement laws

72
73
74
75

Reed, 1941; International City Management Association, 1976
Fogelson, 1977
Eisinger, 1983
Angel, 2011
89

reduce local unemployment, benefit the local economy and increase the tax base.

76

The theory

is based on the notion that salaries paid to civil servants should recirculate within the public and
private economy of the city that pays those salaries.

77

Opponents argue that residency requirements do not actually benefit the public coffer, but
instead encourage an increase in expenditures and government size due to the employees now
living and voting in the municipality. This residency, they argue, increases the demand for
public services and payrolls, and thus expenditures.
Studies investigating the economic effects of residency laws on municipal labor markets
have been contradictory. Hirsch and Rufalo (1986) find no effect of residency requirements on
labor supply or demand, but do find effects on compensation because they alter the bargaining
environment with unions. Gonzales, Mehay, and Duffy-Deno (1991) find that police
employment is approximately 10 percent higher in cities with residency requirements which they
suspect to be either due to the increase in the number of resident voters who are city employees
and who are more likely to support spending increases, or due to a decrease in productivity
caused by the residency requirement. Mehay and Seiden (1986) find that the gain in efficiency
caused by residency requirements appears to have been lost due to the public employees’ high
demand for public services. They also argue that even if such laws do reduce per unit cost, the
main benefit appears to accrue to the local bureaucracy in the form of greater output,
expenditures, and agency size, and not to employees in the form of higher wages or to city
residents in the form of tax relief. O’Brien (1997) finds that residency laws do not affect
compensation and employment for either police or firefighters. I examine the public coffer

76
77

Myers, 1986
Eisinger, 1983
90

theory by estimating the effect of residency requirements on municipal fiscal health ratios,
population, income per capita, and unemployment.
The second type of rationale for residency requirements relates to the efficiency and
quality of service provision. Some policy makers argue that requiring municipal employees to
live within the city boundaries or some other radius will improve their ability to provide services
78

through increased knowledge of the community, 24 hour protection by public safety personnel ,
79

the fostering of employee concerns about the affairs of their city , the reduction of absenteeism
80

and tardiness, and greater social symmetry between social servants and their constituents .
Opponents argue, however, that requiring residency will restrict the pool of the workforce
which decreases the quality of applicants.

81

The quality of public services may decline because

the best employees will have better outside options for employment and may choose to find a
new job rather than relocate into the city.
Two papers have examined this relationship. Smith (1980) finds a positive relationship
between the efficiency of a police department and the percentage of officers living in the
community they serve, though this is likely correlation rather than causation. Gonzales, Mehay,
and Duffy-Deno (1991) find that the actual crime rate tends to be below the predicted rate in
cities that have residency requirements. I examine the service provision theory by estimating the
effect of residency requirements on crime.
The third rationale for residency requirement laws is based on minority representation
issues. In Milwaukee, for example, the firefighters union is arguing for the preservation of their
78
79
80
81

Smith, 1980
City of Cincinnati, 1977
Ector Versus City of Torrance, 1973
Eisinger, 1983
91

residency requirements partly in order to encourage the hiring of more black and female
firefighters.

82

Residency requirements are argued to make municipal hiring representative of the

demographics of a city. Stein (1986) studies the representativeness of local governments and
finds that the presence of a residency requirement can offset the negative impact on
representative hiring caused by civil service commissions, namely a 3% change. However,
residency requirements may actually be discriminatory if the majority of minorities lives outside
of the city. The National Association for the Advancement of Colored People (NAACP) has
been opposed to residency requirements in cities where this is the case, but a proponent of them
where minorities live within the cities rather than outside of them. I do not have data on the
racial makeup of municipal employees and can therefore not test this theory directly. See
Plerhoples (2013) for an analysis of whether racial distribution is a significant factor for
implementation of residency laws.
The final rationale for residency requirements is that they improve the overall population
and demographics of a city. Eisinger (1983) argues that this is the true motivation for all
residency requirement laws and that they are not actually a tool of public personnel management
but an attempt to combat local unemployment and encourage the spending of city salaries in the
local economy. He finds that the adoption of residency laws is associated with unemployment,
fiscal hardship, population loss, Frost Belt location, and mayor-council government. He also
finds that, on average, cities that rescind residency laws generally displayed stable or high
population growth and low unemployment. I examine this effect using panel techniques and an
instrumental variables approach to remove the endogeneity caused by the choice of when to
implement residency laws.

82

Laasby, 2011
92

No study has analyzed the overall impact of the laws on population or demographic
changes. This paper fills this gap in the literature by providing the first causal estimates of how
residency requirements affect the overall population and demographics within a city.

Empirical Methodology
To estimate the causal impact of residency requirements on the population and
demographic characteristics of cities, I employ a linear fixed effects panel data model. I analyze
a variety of dependent variables relating to crime, fiscal health, and demographics which will be
discussed in detail in the next section.
The basic model is as follows:

yit = β 0 + β1x it −1 + θ i + γ t + u it
where

x it

yit

(1)

are the different population and demographic characteristics (estimated separately);

is a vector of independent variables including whether or not the city has a residency

requirement;

θi

are individual fixed effects;

γt

are time fixed effects; and

u it

is the error

term.
In order to examine and account for time trends, I estimate both a contemporaneous
model with residency laws specified as a zero one indicator variable and a dynamic model with
the residency laws specified as a flow variable defined as the first difference of the indicator
variable. In the dynamic model, I include six leads and five lags of residency requirements in
93

order to examine the varying effects over time and to check for spikes in independent variables
before a residency law is implemented. I also include a variable that equals the one if the city
ever had a residency law six or more years before year t. This allows me to examine the effect of
a residency law in period t compared to seven years before the law was implemented, as well as
six years before and after implementation of the law.

Data
To the best of my knowledge, no comprehensive historical data on residency
requirements exists. To collect these data, I first searched through available city databases for
the 100 largest cities in the United States to find historical city codes and charters to identify the
years in which comprehensive residency requirements were implemented and removed as well as
details of each law. However, most cities do not have electronic historical data available, so I
then combined this search with a search of law articles and historical newspapers. This also only
produced data on the largest cities. To complete the data set, I then requested information from
the missing cities individually through of series of emails and phone calls. I attempted to collect
data on residency requirements back to 1900, but many cities do not have these data on record
before the 1970s. Some cities were able to provide documentation to back up their facts; others
simply provided the information based on memory. Out of the 100 cities from which I attempted
to collect data, I was able to collect the history of residency requirement laws back to 1970 for
73 cities. Table 3.1 shows the number of cities in my sample that reported having a residency
requirement law in each decade.
To account for city and census tract boundary changes over time, I use Brown
University’s Longitudinal Tract Database (LTDB) which standardizes census data from 1970

94

through 2000 to 2010 census tract boundaries. The LTDB provides estimates using 2010
boundaries for a standard set of variables from 1970 through 2000 for both full count variables
such as population, age, and race, as well as sample count variables based on the one-in-six
samples from the decennial census and the American Community survey such as income and
employment.
I then take the census tracts that are within the city boundaries in 2010 and compare these
over time. Finally, I merge these decennial city demographics with yearly fiscal data from the
Census of Governments and yearly crime data from the Federal Bureau of Investigation’s
Uniform Crime Reports. Descriptive statistics for all of the variables can be found in Table 3.2.
Table 3.3 show the number of cities that responded to the inquiry and the number that had and
removed a law during the study period.

Results
The Effect of Residency Requirement Laws on Crime
In order to examine the effect of residency requirement laws on crime, I merged data
from the FBI’s Uniform Crime Reports on offenses known and clearances by arrest for each year
from 1970 through 2006. The only crime data that are available for all years are what are known
as the crime index offenses: murder, forcible rape, robbery, aggravated assault, burglary,
larceny-theft, and motor vehicle theft. I sum all of these crimes into one overall crime level for
each city/year.
There is a large literature pointing to inaccuracies in this FBI data.

83

Namely, many

city/years are endogenously missing because the data are self-reported. Some cities choose not

83

Maltz, 1999; Lynch and Jarvis, 2008; Akiyama and Propheter, 2005 to name a few
95

to report their crime rates in years where they had a particularly high level of crime. Maltz
(1999) advocates for the use of longitudinal rather than cross-section imputation techniques to
correct for this missing data rather than the often used technique of using contemporaneous data
from similar jurisdictions to impute the missing values. I, therefore, use longitudinal imputation
techniques to interpolate the missing values in my data set.
Table 3.4 presents the results of the contemporaneous effect of residency requirement
laws on crime. The results suggest that residency laws are not correlated with changes in the
sum of the seven Crime Index offenses. To check whether time trends are confounding these
results, I run the same model including individual city by year variables. These results can be
seen in Table 3.5. Residency laws remain insignificant in this model.
Table 3.6 presents the results of the dynamic model which includes six lags and six leads
of residency laws. Residency laws appear to be correlated with a slight increase in crime two
years after the law is in place (by 4.2%), but then a decline in crime six or more years after the
law is implemented (by 14.5%). This may imply that there are long run decreases in crime after
a residency law is implemented. These numbers are large, but so are the confidence intervals
which implies that there is an effect but we do not know whether it is large or small. The
coefficients are imprecisely measured so we should not read too much into the point estimates.
In these regressions I control for lagged population, lagged income per capita, the lagged
percentage of residents who are between the ages of 18 and 25 (because this is considered to be
the age range that contains the highest crime rates), lagged race variables (specifically percent
black and percent not black nor white because these are the only breakdowns available in 1970),
the percent of residents with a high school degree, and the percent of vacant housing units in the
city. All of the covariates that are significant are of the expected sign. I chose not to include

96

police expenditures as a covariate due to concerns about feedback and endogeneity; police
expenditures are likely to increase when crime increases. This violates the strict exogeneity
requirement for a fixed effects model.
I also examined the effect of removing a residency law on crime in these cities. Table 3.7
presents these results. I remove all city/years from the data set that are before a residency law is
implemented so that I am only comparing the post residency removal period to the residency
period. The sum of the Crime Index offenses is significant and negative, indicating that the
removal of a residency requirement law is associated with a reduction in crime index crimes by
11.6%. However, once I include a time trend in the model, the coefficient becomes insignificant.
Therefore, it is unclear whether the reduction in crime seen after the removal of a residency law
is caused by the removal of the law or because the cities that remove the laws have decreasing
crime around the time of the removal. In the dynamic model results, shown in Table 3.9, the six
lags and leads of the removal of a residency law are not statistically significant, implying that
there is no significant change in crime after the removal of such a law.

The Effect of Residency Requirement Laws on Municipal Fiscal Health
Residency laws are hypothesized to affect the fiscal health of a city. As city employees
move into the city, they have an incentive to vote for increased expenditures. I measure fiscal
stress using the five fiscal health ratios shown in Table 3.10.
I define fiscal health using the definition proposed in Groves and Valente (2003). I then
define fiscal stress as a deficiency in any of these measures. Fiscal health is defined as a
government’s ability to maintain solvency in four measures: cash, budgetary, long run, and
service level. Cash solvency refers to a local government’s liquidity and effective cash

97

management, as well as its ability to pay current liabilities. Budgetary solvency refers to the
ability of the government to generate sufficient revenues to fund its current or desired service
levels. Long run solvency refers to the impact of existing long term obligations on future
resources. Service level solvency refers to the ability of the government to provide and sustain a
service level that citizens require and desire. Fiscal stress is therefore defined as a deficiency in
any of these measures.
Because no cash measure is included in the Census of Governments data, I can only
measure cash solvency using a modified debt service ratio. The typical debt service ratio
includes total interest on debt and principal payments. However, I only have information on
total interest on debt from the Census of Governments. I therefore measure the debt service as
the interest on debt compared to total revenues.
I measure budgetary solvency using an operating deficit ratio which measures the
operating surplus or deficit in comparison to total revenues. A deficit in one year does not
necessarily indicate fiscal stress. A government may have unusually large expenditures in the
current year but have planned for such events by saving money in previous years. Credit rating
agencies are generally concerned when there is a budget deficit in two consecutive years.
Because of this, I include an indicator for a budget deficit in two consecutive years or an
abnormally large deficit in one year.
I also measure budgetary solvency using an intergovernmental revenues ratio, which
measures how dependent the city is on intergovernmental transfers, and a property tax ratio,
which measures the tax base of the city. Property tax is the main revenue source for most cities
and therefore is a good indication of the health of the city’s budget.

98

Unfortunately, the Census of Governments data do not provide any measures of long
term solvency. However, these are partially captured in the demographic variables that are
estimated in the next section. The crime results in the previous section represent service level
solvency analysis.
I find no evidence that lagged residency laws affect the fiscal health of a city as measured
by these five fiscal stress ratios as can be seen in Tables 3.11 and again in Table 3.12 where I
include a time trend. I control for population, income per capita, race, percent of residents above
the age of 60, education, and number of vacant housing units. Income per capita could be either
positive or negative because as income increases, often both expenditures and revenues do as
well. There is also no clear prediction for the effect of race. We might expect that as a city
ages, the government will be more fiscally stressed because there are fewer people working and
providing tax income, and more people getting pensions and retiree health care services.
However, I find this not to be true, controlling for the other covariates. Education and income
per capita are closely correlated and may therefore be hard to interpret.
Table 3.13 presents the results of the dynamic model. Here we see a slight dip in the
probability of having a large operating deficit ratio in the years before a residency law is
implemented, then a decrease again four, five, and six or more years after the law is
implemented. However, we see a slight increase in the probability of having an operating deficit
ratio two years in a row two and one years before the law is implemented, then an increased
chance four, five, and six or more years after the law is implemented. As discussed earlier,
having one or the other of these variables be positive does no alone indicate stress. Therefore,
since these results are of the opposite sign, they do not indicate a clear implication for the effect
of residency laws on fiscal stress. We also see a slight increase in the intergovernmental revenue

99

ratios of cities the year before the implementation of a residency law, and then a reduction four
and six or more years after. This may imply an improvement in the fiscal health of a city after a
residency law is implemented.
I examine the effect of the removal of a residency requirement law on municipal fiscal
health in Tables 3.14 and 3.15. Both models, with and without time trends, show no significant
effect of the removal of residency laws on the fiscal health of cities.
In the dynamic model, presented in Table 3.16, we see an elevated level of property taxes
both before and after the residency law is removed. This may be due to overall trends in the
cities that remove the laws.

The Effect of Residency Requirement Laws on Demographics
Due to the nature of the census data, estimating the effect of residency requirement laws
on demographics is more difficult. Using the city boundaries in 2010 back over time, I can only
obtain demographic variables for the decennial years. I therefore estimate the model with the
data collapsed to the decennial years. This model is not ideal because it is so aggregated, so
results should be only used as a first glance at the effect of the laws on demographics.
Tables 3.17 and 3.18 present the results of the effect of residency requirement laws on
population, unemployment, and per capita income with data collapsed to the decennial years
both with and without including a time trend. I find no evidence that residency laws have a
statistically significant effect on population, unemployment, or income per capita.
In these models, I control for covariates that are thought to affect each dependent
variable. Glaeser, Scheinkman, and Shleifer (1995) find that population growth moves with
income growth, and that they are both positively related to schooling, negatively related to

100

unemployment, and negatively related to employment in manufacturing. I chose to include only
income per capita and not unemployment in my model since they are difficult to interpret when
both are included. I also include other covariates thought to affect population like crime,
property taxes, race, the share of the population that is elderly, and housing status. Similar
covariates are present in my percent unemployed and income per capita models. Sachs (2003)
argues that income per capita is caused by geography which is captured in city fixed effects in
my model, and quality of institutions which is captured in fiscal stress variables in my model.
Tables 3.19 and 3.20 present the results of the effect of the removal of residency
requirements on these variables collapsed to the decennial years, with and without instruments.
Once again, we see no statistically significant effect of the removal of residency laws on
population, percent unemployed, and income per capita.
Because the data is so aggregated in the previous analyses, I further explore the effect of
residency laws on population by estimating the model using yearly population estimates from the
Census of Governments. This model is also not ideal because these population estimates do not
control for city boundary changes. However, the two sets of results combined provide some idea
of the effect.
Tables 3.21 and 3.22 present the results of this analysis with and without a time trend. I
find no statistically significant effect again. When I examine this model using dynamics, I again
find no significant correlation as can be seen in Table 3.23.
Tables 3.24 and 3.25 use these same data to examine the effect of the removal of
residency laws on population. I once again find no significant effect with or without a time
trend, or dynamics. Removing a residency law does not appear to cause mass exodus from the
city as some policymakers fear.

101

Conclusion
My results show that the implementation of a residency law may be correlated with a
slight increase in crime two years after the law is implemented, and then a reduction in crime six
or more years after implementation. The removal of a law is correlated with an increase in
crime, but these results disappear once a time trend is included, indicating that this is likely
caused by city trends rather than by the law itself.
Residency laws may be correlated with a reduction in intergovernmental revenues four
and six or more years after it is implemented. This may imply a reduction in fiscal stress but it is
unclear using that indicator alone. Finally, residency laws and their removal do not appear to be
correlated with any changes in demographics or population.
These results provide no evidence that residency requirement laws remedy cities of high
crime rates, fiscal stress, or demographic ills. In addition, removing a residency laws do not
appear to cause mass exodus from the city.

102

APPENDIX

103

APPENDIX
Tables and Figures
Table 3.1: Number of Cities in dataset with Residency Requirements from 1970 to 2010

Year
1920
1930
1940
1950
1960
1970
1980
1990
2000
2010

Number of Cities with
Residency Requirement
Laws
3
5
9
9
15
15
24
23
20
17

104

Table 3.2: Descriptive Statistics
Variable
Residency Requirement
Population
Crime
% Age 18 to 25
% > Age 60
% Black
% Other Races
% High School Degree
% College Degree
% Vacant Units
% Owner Occupied Units
% Manufacturing Jobs
% Professional Jobs
Per Capita Expenditures
Income Per Capita
Property Tax Per Capita
Debt Ratio
Operating Deficit Ratio > 5%
Operating Deficit Ratio Two
Consecutive years
Intergovernmental Revenue Ratio
% Unemployed

Obs
Mean
Std. Dev.
Min
Max
73
0.589
0.495
0.000
1.000
2701
607,026
989,563
16,478 8,107,637
2701
45,640
68,666
49
682,063
2701
0.131
4%
6%
38%
2701
0.145
4%
3%
37%
2701
0.209
17%
0%
86%
2701
0.044
5%
0%
37%
2701
0.568
14%
16%
90%
2701
0.209
9%
4%
58%
2701
0.074
3%
-1%
22%
2701
0.534
10%
18%
83%
0.151
7%
0%
42%
2701
2701
0.281
7%
10%
64%
2701
$2.39
$1.78
$0.40
$16.78
2701 $10,984.00 $7,095.50 $235.91 $40,813.89
0
$0.36
$0.00
$2.51
2701
0.069
0.042
0.000
0.282
2701
2701
0.544
0.498
0.000
1.000
2650
2694
2701

0.317
0.236
7%

0.466
0.141
3%

0.000
0.000
2%

Table 3.3: City Residency Laws
Residency Requirement Law
Number of Cities Surveyed
Number of Responses
Number of Cities that had a Residency Law Between 1970 and
2006
Number of Cities that Removed a Residency Law Between 1970
and 2006
Number of Cities that Removed a Residency Law Between 1970
and 2006 because of a State Ban

105

100
73
42
23
13

1.000
0.913
20%

Table 3.4: Mean Effect of Residency Requirement Laws on Crime
(1)
Crime
Lagged Residency Requirement Law
0.001
(0.045)
Lagged Log of Population Interpolated
0.995***
(0.159)
Lagged Log of Income Per Capita Interpolated
0.017
(0.122)
Lagged % Age 18 to 25 Interpolated
0.417
(0.836)
Lagged % Black Interpolated
0.636
(0.384)
Lagged % Other Races Interpolated
-3.956***
(0.708)
Lagged % High School Degree Interpolated
-0.798
(0.657)
Lagged % Vacant Units Interpolated
0.394
(0.886)
Observations
2,628
R-squared
0.617
Number of Cities
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

106

Table 3.5: Mean Effect of Residency Requirement Laws on Crime with Time Trend
(1)
Crime
Lagged Residency Requirement Law
-0.003
(0.043)
Lagged Log of Population Interpolated
1.008***
(0.308)
Lagged Log of Income Per Capita Interpolated
0.147
(0.130)
Lagged % Age 18 to 25 Interpolated
-0.473
(0.504)
Lagged % Black Interpolated
0.213
(1.202)
Lagged % Other Races Interpolated
0.850
(2.349)
Lagged % High School Degree Interpolated
0.132
(0.966)
Lagged % Vacant Units Interpolated
0.368
(0.937)
Observations
2,628
R-squared
0.778
Number of Cities
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

107

Table 3.6: Dynamic Effects of Residency Requirement Laws on Crime

Six years before Residency Law Implementation
Five years before Residency Law Implementation
Four years before Residency Law Implementation
Three years before Residency Law Implementation
Two years before Residency Law Implementation
One year before Residency Law Implementation
Year of Residency Law Implementation
One year after Residency Law Implementation
Two years after Residency Law Implementation
Three years after Residency Law Implementation
Four years after Residency Law Implementation
Five years after Residency Law Implementation
Six or more years after Residency Law Implementation
Lagged Log of Population Interpolated
Lagged Log of Income Per Capita Interpolated
Lagged % Age 18 to 25 Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % High School Degree Interpolated

108

(1)
Crime
0.026
(0.032)
0.033
(0.042)
0.024
(0.045)
-0.006
(0.055)
-0.018
(0.072)
-0.025
(0.076)
0.034
(0.036)
0.040
(0.026)
0.042*
(0.021)
Omitted
0.010
(0.021)
-0.033
(0.031)
-0.145***
(0.048)
1.328***
(0.369)
-0.137
(0.337)
1.930
(1.430)
-0.057
(0.976)
-9.993***
(3.478)
-1.328
(1.210)

Table 3.6 (cont’d)
(1)
Crime
Lagged % Vacant Units Interpolated
4.984**
(1.878)
Observations
578
R-squared
0.477
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

Table 3.7: Mean Effect of the Removal of Residency Requirement Laws on Crime
(1)
Crime
Lagged Removal of Residency Requirement Law
-0.116*
(0.060)
Lagged Log of Population Interpolated
0.833***
(0.146)
Lagged Log of Income Per Capita Interpolated
0.07
(0.116)
Lagged % Age 18 to 25 Interpolated
-0.139
(0.803)
Lagged % Black Interpolated
0.381
(0.388)
Lagged % Other Races Interpolated
-3.476***
(0.558)
Lagged % High School Degree Interpolated
-1.136*
(0.654)
Lagged % Vacant Units Interpolated
0.481
(0.862)
Observations
2,392
R-squared
0.657
Number of Cities
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

109

Table 3.8: Mean Effect of the Removal of Residency Requirement Laws on Crime with
Time Trend
(1)
Crime
Lagged Removal of Residency Requirement Law
-0.037
(0.068)
Lagged Log of Population Interpolated
0.831**
(0.328)
Lagged Log of Income Per Capita Interpolated
0.153
(0.137)
Lagged % Age 18 to 25 Interpolated
-0.965*
(0.492)
Lagged % Black Interpolated
0.122
(1.315)
Lagged % Other Races Interpolated
0.909
(2.055)
Lagged % High School Degree Interpolated
-0.1
(1.034)
Lagged % Vacant Units Interpolated
0.689
(0.959)
Observations
2,392
R-squared
0.809
Number of Cities
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

110

Table 3.9: Dynamic Effects of the Removal of Residency Requirement Laws on Crime

Six years before Residency Law Removal
Five years before Residency Law Removal
Four years before Residency Law Removal
Three years before Residency Law Removal
Two years before Residency Law Removal
One year before Residency Law Removal
Year of Residency Law Removal
One year after Residency Law Removal
Two years after Residency Law Removal
Three years after Residency Law Removal
Four years after Residency Law Removal
Five years after Residency Law Removal
Six or more years after Residency Law Removal
Lagged Log of Population Interpolated
Lagged Log of Income Per Capita Interpolated
Lagged % Age 18 to 25 Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % High School Degree Interpolated

111

(1)
Crime
-0.035
(0.039)
-0.034
(0.050)
-0.029
(0.055)
0.015
(0.060)
0.042
(0.070)
0.042
(0.065)
0.039
(0.068)
0.032
(0.068)
0.024
(0.062)
0.010
(0.064)
-0.026
(0.088)
-0.005
(0.062)
-0.029
(0.067)
0.851***
(0.168)
0.034
(0.148)
-0.213
(0.744)
0.187
(0.436)
-3.814***
(0.764)
-1.828**
(0.786)

Table 3.9 (cont’d)
(1)
Crime
Lagged % Vacant Units Interpolated
2.554**
(1.206)
Observations
1,785
R-squared
0.595
Number of Cities
72
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

Table 3.10: Fiscal Stress Ratios

84

Ratio

Calculation
= ((Expenditures - Revenues)/
Revenues)>.05

Large Operating Deficit Ratio
Operating Deficit Ratio for Two Consecutive
Years
Intergovernmental Revenue Ratio
Debt Ratio
Property Tax Per Capita

84

= (Expenditures - Revenues)/ Revenues>0
= Intergovernmental Revenues/ Revenues
= Debt Interest/ Revenues
= Property Tax Revenue/ Population

Each of these variables is converted into 2010 dollars
112

Table 3.11: Mean Effect of Residency Requirement Laws on Municipal Fiscal Health
(1)
Operating
Deficit
Ratio > 5%
Lagged Residency
Requirement Law
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races Interpolated
Lagged % > Age 60
Interpolated
Lagged % High
School Degree
Interpolated

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

-0.020
(0.037)

0.025
(0.047)

-0.001
(0.010)

-0.001
(0.003)

0.001
(0.048)

0.003
(0.155)

0.094
(0.169)

0.006
(0.032)

-0.011
(0.017)

0.024
(0.168)

0.257*
(0.133)

-0.360**
(0.136)

0.040
(0.029)

0.005
(0.017)

0.285*
(0.153)

0.210
(0.670)

-0.570
(0.738)

0.486***
(0.099)

-0.153***
(0.058)

-0.463
(0.683)

-1.711*
(0.863)

1.322
(0.952)

-0.001
(0.196)

0.035
(0.100)

-0.789
(1.047)

-0.398
(0.983)

0.297
(1.040)

-0.252
(0.222)

-0.271**
(0.117)

-0.167
(1.046)

0.920
(0.636)

-1.131*
(0.579)

0.301***
(0.109)

-0.117*
(0.068)

-0.854
(0.651)

Lagged % Vacant
Units Interpolated

-0.541
0.823
0.326
0.138
1.027
(1.132)
(1.226)
(0.231)
(0.094)
(0.919)
Observations
2,628
2,628
2,622
2,622
2,617
R-squared
0.084
0.081
0.411
0.246
0.241
Number of Cities
73
73
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

113

Table 3.12: Mean Effect of Residency Requirement Laws on Municipal Fiscal Health with
Time Trend
(1)
Operating
Deficit
Ratio > 5%
Lagged Residency
Requirement Law
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races Interpolated
Lagged % > Age 60
Interpolated
Lagged % High
School Degree
Interpolated

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

-0.008
(0.047)

0.021
(0.052)

-0.001
(0.009)

0.002
(0.004)

0.007
(0.039)

-0.371
(0.487)

0.388
(0.437)

-0.075
(0.087)

0.062
(0.046)

0.852**
(0.349)

0.157
(0.226)

-0.192
(0.245)

-0.004
(0.032)

-0.006
(0.019)

0.320***
(0.120)

1.211
(1.236)

-1.421
(1.280)

0.240
(0.189)

-0.114
(0.110)

-0.510
(0.969)

2.703
(2.819)

-3.497
(3.048)

0.049
(0.504)

0.527
(0.369)

2.848
(3.235)

-0.065
(2.771)

-0.435
(2.783)

-0.903
(0.660)

-0.122
(0.287)

2.841**
(1.392)

-0.792
(1.479)

0.855
(1.651)

0.162
(0.254)

-0.145
(0.197)

-0.234
(0.906)

Lagged % Vacant
Units Interpolated

0.612
-0.362
0.217
0.307**
1.202
(1.338)
(1.496)
(0.198)
(0.144)
(1.031)
Observations
2,628
2,628
2,622
2,622
2,617
R-squared
0.151
0.177
0.585
0.439
0.492
Number of Cities
73
73
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

114

Table 3.13: Dynamic Effects of Residency Requirement Laws on Municipal Fiscal Health
(1)
Operating
Deficit
Ratio > 5%
Six years before
Residency Law
Implementation
Five years before
Residency Law
Implementation
Four years before
Residency Law
Implementation
Three years before
Residency Law
Implementation
Two years before
Residency Law
Implementation
One year before
Residency Law
Implementation
Year of Residency
Law
Implementation
One year after
Residency Law
Implementation
Two years after
Residency Law
Implementation

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

-0.081
(0.145)

0.064
(0.124)

0.008
(0.012)

-0.022
(0.014)

-0.073
(0.050)

-0.336**
(0.161)

0.120
(0.143)

0.014
(0.016)

-0.017
(0.014)

-0.116
(0.072)

-0.316*
(0.157)

0.079
(0.146)

0.002
(0.021)

-0.019
(0.016)

-0.099
(0.084)

-0.167
(0.180)

0.124
(0.132)

0.026
(0.022)

-0.014
(0.016)

-0.047
(0.061)

-0.418***
(0.115)

0.182*
(0.102)

0.034
(0.023)

-0.017
(0.014)

-0.083
(0.065)

-0.253*
(0.142)

0.189*
(0.105)

0.041*
(0.023)

-0.019
(0.016)

-0.081
(0.072)

-0.145
(0.151)

0.115
(0.109)

-0.015
(0.016)

0.000
(0.005)

0.061
(0.077)

-0.179
(0.187)

0.192
(0.139)

-0.013
(0.014)

-0.001
(0.005)

-0.021
(0.033)

-0.033
(0.163)

0.039
(0.113)

-0.018
(0.012)

0.003
(0.004)

0.003
(0.040)

115

Table 3.13 (cont’d)
(1)
Operating
Deficit
Ratio > 5%
Three years after
Residency Law
Implementation
Four years after
Residency Law
Implementation
Five years after
Residency Law
Implementation
Six or more years
after Residency Law
Implementation
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races Interpolated
Lagged % > Age 60
Interpolated

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

Omitted

Omitted

Omitted

Omitted

Omitted

-0.319***
(0.116)

0.184**
(0.090)

-0.015*
(0.009)

0.000
(0.003)

-0.033
(0.031)

-0.287*
(0.165)

0.282**
(0.119)

-0.001
(0.010)

0.001
(0.005)

-0.032
(0.031)

-0.228*
(0.128)

0.246**
(0.098)

-0.034***
(0.011)

0.004
(0.007)

0.008
(0.042)

0.360
(0.615)

-0.753
(0.906)

-0.034
(0.121)

0.076
(0.081)

0.929**
(0.358)

-0.436
(0.593)

0.252
(0.652)

0.236**
(0.092)

-0.056
(0.041)

-0.142
(0.432)

-1.588
(1.001)

1.118
(1.326)

0.539
(0.323)

-0.339**
(0.144)

-0.545
(1.379)

2.018
(4.960)

-1.815
(5.488)

3.214*
(1.788)

-0.246
(0.509)

-5.285
(4.818)

-0.828
(4.266)

0.268
(5.256)

-0.618
(1.162)

-0.448
(0.497)

2.511
(3.861)

116

Table 3.13 (cont’d)
(1)
Operating
Deficit Ratio
> 5%
Lagged % High
School Degree
Interpolated

3.193
(2.225)

(2)
Operating
Deficit Two
Consecutive
Years

-2.284
(2.048)

Lagged % Vacant
Units
Interpolated

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

1.017*
(0.549)

-0.003
(0.243)

(5)
Log of
Property
Tax Per
Capita

-2.607
(1.690)

-4.791*
0.891
0.135
0.643**
2.510
(2.822)
(3.028)
(0.860)
(0.276)
(3.744)
Observations
578
578
578
578
578
R-squared
0.146
0.125
0.447
0.359
0.113
Number of Cities
34
34
34
34
34
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

117

Table 3.14: Mean Effect of the Removal of Residency Requirement Laws on Municipal
Fiscal Health
(1)
Operating
Deficit
Ratio > 5%
Lagged Removal of
Residency
Requirement Law
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races Interpolated
Lagged % > Age 60
Interpolated
Lagged % High
School Degree
Interpolated

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

0.038
(0.064)

-0.018
(0.074)

-0.018
(0.015)

0.006
(0.006)

0.043
(0.045)

0.056
(0.177)

0.053
(0.183)

0.003
(0.037)

-0.006
(0.019)

0.082
(0.161)

0.23
(0.145)

-0.330**
(0.144)

0.042
(0.032)

0.002
(0.019)

0.233
(0.152)

0.314
(0.723)

-0.77
(0.782)

0.524***
(0.103)

-0.154**
(0.064)

-0.326
(0.694)

-1.988**
(0.827)

1.558*
(0.926)

-0.034
(0.180)

0.061
(0.101)

-0.254
(1.021)

-0.321
(1.078)

0.006
(1.143)

-0.357*
(0.211)

-0.213*
(0.114)

0.029
(1.190)

1.072
(0.683)

-1.192*
(0.621)

0.296**
(0.118)

-0.106
(0.071)

-1.098
(0.685)

Lagged % Vacant
Units Interpolated

-0.62
0.955
0.287
0.166*
0.951
(1.175)
(1.249)
(0.224)
(0.096)
(0.892)
Observations
2,392
2,392
2,386
2,386
2,381
R-squared
0.086
0.082
0.395
0.227
0.254
Number of Cities
73
73
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

118

Table 3.15: Mean Effect of the Removal of Residency Requirement Laws on Municipal
Fiscal Health with Time Trend
(1)
Operating
Deficit
Ratio > 5%
Lagged Removal of
Residency
Requirement Law
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races Interpolated
Lagged % > Age 60
Interpolated
Lagged % High
School Degree
Interpolated

(2)
Operating
Deficit Two
Consecutive
Years

(3)

(4)

Intergovernmental
Revenue Ratio

Debt
Ratio

(5)
Log of
Property
Tax Per
Capita

0.041
(0.078)

-0.024
(0.084)

-0.007
(0.013)

-0.007
(0.008)

-0.033
(0.041)

-0.272
(0.566)

0.394
(0.499)

-0.078
(0.090)

0.074
(0.048)

0.888**
(0.383)

0.089
(0.250)

-0.103
(0.275)

-0.003
(0.034)

-0.010
(0.021)

0.298**
(0.121)

2.500*
(1.472)

-2.420*
(1.362)

0.290
(0.252)

-0.098
(0.145)

-1.111
(0.956)

3.688
(2.833)

-4.531
(3.004)

-0.334
(0.509)

0.281
(0.327)

2.685
(3.307)

-0.963
(3.058)

-0.172
(3.085)

-0.731
(0.664)

-0.057
(0.323)

2.556*
(1.445)

-1.196
(1.775)

1.930
(1.883)

0.149
(0.285)

-0.159
(0.233)

-0.084
(0.987)

Lagged % Vacant
Units Interpolated

0.911
-0.739
0.191
0.328**
0.642
(1.412)
(1.557)
(0.212)
(0.159)
(1.042)
Observations
2,392
2,392
2,386
2,386
2,381
R-squared
0.161
0.188
0.59
0.425
0.483
Number of Cities
73
73
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

119

Table 3.16: Dynamic Effect of the Removal of Residency Requirement Laws on Municipal
Fiscal Health
(1)
Operating
Deficit
Ratio >
5%
Six years before
Residency Law
Removal
Five years before
Residency Law
Removal
Four years before
Residency Law
Removal
Three years before
Residency Law
Removal
Two years before
Residency Law
Removal
One year before
Residency Law
Removal
Year of Residency
Law Removal
One year after
Residency Law
Removal
Two years after
Residency Law
Removal

(2)

(3)

(4)

(5)

Operating Deficit
Two Consecutive
Years

Intergovernmental Revenue
Ratio

Debt
Ratio

Log of
Property Tax
Per Capita

-0.028
(0.138)

-0.006
(0.113)

-0.019*
(0.010)

0.024*
(0.013)

0.085**
(0.041)

0.233
(0.143)

-0.046
(0.130)

-0.018
(0.013)

0.019
(0.014)

0.102**
(0.046)

0.207
(0.147)

-0.035
(0.141)

-0.005
(0.017)

0.022
(0.015)

0.090
(0.066)

0.129
(0.128)

-0.083
(0.104)

-0.030
(0.021)

0.013
(0.014)

0.117*
(0.061)

0.331***
(0.107)

-0.132
(0.100)

-0.014
(0.022)

0.018
(0.012)

0.122*
(0.062)

0.108
(0.092)

-0.166*
(0.089)

-0.019
(0.023)

0.023*
(0.013)

0.120*
(0.068)

0.076
(0.109)

0.050
(0.113)

-0.035*
(0.020)

0.021
(0.014)

0.192***
(0.059)

0.181
(0.109)

0.012
(0.106)

-0.019
(0.021)

0.017
(0.015)

0.178***
(0.063)

0.123
(0.108)

-0.037
(0.097)

-0.024
(0.027)

0.021
(0.014)

0.167**
(0.065)

120

Table 3.16 (cont’d)
(1)
Operating
Deficit
Ratio >
5%
Three years after
Residency Law
Removal
Four years after
Residency Law
Removal
Five years after
Residency Law
Removal
Six or more years
after Residency
Law Removal
Lagged Log of
Population
Interpolated
Lagged Log of
Income Per
Capita
Interpolated
Lagged % Black
Interpolated
Lagged % Other
Races
Interpolated
Lagged % > Age
60 Interpolated

(2)

(3)

Operating Deficit
IntergovernTwo Consecutive mental Revenue
Years
Ratio

(4)

(5)

Debt
Ratio

Log of
Property Tax
Per Capita

0.157
(0.111)

0.024
(0.106)

-0.021
(0.028)

0.016
(0.014)

0.163**
(0.064)

0.110
(0.123)

-0.080
(0.111)

-0.033
(0.026)

0.017
(0.013)

0.198***
(0.071)

0.140
(0.113)

0.017
(0.130)

-0.015
(0.028)

0.017
(0.013)

0.030
(0.096)

0.154
(0.122)

-0.110
(0.126)

-0.043
(0.029)

0.017
(0.013)

0.116
(0.073)

-0.009
(0.238)

0.110
(0.247)

0.010
(0.051)

0.008
(0.024)

0.196
(0.183)

0.425**
(0.207)

-0.528**
(0.224)

0.051
(0.044)

0.003
(0.023)

0.174
(0.181)

0.816
(0.873)

-1.275
(0.884)

0.418***
(0.142)

-0.215**
(0.082)

-0.672
(0.752)

-2.146**
(1.049)

2.106*
(1.163)

-0.080
(0.250)

0.057
(0.126)

-1.083
(1.305)

-1.310
(2.212)

0.937
(2.632)

-0.373
(0.337)

-0.236
(0.161)

-0.348
(1.431)

121

Table 3.16 (cont’d)
(1)
Operating
Deficit
Ratio >
5%
Lagged % High
School Degree
Interpolated

2.788**
(1.111)

(2)

(3)

(4)

(5)

Operating Deficit
Two Consecutive
Years

Intergovernmental Revenue
Ratio

Debt
Ratio

Log of
Property Tax
Per Capita

-0.071
(0.099)

-1.184
(0.735)

-2.795**
(1.205)

0.282*
(0.151)

Lagged %
Vacant Units
Interpolated

-1.822
2.366
0.252
0.303**
1.789
(1.516)
(1.539)
(0.353)
(0.138)
(1.357)
Observations
1,785
1,785
1,785
1,785
1,785
R-squared
0.07
0.072
0.452
0.269
0.22
Number of Cities
72
72
72
72
72
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

122

Table 3.17: The Effect of Residency Requirement Laws on Municipal Demographics
Collapsed to Decennial Years

Lagged Residency Requirement Law
Lagged Log of Crime
Lagged Log of Property Tax Per
Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures
Lagged Log of Income Per Capita
Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % > Age 60 Interpolated
Lagged % High School Degree
Interpolated
Lagged % College Degree Interpolated
Lagged % Manufacturing Jobs
Interpolated
Lagged % Professional Jobs
Interpolated
Lagged % Vacant Units Interpolated

(1)
Log
Population
-0.007
(0.023)
0.132***
(0.032)

(2)
Percent
Unemployed
0.003
(0.003)
0.006**
(0.002)

(3)
Income Per
Capita
-0.007
(0.032)
0.273***
(0.065)

-0.021
(0.032)
0.132
(0.198)
-0.082**
(0.041)

-0.003
(0.004)
0.067**
(0.030)
0.002
(0.006)

0.126*
(0.072)
0.076
(0.436)
-0.168*
(0.095)

0.560***
(0.064)
-0.855***
(0.258)
1.531***
(0.536)
0.182
(0.749)

0.078***
(0.021)
-0.131**
(0.051)
-0.010
(0.034)

-1.304***
(0.288)
2.450**
(1.013)
5.769***
(1.249)

-0.453
(0.486)
-1.025
(0.749)

0.264***
(0.034)
0.071
(0.074)

-1.349
(0.932)
2.681**
(1.309)

0.042
(0.340)

0.006
(0.031)

0.603
(0.634)

-1.084*
(0.586)
-0.671*
(0.397)

0.150**
(0.071)

-1.229
(1.388)

123

Table 3.17 (cont’d)
(1)

(2)

Log Population

Percent Unemployed

Lagged % Owner Occupied Units
Interpolated

(3)
Income Per
Capita

0.284
(0.314)
Observations
291
291
291
R-squared
0.912
0.421
0.951
Number of Cities
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970, 1980, 1990, 2000, and 2006.

124

Table 3.18: The Effect of Residency Requirement Laws on Municipal Demographics
Collapsed to Decennial Years with Time Trend

Lagged Residency Requirement Law
Lagged Log of Crime
Lagged Log of Property Tax Per
Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures
Lagged Log of Income Per Capita
Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % > Age 60 Interpolated
Lagged % High School Degree
Interpolated
Lagged % College Degree Interpolated
Lagged % Manufacturing Jobs
Interpolated
Lagged % Professional Jobs
Interpolated
Lagged % Vacant Units Interpolated

(1)
Log
Population
-0.018
(0.015)
0.054**
(0.021)

(2)
Percent
Unemployed
0.004
(0.007)
0.015**
(0.006)

(3)
Income Per
Capita
0.020
(0.046)
0.013
(0.041)

0.004
(0.018)
0.052
(0.134)
-0.051
(0.034)

-0.009
(0.007)
0.106**
(0.044)
0.009
(0.011)

0.102
(0.074)
0.325
(0.339)
-0.006
(0.099)

0.125***
(0.041)
-1.159**
(0.548)
-0.020
(0.760)
-3.629***
(0.870)

0.407***
(0.089)
0.221
(0.330)
0.347
(0.269)

-1.136
(0.852)
-5.379**
(2.168)
-3.172
(2.150)

-1.687***
(0.510)
-1.898**
(0.796)

0.134
(0.116)
0.293
(0.218)

-0.546
(1.188)
5.186***
(1.757)

0.381
(0.553)

0.203
(0.126)

-1.813
(1.189)

0.762
(0.517)
-0.482*
(0.289)

0.103
(0.196)

-1.621
(1.715)

125

Table 3.18: The Effect of Residency Requirement Laws on Municipal Demographics
Collapsed to Decennial Years with Time Trend

(1)

(2)

Log Population

Percent Unemployed

Lagged % Owner Occupied Units
Interpolated

(3)
Income Per
Capita

0.456
(0.326)
Observations
291
291
291
R-squared
0.989
0.636
0.991
Number of Cities
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970, 1980, 1990, 2000, and 2006.

126

Table 3.19: The Effect of the Removal of Residency Requirement Laws on Municipal
Demographics Collapsed to Decennial Years
(1)
Log
Population
Lagged Removal of Residency
Requirement Law
Lagged Log of Crime
Lagged Log of Property Tax Per Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures
Lagged Log of Income Per Capita
Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % > Age 60 Interpolated
Lagged % High School Degree
Interpolated
Lagged % College Degree Interpolated
Lagged % Manufacturing Jobs
Interpolated
Lagged % Professional Jobs
Interpolated
Lagged % Vacant Units Interpolated

(2)
Percent
Unemployed

(3)
Income Per
Capita

0.014
(0.033)
0.124***
(0.031)
-0.029
(0.035)
0.157
(0.187)
-0.087*
(0.047)

-0.004
(0.005)
0.006**
(0.003)
-0.003
(0.004)
0.069**
(0.032)
-0.001
(0.006)

-0.027
(0.049)
0.276***
(0.069)
0.145*
(0.079)
0.095
(0.465)
-0.140
(0.106)

0.565***
(0.066)
-0.926***
(0.283)
1.469**
(0.555)
0.481
(0.777)

0.086***
(0.023)
-0.135**
(0.057)
-0.013
(0.037)

-1.185***
(0.326)
2.126*
(1.083)
6.014***
(1.316)

-0.621
(0.498)
-1.529*
(0.786)

0.273***
(0.039)
0.101
(0.078)

-1.152
(0.918)
2.946**
(1.448)

0.038
(0.392)

0.033
(0.035)

0.789
(0.699)

-0.790
(0.646)
-0.598
(0.398)

0.139*
(0.070)

-1.413
(1.473)

127

Table 3.19 (cont’d)
(1)

(2)

Log Population

Percent Unemployed

Lagged % Owner Occupied Units
Interpolated

(3)
Income Per
Capita

0.239
(0.322)
Observations
278
278
278
R-squared
0.917
0.400
0.949
Number of Cities
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970, 1980, 1990, 2000, and 2006.

128

Table 3.20: The Effect of the Removal of Residency Requirement Laws on Municipal
Demographics Collapsed to Decennial Years with Time Trend
(1)
Log
Population
Lagged Removal of Residency
Requirement Law
Lagged Log of Crime
Lagged Log of Property Tax Per Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures
Lagged Log of Income Per Capita
Interpolated
Lagged % Black Interpolated
Lagged % Other Races Interpolated
Lagged % > Age 60 Interpolated
Lagged % High School Degree
Interpolated
Lagged % College Degree Interpolated
Lagged % Manufacturing Jobs
Interpolated
Lagged % Professional Jobs
Interpolated
Lagged % Vacant Units Interpolated

(2)
Percent
Unemployed

(3)
Income Per
Capita

0.021
(0.021)
0.063***
(0.023)
-0.006
(0.022)
0.068
(0.143)
-0.042
(0.038)

-0.005
(0.009)
0.011*
(0.006)
-0.010
(0.007)
0.105**
(0.048)
0.006
(0.012)

-0.007
(0.049)
0.007
(0.039)
0.144**
(0.070)
0.287
(0.334)
0.046
(0.113)

0.137***
(0.050)
-1.170**
(0.533)
0.211
(0.737)
-3.717***
(0.969)

0.457***
(0.090)
0.182
(0.294)
0.372
(0.275)

-0.514
(0.839)
-5.051**
(2.085)
-3.263
(2.254)

-1.666***
(0.517)
-2.059**
(0.868)

0.088
(0.125)
0.439**
(0.199)

-0.661
(1.168)
5.212***
(1.772)

0.516
(0.583)

0.281**
(0.123)

-2.261*
(1.176)

0.890
(0.583)
-0.532*
(0.304)

-0.057
(0.225)

-1.329
(1.546)

129

Table 3.20 (cont’d)
(1)
Log Population
Lagged % Owner Occupied Units
Interpolated

(2)
Percent
Unemployed

(3)
Income
Per Capita

0.489
(0.321)
Observations
278
278
278
R-squared
0.99
0.669 0.991
Number of Cities
73
73
73
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970, 1980, 1990, 2000, and 2006.

130

Table 3.21: Mean Effect of Residency Requirement Laws on Population Using Yearly
Census of Governments Estimates
(1)
Log Population
Lagged Residency Requirement Law
-0.031
(0.024)
Lagged Log of Crime
0.157***
(0.035)
Lagged Log of Property Tax Per Capita
-0.029
(0.027)
Lagged Debt Ratio
0.194
(0.174)
Lagged Per Capita Expenditures
-0.059
(0.036)
Lagged Log of Income Per Capita Interpolated
0.702***
(0.078)
Lagged % Black Interpolated
-1.118***
(0.310)
Lagged % Other Races Interpolated
1.629***
(0.591)
Lagged % > Age 60 Interpolated
0.098
(0.882)
Lagged % High School Degree Interpolated
-1.371**
(0.659)
Lagged % College Degree Interpolated
-2.263**
(0.983)
Lagged % Manufacturing Jobs Interpolated
-0.348
(0.384)
Lagged % Professional Jobs Interpolated
-1.676**
(0.772)
Lagged % Vacant Units Interpolated
-0.804
(0.755)
Lagged % Owner Occupied Units Interpolated
-0.015
(0.211)
Observations
2,618
Number of Cities
73
R-squared
0.861
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

131

Table 3.22: Mean Effect of Residency Requirement Laws on Population Using Yearly
Census of Governments Estimates with Time Trend
(1)
Log Population
Lagged Residency Requirement Law
-0.030
(0.025)
Lagged Log of Crime
0.047**
(0.019)
Lagged Log of Property Tax Per Capita
-0.010
(0.015)
Lagged Debt Ratio
0.226
(0.152)
Lagged Per Capita Expenditures
-0.008
(0.024)
Lagged Log of Income Per Capita Interpolated
0.159***
(0.047)
Lagged % Black Interpolated
-0.827***
(0.312)
Lagged % Other Races Interpolated
-0.376
(1.394)
Lagged % > Age 60 Interpolated
-4.384***
(0.794)
Lagged % High School Degree Interpolated
-0.965*
(0.506)
Lagged % College Degree Interpolated
0.134
(1.226)
Lagged % Manufacturing Jobs Interpolated
0.543
(0.791)
Lagged % Professional Jobs Interpolated
0.683
(0.609)
Lagged % Vacant Units Interpolated
-0.397
(0.398)
Lagged % Owner Occupied Units Interpolated
0.601**
(0.255)
Observations
2,618
Number of Cities
73
R-squared
0.955
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

132

Table 3.23: Dynamic Effect of Residency Requirement Laws on Population Using Yearly
Census of Governments Estimates

Six years before Residency Law Implementation
Five years before Residency Law Implementation
Four years before Residency Law Implementation
Three years before Residency Law Implementation
Two years before Residency Law Implementation
One year before Residency Law Implementation
Year of Residency Law Implementation
One year after Residency Law Implementation
Two years after Residency Law Implementation
Three years after Residency Law Implementation
Four years after Residency Law Implementation
Five years after Residency Law Implementation
Six or more years after Residency Law Implementation
Lagged Log of Crime
Lagged Log of Property Tax Per Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures
Lagged Log of Income Per Capita Interpolated

133

(1)
Log Population
0.019
(0.015)
0.003
(0.019)
0.011
(0.019)
0.003
(0.018)
0.006
(0.022)
0.005
(0.024)
-0.015
(0.014)
-0.010
(0.009)
-0.007
(0.010)
Omitted
0.004
(0.009)
0.005
(0.009)
0.019
(0.014)
0.098***
(0.032)
-0.004
(0.004)
0.182
(0.167)
0.045*
(0.023)
0.571**
(0.220)

Table 3.23 (cont’d)
(1)
Log Population
(0.264)
Lagged % Other Races Interpolated
2.136
(1.695)
Lagged % > Age 60 Interpolated
0.772
(1.696)
Lagged % High School Degree Interpolated
0.507
(0.732)
Lagged % College Degree Interpolated
-1.035
(1.743)
Lagged % Manufacturing Jobs Interpolated
1.012**
(0.376)
Lagged % Professional Jobs Interpolated
-0.837
(1.089)
Lagged % Vacant Units Interpolated
-1.198*
(0.597)
Lagged % Owner Occupied Units Interpolated
-0.766
(0.590)
Observations
578
Number of Cities
34
R-squared
0.667
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the 100 largest cities in the
US (as of 2010) from 1970- 2006.

134

Table 3.24: Mean Effect of the Removal of Residency Requirement Laws on Population
Using Yearly Census of Governments Estimates
(1)
Log Population
Lagged Removal of Residency Requirement Law
0.026
(0.037)
Lagged Log of Crime
0.142***
(0.040)
Lagged Log of Property Tax Per Capita
-0.029
(0.028)
Lagged Debt Ratio
0.195
(0.174)
Lagged Per Capita Expenditures
-0.070
(0.042)
Lagged Log of Income Per Capita Interpolated
0.708***
(0.081)
Lagged % Black Interpolated
-1.258***
(0.353)
Lagged % Other Races Interpolated
1.501**
(0.630)
Lagged % > Age 60 Interpolated
0.472
(0.879)
Lagged % High School Degree Interpolated
-1.630**
(0.754)
Lagged % College Degree Interpolated
-2.752**
(1.134)
Lagged % Manufacturing Jobs Interpolated
-0.509
(0.468)
Lagged % Professional Jobs Interpolated
-1.276
(0.888)
Lagged % Vacant Units Interpolated
-0.763
(0.770)
Lagged % Owner Occupied Units Interpolated
-0.106
(0.226)
Observations
2,382
Number of Cities
73
R-squared
0.87
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the largest cities in the US
(as of 2010) from 1970- 2006.

135

Table 3.25: Mean Effect of the Removal of Residency Requirement Laws on Population
Using Yearly Census of Governments Estimates with Time Trend
(1)
Log Population
Lagged Removal of Residency Requirement Law
0.054
(0.036)
Lagged Log of Crime
0.049**
(0.023)
Lagged Log of Property Tax Per Capita
-0.011
(0.015)
Lagged Debt Ratio
0.248
(0.156)
Lagged Per Capita Expenditures
-0.011
(0.024)
Lagged Log of Income Per Capita Interpolated
0.178***
(0.047)
Lagged % Black Interpolated
-1.146***
(0.362)
Lagged % Other Races Interpolated
-0.440
(1.380)
Lagged % > Age 60 Interpolated
-4.568***
(0.907)
Lagged % High School Degree Interpolated
-0.878
(0.579)
Lagged % College Degree Interpolated
0.124
(1.183)
Lagged % Manufacturing Jobs Interpolated
0.418
(0.787)
Lagged % Professional Jobs Interpolated
0.649
(0.664)
Lagged % Vacant Units Interpolated
-0.426
(0.418)
Lagged % Owner Occupied Units Interpolated
0.563*
(0.305)
Observations
2,382
Number of Cities
73
R-squared
0.959
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the largest cities in the US
(as of 2010) from 1970- 2006.

136

Table 3.26: Dynamic Effect of the Removal of Residency Requirement Laws on Population
Using Yearly Census of Governments Estimates

Six years before Residency Law Removal
Five years before Residency Law Removal
Four years before Residency Law Removal
Three years before Residency Law Removal
Two years before Residency Law Removal
One year before Residency Law Removal
Year of Residency Law Removal
One year after Residency Law Removal
Two years after Residency Law Removal
Three years after Residency Law Removal
Four years after Residency Law Removal
Five years after Residency Law Removal
Six or more years after Residency Law Removal
Lagged Log of Crime
Lagged Log of Property Tax Per Capita
Lagged Debt Ratio
Lagged Per Capita Expenditures

137

(1)
Log Population
-0.011
(0.023)
0.001
(0.026)
-0.013
(0.025)
-0.013
(0.025)
-0.011
(0.024)
-0.038
(0.036)
-0.035
(0.029)
-0.033
(0.032)
-0.022
(0.036)
-0.017
(0.037)
-0.015
(0.040)
-0.022
(0.042)
-0.021
(0.043)
0.145***
(0.038)
-0.027
(0.022)
0.042
(0.122)
0.008
(0.028)

Table 3.26 (cont’d)
(1)
Log Population
(0.076)
Lagged % Black Interpolated
-0.842**
(0.370)
Lagged % Other Races Interpolated
1.472**
(0.705)
Lagged % > Age 60 Interpolated
0.164
(0.962)
Lagged % High School Degree Interpolated
-1.122
(0.721)
Lagged % College Degree Interpolated
-2.698**
(1.163)
Lagged % Manufacturing Jobs Interpolated
0.335
(0.426)
Lagged % Professional Jobs Interpolated
-1.793*
(0.966)
Lagged % Vacant Units Interpolated
-1.271
(0.918)
Lagged % Owner Occupied Units Interpolated
0.117
(0.232)
Observations
1,785
Number of Cities
72
R-squared
0.865
Robust standard errors in parentheses, clustered at the city level. * significant at 10%; **
significant at 5%; *** significant at 1%. Sample is a panel of 73 of the largest cities in the US
(as of 2010) from 1970- 2006.

138

BIBLIOGRAPHY

139

BIBLIOGRAPHY

Akiyama, Yoshio, and Sharon K. Propheter. (2005) “Methods of Data Quality Control: For
Uniform Crime Reporting Programs.” Criminal Justice Information Services Division. Federal
Bureau of Investigation.
Angel, Cecil. “Detroiters tell legislators how ban on residency laws is hurting city.” Detroit
Free Press (Aug. 18, 2011).
http://www.freep.com/article/20110819/NEWS01/108190413/Detroiters-tell-legislators-howban-residency-laws-hurting-city
Boies, Christopher G. “Fleming v. Giuliani.” New York Law School Law Review. (2005/2006)
50:4 pp. 1003-1014.
City of Cincinnati. (1977). Ordinance of 29 June 1977.
Duncan, Brian. “Using Municipal Residency Requirements to Disguise Public Policy.” Public
Finance Review. 33: 84. (2005)
“Ector v. City of Torrance.” (1973) 514 P 2nd 433.
Eisinger, Peter K. “Municipal Residency Requirements and the Local Economy.” Social Science
Quarterly. 64:1 (March 1983) p. 85
Fogelson, Robert. “Big City Police.” Cambridge, Ma. Harvard University Press. (1977)
Glaeser, Edward L., Jose A. Scheinkman, Andrei Shleifer. (1995)“Economic Growth in a CrossSection of Cities.” National Bureau of Economic Research Working Paper No. 5013.
Gonzalez, Rodolfo A., Stephen L. Mehay, Kevin Duffy-Deno. “Municipal Residency Laws:
Effects on Police Employment, Compensation, and Productivity.” Journal of Labor Research
(Fall 1991) 13:4
Hirsch, Werner Z., Anthony M. Rufolo. “Residence Laws and Unionization in Municipal Labor
Markets: The Case of Firefighters.” Journal of Labor Research (Winter, 1986) 7:1
International City Management Association. “Personnel Practices in the Municipal Police
Service: 1976.” Urban Data Service Report (Dec., 1976) pp. 2-14
Laasby, Gitte. “Firefighter group backs keeping residency rule.” Journal Sentinel Online.
(2011) http://www.jsonline.com/news/milwaukee/119739199.html
“Local hire laws: Alaska’s futile attempts at preferential treatment.” Alaska Law Review (Dec.,
1987) pp. 359-386.

140

Lynch, James P. and John P. Jarvis. (2008). “Missing Data and Imputation in the Uniform Crime
Reports and the Effects on National Estimats.” Journal of Contemporary Criminal Justice.
24:69.
Maltz, M. (1999). “Bridging gaps in police crime data.” Washington, DC: U.S. Bureau of Justice
Statistics.
Mehay, Stephen L., Kenneth P. Seiden. “Municipal Residency Laws and Local Public Budgets.”
Public Choice (1986) 48:1 pp.27-35
Myers, Ross S. “The constitutionality of continuing residency requirements for local
government employees: a second look.” California Western Law Review (Fall 1986) 23 p 24-41
Nichols, Darren A. “Bank of America donating houses to entice Detroit cops to live in city.”
Detroit News (2011) http://detnews.com/article/20110323/METRO01/103230415/Bank-ofAmerica-donating-houses-to-entice-Detroit-cops-to-live-in-city#ixzz1Jz84pw1v
New York News. (2011) “NYPD Facebook flap prompts call for residency rule.”
http://abclocal.go.com/wabc/story?section=news/local/new_york&id=8463511
O’Brien, Kevin M. “Do Municipal Residency Laws Affect Labour Market Outcomes?” Urban
Studies (1997) 34:11 pp. 1759-1769
Plerhoples, Christina. (2013) “Motivations for the Implementation of Municipal Residency
Requirement Laws.” PhD Dissertation.
Reed, Thomas. “Municipal Management.” New York: McGraw Hill. (1941)
Sachs, Jeffrey D. (2003) “Institutions Don’t Rule: Direct Effects of Geography on per Capita
Income.” National Bureau of Economic Research Working Paper 9490.
Smith, Dennis C. “Police Attitudes and Performance: The Impact of Residency.” Urban Affairs
Review (1980) 15:317
Stein, Jason, and Don Walker. (2013). “Local officials, some GOP lawmakers sound off against
residency change.” http://www.jsonline.com/news/statepolitics/local-officials-some-goplawmakers-sound-off-against-residency-change-179gnlk-202418121.html
Stein, Lana. “Representative Local Government: Minorities in the Municipal Work Force.” The
Journal of Politics (Aug., 1986) 48:3 pp. 694-713
Wooldridge, Jeffrey. 2002. Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA: MIT Press.

141