Shiva Prasad Acharya

Journal of Housing Economics 21 (2012) 187–194

Contents lists available at SciVerse ScienceDirect

Journal of Housing Economics

journal homepage: www.elsevier.com/locate/jhec

Voter decisions on eminent domain and police power reforms

Kwami Adanu a,⇑, John P. Hoehn b, Patricia Norris b, Emma Iglesias c
a
Department of Economics and Finance, GIMPA Business School, Accra, Ghana
b
Department of Agricultural, Food, and Resource Economics, Michigan State University, East Lansing, USA
c
Department of Economics, Michigan State University, East Lansing, USA

a r t i c l e i n f o a b s t r a c t

Article history: One unresolved issue arising from the use of eminent domain power involves how the
Received 16 March 2011 perceived benefits and costs of eminent domain power affect people’s positions on the
Available online 7 May 2012 reform of eminent domain and police power law. The paper addresses this issue by esti-
mating a voting model that explains voters’ decisions on eminent domain and police
JEL classification: power reform referenda in the US. Estimates indicate that eminent domain referendum
Q15 outcomes hinged on voters’ fundamental values and ideology, and voters’ immediate
Q24
self-interest. Voters’ fundamental values and ideology affects referendum outcomes
R38
R52
insofar as educational attainment in a county has a statistically significant effect on sup-
port for reform. Despite the greater incidence of eminent domain in low income and
Keywords: poorer communities, success of reform referenda in this study was found to be greater
Eminent domain in counties with higher incomes and lower unemployment rates. This implies that
Police power whatever asymmetry exists in the exercise of eminent domain law across income
Reforms groups does not affect voter reaction to eminent domain reforms. Moreover, counties
Voting
with high unemployment rates consider the larger potential benefits from urban
Logistic regression
renewal projects in vote decision-making providing a link between self-interest and vot-
ing behavior.
Ó 2012 Elsevier Inc. All rights reserved.

1. Introduction regulations that impose limits on land use by private own-

ers without depriving them of ownership rights over the
Eminent domain and police power are two principal property.
avenues by which governments exercise control over land Court decisions have however gradually broadened the
resources. Eminent domain refers to the power of govern- definition of public use to include development initiatives
ment to take private property for public use. Public use undertaken to provide public benefit (US Supreme Court,
here refers to public services like highways, public utilities, 1954, 2005; Michigan Supreme Court, 1981). In 1954, the
community centers, schools, and other facilities that can be US Supreme Court affirmed the use by the District of
made available for use of the entire community (Merrill, Columbia of eminent domain to eliminate blight and rede-
1986). Police power on the other hand describes the right velop blighted area, including the sale or lease of con-
of government to enact and enforce laws that restrict land demned properties to private entities that would
use to ensure orderly development, safety, health, and pro- undertake redevelopment (US Supreme Court, 1954). Then,
tection of the general welfare of the public (Sax, 1964). in 2005, the US Supreme Court upheld the decision of the
Good examples of the use of police power include zoning Connecticut Supreme Court in the famous Kelo v. New
laws, building and health codes, and environmental London case that under the US Constitution governments
are permitted to use eminent domain to take property
and transfer its use to other private parties as long as there
⇑ Corresponding author. is a public benefit, such as economic development (US
E-mail address: dkadanu@gmail.com (K. Adanu). Supreme Court, 2005).

1051-1377/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jhe.2012.04.005
188 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194

The broadening of the definition of public use has gen- Table 1

erated considerable political response. Opinion polls on the Summary of results for all eminent domain ballots in 2006. Source: National
Conference of State Legislatures: Property Rights Issues on the 2006 Ballot.
Kelo decision for instance show that more than 80% of
respondents disagreed with the decision of the Court (Na- State Measure # Topic area Pass/fail
dler et al., 2008). Nadler et al. (2008) review opinion polls Arizona Prop. 207 Eminent domain Pass
over the last three decades that suggest that Kelo struck at & (64.8%)
core American values. Nadler et al. cite poll finding that police power
California Prop. 90 Eminent domain Fail (47.6%)
70% of respondents agree with the statement that ‘‘The &
right of property is sacred’’ and 88% agree that ‘‘‘allowing police power
people to own private property’ is a major contributor to Florida Amendment Eminent domain Pass (69%)
making America great’’ (p. 291). Consistent with these val- 8
Georgia Amendment Eminent domain Pass
ues, polls data show that disapproval of Kelo was indepen-
1 (82.7%)
dent of political affiliation. Idaho Prop. 2 Eminent domain Fail (23.9%)
Although eminent domain and police power are related &
in the sense that both affect land use, they represent two police power
fundamentally different perspectives about property Louisiana Amendment Eminent domain Pass (55%)
5
rights. The exercise of eminent domain involves forceful
Michigan Proposal 06-4 Eminent domain Pass
transfer of property rights and, as established by the fifth (80.1%)
amendment of the US Constitution, requires payment of Nevada Question 2 Eminent domain Pass
compensation. Police power is exercised to prevent the ac- (63.1%)
New Question 1 Eminent domain Pass
tions of property owners from creating some public harm.
Hampshire (85.7%)
Because affecting others in some negative way is not part North Dakota Measure 2 Eminent domain Pass
of land ownership rights, regulatory action to protect the (67.5%)
public does not require compensation (Flick et al., 1995; Oregon Measure 39 Eminent domain Pass
Goldstein and Watson, 1997). Nevertheless, efforts to make (67.1%)
South Carolina Amendment Eminent domain Pass (86%)
compensation for the exercise of police power a legal
5
requirement began in 1995 when the US House of Washington Initiative 933 Police power Fail (41.2%)
Representatives passed a property rights bill calling for
compensation of property owners whenever federal regu-
latory actions decrease property values by more than
20%. The bill however failed to pass the Senate (Goldstein ballot. Estimates indicate that eminent domain referen-
and Watson, 1997). Subsequently, the issue was addressed dum outcomes depend on voters’ fundamental values
in some states through legislation and ballot initiatives and ideology insofar as educational attainment in a county
requiring compensation for police power. has a statistically significant effect on support for reform.
Following the 2005 US Supreme Court ruling in Kelo v. However, the results also show that counties with high
New London, several more states initiated referenda to unemployment rates consider the larger potential benefits
ban the use of eminent domain for economic development from urban renewal projects in vote decision-making thus
purposes or restrict the circumstances under which emi- providing a link between self-interest and voting behavior.
nent domain could be used (Orthner, 2007; Sandefur, Moreover, despite the greater incidence of eminent do-
2006; Berliner, 2003). Several states also proposed limiting main in low income and poorer communities (Carpenter
the exercise of police power by requiring compensation in and Ross, 2009), success of reform referenda in this study
certain instances. In November 2006, 13 states (Arizona, was found to be greater in counties with higher incomes.
California, Florida, Georgia, Idaho, Louisiana, Michigan, The remainder of the paper is ordered as follows. The
New Hampshire, Nevada, North Dakota, Oregon, South Car- next section presents the conceptual framework and re-
olina, and Washington) presented special ballots on search hypotheses of the paper. This is followed by the
reforming eminent domain and/or police power to voters. research data description and economic model sections.
All but three (California, Idaho, and Washington) were ap- Discussion of the research results and conclusions then fol-
proved (Table 1). In general, two main types of ballot mea- low in that order.
sures were presented: eminent domain only ballots, and
eminent domain and police power compensation ballots. 2. Framework and hypotheses
States with eminent domain only ballots generally called
for banning or restricting the use of eminent domain The rational voters model suggests that voters’ deci-
power for economic development purposes while eminent sions on public good provision can be treated as a derived
domain and police power compensation ballots combine demand of how much public good voters want to consume
restricted use of eminent domain power with requirement at the optimum (Downs, 1957; Deacon and Shapiro, 1975;
for police power compensation. Matsusaka, 1993; Kotchen and Powers, 2006). This implies
The research reported here examines voter decisions to that voters make voting decisions on the provision of pub-
identify the factors that influenced eminent domain and lic goods to maximize utility derived from the consump-
regulatory compensation referenda outcomes. The empiri- tion of private and public goods subject to an income
cal analysis applies logistic regression to county-level voter constraint. The analysis of vote outcomes on eminent do-
returns in 10 states with reform measures on the 2006 main and police power compensation is therefore treated
 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194 189

here as one of unveiling the derived demand for public measure since their net benefits from voting yes to re-
goods provided by these two institutions. stricted use of eminent domain and police power compen-
Voter decision in this study is conditioned not only on sation exceeds that for renters. There are, however, equally
self-interest of voters but also on their ideological posi- relevant reasons to expect the average homeowner to vote
tions. Ideological position here captures voter response no. For instance, given that a common rationale for emi-
that cannot be explained by benefits and costs associated nent domain takings for economic development is to com-
with eminent domain and regulatory compensation bat blight (Sandefur, 2006), property price increases that
requirements. In addition to ideological satisfaction, voters may be expected to come with neighborhood improve-
can expect direct benefits from their vote choices ments associated with the use of eminent domain to clean
(Sandefur, 2006; Lazzarotti, 1999). The direct benefits ex- blight provide good reason for a class of property owners
pected from voting on restricted use of eminent domain to vote no to restricting the use of eminent domain. The
power and police power compensation ballots include hypothesized positive relationship between homeowner-
the value at risk (e.g. home values) that voters seek to pro- ship rate and eminent domain and regulatory compensa-
tect (Sandefur, 2006; Riddiough, 1997), direct transfers tion could therefore be neutralized or reversed by
(e.g. police power compensation) to landowners as a result benefits expected from use of eminent domain and regula-
of government regulatory action (Miceli and Segerson, tory action.
1994), and nonmarket values (e.g. open space) provided Household income level is linked to vote outcomes on
by regulatory actions (Bengston et al., 2004). On the other police power compensation based on the Kuznets in-
hand, there are costs attributable to voter decisions on the verted-U relationship between income and environmental
use of eminent domain and police power. Such costs often quality. This relationship has been examined in several
take the form of higher tax obligations that can be ex- studies on vote behavior and environmental quality (Dea-
pected from some of these decisions (Deacon and Shapiro, con and Shapiro, 1975; Kahn and Matsusaka, 1997; Popp,
1975). For instance, in order to pay the increased 2001; Kotchen and Powers, 2006). This implies that high
compensation for eminent domain or police power when income voters may support police power actions because
the average voter supports a ballot on unrestricted use of of their relatively high demand for environmental quality
eminent domain or a requirement for police power com- and open space in urban and congested areas. Thus, it is ex-
pensation, voters may have to pay higher taxes to raise pected that high income voters would reject requirements
the necessary revenue. The increase in tax obligation re- for police power compensation, and vote no on the initia-
duces the disposable income of voters and changes the in- tive, in order to promote its use. The hypothesized rela-
come constraint of their utility maximization problem. tionship between income and eminent domain is based
Explanatory variables included in the study include on the assertion that the exercise of eminent domain
ballot measure type presented to voters, homeownership power tends to disproportionately affect poor communi-
rates, household income, education, unemployment rate, ties (Dana, 2007; Somin, 2007; Sandefur, 2003, 2006). This
population density, and voter turnout. These variables con- implies that lower income earning voters would be more
trol for the probable incentives and disincentives associ- supportive of a ballot measure that calls for banning or
ated with vote decisions at the polls and differences in restricting the use of eminent domain power for economic
the ballot measure types presented to voters. On ballot development.
measure type, summary results on eminent domain and Past studies on factors affecting attitude towards the
police power ballots in the 2006 midterm election (see Ta- environment and natural resource use also consistently
ble 1 in Appendix) suggest that voter support may be show that the level of education of voters positively affects
declining as the ballot measure type presented to voters voters’ attitudes towards resource protection (see, Deacon
extends from restricted use of eminent domain to re- and Shapiro, 1975; Kahn and Matsusaka, 1997; Kahn,
stricted use of eminent domain and requirement for police 2002; Fischel, 1979). This is because knowledge about the
power compensation. Support for the ballot measure is value of environmental quality and open space and how
therefore hypothesized to decline as the ballot measure these can be improved, and exposure to research findings
extends from restricted use of eminent domain to re- on the impact of environmental quality and open space
stricted use of eminent domain and police power compen- on property values and human health are important deter-
sation. If this relationship turns out to be positive instead, minants of voters’ positions on the environment. Research
then the result will be suggestive of two things: that voters in psychology and related fields has also tied educational
supporting restricted use of eminent domain power also attainment to socio-political attitudes and ideological posi-
tend to support compensation for police power, and voters tions in general. For instance, high educational attainment
who are not supportive of restricted use of eminent do- has been determined to be positively correlated to liberal
main power tend to support compensation for police ideological positions (Ekehammar et al., 1987; Zakrisson
power strongly enough to vote yes instead of no given that and Ekehammar, 1998; Weisenfeld and Ott, 2011). Educa-
their decision on these two issues conflict. tion is therefore one factor that can affect the ideological
Next, homeowners are expected to be more concerned position and subsequent choice of voters on natural re-
about use of eminent domain power and property regula- source-related ballot measures. It is hypothesized, based
tory action than voters living in rented properties. This is on previous findings on education-environment-natural re-
because homeowners have more value at risk than renters. source relationships that support for the ballot measure is
The implication here is that counties with high homeow- decreasing in educational level of voters. A decrease in
nership rates would be more supportive of the ballot yes votes here means increased support for use of eminent
190 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194

domain to promote desirable land use, and a decline in sup- in referenda (Stone, 1965; Knox et al., 1984; Hadwiger,
port for regulatory compensation to promote its use to im- 1992). As turnout rises the proportion of favorable votes
prove environmental quality. declines. One explanation offered for this result is that
Next, counties with high unemployment rates are ex- qualified voters who oppose ballot propositions tend to ex-
pected to be more supportive of eminent domain since press their protest by boycotting elections (Stone, 1965).
use of eminent domain power for economic development Support for the ballot measure is thus hypothesized to be
purposes can be valuable for economically depressed areas decreasing in voter turnout. As noted by Hadwiger, this is
that are looking forward to economic expansion and job a result that requires further research to better explain
creation (Clarke and Kornberg, 1994; Bowler and Donovan, the finding.
1994; Sandefur, 2006). Several of these studies indicate
that voter dissatisfaction with bad economic conditions
erodes support for ballot proposals because of low support 3. Data and variables
for government (Clarke and Kornberg, 1994; Bowler and
Donovan, 1994). On the other hand, given that police The paper uses cross-sectional county level data cover-
power does not involve any subsequent use of the property ing yes/no vote outcomes on eminent domain and police
to provide jobs or any collective good, unemployment rate power propositions in the 2006 mid-term elections. The
may not have a significant effect on how voters react to po- sample size of 189 covers 10 of 13 US states shown in Table
lice power compensation ballots. This implies that high 1. Washington State is not included in the data set because
unemployment rate can be expected to decrease the pro- the vote initiative involves only police power. Nevada and
portion of yes votes cast on restricted use of eminent do- North Dakota are dropped from the study due to missing
main and police power compensation. data problems. A county is dropped if it has missing data
Population density is another variable that can be for one or more variables in the model. High incidence of
linked to the potential direct benefits of eminent domain missing values was observed for educational attainment
and police power. Limited land availability and high land and household median income variables. The dependent
prices in high population density areas often imply that variable in the model is the logodds of yes votes as defined
some public services may only be provided by taking some in Eq. (1). The source of the vote data is the University of
existing properties and converting them to alternative Michigan library government documents center (Univer-
uses. For instance, single family homes at good locations sity of Michigan, 2006).
may be taken and converted to multi-story apartment Fig. 1 below shows the contribution of the various
complexes to serve more people and expand the tax base. states to the total sample size. Florida has the highest
More densely populated counties are thus expected to county contribution of 18% followed by California with
show more support for measures like eminent domain that 17%. On the lower end New Hampshire and Idaho contrib-
promise the provision of these much needed services by ute 3% each to the total sample size.
voting against restrictions on the use of eminent domain. Variations in state contribution to total sample size is
Further, since properties in urban areas tend to be more influenced by the number of reporting counties in a state,
expensive than comparable properties in rural or low pop- and availability of data for variables in the model for the
ulation density areas, voters in high population density county concerned. Summary statistics for all variables used
counties may be more inclined towards voting no to re- in the model is shown in Table 2 below.
quire compensation for use of police power. This reduces The table provides information on the sample size, unit
the budgetary burden that voters have to face in the form of measure, mean, standard deviation, minimum, and max-
of additional tax payments to finance such compensation imum values for each variable.
payments. In summary, voters in counties with high popu- Independent variables included in the model are ballot
lation density are likely to vote no on restricted use of emi- type, homeownership rate, education, income, unemploy-
nent domain, and police power compensation. ment rate, population density, and turnout. The ballot type
Finally, previous empirical studies indicate that low vo- variable (ballottype) is binary and is defined to equal 0 if
ter turnout correlates strongly with approval of initiatives the ballot question in a given state calls for restricted use

County Contribution of States to Total Data Sam ple

Michigan, 25, 14% Georgia, 24, 13%
California, 31, 17%

Idaho, 6, 3%
S.Carolina, 20, 11%
Arizona, 10, 5%
Louisiana, 15, 8%

N.Hampshire, 6, 3%
Oregon, 14, 8% Florida, 33, 18%

Fig. 1. Composition of the data by contributing states.

 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194 191

Table 2 (2004). The source of the incidence data is Berliner (2003)

Summary statistic of variables. while Takelaw is taken from Castle Coalition (2007).
Variable Unit Mean Standard Min Max
deviation
Yes votes 0.931 0.806 1.537 2.223 4. Model
(logodds)
Yes votes Percent 69.600 16.300 17.7 90.200 The logistic regression model estimated is specified as,
Ballot type 0.249 0.433 0 1
Homeownership Percent 69.740 8.247 45.200 87.800
 
rate F
High school Percent 83.971 6.396 62.300 96.000 In ¼ b0 þ b1 ballottype þ b2 home þ b3 home
1F
Bachelors Percent 23.412 8.166 10.000 52.600
Income ‘000$ 46.576 9.551 23.119 81.761  ballottype þ b4 highschool
Unemployment Percent 5.048 2.014 2.400 15.300
rate þ b5 highschool  ballottype
Population Pop/ 2.523 1.988 0.063 9.462
þ b6 bachelor þ b7 bachelor  ballottype
density miles2
(‘00) þ b8 income þ b9 income  ballottype
Turnout Percent 50.008 12.712 2.032 75.700
þ b10 unemp þ b11 unemp  ballotype
þ b12 density þ b13 density  ballottype
of eminent domain only and 1 if a requirement for police þ b14 turnout þ b15 Ariz þ b16 Calif
power compensation is added to restricted use of eminent
þ    b23 Ore þ n ð1Þ
domain. The source of this data is the National Conference
of State Legislatures.

Number of Yes votes F
Homeownership rate (home) is defined as the percent- where F ¼ Number of valid votes
, and 1F represents the odds
age of occupied housing units that were owner-occupied of a supporting vote at the polls. The dependent variable
in 2006. Data on homeownership rate is obtained from in Eq. (1) is therefore the logodds of a supporting vote.
the 2006 US Census Bureau’s American Community Survey. The model specification is informed by results of a Chow
The data is limited to household population and excludes test (Chow, 1960) conducted to test for parameter stabil-
population living in institutions, college dormitories, and ity across the two ballot types covered by the data. The
other group quarters. returned F-statistic value of 23.157 for the Chow test is
Level of education is represented by two variables in compared to a 5% critical value of 1.94 to reject the null
this study: percent of county population 25 years and over hypothesis that the estimated parameters are equal
who have completed at least high school education (in- across the two data sub-samples. Ballot measure type
cludes equivalency) by 2006 (highschool) and percent of interaction terms are thus included in the model to ac-
county population 25 years and over who have completed count for shifts in the impact of the explanatory variables
at least bachelors degree by 2006 (bachelors). The source of across the two sub-samples. The model is estimated using
the education variables is the 2006 American Community Weighted Least Squares (WLS) to control for inherent
Survey data tables (US Census Bureau, 2006). The income Heteroskedasticity.
variable (income) is represented by the county-level med- The decision to reform eminent domain and police
ian household income (in 2006 Inflation-Adjusted Dollars). power law in a given state may be influenced by several
The income data is also drawn from the 2006 American factors including presence of special interest groups, and
Community Survey data tables (US Census Bureau, 2006). incidence of eminent domain takings. The Heckman
Population density (density) is measured by the number (1976) two-step procedure is employed to test for this
of people per square mile living in each county and is com- selection bias problem. To implement the Heckman proce-
puted using July 1 2006 population estimates and county dure three other variables are included in the selection
area (land area in square miles) data. The source of these equation to control for the effect of special interest groups
data is the population division of US Census Bureau (US (ingroup), existing restrictions on state eminent domain
Census Bureau, 2006). For the unemployment variable (un- law (takelaw), and incidence of eminent domain takings
emp), county level unemployment rate data is taken from (incidence). Incidence here is an example of the so-called
the Bureau of Labor Statistics’ local area unemployment backlash variables that have been shown to correlate with
statistics. The title of this data at source is ‘‘Labor Force the decision to reform eminent domain law across US
Data by County, 2006 Annual Average (US Department of states (Lopez and Totah, 2007; Lopez et al., 2009). It is ex-
Labor, 2006)’’. Election turnout (turnout) is measured as pected that states with higher per capita takings are more
the percentage of qualified voters who voted in the likely to present voters with a proposition to restrict the
referendum. use of eminent domain power and require compensation
Additional explanatory variables included in the dataset for police power. A test of the inverse Mills ratio parameter
to allow for the implementation of Heckman sample selec- from the Heckman regression indicated that the variable is
tion regression are special interest groups (ingroup), exist- not statistically significant at the 10% level.1 The model is
ing restrictions on state eminent domain law (takelaw), thus estimated without the Mills ratio variable.
and incidence of eminent domain takings (incidence). The
interest group variable is taken from Thomas and Hrebenar 1
Results from the Heckman regression are available on request.
192 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194

5. Results Table 3
Estimated logit and odds elasticity results.a

Results of the regression analyses are presented in Table Variable Model 1: with Model 2: without
3. Two sets of results are presented: logit estimates and income and income and
education controls education controls
odds elasticity values. The logit coefficients measure the
change in the logodds of a yes vote given a one unit change Logits Odds Logits Odds
elasticity elasticity
in the explanatory variables.
The computed odds elasticity2 values are evaluated at Ballot type 3.260** 0.812 1.596** 0.397
the mean of the explanatory variables and represent the per- (0.976) (0.512)
Homeownership 0.003 0.21 0.016** 1.116
centage change in the odds of a yes vote for a 1% change in (0.004) (0.003)
an explanatory variable. Table 3 presents results for two Homeownership ⁄ ballot 0.006 0.097 0.010 0.161
models. Model 1 shows results for the model specification type (0.011) (0.007)
in Eq. (1) while model 2 shows results for the same specifi- High school 0.009 0.756
(0.009)
cation without the income and education controls. Dropping
High school ⁄ ballot type 0.031* 0.626
the income and education controls allows for some assess- (0.015)
ment of effects of any collinearities between unemployment Bachelors 0.011* 0.258
rate and the two dropped variables. The results however ap- (0.006)
pear to be reasonably stable with consistent coefficient signs Bachelors ⁄ ballot type 0.014 0.078
(0.014)
across the two model estimations. Discussion of results is Income 0.015** 0.699
thus focused on results from model 1. (0.004)
The ballot measure type variable is statistically signifi- ⁄
Income ballot type 0.017* 0.207
cant at the 1% level in explaining the vote outcomes. The (0.01)
Unemployment rate 0.106* 0.535 0.104** 0.525
result indicates that moving from a measure that covers
(0.035) (0.029)
eminent domain only to one that covers eminent domain Unemployment 0.096* 0.141 0.107** 0.157
and police power compensation reduces the odds of pass- rate ⁄ ballot type (0.04) (0.032)
ing the ballot measure by 0.8%. This finding is consistent Population density 0.01 0.025 1.596 4.027
with the hypothesized relationship which suggests that (0.014) (0.512)
Population 0.044* 0.019 0.019 0.008
combining these two issues on one ballot increases the density ⁄ ballot type (0.023) (0.021)
proportion of voters rejecting the initiative relative to pre- Turnout 0.009** 0.45 0.013** 0.65
senting a ballot on restricted use of eminent domain only. (0.004) (0.004)
Homeownership rate is found to be statistically Constant 1.597** 1.597 0.761** 0.761
(0.658) (0.308)
insignificant at the 10% level in explaining the chances that
R-square 0.931 0.918
a voter would vote yes at the polls. This implies that voters’
position on restricting the use of eminent domain and Standard errors are given in parentheses below the estimated coefficients.
a
State-dummy variables were included in the regression but do not
requiring police power compensation is independent of
appear in the results shown. Full results are available upon request.
whether they own a home or live in rented property. A *
Statistically different from zero at the 5% level.
joint zero restriction test of homeownership rate and its **
Statistically significant at 1% level.
interaction with the ballot type variable produced F-statis-
tic of 0.38 with an associated p-value of 0.69. This joint test
represents a test of the statistical significance of homeow- is statistically significant in explaining voter response irre-
nership rate on vote outcomes on eminent domain and spective of ballot measure type.4 A one percentage point in-
regulatory compensation reforms (ballottype = 1). The test crease in the percentage of people 25 and over that hold at
result confirms the statistical insignificance of homeow- least a bachelor’s degree decreases the odds of voting yes
nership rate in the model. The finding is inconsistent with by 0.26% for eminent domain reforms and by 0.34% for emi-
the hypothesized relationship in this study which indicates nent domain and regulatory compensation ballots. Overall, it
that homeowners are more likely to vote yes. This finding takes a level of education above highschool for education to
is however consistent with the November 2005 survey re- have the hypothesized negative impact on the odds of voting
sults by the Saint Index organization (Somin, 2007) on the yes on restricted use of eminent domain and regulatory
Kelo ruling which indicate that renters reject the Kelo compensation.
ruling almost as strongly as homeowners. The regression analysis addresses local economic im-
Two measures of education are included in the model pacts in two ways: household income and unemployment
(highschool and bachelor) to control for knowledge of vot- rate. On eminent domain reforms, a 1 percentage point
ers. Highschool is statistically insignificant at the 10% level increase in median household income increases the odds
in explaining vote outcomes on eminent domain and police of voting yes by 0.70%.5 This finding of positive relationship
power initiatives irrespective of ballot measure type.3 On
the other hand, the higher measure of education (bachelor)
4
A joint zero restriction test of bachelor and its interaction with
ballottype produced F-statistic of 3.42 with an associated p-value of 0.035.
2
The odds elasticity with respect to a given attribute (where [Zi, hi] 2 Xi) is 5
Median household income is statistically significant in the estimated
given by, @½ðF i Þ=ð1F
@X i
i Þ
= ðF i Þ=ð1F
Xi
iÞ
¼ bi X i . See Fridstrom and Elvik (1997). model irrespective of ballot measure type. A joint zero restriction test of
3
A joint zero restriction test of highschool and its interaction with income and its interaction with ballottype produced F-statistic of 6.38 with
ballottype produced F-statistic of 2.24 with an associated p-value of 0.11. an associated p-value of 0.00.
 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194 193

between income and yes votes for restricted use of eminent vote decision-making thus providing a link between
domain is inconsistent with the hypothesized relationship self-interest and voting behavior. Next, voters show a
that the exercise of eminent domain power tends to dispro- nuanced understanding of eminent domain takings ver-
portionately affect poor communities. When the ballot mea- sus property value diminution due to regulatory action.
sure is broadened to include a requirement for regulatory Including compensation for regulatory takings in a re-
compensation the effect of income declines but remains po- form initiative reduces the odds of passage drastically.
sitive; a 1% increase in median household income results in a Here, voters may be concerned about the costs of requir-
0.49% increase in the odds of voting yes. Unemployment ing compensation for police power actions, as well as the
rate, the second measure of economic strength is statisti- uncertainties about just what the compensation would
cally significant at the 5% level in explaining vote outcomes entail. While it is relatively easy to determine compensa-
irrespective of ballot measure type.6 Higher unemployment tion values for eminent domain takings, determining
rates have a negative effect on the odds of voting yes at the such value for police power actions is a much more com-
polls. Specifically, a 1% increase in unemployment rate re- plicated and expensive task.
duces the odds of voting yes by 0.54% for eminent domain Turnout has a negative impact of on passing eminent
only ballots. When the ballot measure is broadened to in- domain and police power referenda outcome as hypoth-
clude a requirement for regulatory compensation a 1% esized. As noted by Hadwiger (1992) the effect of turn-
increase in unemployment rate decreases the odds of voting out on referenda outcomes requires more research and
yes by 0.39%. The result shows that voters in counties that explanations. In this study the negative impact of turn-
are relatively weak economically tend to support relaxation out on passing eminent domain and police power refer-
of restrictions on use of eminent domain power and a enda appears to be capturing the relatively higher
requirement for police power compensation. incentive to vote by voters supporting the ballot measure
Population density is statistically insignificant at the (reforms) as compared to voters who support the status
10% level in explaining changes in the vote outcome irre- quo. This is because a failure by ballot-supporting voters
spective of ballot measure type.7 This is consistent with to vote means the status quo stays. The implication then
Lanza’s finding of no relationship between population den- is that voters supporting the ballot measure tend to be
sity and actual eminent domain takings. The results from over-represented at low turnout levels while those sup-
this study as well as those from Lanza (2006) suggest that porting the status quo are over-represented at higher
population density is not an important variable in explain- turnout levels.
ing both actual eminent domain takings and voter deci-
sions on eminent domain reforms. References
Turnout in this model has the expected negative and
statistically significant (at 5% level) effect on the vote Bengston, D.N., Fletcher, J.O., Nelson, K.C., 2004. Public policies for
outcome. Higher turnout at the polls is associated with managing urban growth and protecting open space. Policy
instruments and lessons. Landscape and Urban Planning 69 (2–3),
rejection of the ballot measure as hypothesized. In par- 271–286.
ticular, a 1% increase in turnout results in 0.45% decline Berliner, 2003. Public Power, Private Gain: A Five Year, State by State
in yes votes. Report Examining the Abuse of Eminent Domain Power, Castle
Coalition Report (April 2003). Available at <castlecoalition.org>.
Bowler, S., Donovan, T., 1994. Economic conditions and voting on ballot
propositions. American Politics Research 22 (27), 27–39.
6. Conclusions
Carpenter, D.M., Ross, J.K., 2009. Testing O’Connor and Thomas: does the
use of eminent domain target poor and minority communities? Urban
Results from this study indicate that eminent domain Studies 46, 2447–2461.
referendum outcomes hinged on voters’ fundamental val- Castle Coalition, 2007. 50 State Report Card: Tracking Eminent Domain
Reform Legislation since Kelo.
ues and ideology, and voters’ immediate self-interest. Chow, G., 1960. Tests of equality between sets of coefficients in two
Voters’ fundamental values and ideology affects referen- linear regressions. Econometrica 28 (3), 591–605.
dum outcomes insofar as high educational attainment Clarke, H.D., Kornberg, A., 1994. The politics and economics of
constitutional choice. Voting in Canada’s 1992 National
in a county has a statistically significant and negative ef- Referendum. The Journal of Politics 56 (4), 940–962.
fect on support for reform. Moreover, despite the greater Dana, D., 2007. The law and expressive meaning of condemning the poor
incidence of eminent domain in low income and poorer after Kelo. Northwestern University Law Review 101 (365).
Deacon, R., Shapiro, P., 1975. Private preference for collective goods
communities, success of reform referenda in this study revealed through voting on referenda. The American Economic
was found to be greater in counties with higher incomes, Review 65 (5), 943–955.
and lower unemployment rates. This implies that what- Downs, A., 1957. An Economic Theory of Democracy. Harper, New York.
Ekehammar, B., Nilsson, I., Sidanius, J., 1987. Education and ideology:
ever asymmetry exists in the exercise of eminent domain basic aspects of education related to adolescents’ sociopolitical
law across income groups does not affect voter reaction attitudes. Political Psychology 8 (3), 395–410.
to eminent domain reforms. Further, it implies that Fischel, W.A., 1979. Determinants of voting on environmental quality. A
study of a New Hampshire Pulp Mill Referendum. Journal of
counties with high unemployment rates consider the
Environmental Economics and Management 6, 108–118.
larger potential benefits from urban renewal projects in Flick, W.A., Barnes, A., Tuft, R., 1995. A public purpose and private
property: the evolution of police power. Journal of Forestry 93 (6),
6
21–24 (4).
A joint zero restriction test of the unemployment rate and its interaction Fridstrom, Elvik, 1997. The barely revealed preference behind road
with ballottype produced F-statistic of 4.80 with an associated p-value of investment priorities. Public Choice 92, 145–168.
0.01. Goldstein, J., Watson, W., 1997. Property rights, regulatory taking, and
7
A joint zero restriction test of the density and its interaction with compensation: implications for environmental protection.
ballottype produced F-statistic of 2.02 with an associated p-value of 0.14. Contemporary Economic Policy 15, 32–42.
194 K. Adanu et al. / Journal of Housing Economics 21 (2012) 187–194

Hadwiger, D., 1992. Money, turnout, and ballot measure success in Opinion and Constitutional Controversy. Oxford University Press,
California cities. The Western Political Quarterly 45 (2), 539–547. New York.
Heckman, J.J., 1976. The common structure of statistical models of Orthner, D., 2007. Toward a more ‘‘Just’’ compensation in eminent
truncation, sample selection and limited dependent variables and a domain. MCGeorge Law Review 38, 429–455.
simple estimator for such models. Annals of Economics and Social Popp, D., 2001. Altruism and the demand for environmental quality. Land
Measurement 5, 475–492. Economics 77 (3), 339–349.
Kahn, M.E., 2002. Demographic change and the demand for Riddiough, T.J., 1997. The economic consequences of regulatory taking
environmental regulation. Journal of Policy Analysis and risk on land value and development activity. Journal of Urban
Management 21 (1), 45–62. Economics 41 (1), 56–77.
Kahn, M.E., Matsusaka, J.G., 1997. Demand for environmental goods: Sandefur, T., 2006. The backlash so far: will Americans get meaningful
evidence from voting patterns on California initiatives. Journal of Law eminent domain reform? Michigan State Law Review 709, 711–746.
and Economics 40 (1), 137–173. Sandefur, T., 2003. Rights perspective on eminent domain in California: a
Kotchen, M.J., Powers, S.M., 2006. Explaining the appearance and success rationale for meaningful judicial. Scrutiny of ‘‘Public Use’’.
of voter referenda for open-space conservation. Journal of Southwestern University Law Review 32, 569–676.
Environmental Economics and Management 52 (1), 373–390. Sax, J.L., 1964. Takings and the police power. Yale Law Journal 74, 36.
Knox, J., Landry, C., Payne, G., 1984. Local initiative. A study of the use of Somin, I., 2007. Is post-kelo eminent domain reform bad for the poor?
municipal initiatives in the San Francisco Bay area. Coro Foundation, Northwestern University Law Review 101 (4), 1931–1943.
San Francisco. Stone, C.N., 1965. Local referendums: an alternative to the alienated-voter
Lanza, S., 2006. An offer you can’t refuse: why do Connecticut and other model. The Public Opinion Quarterly 29 (2), 213–222.
states use eminent domain? The Connecticut Economy 9, 12. Thomas, C., Hrebenar, R., 2004. Interest Groups in the States. In: Gray, V.,
Lazzarotti, J., 1999. Public use or public abuse. University of Missouri- Hanson, R. (Eds.), Politics in the American States: a Comparative
Kansas City Law Review 68, 49–53. Analysis, CQ Press, Washington, DC.
Lopez, J.E., Totah, S., 2007. Kelo and its discontents: the worst (or best) University of Michigan, Election Results and Voting, Election 2006.
thing to happen to property rights. The Independent Review XI (3), Available at: <http://www.lib.umich.edu/govdocs/elec2006.html>.
397–416. US Census Bureau, American Community Survey (2006). Available at:
Lopez, J.E., Jewell, R.T., Campbell, N.D., 2009. Pass a law, any law, fast! <http://factfinder.census.gov/home/saff/main.html?_lang=en>.
State legislative responses to the Kelo backlash. Review of Law and US Department of Labor, Local Area Unemployment Statistics (LAUS)
Economics 5I (1). program, 2006.
Matsusaka, J.G., 1993. Election closeness and voter turnout: evidence US Supreme Court, 2005. Susette Kelo et al., Petitioners v. City of New
from California Ballot Propositions. Public Choice 76, 313–334. London, Connecticut et al. 545 US_2005.
Merrill, T.W., 1986. The economics of public use. Cornell Law Review 61, US Supreme Court, 1954. Berman v. Parker. 348 US 26.
74–75. Weisenfeld, U., Ott, I., 2011. Academic discipline and risk perception of
Miceli, T.J., Segerson, K., 1994. Regulatory takings: when should technologies: an empirical study. Research Policy 40 (3), 487–499.
compensation be paid? Journal of Legal Studies Review 23, 749–776. Zakrisson, I., Ekehammar, B., 1998. Social attitudes and education: self-
Michigan S.C., 1981. Mich 616 (en banc). selection or socialization? Scandinavian Journal of Psychology 39 (2),
Nadler, J., Diamond, S.S., Patton, M.M., 2008. Government takings of 117–122.
private property. In: Persily, N., Citrin, J., Egan, P. (Eds.), Public