Content uploaded by Mehmet Serkan Tosun
Author content
All content in this area was uploaded by Mehmet Serkan Tosun
Content may be subject to copyright.
DISCUSSION PAPER SERIES
Forschungsinstitut
zur Zukunft der Arbeit
Institute for the Study
of Labor
Property Value Assessment Growth Limits,
Tax Base Erosion and Regional In-Migration
IZA DP No. 4906
April 2010
Mark Skidmore
Mehmet S. Tosun
Property Value Assessment Growth
Limits, Tax Base Erosion and
Regional In-Migration
Mark Skidmore
Michigan State University
Mehmet S. Tosun
University of Nevada, Reno
and IZA
Discussion Paper No. 4906
April 2010
IZA
P.O. Box 7240
53072 Bonn
Germany
Phone: +49-228-3894-0
Fax: +49-228-3894-180
E-mail: iza@iza.org
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in
this series may include views on policy, but the institute itself takes no institutional policy positions.
The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
organization supported by Deutsche Post Foundation. The center is associated with the University of
Bonn and offers a stimulating research environment through its international network, workshops and
conferences, data service, project support, research visits and doctoral program. IZA engages in (i)
original and internationally competitive research in all fields of labor economics, (ii) development of
policy concepts, and (iii) dissemination of research results and concepts to the interested public.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 4906
April 2010
ABSTRACT
Property Value Assessment Growth Limits,
Tax Base Erosion and Regional In-Migration
In 1994 a limit on the growth of property values for tax purposes was imposed in Michigan.
One consequence of the newly imposed assessment growth cap was an emerging
differential in tax prices between potential new property owners and long-time property
owners. The purpose of this paper is to examine the impact of this growing tax price
differential on migration patterns. Using county level data on migration activity over the 1994-
2006 period, we present evidence that differential tax prices resulting from the assessment
growth cap have reduced in-migration.
JEL Classification: H71, H73, J61
Keywords: property tax, tax base erosion, regional migration, Michigan
Corresponding author:
Mehmet Serkan Tosun
Department of Economics
College of Business
University of Nevada, Reno
Mail Stop 0030
Reno, NV 89557
USA
E-mail: tosun@unr.edu
2
I. Introduction
With the passage of Michigan’s Proposal A in 1994, a property value assessment growth limit1
was implemented which restricts the growth of property valuation for tax purposes to the lesser of
rate of inflation or five percent.2 Since that time, every year local authorities are required to
report for each property within their jurisdiction the actual state equalized value (SEV)3 as well as
taxable value (TV). In 1994 each property’s state equalized value equaled its taxable value.
Through time, however, due to rapidly rising housing prices the growth of SEV has outpaced the
growth of TV, resulting in tax base erosion.4 Further, there is significant variation across
communities in the rate at which TV has fallen behind SEV. To illustrate the differences across
the state in tax base erosion, consider the ratio of state equalized value to taxable value (SEV/TV)
for all Michigan counties. According to the Michigan Department of Treasury State Tax
Commission, in 2006 the average SEV/TV for all counties was 1.3. Leelanau County had the
highest ratio of state equalized to taxable value (1.75) and Midland County had the lowest ratio of
state equalized to taxable value (1.15). Figure 1 shows how SEV/TV varies across Michigan
counties, but the story of tax base erosion across regions requires some explanation. Consider the
lake counties in the northwest region of the Lower Peninsula. This region experienced substantial
growth in prime lakefront property values, and this has led to significant tax base erosion. On the
other hand, Bay and Genesee counties in eastern Michigan have experienced very slow growth in
property values due to poor economic conditions, and thus have experienced relatively little tax
base erosion resulting from the taxable value cap. In contrast, the counties surrounding Grand
Rapids have experienced population growth and a substantial amount of new construction, which
has expanded the tax base. Like Bay and Genesee Counties, these counties have experienced
1 Such limitations have been referred to as “assessment growth caps”, “taxable value caps”, “assessment
growth limits”, or property tax assessment limits”. These terms are used interchangeably throughout the
manuscript.
2 Feldman, Courant, and Drake (2003) provide an excellent review of property taxation in Michigan.
3 SEV is equal to one half of estimated market value.
4 In 2006 housing prices began to fall across the state.
3
relatively little tax base erosion, but the underlying economic and demographic conditions are
very different. The Grand Rapids area has experienced significant new residential development
and this has served to keep tax base erosion in check. Bay and Genesee Counties, however, have
experienced long-term economic decline and flat housing prices. In this region, tax base erosion
is limited because housing prices have been stable or in decline. It should also be noted that over
the past two years average property values in Michigan have actually declined so that tax base
erosion is now being reversed.
According to a new report by the Lincoln Institute of Land Policy (Haveman and Sexton,
2008), property value assessment growth limits exist in 19 states and the District of Columbia, of
which 15 are imposed statewide.5 A number of these limitations were imposed in the 1970s and
1980s in the wake of the tax revolt. However, in response to rapidly rising property values across
the nation during the late 1990s and running through 2007, interest among policymakers in
assessment growth limits was renewed and several states have imposed new restrictions.6
Although property tax limitations generally and assessment growth caps specifically has received
the attention of researchers, several important dimensions of this particular type of property tax
limitation remain largely unexplored. In particular, the objective of this research is to determine
the degree to which tax base erosion resulting from the taxable value cap may have affected
regional migration patterns. A key feature of the Michigan’s taxable value cap is that upon sale
of a property, the taxable value increases to state equalized value. What has become known as
the “pop up” has resulted in a growing tax price differential between long-time and new property
owners. We therefore expect tax base erosion resulting from the taxable value cap to discourage
in-migration (due to the high cost of public services imposed on new property owners) and deter
out-migration (due to the substantial loss in the tax benefit associated with moving). To our
knowledge, the relationship between taxable value caps and migration decisions has not yet been
5 We note that Michigan’s assessment growth cap applies to all property types, whereas in many states the
assessment growth caps apply to residential property only.
6 Illinois and Washington are two recent examples.
4
examined by researchers. In this paper, we focus on in-migration although we also present some
analysis of out-migration as well.
The remainder of the paper is organized as follows. In the next section we provide an
outline of Michigan’s property value assessment growth limit and other pertinent property tax
policies. We then provide a review of the most relevant research regarding property value
assessment growth limits as well as related research on migration patterns. In section IV, we
present a simple framework to illustrate the determinants of tax base erosion and show how tax
prices are altered as result of tax base erosion. Section V presents the data and empirical analysis,
and section VI concludes.
II. The Property Tax in Michigan
During the latter part of the 20th century Michigan legislators as well as citizens via referenda
processes have implemented several measures designed to protect targeted property owners from
the property tax. In addition to the taxable value cap, it is important to also note other key
property tax policies such as the homestead exemption, the income tax credit, the mobile home
park exemption, and the Headlee amendment.7 Of particular relevance to the present study is the
homestead exemption which was implemented jointly with the taxable value cap under
Proposition A, and Headlee amendment which was imposed in 1978.
In Table 1 we present average statewide property tax millage rates from 1990 through
2006. Over this period, the only major change occurred in 1994 with the passage of Proposal A.
Proposal A reduced average millage rates for all homestead properties by 34 mills. This has
become known as the “homestead exemption.” The state government then added a 6 mill state
education tax and increased sales taxes and cigarette taxes to provide for the financing of k-12
public education. With the exception of this major policy shift, average millage rates across the
state have been stable, although it should be noted that property tax rates vary considerably across
7 These policies are clearly articulated and discussed in Feldman, Courant, and Drake (2003).
5
jurisdictions. This stability is largely due to the implementation of the Headlee Amendment in
1978, which restricts property tax revenue growth to the rate of inflation plus new construction.
Importantly, under Headlee any potential revenue increases beyond the limit resulting from
property value growth would require rate reductions to bring revenues into line with the revenue
growth restriction. This feature of the property tax revenue growth limit is known as the
“Headlee Rollback.” Local residents can, however, choose to exceed the Headlee limitation by
referendum. Once the taxable value cap was imposed, Headlee Rollbacks were greatly reduced in
numbers and magnitudes.
Each year the State of Michigan Department of Treasury provides estimates of tax
expenditures for all major sources of tax revenues. The single largest property tax expenditure
arises from the taxable value cap, and this is followed by the homestead exemption. Tax exempt
property is a distant third. These three property tax expenditures make up more than three
quarters of the total. As of 2007, the tax expenditure associated with the taxable value cap had
become so large that its repeal, holding total property tax revenues constant, would result in a
reduction in the statewide average statutory tax rate of 23 percent. In some counties the average
statutory tax rate could be reduced by more than 40 percent.
The Citizens Research Council (2001) was among the first organizations in Michigan to
highlight the growing differential between SEV and TV. In a statewide summary, they showed
how the ratio of taxable value to assessed value differed across regions and across types of
property. SEV/TV for commercial and industrial property in 2000 remained relatively high at
1.15 and 1.08, respectively. SEV/TV for agricultural, timber cutover and developmental
properties, however, had all risen above 1.45. SEV/TV for residential property had dropped to
about 1.22.8 While the Citizens Research Council report highlighted the potential shift in tax
8 The Citizens Research Council reports on changes in TV/SEV. We use SEV/TV to be consistent with our
empirical results and interpretations later in our paper. These numbers are simply the inverse of the
numbers reported in Citizens Research Council (2001). For purposes of interpreting estimation results,
in our empirical analysis we estimate the determinants of the inverse of TV/SEV, or SEV/TV.
6
burdens across property types, it did not provide a discussion of the broader implications
regarding the horizontal and vertical inequities that were beginning to emerge. Given that taxable
value caps have now existed in a number of states over a period of rapidly rising property values,
there is now a growing body of research that has examined the effects of property value
assessment growth limits, and it is on this work that we now focus our attention.
III. Literature Review
The early empirical research on property tax limits, including property value assessment growth
constraints tended to focus on determining the degree to which these emerging fiscal institutions
actually constrained property tax revenue growth.9 More recently, however, researchers have
turned their attention to the distributional consequences of property value assessment growth
limits. In 2006 the National Tax Journal devoted an entire issue to the property tax, and several
articles evaluate the potential consequences of property tax limitations, and in particular the
peculiar distributional consequences of property value assessment growth limits. We review
several key articles from this issue as well as some other recent research.
Dye, McMillen, and Merriman (2006) assess the implications of the recently imposed
assessment growth cap which was introduced in Cook County, Illinois. They demonstrate that a
taxable cap of this nature cannot be introduced without having comparative increases in taxes for
others. For example, a taxable value cap that protects residential owners as in Cook County,
Illinois will simply lead to increased taxes for industrial and commercial property owners. They
also show that homeowners with property that appreciates at a rate less than the cap will
experience higher rates to make up for those who are protected. Of particular relevance to the
present study, they indicate that the “extent that the cap reduces the property tax payments for
9 See Dye and McGuire and McMillen (2005), Mullins and Joyce (1996) and Skidmore (1999) for a review
of this literature.
7
rapidly growing areas… may discourage mobility, since the expanded exemption is lost when real
estate is sold, and, thus, may decrease the efficiency of the residential real estate market.”
Dye and McMillen (2007) extended work of Dye, McMillen and Merriman (2006) by
developing a formal theoretical framework to evaluate the effects of assessment caps on property
taxes. Two key conclusions are: 1) “reassessment upon sale can make it expensive for
homeowners to move, and may depress real estate markets”, and 2) the assessment limits can
“lead to higher taxes for some property owners whose assessments had been lowered.” To
illustrate, Dye and McMillen report that in Minnesota 78 percent of all residential homesteads
had to pay a higher tax as a result of taxable value cap than they would have had if taxable values
remained unrestricted.10 Dye and McMillen conclude, “taxes must rise for some properties in
order to provide relief to others” and “it is obvious…that properties in the cap-eligible group with
appreciation rates below the assessment cap will always come out behind.”
The recent work of Muhammad (2007) evaluates the horizontal and vertical inequities
resulting from the District of Columbia’s taxable value cap policy which was imposed in 2002.
District of Columbia’s tax cap policy (TCP) imposed a property value assessment growth limit set
at 12 percent annually between 2001 and 2005 and 10 percent thereafter. Over the 2001-2007
period, median homestead property values more than tripled in the District of Columbia
($128,499 to $400,050), and this represented an average annual increase of 20.8 percent. By
2007 the median property’s final taxable value differed from the actual market value by over 60
percent. Muhammad demonstrates the degree of horizontal and vertical inequity resulting from
the assessment growth limit using data on taxable value and estimated market value for all homes
in the District of Columbia. He finds that for homesteads with a value of $600,000, the effective
millage rate can be as high as $0.79 or as low as $0.01.
10 Dye and McMillen refer to a November 2006 presentation given by Mark Haveman and Paul Wilson at
the Lincoln Institute of Land Policy.
8
In 2008 the Lincoln Institute of Land Policy (Haveman and Sexton, 2008) published a
comprehensive report on property tax assessment limits and their use across the U.S. states. The
report covers the institutional/legal aspects of such limits, the implications for the tax base and
local government autonomy, equity issues, and the inefficiencies that arise. The report concludes
by offering potential alternatives to provide property tax relief to those in need. The authors
assert that property tax assessment limits are “…the least effective, least equitable, and least
efficient strategies available for providing tax relief.”
We have focused on the most recent literature evaluating the distributional consequences
of property value assessment growth limits. We note, however, that the work of Anderson and
McGuire (2007), Giertz (2006), Bowman (2006), and Youngman (2007), among others provide
excellent discussions of important issues surrounding the property tax and property tax
limitations, including assessment growth caps. There is also a body of research that has
examined the effects of assessment growth caps on mobility, and more generally there is an
extensive body of work that has examined interstate migration patterns. We discuss this research
next.
Mobility and Migration
In a recent study, Wasi and White (2005) examined the potential lock-in effect for
housing choice from Proposition 13, using data from 1970 to 2000, and found a significant effect.
The average tenure length of California homeowners increased by 0.66 years or six percent
relative to homeowners in Texas and Florida, which were chosen as comparison states. The
increase was as high as two to three years in places like San Francisco and San Jose, areas in
which stable homeowners received the highest subsidies from the assessment growth limit
embedded in Proposition 13.11 Ferreira (2004) also examined residential mobility after
Proposition 13, but focused on the two amendments that allowed transferability of the implicit tax
11 Nagy (1997) also examined the change in household mobility after California’s Proposition 13, using the
Annual Housing Surveys from 1975, 1978 and 1982. While he found evidence of a decline in mobility,
this was not significantly different from similar declines in other parts of the country.
9
benefits to a new home for head of households who are 55 or older. In a comparison with two
age groups, he found that mobility for the 55-year old group is about 25 percent higher than the
mobility for the 54-year old group. In summary, these studies present empirical evidence for a
link between California’s assessment growth limit and household mobility.
Another strand of the migration literature examines whether state fiscal variables affect
interstate migration patterns. The study by Cebula and Alexander (2006) examined the impact of
economic and non-economic factors on the net state in-migration between 2000 and 2004. They
found government spending on primary and secondary education, and state individual income tax
burden as positive and negative factors, respectively. Clark, Knapp and White (1996) found that
location characteristics, including fiscal variables, are important determinants of elderly interstate
migration, but the explanatory power of those variables declines with older age groups. Conway
and Houtenville (1998, 2001) used state-level data to examine determinants of elderly interstate
migration and found that fiscal variables play an important role in both in-migration and out-
migration. Somewhat related is the work of Farnham and Sevak (2002) who provided evidence
from the Health and Retirement Study (HRS) that moving households reduce their property tax
liability on average by $115. This suggests that property taxation plays a role in decisions to
move.
In this section we provided a discussion of the literature on property value assessment
growth caps and their distributional consequences, and the relationship between state/local fiscal
variables and mobility and migration. While the mobility research is most closely related to the
present work, to our knowledge no studies have examined the potential impact of the taxable
value cap on regional migration patterns. We now turn to the theoretical considerations of
assessment growth caps in the context of tax base erosion, tax price differentials between long-
time property owners and potential new property owners, and migration patterns.
IV. Theoretical Considerations—Tax Base Erosion and In-Migration
10
When the Michigan assessment growth limit was imposed in 1994, aggregate taxable value for
each individual property in the state equaled its state equalized value.12 From 1995 and on, the
growth of taxable value for any existing property not sold during the period is allowed to grow at
the rate of inflation as measured by the consumer price index or five percent, whichever is lower.
If the state equalized value of properties in a jurisdiction increase at a rate faster than the
inflation, the ratio of state equalized value (SEV) to taxable value (TV) will begin to rise. An
exception is when a property is sold: in this case TV returns to SEV. If one aggregates to the
county level, changes over time in SEV/TV will depend on the rates of change in property
turnover, property prices, and housing stock. However, property turnover, property prices, and
housing stock depend on economic and demographic factors.
We consider a range of socioeconomic variables that vary by county and over time may
determine changes in property turnover, property prices, and housing stock and thus tax base
erosion: Percent of population between the ages of 0 and 17 (0-17), percent of population
between 18-24 (18-24), percent of population between 25-44 (25-44), percent of population
between 45-64 (45-64)13, , the marriage rate (MARRIAGE), the divorce rate (DIVORCE), the
mortality rate (MORTALITY), the birth rate (BIRTH), per capita income (PCINC), the rate of
unemployment (UNEMP), manufacturing employment (MANU), government employment
(GOV), retail employment (RETAIL), and all other employment (OTHER).14
Age distribution determines housing turnover as well as housing demand, and therefore
may affect SEV/TV. For example, generally the young are more likely to move than middle-age
or older demographic groups. Major life changing events such as marriage, a birth, divorce, and
family loss determine both the rate of turnover and housing demand. To illustrate, the work of
12 State equalized value is defined as 50 percent of estimated market value.
13 The percent of population that is over the age of 65 is omitted in our empirical analysis to avoid perfect
multicollinearity.
14 An advantage of county level data is that annual series are available for a wide range of socioeconomic
and demographic variables, which are unavailable for smaller local government jurisdictions. It must be
acknowledged that a disadvantage of using county level data is that aggregation to the county level
abstracts from tax base erosion differentials across communities within a county.
11
Yu and Liu (2007) shows that divorce increases the number of homes occupied. While their
focus was on evaluating the environmental impact of divorce, there is a link to tax base erosion:
Changes in divorce rates can lead to changes in the demand for homes and housing turnover. We
therefore hypothesize that variables that capture major life events are negatively correlated with
SEV/TV. We also suggest that these life change variables are appropriate instruments and can be
used to identify the causal relationship between tax base erosion and migration activity. The
validity of these variables as instruments is evaluated and discussed later.
Economic conditions are also important. We expect counties with increasing income
and reductions in unemployment to experience greater tax base erosion. Also, the expected
relationship between employment and SEV/TV is mixed: Counties experiencing declines in
manufacturing employment and other employment categories may experience higher turnover
rates, but home values may also be falling. The expected impact on SEV/TV is ambiguous.
Tax base erosion results in higher tax prices for potential new property owners, relative to
long-time property owners. Prior to the imposition of the assessment growth cap in 1994, the
Headlee revenue growth limit required property tax rate rollbacks (Headlee rollbacks) whenever
revenues grew faster than the rate of inflation plus new construction. In fact, Headlee rollbacks
were very common prior to 1994, and this insured that all property owners enjoyed lower tax
rates as property values grew. After 1994, however, rollbacks were reduced both in number and
magnitude. Thus, over time new property owners were increasingly required to pay higher tax
prices than long-time property owners. Increasing tax prices for new property owners raises the
cost of public services for this group, and this has the potential to reduce in-migration. The
impact of the assessment growth cap, in part, depends on the time preferences and the expected
time horizon of property ownership. A new property owner with a low discount rate and who
expects to hold onto the property for many years may perceive a long-run benefit of the
assessment growth cap: He may be willing to pay more today knowing that he is protected from
future increases in taxes. On the other hand, new property owners with a high discount rate who
12
expects to own the property for only several years will not enjoy the potential tax benefits
associated with the assessment growth cap. Therefore, the overall effect of the assessment
growth cap on in-migration is indeterminate; however, we hypothesize that in-migration will be
reduced by the assessment growth cap particularly in areas where the tax prices for new property
owners could have been substantially lower in the absence of the assessment growth cap.15 All
else equal, in-migration could be affected in counties that experienced significant tax base
erosion.16 We now turn our attention to a critical issue in our evaluation of tax prices and in-
migration patterns—endogeneity.
In order to determine whether tax base erosion as measured by
it
it
TV
SEV reduces in-
migration, we must properly address the issue of endogeneity. While our interest is in
determining the effect of tax base erosion on in-migration, it must be acknowledged that
migration activity potentially affects turnover rates and thus
it
it
TV
SEV . In order to estimate the
effect of
it
it
TV
SEV on in-migration, we must identify instruments and use appropriate estimation
procedures to address endogeneity. Keeping in mind that our ultimate goal is to examine the
relationship between tax base erosion and in-migration, consider the following first-stage reduced
form tax base erosion equation:
()
itiitnitn
it
it LIFEbXa
TV
SEV
εµ
+++=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛))(ln()ln(ln (1)
15 Recall that in absence of the assessment growth cap, areas of high property value growth would have
experience tax rate rollbacks under the Headlee revenue growth limit.
16 Note that the difference in the tax rates for new comers and current residents is based on the average rate
for existing residents. The marginal difference between a new comer and a given current resident depends
on the length of time a current resident has lived in his/her home. In some cases, an existing resident who
has lived in his/her home since 1994 (the date the assessment growth limit was imposed) might enjoy a tax
bill that is half that of new comers for similarly valued properties.
13
where Xit is a vector of exogenous explanatory variables as described above, and LIFEit
represents a vector of life change event variables, µi are county fixed effects, and εit is the error
term. In some specifications we also include time indicator variables and county-specific time
trends.
The coefficients from this regression are generated from the within county changes in the
variables across Michigan counties over the 1994-2006 period. These estimates are provided for
the interested reader in the Appendix A. As we discuss in more detail in the empirical section,
the vector LIFEit and/or subsets of variables in this vector serve as appropriate instruments in our
econometric analysis.
While we are also interested in the degree to which the taxable value cap has affected
out-migration, we focus on in-migration because our instruments are most compelling in the case
of in-migration (the in-migration decision is made by those not currently living within a given
jurisdiction). In order to estimate the impact of tax base erosion and on in-migration, we must
identify appropriate instrument(s) in order to test for endogeneity, and if present use appropriate
estimation procedures. Appropriate instruments must be correlated with SEV/TV but not be
directly correlated with in-migration. Our analysis suggests that life change events as
characterized by MARRIAGE, BIRTH, DIVORCE and MORTALITY are appropriate
instruments. The identification strategy is based on the idea that these life change variables
trigger turnover rates and demand for housing, thereby affecting the degree of tax base erosion,
but these variables are not direct determinants of in-migration. The analysis also shows that at
least some of these variables are strong instruments in that they are statistically significant and
meaningful determinants of tax base erosion, but these variables do not directly determine in-
migration. Also, as discussed in greater detail later, we use alternative combinations of the
instruments to examine the sensitivity of our findings. Although we find no empirical evidence to
support the following notion, one could argue that, particularly, decisions to marry, divorce or
child birth are influenced by tax base erosion or other economic factors that influence both tax
14
base erosion and in-migration decisions that are not controlled for in our econometric analysis.17
If this is the case, these variables would not serve as appropriate instruments.
To examine the appropriateness of our instruments, we use the Sargon’s Test of
Overidentifying Restrictions. Using different sets of instruments, in each case we fail to reject
the null hypothesis that our instruments are valid. These examinations suggest that we have
identified appropriate instruments for the measure of tax base erosion for in-migration analysis.18
We also conduct similar tests in the context of out-migration, and again these examinations
indicate that we have identified appropriate instruments, although we acknowledge we a have a
weaker theoretical case with out-migration.
Given that we have appropriate instruments, we next conduct a Hausman test of
endogeniety. If the Hausman test of endogeneity indicates that endogeneity is a concern, we must
use a simultaneous equation procedure. To complete the Hausman test, the residual generated
from equation (2) is included as an explanatory variable in the in-migration regression. If the
measure of tax base erosion is endogenous, then the coefficient on this residual should be
significantly different from zero. The Hausman test for examining the endogeneity of SEV/TV
suggests that the null hypothesis that SEV/TV is exogenous is rejected. We therefore proceed
with estimating the in-migration equation with a correction for simultaneity using a two-stage
least squares procedure, but for comparison we also present regressions without the correction for
simultaneity.
Before presenting the specification of the second stage regression some additional
explanation is required. The specification we use is guided by the following framework. Let in-
migration be influenced by π, the tax price faced by potential new property owners and a vector
X, a series of economic and demographic factors that vary within counties over time that may
17 Our analysis also suggests that these instruments are appropriate for out-migration as well.
18 Specifically, we examined the following combinations of instruments: 1) DIVORCE, MARRIAGE,
MORTALITY, and BIRTH; 2) MORTALITY, BIRTH, MARRIAGE; 3) MORTALITY, BIRTH; and 4)
MORTALITY. We also included TV(-1) as an instrument in a different set of estimations. In all
combinations, the null hypothesis that our instruments are valid is not rejected.
15
affect in-migration. Our primary interest is in determining the degree to which in-migration is
affected by the tax price faced by the potential in-migrant, which depends on the degree of tax
base erosion as measured by SEV/TV However, the tax price faced by new comers is also
determined by the statutory tax rate as well as the degree to which housing prices, P, differ from
two times SEV. 19 Consider the following equation to illustrate the tax price faced by a potential
new property owner. Let πit be the tax price faced by a property owner in county i in period t:
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
⋅
=
it
it
it
it
itit TV
SEV
SEV
P
r2
π
(2)
where the rit is the statutory tax rate and Pit is the ratio of the average housing price in the county.
Our core hypothesis is that, holding r and P/(2⋅SEV) constant, tax base erosion as measured by
SEV/TV leads to a higher tax price for new comers and thus deters in-migration.
This tax price framework informs the econometric specification because equation (2)
makes clear that the tax base erosion (SEV/TV) component of the tax price will need to be
isolated to estimate its effect on in-migration. Previous research evaluating the role of tax policy
on migration patterns has used several different specifications. Conway and Rork (2006), for
example, use the elderly migration rate as the dependent variable in a linear specification. On the
other hand, Conway and Houtenville (2001) use migration flow as the dependent variable in a
logarithmic specification. Cebula (1990) and Cebula and Alexander (2006) also use migration
flow as the dependent variable but in a linear framework. In the present study, it is convenient to
use a semi-logarithmic specification in order to separate the components of π and thus isolate the
effect of tax base erosion associated with the taxable value cap. The core in-migration
specification is therefore:
itit
it
it
it
it
itiit eXe
TV
SEV
predictedd
SEV
P
crbaInmig ++
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
+
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
⋅
++= )ln('ln
2
ln)ln( (3)
19 Recall that SEV is defined as one half of market value.
16
where Inmigit is the migration rate for county i in period t. In all regressions we control for
county fixed effects (ai) and a vector of exogenous factors, X it, that vary within counties over
time that may affect in-migration. In different sets of regressions we include a set of time
indicator variables and county-specific time trends to control for rit,
it
it
SEV
Pand other possible
omitted trends and factors that might be correlated with tax base erosion.20 The variable
predicted
it
it
TV
SEV represents the predicted value of (
it
it
TV
SEV ) generated from the first-stage
regression and its effect on in-migration measured by the coefficient d.
We use the two-stage generalized method of moments estimation procedure which is
robust to both heteroskedasticity and autocorrelation. The interested reader may review the first
stage tax base erosion regression results in the Appendix A, but these are not discussed in detail.
We note, however, that two of the variables that indicate major life events (MARRIAGE and
MORTALITY) are statistically significant three of four specifications. The exception is in the
regression in which when both the time indicator variables and county-specific time trends are
included in the specification. Higher rates of marriage and mortality result in less tax base
erosion, perhaps because the rate of home purchases and property turnover is higher. These
estimates provide some assurance that these variables will serve as strong instruments. Further,
an F-test shows that the combination of the instruments is jointly significant in all regressions,
and is thus effective as a set.
Previous studies of migration flows in an interstate context have focused on the role of
amenities such as weather conditions, etc… which are largely fixed over our timeframe. We
acknowledge that amenities play an important role in intra-state migration patterns. However,
20 Unfortunately, we do not have a reliable time varying county level measure of Pit. Also, in specifications
available from the authors upon request we explicitly include the statutory tax rate (r) as an explanatory
variable, but the inclusion of this variable has virtually no effect our core findings, and r is never
statistically significant. We omit r in the specifications presented due to concerns about endogeneity of the
tax rate variable.
17
such amenities are largely fixed over the period of analysis and therefore we control for amenity
factors with county fixed effects. In addition, in some specifications we included time indicator
variables and county-specific time trends to examine the robustness of our findings. This
specification allows us to focus on conditions that have changed within counties over time that
may affect in-migration: Socio-economic and fiscal conditions. Additional details regarding the
methods of analyses used are discussed next.
V. Empirical Analysis
The data are a panel of 982 observations21 that include all 83 Michigan counties for years 1994
through 2006. We estimate fixed effects regressions using a cluster approach in which we cluster
our standard errors at the county level to address temporal autocorrelation. Cluster-standard
errors perform well when the number of clusters is reasonably large (Bertrand et. al., 2004;
Kezdi, 2004). However, note that we also used as an alternative estimation approach spatial
econometric methods in the context of panel data. In regressions that are not presented but are
available upon request from the authors, we find that these estimates are qualitatively similar to
those presented here.22 Below, we present the regression results which examine the relationship
between tax base erosion and in-migration.
Summary statistics for these and all variables used in the estimations are presented in
Table 2, and detailed definitions and sources of all variables used in the analysis are shown in the
Appendix B. The regressions in columns 1-4 of Table 3 present the estimates without addressing
21 We do not have a complete panel due to some missing/unreported data in our employment variables. In
addition the number of observations decreases to 812 in the two-stage least squares estimates because of
missing data in the life change variables.
22 While spatial considerations are important, we do not focus on this issue for several reasons. First, the
fixed effects account for any spatial patterns that are fixed over time. Second, we believe temporal
correlation is a more serious concern. Last, in our in-migration regressions endogeneity is a potentially
serious econometric issue. We therefore believe a fixed effects procedure that addresses temporal serial
correlation, heteroskedasticity, and endogeneity is the best overall econometric approach in the context of
this study.
18
endogeneity, whereas columns (5-8) present the two-stage least squares regressions, using as
instruments the major life change variables: MARRIAGE, BIRTH, DIVORCE, MORTALITY.
The results presented in columns (1-4) consistently show that the
it
it
TV
SEV variable affects
in-migration negatively. These regressions provide evidence that tax base erosion is inversely
correlated with in-migration. However, the standard error of the tax base erosion variable is
somewhat sensitive to the inclusion of time indicator variables and county-specific time trends.
it
it
TV
SEV maintains statistical significance when either time indicator variables or county-specific
time trends are included, but the coefficient falls below the standards of statistical significance
when both are included. Regardless, the magnitude of the coefficient is reasonably consistent
across this set of regressions. It is not too surprising that once statewide trends and county-
specific trends are controlled for that precision of the estimates fall. Overall, these estimates
suggest that counties with greater tax base erosion experience less in-migration, controlling for
other factors. While the inclusion of count-specific time trends reduces the potential for omitted
variable bias and endogeneity, these estimates may still be plagued by simultaneity. We therefore
present a series of two-stage least squares estimates in columns 5-8.
The two-stage least squares estimates are very similar accept the magnitude of the
coefficient on
it
it
TV
SEV is now considerably larger, and indication that any potential bias resulting
from simultaneity appears to be toward zero. When both time indicator variables and county-
specific time trends are included in the regressions, the coefficient on
it
it
TV
SEV again falls below
typical standards statistical significance.
19
To provide a sense of the estimated effects of tax base erosion on in-migration, consider
the following calculation. Based on the coefficient estimate on ⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
it
it
TV
SEV
ln from column (6), a
one standard deviation increase in
it
it
TV
SEV from its mean value is predicted reduce in-migration
by about 32 percent.23 A one standard deviation increase in
it
it
TV
SEV represents an 17 percent
differential in tax price between a older protected homeowner and a potential new homeowner.
In this context, the estimated in-migration response seems quite large. The coefficient estimates
generated from the regressions that do not correct for endogeneity (columns 1-4) are much
smaller, however. Based on these coefficients, a one standard deviation increase in
it
it
TV
SEV is
predicted to reduce in-migration between 8 and 15 percent. Although our examinations suggest
that our instruments are valid and that endogeneity is present, we still have concerns about the
instruments. Furthermore, as illustrated above the instrumental variables approach generates
estimates that seem too large. For these reasons, we tend to favor the more conservative OLS
estimates.
Turning to the control variables, we see that the coefficients on per capita income are
positive and consistently significant. The coefficients on unemployment are negative but not
always significant. Similarly, some of the coefficients on the employment variables, particularly
retail and other employment, are consistently negative and sometimes significant. While a high-
concentration of a particular population age group could be a contributing factor to county in-
23 From column (6), the coefficient on ln(SEV/TV) is 0.12. The mean value and standard deviation of
SEV/TV are 1.22 and 0.167, respectively. The average annual in-migration rate is 0.048. Therefore, a one
standard deviation increase in SEV/TV is predicted to reduce in-migration by the coefficient estimate
multiplied by the (initial ln(SEV/TV) minus new ln(SEV/TV)), or 0.12x0.127=0.0152. This is roughly a
32 percent reduction in in-migration.
20
migration (e.g. retiree in-migration to counties with more elderly population), we generally do not
find the demographic variables to be significant determinants of in-migration.
Robustness
To examine the robustness and sensitivity of the core finding, we consider a range of
issues. First, in regressions not reported but are available from the authors upon request, we
included the aggregate statutory tax rate as an additional explanatory variable in the regressions.24
None of our core conclusions are altered with the inclusion of the statutory rate variable. We also
examined the effect of
it
it
TV
SEV interacted with the statutory rate variable, treating the interaction
term as endogenously determined. In these regressions we find that the effect of tax base erosion
on in-migration is most pronounced in areas with relatively high tax rates.25 In other regressions
we use lagged explanatory variables; again our core finding is unaltered. We examined the
sensitivity of the two-stage least squares estimates by using different combinations of
instruments. These findings are generally consistent with those presented here. We estimated
two alternative specifications that have been used in previous research on migration: 1) a linear
specification; and 2) a log-log specification. Again, our core findings are robust to these
considerations. Finally, we examined the effect of tax base erosion on in-migration in the context
of a system of equations that include in-migration, out-migration, and new construction. We
estimate this system using a three-stage least squares estimation procedure, and these are
available from the authors upon request.26 The in-migration results are very similar to those
presented here.
24 We exclude the statutory rate variable in the regressions we present because we recognize the difficulties
associated with the potential endogeneity of property tax rate changes and migration decisions.
25 Conversely, one could interpret this result as indicating the higher tax rates reduce in-migration more in
areas where tax base erosion is highest. These results are available upon request from the authors.
26 One drawback of this econometric approach, however, is that we are unable to address autocorrelation
with the clustering technique. Thus, in these estimates the standard errors are biased downward, leading to
inflated t values.
21
Although the focus of this research is on the relationship between tax base erosion and in-
migration, in Appendix C we also present a series of out-migration regressions. This set of
regressions is analogous to those in Table 3 except that the dependent variable is out-migration.
We are more cautious in our interpretation of these findings because the life change variables we
use as instruments are arguably not purely exogenous to out-migration decisions.27 The set of
regressions reveals no consistent relationship between out-migration and tax base erosion,
although we note that our most rigorous specification in which we correct for simultaneity and
control for time indicator variables and county-specific time trends indicates that tax base erosion
has reduced out-migration.
VI. Conclusion
The work presented in this paper evaluates the degree to which tax base erosion and
corresponding differential in tax prices between potential property owners and long-time property
owners has affected in-migration. To our knowledge, this is the first such formal analysis of the
migration effects of differential tax base erosion resulting from a property value assessment
growth cap. Other factors such as changing economic circumstances have arguably played a
much stronger role in location decisions. On the margin, however, home seekers appear to be
influenced by the peculiar fiscal environment associated with property tax assessment growth
limits.
Over the past two and half years, home values have fallen across Michigan, but the rate
of decline varies across regions. The analysis presented here may also be useful for
understanding the differential impacts of falling home values in terms of reversing the tax base
27 Although we are cautious with our interpretation, note that Sargon’s Test of Overidentifying Restrictions
suggests that the instruments are valid for the out-migration analysis. The Hausman test provides evidence
of endogeneity in the context of out-migration. Thus, instrumental variable procedures are appropriate.
22
erosion.28 Legislators, aware of the high effective property tax rates associated with tax base
erosion caused by the taxable value cap, have considered implementing a new income tax credit
for new homeowners to offset the “pop up” property tax increase. While not an ideal solution,
the debate suggests that legislators are becoming aware of the implications of the taxable value
cap. In Michigan, however, substantial revisions or possible repeal of the taxable value cap
would require voter approval.
One implication of falling home values is that the distortions caused by the taxable value
cap are being reduced. It would seem then that analysis of such distortions is less relevant.
However, our examination sheds light on the effects of the taxable value cap at an opportune
time. Reductions in property values mean that long-time homeowners who might otherwise have
much to lose by the elimination of the taxable value cap may not object to its repeal in the current
environment. In a political economy framework, there is no time better than now to consider the
elimination of poorly conceived constraints such as the taxable value cap. In 1994, the taxable
value cap was approved by voters under the premise that it would further limit property taxation.
However, it is very likely that many voters did not fully understand that such a constraint would
ultimately lead to significant tax reductions for some at the expense of higher taxes for new
property owners. Under current conditions, fewer voters have immediate losses resulting from a
change in the law, and its repeal would avoid potential future distortions.
There are two limitations of this analysis that should be acknowledged. First, due to data
limitations the unit of analysis is conducted at the county level and therefore necessarily abstracts
from tax base erosion differentials that may exist in more narrowly defined jurisdictions. One
advantage of county level data, however, is that we are able to incorporate into the analysis more
detailed socio-economic variables that are not available for smaller jurisdictions on an annual
basis. A second limitation is that it is possible that assessment growth caps may have a
28 Since 2007, Michigan counties have experienced significant losses in property value and there has been a
reversal such that SEV/TV is now decreasing.
23
capitalization effect on property values, but the direction and degree of capitalization is unclear.
For property owners who expect to hold onto their property for many years, the taxable value cap
represents a potential long-run benefit: They may pay higher taxes today, but tax liabilities are
guaranteed to grow at the rate of inflation and will not exceed five percent.29 The capitalization
of tax payments, including lower future payments, for the duration of tenure in the home may
lead to a higher willingness to pay for such property owners. However, for property owners who
do not expect to hold on to property for an extended number of years or are not able to anticipate
how long he or she may own property, the taxable value growth cap merely represents higher
overall tax prices (O’Sullivan, Sexton, and Sheffrin, 1995). These property owners may very
well be willing to pay less for property under the taxable value growth cap. In addition, if
mobility is reduced because of the taxable value growth cap as suggested by previous studies,
fewer homes might be available on the market and this may lead to higher housing prices. It
should also be noted that because of the Headlee revenue growth limit, it is likely that the overall
revenue growth and thus average tax prices would not be significantly different in the absence of
the taxable value cap.
Despite these limitations, the current work provides new evidence regarding the
implications of tax base erosion, emerging differential in tax prices between potential new
property owners and long-time property owners and in-migration. Future work aimed at
explicitly identifying the implications of the taxable value cap on the redistribution of property
tax payments across demographic groups will be a valuable contribution.
29 In an environment of rising home values, this may be very comforting. However, recently home values
have actually fallen, and taxable values are by law increasing by the rate of inflation. An exception is for
homeowners whose state equalized value no longer exceeds taxable value: For these properties, taxable
value falls with state equalized value.
24
References
Anderson, Nathan B., and Therese J. McGuire. 2007. “An Unfettered Property Tax in Illinois,”
Working Paper.
Bertrand, Marianne, Esther Duflo and Sendhil Mullainathan. 2004. “How Much Should
we Trust Difference-in-Difference Estimators?” Quarterly Journal of Economics, 119, 1: 249-75
Bowman, John H. 2006. "Property Tax Policy Responses to Rapidly Rising Home Values:
District of Columbia, Maryland, and Virginia." National Tax Journal, LIX.3: 717-33.
Cebula, Richard J. 1990. “A Brief Empirical Note on the Tiebout Hypothesis and State Income
Tax Policies,” Public Choice, 67 (1): 87-89.
Cebula, Richard J. and Gigi M. Alexander. 2006. “Determinants of Net Interstate Migration,
2000-2004,” Journal of Regional Analysis and Policy, 36 (2): 116-123.
Citizens Research Council. 2001. CRC Memorandum Number 1058: “The Growing Difference
between State Equalized Value and Taxable Value in Michigan,” Livonia, Michigan: Citizens
Research Council.
Clark, David E., Thomas A. Knapp and Nancy E. White. 1996. “Personal and Location-Specific
Characteristics and Elderly Interstate Migration,” Growth and Change, 27: 327-351.
Conway, Karen S. and Andrew J. Houtenville. 1998. “Do the Elderly Vote with Their Feet”?
Public Choice 97, (1): 63-85.
Conway, Karen S. and Andrew J. Houtenville. 2001. “Elderly Migration and State Fiscal Policy:
Evidence from the 1990 Census Migration Flows,” National Tax Journal, 54 (1): 103-123.
Conway, Karen S, and Jonathon Rork. 2006. “State" Death" Taxes and Elderly Migration-The
Chicken or the Egg?” National Tax Journal, 59 (1): 97–128.
Dye, Richard F., Therese J. McGuire, and Daniel P. McMillen. 2005. “Are Property Tax
Limitations more binding over time?” National Tax Journal, 58(2): 215-25.
Dye, Richard F. and Daniel P. McMillen. 2007. “The Algebra of Tax Burden Shifts from
Assessment Limitations”. Lincoln Institute of Land Policy.
Dye, Richard F., Daniel P. McMillen, and David F. Merriman. 2006. "Illinois' Response to
Rising Residential Property Values: An Assessment Growth Cap in Cook County." National Tax
Journal, LIX.3: 707-16.
Feldman, Naomi E., Paul N. Courant, and Douglas Drake. 2003. “The Property Tax in
Michigan.” in Michigan at the Millennium, edited by Charles L. Ballard, Paul N. Courant,
Douglas C. Drake, Ronald C. Fisher, and Elisabeth R. Gerber. East Lansing: Michigan State
University Press, 577-602.
Farnham, Martin and Purvi Sevak. 2002. “Local Fiscal Policy and Retiree Migration: Evidence
from the Health and Retirement Study,” Hunter College Department of Economics Working
Papers 02/7, Hunter College, Department of Economics.
25
Ferreira, Fernando. 2004. “You Can Take It with You: Transferability of Proposition 13 Tax
Benefits, Residential Mobility, and Willingness to Pay for Housing Amenities.” Working Paper
72, Center for Labor Economics, University of California, Berkeley.
Giertz, J Fred. 2006. "The Property Tax Bound." National Tax Journal, LIX.3: 695-705.
Haveman, Mark, and Terri Sexton. 2008. “Property Tax Assessment Limits: Lessons from
Thirty Years of Experience.” Lincoln Institute of Land Policy.
Kezdi, Gabor. 2004. “Robust Standard-Error Estimations in Fixed-Effect Panel Models.”
Hungarian Statistical Review, 9: 95–116
Muhammad, Daniel. 2007. “Horizontal Inequity, Vertical Inequity and the District of
Columbia’s Property Assessment Limitation.” Presented at the National Tax Association’s 100th
Annual Conference on Taxation.
Mullins, Daniel R. and Philip G. Joyce. 1996. “Tax and Expenditure Limitations and State
Fiscal Structure: An Empirical Assessment,” Public Budgeting and Finance, 16: 75-101.
Nagy, John. 1997. “Did Proposition 13 Affect the Mobility of California Homeowners”? Public
Finance Review, 25 (1): 102-116.
O’Sullivan, Arthur, Terri A. Sexton and Steven M. Sheffrin. 1995. “Property Taxes, Mobility and
Home Ownership,” Journal of Urban Economics, 37: 107-129.
Skidmore, Mark. 1999. “Tax and expenditure limitations and the fiscal relationships between
state and local governments, Public Choice, 99(1/2): 77-102.
Wasi, Nada and Michelle J. White. 2005. “Property Tax Limitations and Mobility: Lock-in Effect
of California’s Proposition 13,” Brookings-Wharton Papers on Urban Affairs: 2005: 59-97.
Youngman, Joan. “The Variety of Property Tax Limits: “Goals, Consequences, and
Alternatives.” State Tax Notes, November, (2007): 541-57.
Yu, Eunice, and Jianguo Liu. 2007. "Environment Impacts of Divorce." PNAS, 104.51: 20629-
34.
26
Figure 1
Ratio of State Equalized Value to Taxable Value, 2006
Source: State of Michigan Department of Treasury.
Legend
1.00 - 1.27
1.28 - 1.38
1.39 - 1.47
1.48 - 1.55
1.56 +
27
Table 1
Average Statewide Millage Rates
Calendar Homestead Nonhomestead All
Year Property Property Property
1990 57.17 57.17 57.17
1991 57.34 57.34 57.34
1992 58.09 58.09 58.09
1993 56.64 56.64 56.64
1994 30.22 48.17 38.19
1995 31.00 48.79 38.88
1996 31.36 49.54 39.32
1997 31.36 49.63 39.25
1998 31.43 49.68 39.27
1999 31.40 49.76 39.16
2000 31.54 50.10 39.32
2001 32.12 50.72 39.78
2002 32.60 51.00 40.17
2003 31.52 50.06 39.00
2004 32.70 51.20 40.00
2005 32.60 51.38 39.88
2006 32.66 51.38 39.96
Source: All Property Millage Rates from State Tax Commission except 1994; CY 1994 All
Property Rate and Homestead and Non-homestead millage rates from the Tax Analysis Division,
Michigan Department of Treasury.
28
Table 2
Summary Statistics about Michigan Counties
Number
of
Obs. Mean Std. Dev.
In-migration/Population 982 0.048 0.026
Out-migration/Population 982 0.045 0.025
SEV/TV 982 1.23 0.168
Population 982 11,950 27,860
Percent o Population Between 0-17 982 24.32 2.748
Percent of the Population between
the Ages of 18 and 24 982 9.03 3.61
Percent of the Population between
the Ages of 25 and 44 982 27.27 2.99
Percent of the Population between
the Ages of 45 and 64 982 24.15 2.88
Percent of Population Over 65 982 15.22 4.01
Marriage 982 14.81 3.83
Birth 812 11.48 2.01
Divorce 982 8.19 2.24
Mortality 982 9.95 2.44
Per Capita Income* 982 24,011 3,700
Unemployment Rate 982 8.62 1.78
Manuf, Emp. Per Capita 982 0.085 0.131
Gov’t Emp. Per Capita 982 0.080 0.058
Retail Emp. Per Capita 982 0.083 0.058
Other Emp. Per Capita 982 0.268 0.159
Statutory Property Tax Rate 982 36.05 4.60
*Monetary values in thousands of nominal dollars
29
Table 3
In-migration Regression Results
(t-statistics or z-statistics in parentheses)
Dependent Variable: In-migration Rate
Independent Variable OLS Fixed
Effects
Cluster
(1)
OLS Fixed
Effects
Cluster
(2)
OLS Fixed
Effects
Cluster
(3)
OLS Fixed
Effects
Cluster
(4)
2SLS Fixed
Effects
Clustera
(5)
2SLS Fixed
Effects
Clustera
(6)
2SLS Fixed
Effects
Clustera
(7)
2SLS Fixed
Effects
Clustera
(8)
Ln(SEV/TV) -0.04*** -0.03* -0.06*** -0.04 -0.20** -0.12* -0.22* -0.14
(-3.208) (-1.723) (-2.699) (-1.305) (-2.391) (-1.927) (-1.663) (-1.444)
Ln(Per Capita Income) 0.04071*** 0.07452*** 0.08316*** 0.12189*** 0.10187** 0.07865*** 0.08054* 0.04447***
(3.001) (3.980) (2.887) (3.106) (2.397) (4.208) (1.708) (3.691)
Unemployment Rate -0.001** -0.001* -0.001 -0.001 -0.002*** -0.001* -0.001 0.000
(-2.528) (-1.892) (-1.367) (-1.496) (-3.216) (-1.845) (-1.212) (0.263)
Ln(Manuf. Emp. Per Capita) 0.001 -0.002 -0.004 -0.006 -0.008 -0.005 -0.006 -0.002
(0.282) (-0.701) (-0.857) (-1.135) (-1.399) (-1.576) (-1.178) (-0.846)
Ln(Gov’t Emp. Per Capita) 0.004 0.000 -0.010 -0.011 0.012 0.003 0.011 -0.005
(0.513) (0.00680) (-0.664) (-0.753) (0.950) (0.322) (0.541) (-0.435)
Ln(Retail Emp. Per Capita) -0.008** -0.005 -0.011** -0.010 -0.014*** -0.009** -0.004 0.000
(-2.545) (-1.000) (-2.257) (-1.526) (-3.731) (-2.233) (-1.217) (0.0899)
Ln(Other Emp. Per Capita) -0.020*** -0.023*** -0.024** -0.023* -0.019*** -0.020*** -0.010 -0.003
(-3.200) (-2.922) (-2.277) (-1.812) (-3.074) (-4.105) (-1.351) (-0.448)
Ln(Age under 18) 0.016 0.005 0.044 -0.069 -0.063 -0.062* 0.071 -0.095**
(1.015) (0.285) (0.735) (-0.981) (-1.133) (-1.776) (1.435) (-2.233)
Ln(Age 18-24) -0.008 -0.007 -0.011 -0.013 0.031 0.001 0.045 -0.017
(-0.873) (-0.743) (-0.464) (-0.644) (1.170) (0.0765) (1.026) (-1.147)
Ln(Age 25-44) 0.006 0.026 -0.016 -0.057 0.025 0.010 -0.029 -0.070***
(0.353) (1.465) (-0.490) (-1.433) (0.604) (0.356) (-0.697) (-2.678)
Ln(Age 45- 64) -0.011 0.036** -0.030 -0.017 0.022 0.006 0.044 -0.094*
(-0.712) (2.103) (-0.618) (-0.391) (0.670) (0.166) (0.703) (-1.937)
Adjusted R2 0.904 0.908 0.920 0.923 -0.121 0.383 -0.791 0.002
Time Indicator Variables No Yes No Yes No Yes No Yes
County-specific Time Trend No No Yes Yes No No Yes Yes
N 982 982 982 982 812 812 812 812
a Instruments: Marriage, Birth, Divorce, Mortality
Notes: All regressions include county fixed effects; * Indicates significance at the 90 percent confidence level for a two-tailed test.; ** Indicates significance at the 95
percent confidence level for a two-tailed test; *** Indicates significance at the 99 percent confidence level for a two-tailed test.
30
Appendix A
Tax Base Erosion Regression Results
(t-statistics or z-statistics in parentheses)
Dependent Variable: Ln(SEV/TV)
Independent Variable OLS Fixed
Effects Cluster
(1)
OLS Fixed
Effects Cluster
(2)
OLS Fixed
Effects Cluster
(3)
OLS Fixed
Effects Cluster
(4)
Ln(Per Capita Income) 0.46*** 0.11 0.30** 0.01
(6.221) (1.054) (2.470) (0.104)
Unemployment Rate -0.00637*** -0.00729*** -0.00228 -0.00033
(-2.859) (-2.639) (-1.087) (-0.140)
Ln(Manuf. Emp. Per Capita) -0.038* -0.018 -0.033* -0.011
(-1.710) (-0.866) (-1.972) (-0.689)
Ln(Gov’t Emp. Per Capita) 0.063 0.053 0.073 -0.010
(1.179) (1.218) (1.244) (-0.184)
Ln(Retail Emp. Per Capita) -0.043** -0.005 -0.005 0.009
(-2.350) (-0.187) (-0.419) (0.416)
Ln(Other Emp. Per Capita) 0.010 0.007 -0.031 -0.010
(0.311) (0.236) (-1.219) (-0.354)
Ln(Age under 18) -0.430** -0.360** 0.172 -0.038
(-2.533) (-2.355) (0.498) (-0.169)
Ln(Age 18-24) 0.286*** 0.145** 0.336*** 0.092
(3.555) (2.207) (3.706) (1.212)
Ln(Age 25-44) 0.282 0.116 -0.080 -0.005
(1.360) (0.561) (-0.313) (-0.0211)
Ln(Age 45- 64) 0.233 -0.363* 0.350 -0.329*
(1.383) (-1.939) (1.621) (-1.688)
Ln(Marriage) -0.025 -0.023** -0.015 -0.018**
(-1.636) (-2.135) (-1.295) (-2.021)
Ln(Birth) -0.045** -0.051*** -0.018 -0.004
(-2.099) (-2.767) (-1.138) (-0.309)
Ln(Divorce) -0.005 -0.008 0.000 -0.003
(-0.890) (-1.571) (0.0732) (-0.697)
Ln(Mortality) -0.013 0.002 -0.016 0.013
(-0.644) (0.0853) (-1.039) (1.059)
Adjusted R2 0.931 0.950 0.975 0.983
Time Indicator Variables No Yes No Yes
County-specific Time Trend No No Yes Yes
N 812 812 812 812
a Instruments: ln(Marriage), ln(Birth), ln(Divorce), ln(Mortality)
Notes: All regressions include county fixed effects; * Indicates significance at the 90 percent confidence level
for a two-tailed test.; ** Indicates significance at the 95 percent confidence level for a two-tailed test; ***
Indicates significance at the 99 percent confidence level for a two-tailed test.
31
Appendix B
Definitions and Sources of Variables
Variables Definitions Source
In-migration Migration into a county (based on tax return
data)/population CENSUS
and SOI
Out-migration Migration out of a county (based on tax return
data)/population CENSUS
and SOI
SEV County aggregate state equalized value, which is equal ½
of estimated market value MDT
TV
County aggregate taxable value which is allowed to grow
at the rate of inflation unless a property is sold (selling
property returns taxable value to state equalized value)
MDT
SEV/TV Ratio of aggregate state equalized value to aggregate
taxable value in the county MDT
Population County population plus in-migration minus out-migration CENSUS
and SOI
Percent of Population
Between 0-17 Percent of the population between the ages of zero and 17 CENSUS
Percent of Population
between 18-24 Percent of the population between the ages of 18 and 24 CENSUS
Percent of Population
between 25-44 Percent of the population between the ages of 25 and 44 CENSUS
Percent of Population
between 45-64 Percent of the population between the ages of 45 and 64 CENSUS
Percent of Population Over
65 Percent of the population aged 65 and older CENSUS
Marriage Marriage rate per 1,000 population MDCH
Birth Birth rate per 1,000 population MDCH
Divorce Divorce rate per 1,000 population MDCH
Mortality Mortality rate per 1,000 population MDCH
Per Capita Income Per capita income BEA
Unemployment Rate The rate of unemployment BEA
Manuf, Emp. Per Capita Employment in manufacturing per capita BEA
Gov’t Emp. Per Capita Employment in government per capita BEA
Retail Emp. Per Capita Employment in retail per capita BEA
Other Emp. Per Capita Employment in all other categories per capita BEA
Per Capita New
Construction The value of new construction per capita BEA
Statutory Property Tax
Rate= Effective Rate for
New Resident
Average statutory rate in the county for all overlying
taxing jurisdictions (township, village, city, community
college, and county)
MDT
Sources:
BEA: Bureau of Economic Analysis, Regional Accounts Data: http://www.bea.doc.gov/bea/regional/reis/
CENSUS: U.S. Census Bureau, County Population Estimates: http://eire.census.gov/popest/estimates.php
MDT: Michigan Department of Treasury
MDCH: Michigan Department of Community Health
SOI: Statistics of Income, Internal Revenue Service, US Federal Government30
30 The Internal Revenue Service collects detailed annual information on the number of in-migrants and out-
migrants by county on an annual basis. These data are based on address and changes provided by tax filers.
32
Appendix C
Out-migration Regression Results
(t-statistics or z-statistics in parentheses)
Dependent Variable: Out-migration Rate
Independent Variable OLS Fixed
Effects
Cluster
(1)
OLS Fixed
Effects
Cluster
(2)
OLS Fixed
Effects
Cluster
(3)
OLS Fixed
Effects
Cluster
(4)
2SLS Fixed
Effects
Clustera
(5)
2SLS Fixed
Effects
Clustera
(6)
2SLS Fixed
Effects
Clustera
(7)
2SLS Fixed
Effects
Clustera
(8)
Ln(SEV/TV) -0.00 0.01 -0.04** -0.03 0.01 0.00 -0.05 -0.18*
(-0.193) (0.648) (-2.433) (-1.055) (0.217) (0.0356) (-0.691) (-1.705)
Ln(Per Capita Income) 0.03299** 0.03866** 0.06190** 0.07928* 0.00996 0.03901*** 0.00586 0.01179
(2.589) (2.344) (2.331) (1.911) (0.417) (4.132) (0.212) (0.722)
Unemployment Rate -0.001* -0.001 -0.000 -0.000 -0.000 0.000 -0.000 0.000
(-1.779) (-1.210) (-0.0628) (-0.201) (-0.196) (0.289) (-0.177) (0.628)
Ln(Manuf. Emp. Per Capita) -0.004 -0.004 -0.005 -0.008* -0.002 -0.004* -0.001 -0.004
(-1.341) (-1.527) (-1.382) (-1.690) (-0.588) (-1.790) (-0.465) (-1.262)
Ln(Gov’t Emp. Per Capita) 0.009 0.009 0.012 0.014 -0.003 -0.002 0.008 0.004
(0.722) (0.659) (0.531) (0.563) (-0.373) (-0.247) (0.688) (0.277)
Ln(Retail Emp. Per Capita) -0.003 -0.003 -0.003 -0.007* -0.002 -0.009* 0.003 0.002
(-0.762) (-0.383) (-0.903) (-1.683) (-0.665) (-1.781) (0.762) (0.330)
Ln(Other Emp. Per Capita) -0.018** -0.020** -0.014 -0.016 -0.016** -0.015** 0.004 0.004
(-2.402) (-2.021) (-1.587) (-1.323) (-2.249) (-2.239) (0.503) (0.433)
Ln(Age under 18) 0.058*** 0.082** 0.011 0.043 0.040** 0.027 0.017 -0.013
(2.893) (2.536) (0.223) (0.498) (2.048) (1.354) (0.435) (-0.269)
Ln(Age 18-24) 0.014 0.017** -0.005 0.014 -0.002 0.012 0.016 0.037*
(1.625) (2.305) (-0.189) (0.686) (-0.0986) (0.920) (0.498) (1.676)
Ln(Age 25-44) 0.048** 0.052** 0.092*** 0.085* 0.017 0.031 0.094*** 0.070
(2.147) (2.394) (2.787) (1.790) (0.654) (1.550) (2.773) (1.400)
Ln(Age 45- 64) 0.007 0.036* 0.045 0.098** 0.002 0.051** 0.075 0.041
(0.432) (1.848) (0.843) (2.205) (0.0784) (2.067) (1.639) (0.666)
Adjusted R2 0.890 0.893 0.917 0.919 0.112 0.208 0.060 -0.263
Time Indicator Variables No Yes No Yes No Yes No Yes
County-specific Time Trend No No Yes Yes No No Yes Yes
N 982 982 982 982 812 812 812 812
a Instruments: Marriage, Birth, Divorce, Mortality
Notes: All regressions include county fixed effects; * Indicates significance at the 90 percent confidence level for a two-tailed test.; ** Indicates significance at the 95
percent confidence level for a two-tailed test; *** Indicates significance at the 99 percent confidence level for a two-tailed test.