an empirical study of the consolidation of local public health...
TRANSCRIPT
An Empirical Study of the Consolidation of Local Public Health Services in Connecticut
Laurie J. Bates
Bryant University Department of Economics
1130 Douglas Pike Smithfield, RI 02917 Phone: 401-232-6459 Fax: 401-232-6319
Email: [email protected]
Becky A. Lafrancois Syracuse University
Department of Economics 110 Eggers Hall
Syracuse, NY 13244 Phone: 315-443-9067 Fax: 315-443-1075
Email: [email protected]
Rexford E. Santerre University of Connecticut
Department of Finance 2100 Hillside Avenue, Unit 1041
Storrs, CT 06269 Phone: 860-486-6422 Fax: 860-486-0634
Email: [email protected]
Abstract: Only a few studies, mostly in the case of school districts, have empirically examined the factors affecting municipal consolidations. This study contributes to the literature by empirically examining the decision of Connecticut communities to consolidate the delivery of public health services. As theory suggests, the prospect of scale economies is found empirically to increase the likelihood that a community consolidates public health services. In addition, differences across communities are found to inhibit the consolidation of public health services. Overall, the results imply that financial incentives may be necessary to encourage more regional districts because localities may underestimate the true minimum efficient scale for public health services and because heterogeneity among jurisdictions impedes regional cooperation. Keywords: public health, consolidation, JEL Codes: H41; H7; I18
An Empirical Study of the Consolidation of Local Public Health Services in Connecticut
1. INTRODUCTION Special districts in the United States have increased nearly three-fold over the last 50
years or so.1 Special districts provide water, fire protection, sanitation, and public health,
among other important collective services. Economic theory suggests the number of
special districts may have grown over time as individual local governments consolidated
specific municipal functions to benefit from scale economies. Economic theory also
suggests that some local governments may not have merged similar municipal functions
because of the real or perceived political externality costs, transaction costs, and principal
agent problems associated with the consolidation of larger and more heterogeneous
jurisdictions.
While theory seems to offer a good explanation as to why local governments may
or may not consolidate municipal functions, few studies have directly subjected the
theory to empirical testing (Bates and Santerre, 2008; Brasington, 1999 and 2003; and
Gordon and Knight, 2006). Moreover, these empirical studies are mostly limited to the
consolidation of school districts. Yet, it would be beneficial to know the specific factors
driving the creation of other types of special districts, especially if the consolidation of
specific municipal functions is viewed favorably by higher levels of government.
Given the paucity of empirical studies devoted to the consolidation issue, this
paper examines the decision of Connecticut communities to enter into a public health
district relationship. More specifically, this study empirically investigates whether scale
economies and community differences influence the regional consolidation of local
1 Statistical Abstract of the U.S. (2007).
2
public health services. The results may be important because many experts believe that
public health will take on an increasing role in the future given the threat of bioterrorism
attacks, concerns over emerging diseases such as avian flu and SARS, and the seemingly
growing burden from natural disasters such as Katrina (Tilson and Berkowitz, 2006). Yet,
concern has been expressed in Connecticut, the focus of the forthcoming empirical
investigation, and elsewhere, that small, independent departments may lack the necessary
resources to produce public health services cost-effectively (e.g., Hicks, 2004; Hartford
Courant, 2009; and Penny, 2009). Sixty-two percent of all local health departments in the
U.S. fall into the small category (NAACHO, 2006). Many public health policy-makers
argue that regional consolidation offers a solution to this problem. In particular, policy-
makers point out that regional public health departments provide more efficient
administration, broader financial resources, improved personnel management, less
duplication of resources, and improved reporting (Turnock, 2004). If so, from a public
policy perspective it may be important to know why some localities choose to offer
public health services on an independent rather than a consolidated basis.
The next section of this paper develops the conceptual framework behind the
empirical model of regional consolidation. Section III describes the sample and data used
in the empirical test and section IV reports on the findings. A summary and some policy
implications are offered in the final section.
3
2. PREVIOUS STUDIES, CONCEPTUAL FRAMEWORK, AND EMPIRICAL MODEL Numerous studies, such as Adelaja and Racevskis (2005) Alesina and Spolaore (1997),
Alesina, Baqir, and Hoxby (2004), Borck (1998), Brasington (1999, 2003, and 2004),
Ellingsen (1998), Feiock (2007), Gerber and Gibson (2005), Krueger and McGuire
(2005), and Sorensen (2006), have investigated the political economics of regional
cooperation and consolidation. While these studies take different approaches and/or
address slightly different issues, they all share two things in common. First, they agree
that trade-offs are involved when local governments consider cooperating or
consolidating municipal functions with other local governments in the region.2 Second,
all of these studies agree that this trade-off can be couched in terms of costs and benefits.
The benefits of cooperating or consolidating include any cost-savings from scale
economies in production and the internalization of any externality problems in the region.
In this regard, smaller-sized communities have more to gain from cooperating or
consolidating than larger jurisdictions do. Santerre (2009) finds that local public health
costs per capita continue to fall with population until a minimum efficient scale (MES) of
100,000 people is served by a local public health department. However, Martin and
McKenzie (1975) argue that consolidation may not offer tax savings for consumer-voters
because bureaucrats siphon them off in the form of nonmonetary gains. It should be
pointed out that Martin and McKenzie are referring to the consolidation of general
2 Local governments may also turn to contracting-out as an alternative method of delivering services. For relatively recent studies on the contracting-out decision see Boyne (1998), Brown and Potoski (2003), Ferris and Grady (1994), Joassart-Marcelli and Musso (2005) and Nelson (1997) .
4
purpose governments such as cities or counties and not to the consolidation of a specific
municipal function such as the delivery of public health services which is studied below.3
Principal/agent problems stemming from greater centralization, transaction costs
relating to inter-local negotiations and agreement, and the costs associated with losing
control over decision-making make up the cost side of the cooperating and consolidating
calculus. The degree of heterogeneity among communities weighs importantly in
determining the size of these costs. Greater heterogeneity among communities potentially
raises both transaction and decision making costs. In fact, Brasington (2004) finds
empirically that the loss of control over local public school services, because of
consolidation, reduces house values by slightly over $2,900 or 3.5 percent assuming all
other factors remain constant.
The few existing empirical studies model the decision to consolidate municipal
responsibilities as a function of a community’s economic and demographic factors as
well as the difference between the community’s and potential merger partner’s
characteristics (Bates and Santerre, 2008; Brasington, 1999 and 2003; and Gordon and
Knight, 2006). In addition, Brasington (1999 and 2003), Ferris and Graddy (1988), and
Bates and Santerre (2008) allow for the possibility that an inverted-U relationship may
hold between the population in a community and the internal production of municipal
services. Ferris and Grady find that medium-sized communities are more likely to retain
internal control for the provision of public health services, whereas both small and large
local governments contract out to a greater degree. Brasington, (1999, 2003) determines
empirically that medium-sized schools are less likely than small or large schools to form
3 See Honadale (1995, 1998) for a couple of case studies focusing on political economy aspects of general government consolidations.
5
a regional school district with surrounding communities. In contrast, Bates and Santerre
find that medium-sized communities are more likely to consolidate public health
activities. Brasington (1999 and 2003) and Gordon and Knight (2006) find evidence that
consolidation of local public education services involves a trade-off between scale
economies and loss of decision-making. Bates and Santerre (2008) show empirically that
differences among communities inhibit the regional consolidation of public health
services but do not study the effect of scale economies.
To model the decision to consolidate public health activities, we employ the
median-voter framework adopted by Bates and Santerre (2008). A large number of
studies have found the median-voter model represents a useful conceptual device when
examining collective decision-making, particularly at the local level of government (e.g.,
Borcherding and Deacon, 1972; Bergstrom and Goodman, 1973; and Santerre, 1985). In
any case, Fischer (2007) points out similar results are obtained if a dominant party model
is employed where the majority party maximizes the utility of the average voter.
Based on fairly normal assumptions, public choice theory predicts that the median
demand dominates over all other demands when collective political outcomes are decided
by a simple majority voting rule in a direct democratic setting (Downs, 1957).
Interestingly, many communities in Connecticut, the observations used in the
forthcoming empirical analysis, have retained the open-town meeting which is a direct
democratic form of local government. But even in a mayor-council or council-manager
form of government the median-voter model may hold because politicians in a
representative democracy gravitate towards the middle of the preference distribution to
maximize their number of votes (Downs 1957). It follows under these two conditions
6
that the median-voter, as the swing-voter in each community, decides either directly or
indirectly if that particular community should enter into a district relationship with
another community or other communities for the provision of a specific public service.
The median-voter in a particular jurisdiction decides in favor of consolidation if
she expects her utility to increase upon joining a district. Utility increases if the expected
benefits from any economies of scale and internalization of spillovers, , conditioned
upon her current tax-share or price, Pi, income, Ii, and tastes and preferences for a
particular public service, Ti, exceed the expected cost, , associated with consolidation.
As discussed above, expected costs consider the loss of political control, any transaction
costs associated with negotiating with partners, and the potential principal/agent
problems that may arise in larger organizations. She realizes that these “political
externality costs” are likely to be higher when the characteristics or attributes of the
community in which she resides differ significantly from those of potential district
partners, . Political externality costs are higher because heterogeneity may result
in the consolidated entity not providing the collective good at a level demanded by that
particular median-voter. Consequently, she votes in favor of consolidation or merging
i , are positive, or: with potential partners, M , if the expected net benefits,
; , , ; 0. (1)
From th s expression, a re uced-f
, , , . (2)
i d orm version can be written as:
The relationship between each of the independent variables in equation 2 and net
benefits is fairly straightforward. A direct relationship can be expected between the
median-voter’s tax-share, P, and her propensity towards consolidation. Assuming all
7
other factors remain constant, it stands to reason that the median-voter has less to gain
from consolidation when her tax-share for public health services is relatively low.
Oppositely, when the tax-share for public health services is relatively high, consolidation
offers a greater chance of reducing her overall tax burden.
As far as income is concerned, Bates and Santerre (2008) find that public health
services represent a normal good which means the demand for public health services rises
with income. If that finding can be generalized, then the influence of income, I, on the
decision to consolidate depends on whether the median-voter perceives the output of
public health services will rise or fall upon consolidating services. If she believes public
health services will rise, then wealthier individuals will be more likely to vote in favor of
regional consolidation. With respect to tastes, the median-voter favors consolidation
when she has a strong preference for public health services and consolidation improves
its delivery when c pared to independent production. om
Of course, , the median-voter’s net benefit from consolidation is not directly
observable. However, we do observe some information about that will allow us to
determine how P, I, T, or individually influences the median-voter’s choice
behavior because we know or how that median-voter actually voted. Thus while we
cannot distinguish between a strong and weak yes, we do observe if = 1 or = 0. If
we let represent the vector of independent variables in equation (2), our regression
e ten as: mod l can be writ
μ (3)
8
with reflecting the vector of parameters to be estimated and μ capturing the random
error term. The related observable variables are 1 if 0 and 0 if
0.
3. OBSERVATIONS, VARIABLES, AND DATA
With observations for a variety of communities and relevant data, probit analysis can be
used to estimate equation 3.4 All 169 towns and cities in Connecticut serve as the set of
observations used in the empirical test and community level data for 2004 are used in the
cross-sectional analyses. Connecticut municipalities have the option to operate their own
independent health department or form or join a unified health district. In 2004, 92
Connecticut municipalities voluntarily participated in 18 unified health departments. The
number of communities in a district health department ranges from 2 to 19. While this
decentralized structure makes Connecticut relatively unique (NAACHO, 2006) and
therefore the findings may be non-generalizable, the more centralized public health
infrastructures in other states do not allow one to observe the decision to voluntarily join
a local public health district. Hence, the underlying demand for consolidation cannot be
estimated in those other areas. 4 We also experimented with a bivariate probit model because consolidation involves both political entities jointly voting in favor of it and not just one. However, the bivariate probit model would not converge using either the LIMDEP or STATA statistical packages. In any case, we are not totally convinced that a bivariate probit model is necessary in this case given that both equations specify the same right-hand side variables. For example, suppose 2 communities, i and k are deciding to combine into a district. The probability of community i voting to join a district can be written as . Similarly, the probability of community k voting to join a district equals . The probability of each community deciding to join a distinct is a function of its own characteristics, or , the difference between the two sets of characteristics, D, and the potential for scale economies among the two, S. Therefore , , and , , . Of course, we never observe the individual probabilities; we only observe the joint probability or or , , , . Assuming the partial effects of and on are equal (and why not) then , , where X = either or . Thus, the estimation of a bivariate probit equation is unnecessary in this case much like it is unnecessary to use seemingly unrelated regression for a system of equations with the same right hand side variables.
9
In the towns and cities of Connecticut, the authorized legislative body (open-town
meeting or town-council) must vote to form or join a district with neighboring
communities. A community can also decide legislatively prior to the beginning of any
year to withdraw from a district if that community has been a member for at least two
years. While an existing district must approve membership, it receives $2.43 per capita
from the state for any town with population under 5,000 and $2.08 per capita for any
town with 5,000 or more people. The district board also levies a head tax on all of the
participating communities. The amount of the head tax varies widely across the public
health districts in Connecticut. We should also note that the Connecticut State
Department of Public health mandates several services, such as restaurant and septic
system inspections, which are mainly funded by fees. Districts also apply for grants from
the state and federal governments and other private sources to cover the costs of
providing additional public health services.
To implement the test, we need some method of matching up communities for
consolidation purposes. Thus, it is assumed that each community views its potential
merging partner or partners in one of two ways. First, our “single-model” supposes the
median-voter in each community considers the consolidation of public health services
individually with each of its adjacent communities. For this matching model, the
characteristics are simply those of the adjacent community. This matching assumption
results in 878 matches because the typical community faces 5 to 6 adjacent communities
on average. If two adjacent communities were in the same public health district in 2004,
the dependent variable in equation 3 takes on the value of one. Otherwise, the dependent
10
variable takes on the value of zero. About 25 percent of the 878 matches reflect two
communities in the same public health district.
One advantage of this single-model is that all Connecticut communities can be
used as observations in the empirical test. Another advantage of this approach is that a
community may actually consider just the characteristics of a nearby community, and not
the characteristics of all of its potential partners, when deciding to consolidate public
health services. This method of deciding might economize on decision-making costs. For
example, median-voter i may think joining a district would be beneficial because
community j, which shares many similar traits, already belongs to it. The last advantage
of this approach is that a community can choose to form a district with any one
community or all of its neighboring communities.
The alternative “group-model” uses only those communities who currently belong
to a public health district or are located adjacent to an existing public health district.
Thus, each community is matched up with a nearby district or the district to which it
currently belongs. In this case, characteristics are determined by calculating the
weighted averages of all of the characteristics of the communities belonging to a
particular public health district exclusive of the selecting community if it also belongs to
that district. This group model results in 268 observations with roughly 34 percent of the
matches involving a district relationship. Notice that, in effect, the group-model
questions if a particular town will join an existing district whereas the single-model
examines if a community will form a district with a neighboring community.5
5 The results may differ because of the greater fixed costs associated with forming a new district rather than joining an existing one.
11
The equalized or effective mill rate in each community serves as the tax-price of
the median voter. Total property values represents the relevant tax base because property
taxes provide the main source of funding in Connecticut communities and tax rates are
constant across different types of property (e.g., residential, commercial, and industrial).
Following studies on the demands for local public goods such as Borcherding and
Deacon (1972) and Bergstrom and Goodman (1973), the median-voter is assumed to
possess the median level of income in each community. The tastes and preferences of the
median-voter regarding public health services are assumed to be shaped by the
demographic and physical composition of the community in which she resides.
Therefore, the proportion of the population that is elderly (65 years of age and older), the
percentage of the population that is white, population size and land area of the
community are specified as taste variables in the estimation equation. While the rest of
the taste variables are specified in linear form, community population enters the probit
equation in quadratic form because the net benefits from consolidation may be nonlinear
with respect to community size as discussed previously.
According to the conceptual model, differences between communities may
impose political externality costs and thereby impede consolidation. In the empirical
analysis, the absolute differences of several characteristics are specified: tax-price,
median income, population, percentage of the population that is white, and land area. A
negative coefficient estimate is expected on each of these absolute difference variables.
Given that some public health services are devoted to environmental issues such as water
quality and sanitation, the difference in land area is included because a geographically
small community may believe that a geographically larger community will draw more
12
attention and resources from the district public health department.6 To capture the
potential for scale economies, the population of each community is added to the
population of each of its adjacent communities or the district and the resulting combined
population is specified in the estimation equation in quadratic form. An inverted-U
relationship seems appropriate between the likelihood of consolidation and combined
population because economies of scale may at first encourage consolidation but
diseconomies may set in after some level, at least in a perceived sense.
Finally, the residual term in equation (3) may be influenced by spatial
autocorrelation. More specifically, the choice that one community makes might affect the
choices that other communities make with regard to joining a public health district. If so,
the error term in equation (3) will be correlated across observations and would therefore
bias the estimated coefficients in some unpredictable way. One way of dealing with a
spatially correlated error term is to include location variables for each community in the
regression equation (e.g., Pace, Barry, and Sirmans, 1998). We experimented with five
location variables: distance to the nearest central city in Connecticut, distance to New
York City, distance to Boston, distance to Hartford, and distance to the Connecticut
coastline. Experimentation showed statistically that the latter two location variables are
the most meaningful and robust in explaining the consolidation decision.
Data for the market value of all taxable property and population are obtained from
Municipal Fiscal Indicators, which can be accessed on-line at the Connecticut State
Office of Policy and Management website. The Connecticut Department of Economic
and Community Development publishes an on-line version of Connecticut Town Profiles
6 Conversations with J. Robert Galvin, Connecticut State Commissioner of Public Health, and Pamela Kilbey-Fox, Branch Chief of the Local Health Administration Board in Connecticut confirmed that these differences among communities could potentially impede consolidation.
13
which lists the rest of the necessary data including median household income, percentage
of the population that is white, land area, distance to Hartford, and the percent of
population 65 years and older. Table 1 provides descriptive statistics for all of the
variables used in the empirical tests.
4. EMPIRICAL FINDINGS
The results for the probit analysis are reported in Table 2 where the coefficient estimates
and corresponding z-statistics, based on heteroskedasticity-consistent standard errors, are
shown opposite each explanatory variable. Results are displayed for both the single-
model and group-model methods of matching up communities. Focusing first on the
single model, the estimated positive coefficient on tax-price suggests that the median-
voter typically views consolidation as lowering tax burden when her tax share is
relatively high. Alternatively, it may be the case that the median-voter views
consideration as providing better services and thus reducing her quality-adjusted tax-
price.
The estimated negative coefficient on income implies that higher income
communities are less likely to form a district with a neighboring community, ceteris
paribus. That inverse relationship may hold for a number of reasons.7 First, the median-
voter in a wealthy community may be concerned that forming a district with a lower-
income community may result in a lower level of public health services supplied than she
desires. This is particularly true if public health services represent a normal good as
shown empirically by Bates and Santerre (2008). Second, higher-income communities
may be better able to afford the provision of local public health services on an 7 We thank the anonymous referees of this journal for pointing out two of the reasons.
14
independent basis than poorer communities. Third, assuming some substitutability
between private medical care and public health services in the production of health,
wealthier individuals may choose the more expensive option of purchasing their health
services in the private marketplace. Finally, higher-income communities may face a more
difficult time finding a potential merger partner with similarly high income.
Four of the five estimated coefficients on the absolute difference variables possess
negative signs, as expected. A Wald test indicates that the estimated coefficients on the
absolute difference variables are collectively significant at the 0.0054 level with an F-
statistic of 3.34. Moreover, on an individual basis, the findings suggest that differences in
median income and land area matter more at influencing the decision to consolidate on a
statistical basis. Thus, the empirical results support the hypothesis that differences across
communities inhibit consolidation.8
The results for the single-model also show that community population does not
influence the decision to form a public health district. In contrast, the combined
population of the community and partner town does affect consolidation in an inverted-U
fashion. Its inverted U-shape reflects that perceived economies and diseconomies hold
with respect to forming larger public health districts. By taking the first derivative of the
single-model specification with respect to combined population and setting the resulting
expression equal to zero, we can solve for the population at which diseconomies set in
with respect to producing public health services. The calculation indicates that perceived
diseconomies set in at a combined population of roughly 47,000 people, which is slightly
8 Note that both the percentage of the population that is white and absolute difference of the percent white have no statistical impact on the decision to form a district. That result may hold because the percentage of nonwhite population is relatively small at about 9 to 10 percent on average in Connecticut, particularly after controlling for population and population differences because nonwhites live disproportionately in the larger cities.
15
more than twice the size of the typical Connecticut community. As mentioned previously,
Santerre (2009) finds empirically that the MES of a public health department occurs at
roughly 100,000 people. Interestingly, Gordon and Knight (2006) report a relatively low
perceived MES for local public education services in Iowa.
Estimating the marginal effects of the various independent variables may help
determine if differences across municipalities or scale economies matter more at the
margin in terms of influencing the decision to remain independent. Point estimates drawn
from a probit function are highly sensitive to specification so Studenmund (2006)
recommends a rough approximation by multiplying each estimated probit coefficient by
0.4. Following through with his recommendation indicates the marginal effects are
economically tiny. This may mean that any single factor does not, in isolation,
significantly influence the decision to consolidate. Moreover, while collectively the
various factors matter, the relative inertia to a single factor may explain why so many
municipalities continue to operate independent public health departments.
Turning to the results for the group-model matching scheme in Table 2, a
different story is portrayed. Similar to the single-model matching scheme, joining a
public health district is more likely when the tax share is higher, median income is lower,
and more elderly people reside in the community. However, in the case of joining a
district, community population does affect the decision to consolidate. In particular,
compared to an otherwise similar small or large community, the results imply that a
medium-sized community is more likely to join an existing health district. Indeed,
according to the estimated equation, a community with 34,000 people is the most likely
to join a health district, ceteris paribus.
16
Four of the five difference variables possess negative coefficient estimates but
only land area possesses an estimated coefficient that is different from zero at
conventional levels of statistical significance. Thus, community differences do not seem
to matter as much when joining an existing health district as compared to forming one
with a neighboring community.
A noticeable difference between these and the single-model results is the
estimated U-shaped relationship between combined population and the probability of
joining a health district. One possibility is that the U-shaped relationship reflects that
relatively small and large public health districts are more successful at attracting new
member towns than medium-sized districts. Small health districts may mean less losing
out on political decisions whereas large districts may offer huge scale economies. Solving
for the district size at which the probability of joining a health district is minimized
results in a population of 203,000. With only 6 of the 268 matches resulting in combined
populations greater than 203,000, a more prudent interpretation is that an independent
community is less likely to join larger districts, ceteris paribus. Recall that the single-
model estimates the perceived MES occurs at roughly 47,000 people. Combined
population lies below 47,000 for only 33 of the 269 group-model observations and the
average level is nearly 94,000 people (see Table 1). Consequently, this observed inverse
relationship between combined population and the probability of joining a public health
district may simply reflect that independent communities are less likely to join larger
districts because most consolidations result in combined populations beyond the MES.
While a few communities belonged to public health districts as far back as the late
1960s, any community can choose to withdraw from a district in Connecticut after a two
17
year membership period. In fact, three communities withdrew from existing health
districts over the last 12 years. Another district dissolved itself two year prior to the year
under observation (2004) with one of the two formerly participating communities joining
another district and the other remaining independent. Yet, since 2004, another district
has been formed and eleven communities have joined existing districts. The implication
is that the net benefit calculus of belonging to a district, as summarized in equation 1, is
continuously made by each and every community and current rather than past values of
the independent variables are relevant.9
Nevertheless, we also test if “path dependent choices” influence the results by
restricting the sample to communities that consolidated after 1995. A cut-off period after
1995 results in too few observations for pairs of communities belonging to a public health
district. The probit regression results are shown in Table 3 for the two different matching
methods. Notice that the multiple regression results are very similar to those in Table 2
and suggest that the regression results are not greatly affected because the decision to join
or form a health district had been made many years earlier.10
9 One of the authors of this paper served on the board of finance of a town participating in a public health district. It was not unusual for discussion to take place about withdrawing from the district during annual budget deliberations. Discussion involved the current costs and benefits of remaining in the district. 10 Both of the estimated coefficients on the location variables are consistently positive and statistically significant for three out of four specifications. These results suggest that towns and cities located at a greater distance from both Hartford and the coast are more likely to consolidate their public health services, ceteris paribus. A glance at a Connecticut map of independent and consolidated districts shows that is definitely the case with few exceptions. For reasons, other than those already captured in the regression equations, towns and cities in the northeast and northwest corners of the state are more likely to cooperate or coordinate decisions regarding public health services.
18
5. SUMMARY
This paper offers some information regarding the factors influencing the regional
consolidation of public health services. Only Bates and Santerre (2008) empirically
examine the consolidation decision regarding public health services but they did not
consider if communities merge public health activities because of scale economies. This
study finds that a community considers both the potential for scale economies and
political externality costs when forming a public health district with a neighboring
community. In fact, the empirical results suggest that perceived diseconomies set in at
relatively low range of population. This may mean that consumer-voters underestimate
the true scale economies associated with the consolidation of public health services. Also
similar to Bates and Santerre, this study finds, not surprisingly, that towns lean more
towards the regional consolidation of public health services when their tax rates are
relatively high and the people in the town are relatively poor.
This paper also analyzes the decision of a community to join an existing public
health district. The empirical results indicate that the decision to join an existing district
is also directly related to tax share and inversely related to income. Moreover, the
findings indicate that medium-sized communities are more likely than small or large
communities to join an existing health district. Finally, empirical results for both the
single- and group-models, taken together, imply that jurisdictional heterogeneity matters
more for the decision to form a health district with an adjacent community than the
decision to join an existing health district. It may be that the averaging or smoothing of
characteristics across the towns and cities in a district makes a district more appealing
than any one individual town.
19
One interesting aspect of the results is the general similarity of the empirical
findings for both the local public school and public health department consolidation
decision. Both sets of research find that potential scale economies motivate and
heterogeneity hinders consolidation. The decision to consolidate public schools is
politically a much hotter issue than the consolidation of local health departments. More
people expect to benefit from their children being educated in the local school than they
expect to benefit from a local public health initiative such as a bioterrorism threat or
communicable disease outbreak. In addition, a much greater amount of money is spent on
local schools than local public health. Recall the small head tax for public health services
in Connecticut. In addition, Santerre (2009) reports that local public health departments
in the U.S. spent only $45 per person on average in 2005. The similarity of the empirical
findings may attest to the general applicability and richness of the underlying public
choice model.
While the empirical analysis in this paper is limited to health departments
Connecticut, the results may shed some light on the relatively small-sized local public
health departments, particularly in other New England states where county governments
do not exist, but also in other areas of the U.S. As noted earlier, Santerre (2009) finds that
the optimal size of a local public health department is around 100,000 people in terms of
minimizing public health costs per capita. However, NACCHO (2006) reports that 77
percent of all local public health departments, containing about 18 percent of the U.S.
population, operate below this efficient level of population. Various factors have
evidently prevented these small-sized local health departments from consolidating with
20
others to reap the cost-savings from greater size. Clearly, the results of this paper have
relevance beyond the state of Connecticut.
In the case of local public health departments, which are studied here, the
conventional wisdom is: “If you’ve seen one local health department, you’ve seen one
local health department” That is, local governments, lacking a national template
regarding an effective structure, organize their local public health departments in
countless ways (Tilson and Berkowitz, 2006). The myriad of organizational forms may
create challenges for the coordination of the entire national system particularly in a time
period facing bioterrorist threats and natural disasters such as Hurricane Katrina. If so,
and if national uniformity is valued by society, regional consolidation may provide one
way of moving towards that goal. However, the results from this paper suggest voluntary
movement may be limited because of differences across communities and a relatively low
perceived minimum efficient scale for public health services. The implication is that
higher levels of government might want to use financial incentives to induce more local
governments to form regional public health districts.
21
ACKNOWLEDGEMENTS
We thank the discussants and participants at the 2008 annual meeting of the Eastern Economics Association and the anonymous referees of this journal for their helpful comments. REFERENCES
Adelaja, S, Racevskis, LA. Cooperation costs and the economics of intergovernmental partnerships. Political Science Working Group on Interlocal Services Cooperation. Wayne State University, 2005. Alesina, A, Spolaore, E. On the number and Size of Nations. The Quarterly Journal of Economics. 1997; 112: 1027-1056. Alesina, A., Baqir, R., Hoxby, C. Political jurisdictions in heterogeneous communities. Journal of Political Economy. 2004; 112: 348-396. Bates, L.J., Santerre, R.E. The demand for local public health: do unified and independent public health departments spend differently. Medical Care. 2008; 46: 590-596. Bergstrom, TC, Goodman, RP. Private demands for public goods. American Economic Review. 1973; 63: 280-96. Borcherding, TE, Deacon, RT. The demands for the services of non-federal governments. American Economic Review. 1972; 62: 891-901. Borck, R. Centralization of public good supply with majority voting. Finanzarchiv N.F. 1998; 55: 21-40. Boyne, GA. Bureaucratic theory meets reality: public choice and service contracting in U.S. local government. Public Adminstration Review. 1998; 58: 474-484. Brasington, DM. House prices and the structure of local government: an application of spatial statistics. Journal of Real Estate Finance and Economics. 2004; 29: 211-231. Brasington, DM. Size and school district consolidation: do opposites attract? Economica. 2003; 70: 673-690. Brasington, DM. Joint provision of public goods: the consolidation of school districts. Journal of Public Economics. 1999; 73: 373-393. Brown, TL, Potoski, M. Contract-management capacity in municipal and county governments. Public Administration Review. 2002; 63: 153-164.
22
Downs, A. An Economic Theory of Democracy. New York: Harper and Row; 1957. Ellingsen, T. Externalities vs internalities: A model of political integration. Journal of Public Economics. 1988; 68: 251-268. Feiock, RC. Rational choice and regional governance. Journal of Urban Affairs. 2007; 29: 47-63. Ferris, JM, Graddy, E. Organizational choice for public service supply. The Journal of Law, Economics, and Organizations. 1994; 10: 126-141. Ferris, J, Graddy, E. Production choices for local government services. Journal of Urban Affairs. 1988; 10: 251-268. Fischer, R. State and Local Public Finance. Mason, Ohio: Thomson/Southwestern Publishers, 2007. Gerber, ER and Gibson, CC. Cooperative Municipal Service Provision: a political-economy framework for understanding intergovernmental cooperation. Political Science Working Group on Interlocal Services Cooperation. Wayne State University, 2005. Gordon, N., Knight, B. The causes of political integration: an application to school districts. NBER Working Paper No. 12047, 2006. Hartford Courant Editorial. Do more with fewer: regional consolidation of agencies could bring regional efficiency. Hartford Courant, March 1, 2009. www.courant.com/news/opinion/editorials/hc-anning.art.artmar01,0,2368945.story. Hicks, J. Sprague leads the effort to create a new health district. Norwich Bulletin. 2004: c4. Honadle, BW. The barriers to citizen-led consolidations: an analysis of Minnesota’s municipal boundary adjustment law. Hamline Journal of Public Law and Policy. 1995; 17: 63-82. Honadale, BW. Projecting the public services and finance implications of municipal consolation: evidence from a small-city consolidation. The Regionalist. 1998; 41-53. Joassart-Marcelli, P, Musso, J. Municipal service provision choices within a metropolitan area. Urban Affairs Review. 2005; 40: 492-519. Krueger, S, McGuire, M. A transaction costs explanation of interlocal government collaboration. Political Science Working Group on Interlocal Services Cooperation, Wayne State University, 2005.
23
Martin, DT, and McKenzie, RB. Bureaucratic profits, migration costs, and the consolidation of local government. Public Choice. 1975; 23: 95-100. National Association of County and City Health Officers (NACCHO). 2005 National Profile of Local Health Departments. Washington D.C.: NACCHO, 2006. Nelson, MA. Municipal government approaches to service delivery: an analysis from a transactions cost perspective. Economic Inquiry. 1997; 35: 82-96. Pace, RK, R Barry, and Sirmans, CF. Spatial statistics and real estate. Journal of Real Estate Finance and Economics. 1998; 17: 5-13. Penny, J. Health district faces shortfall: department asks towns for additional funds. Norwich Bulletin, February 0, 2009. www.norwichbulletin.com/archive/x84121968/Health-district-faces-shortfall. Santerre, RE. Jurisdiction Size and Local Public Health Spending. Health Services Research (forthcoming 2009) Sorensen, RJ. Local government consolidation: the impact of political transaction costs. Public Choice. 2006; 127: 75-95 Studenmund, A. H. Using Econometrics. Boston: Pearson/Addison Wesley, 2006. Tilson, H., Berkowitz, B. The public health enterprise: examining our twenty-first-century policy challenges. Health Affairs. 2006; 25: 900-910. Turnock, B.J., Public Health: What is it and How It Works. Sudbury, MA: Jones and Bartlett Publishers, 2004.
24
Table 1: Descriptive Statistics Variable Single-Model
(N=878) Group-Model (N= 268)
Mean Value (standard deviation)
Consolidated 0.246 (0.43)
0.340 (0.47)
Tax-Price 16.34 (4.18)
17.02 (4.24)
Median Household Income 64167 (19493)
62376 (17571)
Fraction Old 0.133 (0.04)
0.133 (0.04)
Fraction White 0.907 (0.11)
0.904 (0.11)
Land Area (square miles) 29.44 (12.3)
28.29 (12.1)
Distance to Hartford 28.03 (13.0)
25.85 (12.4)
Distance to Coastline 25.10 (16.5)
27.40 (16.2)
Community Population 20212 (24018)
20902 (16.17)
Absolute Difference of Tax-Price 3.309 (3.4)
2.932 (2.9)
Absolute Difference of Median Income 14999 (14217)
13621 (11332)
Absolute Difference of Population 18119 (23561)
15883 (18640)
Absolute Difference of Fraction White 0.075 (0.11)
0.064 (0.09)
Absolute Difference of Land Area 12.457 (9.7)
9.983 (7.6)
Combined Population 40601 (38023)
93721 (43926)
25
Table 2: Results from the Probit Estimation (Dependent Variable = 1 if Consolidated; 0 if not) Single-Model Group-Model Estimated Coefficient
(Absolute Value of Z-statistic) Constant -4.162*
(3.37) 0.097 (0.04)
Tax-Price 0.083* (3.79)
0.069* (1.69)
Median Household Income -1.18E-05* (2.61)
-1.53E-05* (2.13)
Fraction Old 4.385* (2.85)
7.756* (2.65)
Fraction White 0.534 (0.486)
-0.631 (0.34)
Land Area 0.0007 (0.15)
-0.002 (0.18)
Distance to the Hartford
0.024* (3.88)
0.036* (3.33)
Distance to the Coastline 0.035* (8.39)
0.041* (4.80)
Community Population -1.70E-05 (1.11)
7.10E-05* (2.80)
Square of Community Population 9.48E-11 (0.39)
-1.05E-09* (2.42)
Absolute Difference of Tax-Prices 0.006 (0.22)
-0.070 (1.28)
Absolute Difference of Median Incomes
-1.95E-05* (3.00)
-2.14E-05 (0.17)
Absolute Difference of Populations -4.78E-06 (0.67)
2.01E-06 (0.14)
Absolute Difference of Fractions White -1.308 (1.49)
-1.480 (0.69)
Absolute Difference of Land Areas -0.011* (2.06)
-0.029* (2.17)
Combined Population 5.68E-05* (4.40)
-5.17E-05* (3.73)
Square of Combined Population -5.96E-10* (4.82)
1.27E-10* (1.83)
Number of Observations 878 268 McFadden R-Squared 0.234 0.332 * statistically significant at the 10 percent level or better.
26
27
Table 3: Results from the Logit Estimation (Dependent Variable = 1 if Consolidated; 0 if not) Single-Model
Consolidations After 1995
Group-Model Consolidations After 1995
Estimated Coefficient (Absolute Value of Z-statistic)
Constant -2.589 (1.37)
-1.710 (0.50)
Median Tax-Price 0.064* (2.00)
0.139* (1.98)
Median Household Income -1.44E-05* (2.36)
-3.37E-05* (3.38)
Fraction Old -2.318 (0.92)
4.996 (0.96)
Fraction White 0.287 (0.22)
3.114 (0.98)
Land Area 0.013* (2.10)
0.017 (1.22)
Distance to Hartford 0.002
(0.22) 0.043* (2.26)
Distance to Coastline 0.017* (3.53)
0.030* (2.16)
Community Population -1.20E-05 (0.53)
0.0001* (2.58)
Square of Community Population -1.65E-10 (0.72)
-2.69E-09* (2.02)
Absolute Difference of Tax-Prices 0.0106* (2.78)
0.009 (0.09)
Absolute Difference of Median Incomes -3.81E-05* (3.78)
-2.26E-05 (1.03)
Absolute Difference of Populations -3.30E-06* (3.40)
-3.05E-05 (1.25)
Absolute Difference of Fractions White -0.446 (0.35)
3.353 (1.04)
Absolute Difference of Land Areas -0.019* (2.35)
-0.040* (2.03)
Combined Population 6.55E-05* (3.88)
-0.0001* (4.93)
Square of Combined Population -6.13E-10* (3.88)
4.78E-10* (4.31)
Number of Observations 720 201 Number of Consolidating Pairs 58 24 McFadden R-Squared 0.230 0.435 * statistically significant at the 10 percent level or better.