MUNICIPAL GOVERNANCE AND INTERNAL CONTROL DEFICIENCIES:
DOES STRUCTURE MATTER?

By

Dara Marie Marshall

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Accounting

2012

ABSTRACT

MUNICIPAL GOVERNANCE AND INTERNAL CONTROL DEFICIENCIES:
DOES STRUCTURE MATTER?

By

Dara Marie Marshall

This study examines the association between governance and internal control deficiencies in the
municipal government sector using cities with populations over 50,000. I hypothesize that
municipalities with governance structures consistent with greater allocative efficiency (i.e., how
well services produced match citizens’ preferences) and lower productive efficiency (i.e., how
effectively services are produced and provided given a level of resources) will have more
internal control deficiencies. To test my hypothesis, I develop a new measure of municipal
governance based on the following eight governance mechanisms: form of government, presence
of a chief appointed official, shared budget-setting authority, election or appointment of
department heads, direct democracy, council voting power of chief elected official, presence of
chief elected official on council, and council terms. As predicted, I find that city governments
with a combination of governance mechanisms consistent with greater allocative efficiency are
more likely to report an internal control deficiency than cities that have adopted governance
mechanisms consistent with greater productive efficiency.

DEDICATION

I dedicate this dissertation to my parents William and Joyce, my four sisters: Cheshana, Shelley,
Amina, and Ashira, and my brother William.

iii

ACKNOWLEDGEMENTS

I thank my dissertation committee members for their patience and diligence, but
specifically Matt Anderson for his support and encouragement, Ted Fee for introducing me to
the finance literature, John Jiang (Chair) for guiding me through this process, and Marilyn
Johnson for her insightfulness in the writing process. I thank K. Ramesh for spurring me to
pursue this area and Bill Kinney for his helpful comments on an earlier related project. I am
grateful to Ge Bai, Michelle Lau, Sonja Perakis and Joe Schroeder for their support. I appreciate
the comments from Sanjay Gupta, Chris Hogan, Ranjani Krishnan, Mike Shields, Kris Allee,
Isabel Wang, Andy Acito, and Dan Wangerin. I thank Sue Convery for her help and expertise in
this area. I also thank Matt DeAngelis, Eric Marinich, Colleen Boland, and Tyler Thomas for
their comments on previous versions. I am grateful for the comments from workshop participants
at Miami University, Rutgers, the University of Denver, Ball State University, and the University
of Memphis. I also gratefully acknowledge the financial support provided by Michigan State
University through a Dissertation Completion Fellowship.

iv

TABLE OF CONTENTS

LIST OF TABLES

vi

LIST OF FIGURES

vii

CHAPTER I: INTRODUCTION

1

CHAPTER II: BACKGROUND AND LITERATURE REVIEW
2.1 Background
2.2 Literature Review
2.2.1 Prior Research on Internal Controls
2.2.2 Corporate versus Municipal Governance
2.2.3 Municipal Governance and Efficiency

7
7
8
8
11
12

CHAPTER III: HYPOTHESIS DEVELOPMENT

15

CHAPTER IV: METHODS
4.1 Dependent Variable
4.2 Governance Structure
4.3 Control Variables

23
23
24
27

CHAPTER V: SAMPLE SELECTION AND DATA

29

CHAPTER VI: RESULTS
6.1 Primary Tests
6.2 Supplemental Analyses

31
31
34

CHAPTER VII: CONCLUSION

39

APPENDICES
Appendix A: Municipal Governance Mechanisms
Appendix B: Latent Class Analysis

62
63
69

REFERENCES

73

v

LIST OF TABLES
TABLE 1: Latent Class Analysis

43

TABLE 2: Variable Definitions

44

TABLE 3: Sample Description

46

TABLE 4: Descriptive Statistics of No Reportable Condition Firm-Year
versus Reportable Condition Firm-Years.

47

TABLE 5: Correlation Matrix

49

TABLE 6: Determinants of Internal Control Deficiencies and Tests of
the Effect of Governance Structure

52

TABLE 7: Determinants of Internal Control Deficiencies Using Form
of Government Measure

54

TABLE 8: Comparison of Governance Measures

56

TABLE 9: Incidence of At Least One Reportable Condition over Time

58

TABLE 10: Poisson Regression of Incidence of Reportable Conditions
over a Four-Year Period

60

vi

LIST OF FIGURES
Figure 1: Governance Mechanisms and Influence
on Incentives and Organizational Missions

vii

72

CHAPTER I: INTRODUCTION
In this study, I examine the relation between municipal governance structure and internal
controls. Municipalities receive a significant amount of funding from the federal government. In
the fiscal year 2010, the federal government had direct expenditures to state and local
governments of over $630.2 billion (U.S. Census 2010). With the new Budget Control Act of
2011, municipalities will face decreased federal assistance, which could increase financial
distress. Also, municipalities are facing financial constraints given declining tax bases and
reduced funding from state governments. Increased financial distress could lead to internal
control deficiencies as it does in for-profit firms (Ashbaugh-Skaife et al. 2007; Doyle et al. 2007;
and Ogneva et al. 2007). Poor internal controls provide opportunities for fraud (Garrison et al.
2011; MacDonald 2011).
I examine whether the government structure, which I define as the combination of
governance characteristics, affects the control environment. The city of Detroit recently had a
change in leadership precipitated by numerous scandals involving the former mayor. Some of
these scandals involved the misuse of city resources. When the new mayor was elected, his
business background as the chief executive officer of his own successful firm was expected to
bring professionalism and better management to the city. Despite high expectations, the city
continues to face financial issues. For example, the city's health department has problems with
internal control as evidenced by an employee writing herself a grant for just under $5,000. The
leadership changed but the government structure did not.
Municipalities govern themselves in different ways in order to meet the preferences of
their citizens and other stakeholders. These preferences can be particular preferences of different
groups of citizens or stakeholders or they can be community wide preferences. The preferences

1

of particular groups do not always overlap with each other, community wide preferences, or the
long-term interests of the community. Frant (1996) defines the matching of these two different
types of preferences as allocative efficiency and productive efficiency. Municipalities that are
allocatively efficient meet the particularistic preferences of various citizen or stakeholder groups.
Municipalities that are productively efficient meet preferences of the city as a whole, which
usually includes long-term goals and projects and delivering services in a cost-effective manner.
Cities achieve these different efficiencies by combining different governance mechanisms, which
I define as municipal governance structure.
American municipal governance structures are consistent with different mixes of
allocative efficiency – meeting multiple particularistic preferences – and productive efficiency –
meeting the needs of the city as a whole. For example, Lansing, Michigan has the following
combination of governance characteristics: a strong-mayor form of government, a city council
elected by district, an at-large, budgetary authority that rests with the mayor, a chief appointed
officer (CAO), department heads that are appointed by the mayor, and staggered council terms;
and the mayor is not a member of council nor can vote on council issues. By having a directly
elected mayor with the authority to propose the budget and hire his/her own administration, the
city of Lansing is more responsive to various interest groups. For example, the mayor can
promise to appoint the head of an interest group to a position in his/her administration for
support by that interest group in an election.
This is in contrast to the city of Grand Rapids, Michigan, that has a governance structure
comprising these following governance characteristics: a directly elected mayor with no
appointment or budgetary power, a chief appointed officer – the city manager – that is appointed
by city council, and appointed department heads that serve the city manager; and the mayor is a

2

member of council that can vote in the event of a tie. The operating decisions of the city, such as
the appointment of department heads and setting the budget, are more shielded from the
influence of interest groups since the city manager is appointed by the council and not directly
elected. Even though individual council members may still try to meet particular citizen
preferences in order to get elected, the choice of city manager still rests on a majority vote of
council and thus should reflect community wide preferences versus particularistic preferences of
different citizen or stakeholder groups. By giving the city manager powers such as sole budget
authority and appointment powers for department heads as well as strengthening the power of
council by having election terms staggered and unifying power under council, this governance
structure is less allocatively efficient.
Municipal government structure arises to meet varying preferences for allocative and
productive efficiencies but these structures, the combination of governance mechanisms, affect
components of the control environment. In the Detroit example, the leadership of the city
changed but the structure did not. The structure of the city, with its combination of governance
mechanisms such as the mayor solely proposing the budget and having his/her own staff separate
from council, decreases communication. By the mayor being directly elected and separate from
council, each branch is competing for votes and has lower incentive to monitor the other. This
combination of governance mechanisms, the governance structure, creates a lower control
environment leading to a higher incidence of auditors detecting internal control deficiencies. I
test this hypothesis by examining the incidence of reportable conditions related to the financial
statements as detected during the A-133 audit as a function of four governance structures. I do
this while controlling for other characteristics of cities such as complexity, size, financial risk,
revenue diversity, and auditor characteristics.

3

The examples of Lansing and Grand Rapids provide insight into how governance
structures designed to meet different preferences for allocative and productive efficiencies can
affect internal control. Lansing has had at least one significant deficiency each year for the past
four years, with at least three being material weaknesses. These deficiencies include no
segregation of duties and multiple adjustments that collectively caused material misstatement in
the financial statements. The causes cited include consolidation of positions in the accounting
department and lack of staff in the finance department. Grand Rapids has not had a significant
deficiency in internal control since 2008. In its 2008 Single Audit, the significant deficiency was
related to timeliness of departments reporting to the Comptroller for financial statement
preparation. There wasn’t any evidence that departments did not communicate with one another.
This seems reasonable given that all department heads report to the city manager who reports to
the council. There is no department within the city that is outside of that chain of command. This
is in contrast to Lansing, where some departments are under the control of the mayor or his/her
staff and others are under the council or a directly elected department head. Having different
directly elected officials with their own staffs creates the potential of using appointments as
political favors to gain reelection. This decreases productive efficiency and decreases the control
environment as there are fewer resources dedicated to city wide needs such as financial
reporting.
I test my hypothesis using a governance measure developed in Marshall (2010). Based on
latent class analysis (LCA) of eight governance mechanisms (i.e., form of government, presence
of a CAO, sole budget responsibility of the mayor, election or appointment of department heads,
direct democracy, council voting power of chief elected official (CEO), presence of CEO on the
council, and council terms), I identify four governance classes that reflect voter preferences for

4

different levels of allocative efficiency. Since governance mechanisms develop jointly
(Armstrong et al. 2010), using a measure that accounts for different combinations of mechanisms
addresses the endogeneity problem of relating governance to accounting outcomes. In the case of
municipalities, governance mechanisms arise to meet the needs of the citizenry but there are
tradeoffs. I posit that there is a tradeoff between responsiveness to the citizenry through
allocative efficiency and investment in control systems through productive efficiency.
Using a sample of municipalities with populations over 50,000, I model the probability of
disclosing an internal control deficiency as a function of governance structure and municipal
characteristics analogous to the characteristics examined in the literature on publicly traded
firms. Consistent with my hypothesis, I find that municipalities that have governance structures
comprising a combination of mechanisms that reflect preferences for higher allocative efficiency
are more likely to have internal control deficiencies than municipalities with mechanisms that
reflect lower allocative efficiency. I also find that municipalities that are more complex, larger,
and riskier are more likely to have an internal control deficiency consistent with prior literature
on firms and not-for-profit organizations.
I contribute to the accounting literature in two ways. First, I examine the relation between
governance and internal control using a measure of governance structure that is consistent with
preferences for various levels of allocative and productive efficiencies. Prior studies that
examine governance and internal control use governance indexes or individual governance
mechanisms. This ignores the endogenous nature of governance, since governance mechanisms
and an organization’s information environment are simultaneously determined (Armstrong et al.
2010). Since different combinations of governance mechanisms are observed in different

5

information environments, my use of a governance measure based on LCA addresses the issue of
combinations of municipal governance mechanisms being endogenously determined.
Second, my analysis provides a theoretical explanation of how governance affects
internal controls through the efficiency issues that governance structure solves. Political science
theory provides an explanation for the broad set of governance mechanisms observed in the
municipal setting, whereas explanations for governance in the corporate or not-for-profit settings
are varied and still being developed. Most U.S. municipalities modeled themselves after the
federal government, but many have changed their governance structures in response to various
problems, including corruption and patronage (Frederickson et al. 2003). Municipalities
continue to adopt governance mechanisms in response to tradeoffs between levels of allocative
efficiency, (i.e., how much its citizens will value the services produced) and productive
efficiency (i.e., how effectively services are produced and provided given a level of resources)
(Frant 1996; Frederickson et al. 2004; Svara and Watson 2010). An analogue in corporate
governance today is the call by institutional investors and other stakeholders of corporations such
as unions for more influence in corporate governance. This is one of the many reforms included
in the Dodd-Frank Wall Street Reform and Consumer Protection Act which gives the Securities
and Exchange Commission authority to make rules to give shareholders a greater role in
corporate governance.
The rest of the study is organized as follows. Section two describes background
information on municipal governance and internal control. Section three reviews relevant
literature and develops my hypotheses. Section four describes the methods used, and section five
outlines my sample selection and data. Section six presents the results, and section seven
concludes.

6

CHAPTER II: BACKGROUND AND LITERATURE REVIEW
2.1 Background
For cities, the evaluation and disclosure of internal controls is mandated by the Single
Audit Act of 1984, its subsequent amendments, and the Office of Management and Budget
(OMB) Circular A-133. If a city receives enough federal funding ($100,000 from 1984 until
1996; $300,000 from 1996 until 2003; and $500,000 from 2003 until the present), it is subject to
provisions in the Act that mandate an audit. This audit includes a report on internal control over
financial reporting and major programs. The report on internal controls for a Single Audit is not
the same as what is disclosed under Sections 302 and 404 of the Sarbanes-Oxley Act. Under the
Single Audit Act, auditors are not required to opine on the internal controls of an organization
but rather describe:
“The scope of the auditors’ testing of internal control over financial reporting and
compliance with laws, regulations, and provisions of contracts or grant agreements…
Auditors should state in the reports whether the tests they performed provided sufficient,
appropriate evidence to support an opinion on the effectiveness of internal control over
financial reporting and on compliance with laws, regulations, and provisions of contracts
or grant agreements.” (GAO Government Audit Standards 2007)
In addition to describing the scope of the testing of internal controls, the report must also contain
any significant deficiencies, material weaknesses, and non-compliance.
Though the Single Audit Act has existed for almost three decades, there have been only a
handful of studies in accounting examining internal controls in cities (Wallace 1981; Raman and
Wilson 1992; Lopez and Peters 2010). Wallace’s (1981) study examines information about
internal controls contained in the financial disclosures of municipalities. The author finds that

7

disclosures about internal controls generally do not provide information on risk exposure or costs
and benefits, but her study came before the Single Audit Act and the formation of the
Government Accounting Standards Board. Raman and Wilson (1992) examine the effect of
mandating Single Audits on audit fees and find that cities that have internal control deficiencies
and material non-compliance have higher audit fees, but they do not find consistent results for
single audits increasing audit fees. Lopez and Peters (2010) use internal control deficiencies and
material weaknesses disclosed in Single Audits as a measure of audit quality and examine the
relationship between auditor type and audit quality. The authors find that audit quality in cities
and counties is higher when city and county clients use Big-4 CPA firms or large audit firms,
rather than government auditors. None of these studies examine the determinants of internal
control deficiencies in cities.
2.2 Literature Review
In this section, I review the literature on governance and internal controls. I compare
corporate and municipal governance. I also explain how U.S. municipal governance has changed
through time and the reasons for the changes. I conclude by developing my hypothesis about
how governance structure is consistent with levels of allocative and productive efficiencies,
which in turn affect internal control.
2.2.1 Prior Research on Internal Controls
In the for-profit setting, recent accounting research examines the evaluation and reporting
of internal controls required under Sections 302 and 404 of the Sarbanes-Oxley Act of 2002
(Schneider et al. 2009). Section 302 went into effect in August of 2002 and includes a mandate
that management certify that they “have evaluated the effectiveness of internal controls and that
they include in their report their conclusions on the effectiveness of internal controls” (Schneider

8

et al. 2009). Ge and McVay (2005) use data from the newly implemented disclosures under
Section 302 to find that firms with material weaknesses disclosed under Section 302 tend to be
more complex, have a large auditor, are smaller, and have lower return on assets. AshbaughSkaife et al. (2007) find results similar to Ge and McVay (2005). In addition to concluding that
smaller and more complex firms that engage a large auditor are more likely to have internal
control deficiencies, they also find that financially distressed firms with prior restatements and
merger and acquisition activity have a higher incidence of internal control deficiencies.
After the reporting requirements under Section 404 of the Sarbanes-Oxley Act (SOX)
went into effect at the end of 2004 for large accelerated filers, other studies examined the
determinants of internal control deficiencies. These studies found results similar to past studies;
firms with losses, financial distress or risk, greater complexity (more segments), and high sales
growth had higher incidence of internal control deficiencies (Krishnan and Visvanathan 2007;
Ogneva et al. 2007).
There is also limited research on the determinants of internal control quality in the notfor-profit setting. Required internal control evaluations for state and local government and
governmental agencies are mandated in the Single Audit Act of 1984. Through a 1990
administrative action, the OMB extended the scope of the Single Audit Act to not-for-profit
organizations with the issuance of OMB Circular A-133. Two accounting studies have examined
audits under the Single Audit Act of not-for-profits. Keating et al. (2005) conduct univariate
analysis to determine the characteristics associated with adverse Single Audit Findings in notfor-profits. Their study finds that smaller not-for-profits and those not qualified as low-risk have
more adverse Single Audit Findings. They also find that not-for-profits employing a Big-5 CPA
firm had fewer reportable conditions, but those employing a large regional firm disclosed more

9

reportable conditions. Petrovits et al. (2011) extend the Keating et al. (2005) analysis with
multivariate tests of the determinants of reportable conditions and material weaknesses. They
also examine the impact of having reported internal control problems on donations. Those
authors find that charities with reportable conditions in internal controls over financial reporting
are smaller, complex, growing, and facing financial distress. They also find that charitable
donations and government contributions decrease after these reportable conditions are disclosed.
The relationship between governance and internal controls has also been examined in
both the for-profit and not-for-profit literature. Krishnan (2005) examines the role of audit
committees on internal control quality and finds that having more financial experts on the audit
committee is associated with fewer internal control issues. The author also finds that the more
independent an audit committee, as measured by the proportion of independent directors, the
fewer reported internal control issues there are in the firm’s 8-Ks. Similar to Krishnan (2005),
Zhang et al. (2007) examine audit committees and their relationship with internal control quality
using audit committee independence and expertise. They extend Krishnan (2005) by including
other audit committee characteristics such as audit committee size, audit committee meetings,
audit committee accounting expertise, and audit committee non-accounting financial expertise.
The authors also examine corporate governance characteristics such as board size, board
meetings, and a dichotomous governance measure based on the median of the sum of five
governance characteristics. All governance characteristics except their overall governance
measure are examined individually. They find that firms with more independent audit
committees and greater financial expertise are less likely to have internal control deficiencies.
The authors do not find consistent results for their overall governance measure. Hoitash et al.
(2009) examine various audit committee characteristics individually but also include a

10

governance measure that is a sum of five governance characteristics. Their results are consistent
with prior research. They find that their governance measure is negatively associated with
Section 404 material weakness disclosures. Doyle et al. (2007) also use a governance measure
that is the sum of various indicators for governance mechanisms. Their governance measure is
significant in only one of their analyses.
All of these studies examine individual governance characteristics or use governance
indexes. By examining governance characteristics individually or by using governance indexes
that are sums of various characteristics, the co-occurrence of various mechanisms and the
endogenous nature of combinations of governance mechanisms are ignored. Armstrong et al.
(2010) explain that governance mechanisms and a firm’s information environment are
simultaneously determined and thus different governance mechanisms and information
environments should be seen together for a particular firm. They cite two papers (Larcker et al.
2007; Dey 2008) that create better measures of corporate governance by using factor analysis to
determine the dimensions of corporate governance. I contribute to a better understanding of the
association between governance and internal control by using a governance measure that
attempts to address the issue of various governance mechanisms being endogenously determined.
2.2.2 Corporate versus Municipal Governance
Zimmerman (1977) argues that corporate governance and municipal governance differ in
control and monitoring. Citizens are the residual claimants of a municipality, similar to
shareholders, but due to high transaction costs and low individual benefits of voting, monitoring
and control by municipal residents is low. In addition to shareholders in corporations facing
lower transaction costs than citizens, shareholders can concentrate ownership through buying
multiple shares whereas citizens cannot. Another difference between municipal governance and

11

corporate governance is form. Corporate managers are hired and monitored by boards of
directors that are elected by shareholders. Some municipalities have a similar form where a city
council is elected by the citizens of the city and that council monitors and hires a city manager
for the city. Other municipalities have a form of government where the citizens directly elect a
mayor who manages the city, as well as a council that acts as the legislative branch of the city.
The mayor and council are monitored by each other as well as the citizens. The comparison of
corporate governance to municipal governance highlights the complexity of the governance
construct. Governance is multidimensional (Armstrong et al. 2010), but the current literature on
internal controls does not address the endogenous co-occurrence of governance mechanisms.
2.2.3 Municipal Governance and Efficiency
Various governance mechanisms have developed in cities to deal with different
efficiency issues. In the nineteenth century, most United States municipal governments adopted
governance mechanisms to model the federal government with a separation of powers between
executive and legislative branches (Frederickson et al. 2003). One mechanism is form of
government. One type of form of government is the mayor–council form where there is a directly
elected mayor who manages the city and a separately elected council. One consequence of this
governance mechanism is that it leads to a less powerful government. Toward the turn of the
nineteenth century, many cities responded to this lack of concentration of power in municipal
government by strengthening the power of their mayors through various governance mechanisms
such as giving the mayor power to appoint department heads (Renner and DeSantis 1993).
At the beginning of the twentieth century, the municipal reform movement began, and
many cities adopted reforms such as civil services systems, city manager forms of government,
direct democracy measures such as initiative and referendum, and at-large elections in an attempt

12

to make cities more professional and mirror the corporate form. These mechanisms were meant
to combat corruption, patronage, and political machines, but there are tradeoffs. Frederickson et
al. (2004) point out: “The structural characteristics of the reform project made it more difficult
for disadvantaged groups to participate in local politics.” Some cities that reformed have
subsequently re-adopted mechanisms such as elections by district, directly elected mayors, and
increasing mayoral powers to make their governments more accessible and responsive to all their
citizens (Frederickson et al. 2004; Svara and Watson 2010). In addition, some cities that did not
reform adopted the governance mechanisms of reformed cities to become more professional and
efficient. An example of the latter is the hiring of a CAO.
Prior research that examines the effect of municipal governance on various accounting
outcomes uses a dichotomous measure of governance: whether a city has a mayor–council form
of government or a council–manager form of government (i.e., a government with no separation
of powers between mayor and council and an appointed city manager) (Zimmerman 1977, Evans
and Patton 1987; Ward et al. 1994; Vijayakumar 1995; Giroux and McLelland 2003).
Zimmerman (1977) examines differences in disclosure and auditors between mayor–council and
council–manager forms of government. He finds that council–manager cities are more likely to
be audited, use large national auditors, and have longer annual reports than mayor–council cities.
Evans and Patton (1987) model municipal disclosure and test whether monitoring and signaling
motivate disclosure as proxied by participation in the Certificate of Conformance program. The
authors use several variables including form of government to proxy for monitoring incentives
and find significance for form of government but not the other monitoring proxies. Vijayakumar
(1995) examines the determinants of call decisions for municipal bonds. He uses the form of
government to test whether incentives to pander to interest groups influence call decisions but

13

does not find significant results for that measure. Giroux and McLelland (2003) build upon
Zimmerman’s (1977) study by examining differences in financial performance between different
forms of government in addition to differences in disclosure. Using two time periods, they find
that council–manager forms of government have better quality disclosure and financial
performance.
The evolving nature of municipal governance implies a need for a measure that includes
additional mechanisms other than form of government. In a descriptive study, Frederickson et al.
(2004) find that cities choose from governance mechanisms such as having a CAO, whether the
mayor has sole budget authority, appointment versus elected department heads, and having direct
democracy provisions such as initiative, referendum, or recall, in order to balance allocative
efficiency and productive efficiency. Few accounting studies have used these and other
governance mechanisms. Gore’s (2009) analysis of the determinants of municipal cash balances
includes controls for whether the CEO is on the council, whether that official votes on the
council, and whether the council members’ terms are staggered versus concurrent. She finds that
excess cash balances, a proxy for high agency costs, are higher in cities where the CEO is on the
council, the CEO votes on the council, and the council has staggered instead of concurrent terms.
In a working paper, Baber et al. (2010) use twelve governance mechanisms including voter
oversight and staggered council terms. Many of the measures they use are outside of the control
of cities such as their state oversight measures and political competition measure, which differ
from the mechanisms I use in this paper. Also, the authors’ restatement model examines the
individual effect of each mechanism and does not consider how these governance mechanisms
interact.

14

CHAPTER III: HYPOTHESIS DEVELOPMENT
Several studies (Frederickson et al. 2003; Renner and DeSantis 1994; Frederickson et al.
2004) in public administration have found that “form of government” does not fully describe
governance of municipalities due to cities adopting different combinations of municipal
governance mechanisms. These authors attempt to describe municipal governance structure
through case studies and qualitative clustering of governance mechanisms. Frederickson et al.
(2004) surveyed 200 cities (116 responded) and by using 15 governance mechanisms, grouped
cities into five categories: political, adapted political, conciliated, adapted administrative, and
administrative. The authors did not construct their categories using any sort of quantitative
method such as cluster analysis. Instead, they used historical references, like the Model City
Charter, as a starting point for two of their governance categories, political cities and
administrative cities. The authors developed the other three categories by documenting trends in
how cities are adopting different governance mechanisms and then creating categories that fall
between political and administrative cities. Though Frederickson et al.’s (2004) classification is
informative; it is not based on a quantitative analysis of how governance mechanisms group
together. Thus, it is difficult to determine if their classification truly measures the underlying
construct of municipal governance structure.
My attempt at developing a governance measure using quantitative methods involves
using LCA on eight governance mechanisms in order to identify distinct governance structures.
Three of the mechanisms I use directly overlap with the mechanisms used in the Frederickson et
al. (2004) study: form of government, presence of a CAO, mayor on council. The other four
mechanisms I use have been examined in other studies: budgetary authority of the mayor
(DeSantis and Renner 2002), elected department heads (DeSantis and Renner 2002; Frederickson

15

et al. 2003), mayor votes on council (Gore 2009), and staggered or concurrent council terms
(Gore 2009).
Citizens’ preferences for certain levels of allocative and productive efficiencies affect
municipalities’ governance structures, which in turn affect municipalities’ internal control. The
Committee of Sponsoring Organizations of the Treadway Commission (COSO) defines internal
control as consisting of five components: control environment, risk assessment, control activities,
information and communication, and monitoring (COSO 1992). For each governance structure, I
will first discuss how preferences for allocative and productive efficiencies are expressed
through the combination of governance mechanisms chosen and then how the combination of
governance mechanisms affects two of the components of internal control: communication and
monitoring.
From the eight governance mechanisms, four governance structures, i.e., specific
combinations of mechanisms are the best fit for the LCA. In Table 1, structure one has the
following combination of characteristics: mayor–council form of government, no CAO, mayor
solely prepares the budget, appointed department heads, direct democracy measures, separation
of powers (the mayor is not a member of council and does not vote on council), and concurrent
council terms. This combination of mechanisms most closely resembles Frederickson et al.’s
(2004) political city category. Political cities are consistent with high allocative efficiency and
several of the mechanisms present in structure one (mayor–council form of government, no
CAO, and separation of powers) are consistent with allocative efficiency. In combination, these
eight mechanisms create an allocatively efficient governance structure by strengthening the
power of the mayor. In addition, the governance mechanisms of concurrent council terms
decrease the power of council by decreasing the continuity of council and potentially increasing

16

competition between council members. As a result, mayoral budgetary authority combined with
a weaker council gives the mayor power to appease various interest groups by providing city
resources to those groups. Some examples are the provision of services to certain neighborhoods
or economic development resources for businesses.
The combination of governance mechanisms that create governance structure one is
consistent with preferences for high allocative efficiency but also affects internal control through
its components of monitoring and communication. Governance structure one's combination of
mayor–council form of government with the mayor having sole authority to propose the budget
with no CAO, separation of powers, and a weaker council creates an environment where the
council has less incentive to monitor the delivery of city services. According to Zimmerman
(1977), the explanation for lower monitoring in this environment is that the costs and benefits of
the mayor's decisions are reduced for the council, since the mayor is directly elected by the
people and can thus be held accountable separately for his or her actions. The council can blame
the poor delivery of city services on the mayor and use their weaker power as an excuse for not
being able to do more for their citizens. Since the council’s and mayor's reelection hopes are not
directly tied to each other's performance, communication is lower as well. Svara (2001)
documents this in a survey of council members working under different forms of government.
He finds that council members in mayor–council forms of government feel that communication
is lesser between themselves and the mayor.
Governance structure two differs from structure one in a few ways. Though it has a
mayor–council form of government, there is a CAO and the mayor does not have sole authority
to propose the budget. Similar to structure one, the mayor is not a member of the council but in
this structure the mayor can vote on council. Also, the council's terms are staggered, which

17

increases the power of the council especially with the combination of the mayor not having sole
budget authority but having a CAO. Frederickson et al. (2003) find that in the majority of
mayor–council cities that have a CAO, the decision to hire the CAO is made jointly by the mayor
and the council. The combination of mayor–council form of government with a CAO and a
mayor who can vote on council is similar to either the adapted political or conciliated city
categories of Frederickson et al. (2004). These types of cities try to balance responsiveness to
their citizens with professional management of city services. In terms of efficiency, they would
be allocatively efficient, but less so than purely political cities and would be more productively
efficient than political cities.
This mixture of moderately high allocative efficiency and moderate productive efficiency
affects internal control through communication and monitoring in the following ways.
Communication is moderately high, since the mayor has to work with council on the decision to
hire and fire the CAO and does not have sole budget authority. Also, the mechanisms of the
mayor voting on council increase interaction with the council, which leads to more
communication. Monitoring may still be low in this structure since the council is still not directly
accountable for the actions of the mayor; however, since certain decisions are made jointly, such
as those on the CAO, there is a higher incentive for the mayor and council to monitor one
another compared to structure one.
The combination of governance mechanisms that compose governance structure three
(council–manager form of government with a CAO, no sole budget authority for the mayor, no
separation of powers since the mayor is on the council and votes on the council, and staggered
council terms) differs significantly from the other three governance structures. Structure three is
the only one that has a council–manager form of government and the mayor as a member of

18

council. Though this structure shares governance mechanisms with other structures, such as
having a CAO, having the mayor vote on council, and having staggered council terms, the
combination of these mechanisms with a different form of government changes the mix of
allocative and productive efficiencies. For example, structure two has a CAO, but because it also
has a mayor–council form of government, the CAO may be more of a political appointment by
the mayor and the overall structure of government is more allocatively efficient. On the other
hand, in council–manager forms of government, the council has the power to hire and fire the
CAO, which lowers the level of politics associated with the position. This makes structure three
less allocatively efficient since it is harder for interest groups to affect the choice of CAO. This is
because the hiring and firing of the CAO is a group decision instead of just the mayor's.
Staggered council terms combined with council–manager form of government also decrease
allocative efficiency in this structure by strengthening the power of the CAO. Because council
members’ terms are staggered, it is harder for interest groups to shift power on the council in
order to fire a CAO who is not catering to their particularistic preferences. In terms of the
Frederickson et al. (2004) framework, this structure resembles that of administrative cities. These
types of cities are known for professional management, efficiency in service delivery, and
collaborative government (Newland 1995), which is consistent with high productive efficiency.
The description of administrative cities and the shared combination of mechanisms
common to structure three and administrative cities imply a highly communicative environment
with greater monitoring, which leads to better internal control. Zimmerman (1977) hypothesizes
higher monitoring in council–manager forms of government because the consequences of a
manager’s (i.e., CAO) actions are concentrated in the council and thus the councils have a
greater incentive to monitor city managers. Since there is no separation of powers in this

19

government structure and most decisions are made either by the council or the city managers
(who answer directly to council) and their department heads, communication is higher than in
structures one, two, and four.
Governance structure four is characterized by the following combination of governance
mechanisms: mayor–council form of government with no CAO and no sole budget authority for
the mayor, the mayor is not on council but can vote on council, and the council terms are
concurrent. This structure is similar to structure one except that the mayor does not have sole
budget authority and the mayor can vote on council. By taking away the sole budgetary authority
of the mayor, his ability to make promises to different constituent groups is less even though he
can vote on council. Giving the mayor power to vote on council does enable him some authority;
however, he is still only one vote. Thus, the addition of this governance mechanism with the
deletion of the sole budgetary mechanisms does not make up for the loss of power. In terms of
the Frederickson et al. (2004) classification, this structure is similar to the adapted political city
category, and thus structure four has greater allocative efficiency than productive efficiency.
These cities have lower allocative efficiency than purely political cities because the powers of
the mayor have been decreased through not having sole budgetary authority.
The mix of allocative and productive efficiencies consistent with structure four creates a
governance structure that has decreased monitoring and communication. Structure four is similar
to structure one with separation of powers and no CAO; so, the same issues with monitoring and
communication would be expected. However, structure four does differ from structure one in the
budgetary authority of the mayor and the mayor's role on council. These two mechanisms in
combination with the other six governance mechanisms should decrease the negative effect on

20

communication and monitoring since the mayor shares budgetary and legislative decision
making with others including the council.
As discussed above, the political science and public administration literature argues that
there is variation across municipalities in voter preference for allocative and productive
efficiencies on the part of municipal governments (Lineberry and Fowler 1967; Dye 1973;
Frederickson et al. 2004). Allocative efficiency refers to the value citizens place on the services
produced by the government, while productive efficiency refers to how effectively a government
can provide services with a given level of resources. The choice of which municipal governance
mechanisms to adopt and in what combination affects the levels of both allocative and
productive efficiencies (Frant 1996) as discussed above. Governance structures are comprised of
combinations of governance mechanisms. Thus, variation in governance structure implies
variation in the levels of allocative and productive efficiencies.
From the previous discussion, it is clear that municipal governance structures that are
consistent with greater allocative efficiency are characterized by fragmented power and less
collaboration in the budget process. This leads to less information and communication and lower
monitoring. In contrast, municipal governance structures consistent with high productive
efficiency are characterized by unification of power, which leads to easier communication and
information sharing. These factors positively impact the control environment. This leads to my
primary hypothesis:
Hypothesis: Municipalities that have governance structures comprising a
combination of mechanisms that lead to higher allocative efficiency and lower
productive efficiency will have more internal control deficiencies.

21

From this hypothesis, I would expect that structure one would have a higher incidence of internal
control deficiencies as compared to structure three. Since the mix of allocative and productive
efficiencies of structures two and four is not as certain as that of structure one, I do not make a
prediction about the incidence of internal control deficiencies for those structures.

22

IV: METHODS
To test my hypothesis about the relation between municipal governance structure and
internal control deficiencies, I estimate the probability of disclosing an internal control
deficiency as a function of governance structure, municipal characteristics, and auditor
characteristics using a logistic model as follows:
Prob (RC_FinStat) = Φ (β0+ β1 STRUCTURE1 + β2 STRUCTURE2 + β3 STRUCTURE4
+ β4COMPLEXITY + β5SIZE + β6DEBT_PER_CAPITA
+ β7LOW_RISK + β8LIMITED_REVENUE
+ β9BIG4 + β10MUNICIPAL_SPECIALIST
+ ΣδiYear + ΣγiState) + εit

4.1 Dependent Variable
Consistent with Petrovits et al. (2011), I use the disclosure of a reportable condition
related to financial reporting (RC_FinStat) as my measure of the incidence of internal control
deficiency. A reportable condition related to financial reporting “involves deficiencies in the
design or operation of internal controls that could adversely affect the organization’s financial
reporting” (Petrovits et al. 2011). For example, one reportable condition found in the 2004 Single
Audit of the City of Detroit was: “A lack of communication exists between certain agencies and
the primary government, which has resulted in disagreements of interfund balances.” The
database I use, the Federal Audit Clearinghouse (FAC), compiles Single Audit information based
on form SF-SAC, which only discloses whether there was a reported significant deficiency in
internal control and not the number of deficiencies found. Because of this, my dependent
variable is dichotomous and not continuous.

23

4.2 Governance Structure
The governance structure variables are from a new measure of municipal governance
developed in Marshall (2010). This measure addresses the endogeneity problem of relating
governance to accounting outcomes by allowing for the simultaneous development of various
combinations of governance mechanisms (Armstrong et al. 2010). Municipal governance arises
to meet the needs of the citizenry, but there are tradeoffs between the responsiveness of
municipal government to the citizenry and the efficiency of municipal government control
systems. This measure captures this tradeoff.
The new governance structure measure is determined by LCA on eight municipal
governance mechanisms: form of government, presence of a CAO, shared budget-setting
authority, election or appointment of department heads, direct democracy, council voting power
of CEO, presence of CEO on council, and council terms. Latent class analysis determines the
probability of class membership conditional on the probability of response to each observed
variable. Conditional on class membership, each class is independent from the other. From the
eight observed municipal governance mechanisms, the LCA determines a four class solution.
These four classes form four governance structures with different combinations of the eight
governance mechanisms. Since the analysis gives a conditional probability of class membership,
one class must be omitted from the subsequent analysis. Thus my model includes three of the
four classes.
Table 1, Panel A shows the LCA of governance structure. Using cities over 10,000 in
population – for statistical power, I specify four classes for the LCA because in Marshall 2010
the four class solution had the best fit for these eight governance mechanisms. Each class has a
unique response pattern expressed by the conditional probability of affirmative response to eight

24

governance mechanisms. Governance structure is represented by the probability of class
membership in each of the four classes. For example, municipalities that are most likely to have
governance structure one are likely to have governance mechanisms consisting of a mayor–
council form a government (100%), lack of chief appointed official (23% likely to have a chief
appointed official), sole budgetary authority given to the mayor (60%), appointed department
heads (21% some or all elected), direct democracy provisions (84%), strong separation of powers
manifested by a low probability of the mayor being on council (0%) and voting on council
(13%), and concurrent council terms (35% staggered terms). The combination of these
mechanisms creates a governance structure where the mayor is powerful with little influence
from council since the council’s terms are concurrent and thus the power of the council is
weakened. With a powerful mayor and weak council, allocative efficiency is higher than
productive efficiency since the provision of services is more politicized. This combination of
mechanisms decreases monitoring. Thus, I expect this governance structure to be positively
associated with internal control deficiencies.
Municipalities with a high probability of having governance structure three are most
likely to have governance mechanisms consisting of a council–manager form of government
(0.07% mayor–council), a chief appointed official (100%), shared budgetary authority (2% sole
budget authority), appointed department heads (9%, some or all elected), direct democracy
provisions (85%), little separation of powers (95% CEO member of council and 98% CEO votes
on council), and staggered council terms (88%). This combination of mechanisms leads to
productive efficiency since the provision of services to citizens is less politicized by being
managed by a non-elected official directly monitored by a strong council. Thus, I expect this
governance structure to be negatively associated with internal control deficiencies.

25

Governance structures two and four share mechanisms found in both governance
structures one and three to varying degrees. Conditional on having governance structure two,
municipalities most likely have governance mechanisms consisting of a mayor–council form of
government with a chief appointed official, shared budget authority, appointed department heads,
direct democracy provisions, some separation of powers since the mayor is not likely to be on
council but is likely to vote on council matters, and staggered council terms. This combination of
governance measures may create a balance of power between the mayor and council and
possibly a balance between allocative and productive efficiencies. This could be due to the
professionalism that a chief appointed official brings in addition to having some of the political
incentives that a mayor–council form of government provides. However, Svara and Watson
(2010) report that in two-thirds of mayor–council cities that have a CAO, the mayor appoints the
CAO. The authors report that CAOs in these situations see themselves as working for the mayor.
This structure could be consistent with more allocative versus productive efficiency. Thus I
predict structure two to be positively associated with the incidence of reported internal control
deficiencies.
Municipalities with governance structure four most likely have governance mechanisms
consisting of a mayor–council form of government without a chief appointed official, appointed
department heads, direct democracy provisions, and little separation of powers since the mayor
likely votes on the council. Since the other governance mechanisms (mayor sole budget
authority, mayor member of council, and staggered versus concurrent council terms) are almost
equally likely, I cannot make strong predictions on the efficiency tradeoffs of this governance
structure. All probabilities of class membership and conditional probabilities of response are
found in Table 1.

26

4.3 Control Variables
Prior literature examining the determinants of internal control deficiencies has found
various firm characteristics that are associated with the incidence of reported internal control
deficiencies. These characteristics include complexity, financial risk, auditor type, size, rapid
growth, and auditor firm risk classification. I examine complexity, financial risk, auditor firm
risk classification, and auditor type. Due to data constraints, I do not examine growth. As I will
subsequently discuss, I expect municipalities that are more complex, financially riskier, and
smaller to have a higher incidence of reported internal control deficiencies.
Municipalities that collect revenue from multiple types of taxes or receive many different
types of grants face increased difficulties with internal controls. Greater revenue sources lead to
increased control activities since the amount of policies and procedures increase with each
revenue source, thus increasing the likelihood of an internal control deficiency. I measure
municipal complexity in two ways. First, I calculate the number of different general revenue
accounts as specified by the Census (COMPLEXITY). Second, I include a revenue diversification
index (LIMITED_REVENUE) used in prior government accounting research (Gore 2009;
Copeland and Ingram 1982). This index has higher values when the revenue sources are less
balanced, implying that a municipality receives the majority of its revenue from a few sources.
Similar to the findings from Petrovits et al. (2011), I predict that municipalities with more
general revenue sources (higher values of COMPLEXITY) and more diverse revenue (lower
values of LIMITED_REVENUE) are likely to provide more services and are thus more likely to
report an internal control deficiency.
If a municipality is financially risky, then it is more likely to face financial distress and be
unable to invest in strong internal controls. I use the amount of debt per capita

27

(DEBT_PER_CAPITA) as a proxy for financial risk. I predict that municipalities with higher
financial risk will have more internal control deficiencies.
In the prior literature, many authors argue that larger organizations have more resources
to invest in internal controls and greater experience dealing with internal controls. However,
larger cities usually offer a greater variety of services and thus may be more complex. I measure
size by taking the log of population (SIZE). Because size can either represent more experience
or complexity, I do not make a prediction on its effect on the incidence of internal control
deficiencies.
Prior evidence on the relation between auditor type and internal control deficiencies is
mixed. Ge and McVay (2005), Ashbaugh-Skaife et al. (2007), and Hoitash et al. (2009) find that
material weaknesses are positively associated with the appointment of a Big 6 auditor. In the notfor-profit setting, Petrovits et al. (2011) find that internal control deficiencies are negatively
related to Big 4 auditors and positively related to regional auditors. Lopez and Peters (2010)
examine internal control deficiencies in cities and counties post SOX and find that large CPA
firms are more likely to disclose an internal control deficiency than are government auditors.
Due to the mixed prior results, I do not make a prediction on the direction of the relationship
between the likelihood of an internal control deficiency and auditor type. I measure auditor type
using an indicator variable equal to one if the audit is done by a Big 4 firm (BIG4), a firm listed
by the Government Audit Quality Center as a government audit specialist
(MUNICIPAL_SPECIALIST), or a government auditor (STATE_AUDITOR).

28

CHAPTER V: SAMPLE SELECTION AND DATA
I obtain data on municipalities from three sources: (1) the FAC’s A-133 Single Audit
Database, (2) the Census Bureau Annual Survey of State and Local Government Finances, and
(3) the International City/County Manager Association (ICMA) Municipal Form of Government
Survey. The A-133 data include identifying information on each city, the auditor’s name and
contact information, audit opinions, internal control opinions, and the amount of federal awards.
The Census Bureau conducts an annual survey called the Annual Survey of State and Local
Government, which includes income statement and balance sheet type items. Participation is
required every fifth year, but is voluntary in other years. The ICMA’s Municipal Form of
Government Survey is voluntary and is conducted every five years. It includes a variety of
questions on governance mechanisms including form of government, government official’s
duties and characteristics, provisions for referendum/recall, term limits, and election systems. I
use the years 2001–2005 for the Census data, 2001 for the ICMA survey, and 2001–2005 for the
A-133 data. All variables are defined in Table 2.
Table 3 describes the sample. For the period between 2001 and 2005, merging the FAC
data with the Census data on cities over 50,000 in population on city name results in 980 city–
year observations. Merging these results with the ICMA data results in 966 city–year
observations. For my regression analysis, I omit the 2001 data since that is the year I use to
construct my four governance structure variables, leaving a sample of 800 city–year
observations.
Table 4 shows descriptive statistics for city–years with internal control deficiencies,
measured as a reportable condition related to financial reporting and those without internal
control deficiencies for the years 2002–2005. There are 152 city–years that have an internal

29

control deficiency compared to 648 city–years that do not, giving a rate of 23.5%. In univariate
analysis, city–years that have internal control deficiency have higher probabilities of having
governance structure one (STRUCTURE1), are larger (SIZE), are financially riskier
(DEBT_PER_CAPITA), are less likely to be classified as low risk (LOW_RISK), receive a higher
proportion of revenue from fewer resources (LIMITED_REVENUE), and are more likely to use a
state auditor versus a CPA firm.

30

CHAPTER VI: RESULTS
6.1 Primary Tests
Table 5 reports simple correlations between reported internal control deficiencies,
governance structure measures, and other municipal characteristics. Structure one is positively
correlated with having at least one reportable condition detected. Structures two and three are not
correlated with detected internal control deficiencies. DEBT_PER_CAPITA,
LIMITED_REVENUE, and STATE_AUDITOR are positively correlated with detected internal
control deficiencies, which is consistent with the literature on causes of internal control
deficiencies. In consistency with prior literature, SIZE is also positively correlated with detected
internal control deficiencies. This could be because it does not represent better efficiency as it
does in the corporate literature and it is also negatively correlated with LOW_RISK. Detected
internal control deficiencies are negatively correlated with LOW_RISK.
Table 6 presents initial tests of my hypothesis on the relationship between municipal
governance structure and internal control. The first column presents results of a logistic
regression analysis with RC_FinStat as the dependent variable and control variables used in prior
literature. Consistent with Petrovits el al. (2011), I find that the coefficients for organizational
complexity (COMPLEXITY) and financial risk (DEBT_PER_CAPITA) are positive and
significant. The coefficient on LOW_RISK is significantly negative as predicted; also consistent
with Petrovits et al. (2011) SIZE is positively associated with RC_FinStat, which is opposite to
the relation that has been found in previous literature. SIZE could be a different proxy for
complexity since larger cities have more constituents to serve. Larger values of
LIMITED_REVENUE mean less revenue sources; so, the significantly negative coefficient on
LIMITED_REVENUE means that municipalities with less revenue sources have less reported

31

internal control deficiencies. LOW_RISK was examined in Petrovits et al. (2011) and consistent
with their results, the coefficient is significantly negative.
I also include auditor type variables in the model in Table 6 consistent with Petrovits et
al. (2011) and Lopez and Peters (2010). Both studies find a positive relation between BIG4 and
reported control deficiencies. My results are not significant for BIG4.
MUNICIPAL_SPECIALIST and STATE_AUDITOR are used in Baber et al. where they find a
negative relation between restatements and MUNICIPAL_SPECIALIST and a positive relation
between STATE_AUDITOR and restatements. I do not find significance for
MUNICIPAL_SPECIALIST but the coefficient on STATE_AUDITOR is positive and significant
indicating that the probability of a reported internal control deficiency increases when a state
auditor is used. This suggests that internal control problems may be more prevalent in cities
where the use of governmental auditors is required.
Results from the primary test of my hypothesis are reported in the third column of Table
6. Governance structure three, the structure with lower allocative efficiency but higher
productive efficiency, is the excluded group in my analysis (since the probabilities of being in
each structure sum to one). Consistent with my hypothesis, the coefficient on STRUCTURE1 is
positive and significant suggesting that the incidence of an internal control deficiency is more
likely for municipalities with higher allocative efficiency and lower productive efficiency as
compared to those with lower allocative efficiency and higher productive efficiency. I also find a
positive association between the incidence of internal control deficiencies and governance
structure two (STRUCTURE2). This may reflect higher allocative efficiency and lower
productive efficiency compared to structure three. I do not find results for governance structure

32

four, which may imply that this governance structure has lower allocative efficiency as compared
to structures one and two.
Columns two and four of Table 6 present the marginal effects of the variables on the
probability of a detected control deficiency. LowRisk, LIMITED_REVENUE, and
STATE_AUDITOR have the largest marginal effects on detected reportable conditions. For the
sake of interpretation, I will limit my discussion to the discrete variables: LowRisk and
STATE_AUDITOR. The marginal effect of a discrete variable on the probability of an internal
control deficiency measures the change in probability of a detected internal control deficiency
when that variable’s value changes from zero to one holding all other variables at their means.
The marginal effect of LowRisk is 23.1%, which means that being classified as low risk
decreases the probability of a reported internal control deficiency by 23.1%. Using a state auditor
increases the probability of reported internal control deficiencies by 54.7% holding all other
variables at their means. Column four presents the marginal effects of the full model. The
marginal effect of LowRisk (21.7%) in the full model is similar in magnitude to the base model.
The marginal effect of STATE_AUDITOR is no longer significant in the full model. This could
indicate that cities that use state auditors are also cities with governance structures that enable
internal control problems.
Table 7 presents results from a test using a traditional measure of municipal governance
form of government. MayorCouncil is coded one if the form of government is mayor–council
and zero if it is council–manager. Mayor–council forms of government have traditionally been
considered more allocatively efficient than council–manager forms of government (Frederickson,
Logan, and Wood 2003). Thus, I would expect the sign of the coefficient on MayorCouncil to be
positive, indicating problems with internal control. The coefficient on MayorCouncil is 1.422

33

and significant indicating that municipalities with mayor–council forms of government have an
increased probability of detected internal control deficiencies. The coefficients on the control
variables in this analysis are qualitatively similar to the previous analysis using the newer
measure of governance.
Though inferences from a test that uses the traditional measure appear to be similar to
those from tests based on the newer measure, the analysis using my four governance classes
provides a more nuanced understanding of the relation between municipal governance structure
and internal control. In the analysis using the newer measure, two governance structures are
found to be significant. Both have a mayor–council form of government. However, structure four
has a mayor–council form of government as well, but does not share all of the other governance
mechanisms of governance structures one and two and was not significantly associated with
increased incidence of internal control deficiencies. Thus, municipalities with the combination of
governance mechanisms found in structure one and two may be driving the results found in prior
literature comparing mayor–council and council–manager forms of government. Structure four
has less separation of power and more collaboration in the budget process compared to
governance structure one even though both share the same form of government. Although
structure four is more likely than structure three to have the mayor solely set the budget, the
mayor in structure four is more likely to be a member of council. This increases communication
with the council when setting the budget. The increase in communication leads to better sharing
of information as well as increased monitoring, all of which improve internal controls.
6.2 Supplemental Analyses
To test if the newer measure has explanatory powers beyond what is explained by the
traditional measure of municipal governance, I use both measures in a logistic regression

34

analysis. The results of this analysis are presented in Table 8. I present my analysis with and
without state fixed effects. I present both the fixed effects model and the pooled model because
pooling panel data or using fixed effects with a nonlinear model such as the logit model gives
rise to an incidental parameters problem. (Greene 2004) Columns one and three present the
results of my previous analysis without state fixed effects and columns two and four present the
results of the full models from Tables 6 and 7 for comparison. The last two columns of Table 8
present my analysis on the incremental explanatory power of the newer municipal governance
measure compared to the traditionally used form of government measure. Column five does not
include state fixed effects and the coefficient on MayorCouncil is significant in the expected
direction but none of the newer governance structure measures are significant. In column six I
include state fixed effects and neither governance measure is significant.
From the results in Table 8, it seems that the newer governance measures do not have
explanatory powers beyond the form of government measure. However, there could be another
explanation for the no-significance finding. The logistic regression models I use are nonlinear
1

models and thus are not well suited for the use of fixed effects or panel data (Greene 2004.) A

1

In a simulation analysis, Greene compares three nonlinear models: pooled logit, fixed-

effects logit, and random effects logit. He does this to show the different biases in coefficients so
that researchers are better able to choose between these models when they believe that a fixed
effects binary response model is the most appropriate for their data. Greene finds that the random
effects model is the worst choice but that fixed-effects model tends to be biased upward and the
pooled model tends to be biased downward.

35

better model would be to use a dependent variable that is not a binary response, such as the
number of control deficiencies in a year. At this time, I do not have that data.
I use a balanced panel to explore the persistence of internal control deficiencies by
governance structure. As I am examining the number of audit reports that had at least one
reportable condition in a four-year period per city, I only keep cities that appear in my sample in
all four periods. This reduces my sample to 92 cities. Table 9 presents descriptive statistics on
the number of audit reports with at least one reportable condition in this balanced panel of 92
cities. Panel A shows a breakdown of number of audit reports in a four-year period that have at
least one reportable condition. The majority of cities (68.48%) have no reportable conditions
detected in the 2002–2005 period. Only a few cites (7.61%) have at least one detected reportable
condition every year from 2002 to 2005.
Table 9 also includes a breakdown of cities with at least one detected reportable condition
by year (Panel B) and reportable conditions by city population (Panel C). From Panel B, the year
with the most reportable conditions is 2004 with 22 of the 68 total audit reports with at least one
detected reportable condition. In Panel C, cities with populations below 100,000 make up the
majority (14 cities) of the 29 cities with at least one reportable condition in a four-year period;
however, these cities make up the majority (44.57%) of my sample of 92 cities.
When examining reportable conditions within each population group, the largest cities in
population (over 150,000 people) are more likely to have multiple reportable conditions within
the 2002–2004 period. In cities with population over 150,000, 32.14% will have at least two
audit reports with detected reportable conditions as against 4.35% of cities between 100,000 and
150,000 in population, and 21.96% of cities with population below 150,000. This result is
consistent with the results for size in the multivariate analyses in Tables 6, 7, and 8, and may

36

show that larger cities have more trouble with their accounting systems due to their complexity.
The smallest cities in the sample also have multiple internal control deficiencies in the sample
period, which may be due to the availability of limited resources to invest in their accounting
systems. Overall, Panel C of Table 9 may suggest a U-shaped relationship between internal
control problems and city size.
Table 10 presents the results of a multivariate analysis of the persistence of internal
control deficiencies. The dependent variable in the model is the total number of audit reports
with detected internal control deficiencies during the four-year period. Since the governance
structure measure does not vary in this period, I use the same measure that was used in the
previous analyses. The control variables for municipal characteristics and auditor type change
each year; so, I use the mean of these variables. As mentioned before, I use a sample of 92 cities
that have all four years of single audit data. If I used a sample that included observations that did
not have all four years, I would have to adjust the Poisson model.
From the results in Table 10, I find that cites that are more allocatively efficient are more
likely to have multiple internal control deficiencies within the 2002–2005 period. Columns one
and two show results from the base model and columns three and four show results for the full
model. In column three, the coefficient on STRUCTURE1 is 0.721 and the exponential of the
coefficient is 2.06. This means that a one unit increase in STRUCTURE1 would increase the
number of cumulative reportable conditions by 2.06 times the expected number of cumulative
reportable conditions. So, for example, the mean of the number of cumulative reportable
conditions is 0.74; therefore, increasing STRUCTURE1 by one unit increases the expected
number of cumulative reportable conditions to 1.52. Increasing STRUCTURE2 by one unit
increases the expected number of cumulative reportable conditions to 3.67; for STRUCTURE4,

37

the increase is 2.96. The measure of municipal governance structure is a probability; so, I
interpret these results as the difference between being classified into one type of structure versus
another. From these results, it seems that even though all of these three structures increase the
number of cumulative reportable conditions as expected, STRUCTURE2 is more prone to having
persistent internal control deficiencies.

38

CHAPTER VII: CONCLUSION
This study examines how governance structure impacts the effectiveness of internal
controls. Specifically, I explore this relationship in a municipal setting where each city
receiving federal funding is required to make disclosures about the effectiveness of its internal
controls over financial reporting consistent with A-133 reporting requirements. Consistent with
prior literature, I find that riskier and more complex municipalities have a higher incidence of
internal control deficiencies. Using a new governance measure, I categorize municipal
governance into four structures and find that municipalities with a governance structure
consistent with higher allocative efficiency are more likely to have an internal control deficiency
than municipalities that have governance structures consistent with lower allocative efficiency.
Thus, I contribute to the literature on the determinants of internal controls by explaining how
allocative efficiency in the municipal setting leads to poorer internal controls.
I also contribute to the governance literature by using a measure of governance structure
that reflects the co-occurrence of different combinations of governance mechanisms. Prior
accounting research on governance usually assumes that governance mechanisms have an
additive effect on overall governance. This assumption is reflected in models where various
governance mechanisms are included in regression models with no interactions between
mechanisms. The additive relationship between governance mechanisms is also modeled by the
use of governance measures that are merely the sum of how many different governance
mechanisms are present in an organization. Other than one recent paper (Larcker et al. 2007),
there is no accounting study that uses a governance measure that reflects that organizations’
choice of governance measures are made in combination with other governance measures. The

39

use of a measure that highlights the underlying reasons different combinations of governance
mechanisms are observed addresses the issue of governance being endogenously determined.
My results help to inform various interest groups. Constituents with preferences for
allocative efficiency may encourage their elected officials to invest more in community programs
and other services that directly benefit interest groups at the expense of investments in internal
controls. Understanding that these preferences may lead to underinvestment in accounting
systems may help constituents to be more reasonable in their demands. Being aware of how
governance structure affects internal controls helps government officials to adjust to their
environment and may spur them to invest more in their accounting systems to compensate for
their governance structure type. Auditors can use these results to enhance their risk assessment in
single audits of municipalities.
There are few studies that examine internal controls in the government setting. This
leaves many unaddressed questions. Though I show the determinants of internal control
deficiencies in municipalities and how governance structure affects internal control, I do not
explore the consequences of detected internal control deficiencies. Federal agencies may change
funding levels for cities with persistent internal control problems. Another possible consequence
of detected internal control deficiencies may be violation of bond covenants and increased
interest rates on new bond issues. In my sensitivity analyses, the new measure of municipal
governance did not provide additional explanation over the form of government measure. Power
may be partly to blame; so, a follow up with a larger sample may be appropriate. In my analyses,
I assume that various governance structures are consistent with different levels of allocative and
productive efficiencies. I base this assumption from prior literature in political science and public
administration but I do not directly test this relation. Future research should relate various
40

demographic characteristics with this new governance structure measure as a means of testing
the relation between governance structure and allocative and productive efficiencies.

41

TABLES

42

Table 1
Latent Class Analysis
Panel A : Four-Latent-Class Model of Municipal Governance (ICMA Data N = 1738)
Latent Class

Probability of Membership
Conditional Probability of a Yes* Response
Mayor–Council Form of Government
Chief Appointed Official (CAO)
CEO sole Budget Authority
Some or All Department Heads Elected
Some provisions for initiative,
referendum, popular referendum, or
recall
CEO member of Council
CEO votes on council
Staggered Council Terms

Political City –
Strong Mayor form
of Government
0.12

Strong Council
Elected Mayor –
mixed political city
0.12

City Manager
City – Less
Political no
separation of
powers and
strong council
0.66

1.00
0.23
0.60
0.21

0.67
1.00
0.09
0.11

0.07
1.00
0.02
0.09

1.00
0.28
0.39
0.23

0.84
0.00
0.13
0.35

0.79
0.15
0.68
0.88

0.85
0.95
0.98
0.88

0.75
0.49
0.97
0.53

43

Political City –
Weak Mayor,
little separation
of powers
0.10

Table 2
Variable Definitions
Variable
RC_FinStat

=

STRUCTURE1

=

STRUCTURE2

=

STRUCTURE3

=

STRUCTURE4

=

COMPLEXITY

=

SIZE

=

DEBT_PER_CAPITA

=

LOW_RISK

=

Definition
An Indicator variable that equals 1 if there is a reportable condition/ significant deficiency on
the financial statements; 0 otherwise.
Posterior probability of class membership in latent class 1
Posterior probability of class membership in latent class 2
Posterior probability of class membership in latent class 3
Posterior probability of class membership in latent class 4
Number of general revenue accounts as classified by the U.S. Census
Log of total population
Ratio of total debt outstanding to total population calculated as Account Type 1/ population
An Indicator variable that equals 1 if the auditee is a low risk auditee

(continued on next page)

44

Table 2 continued
Variable Definitions
LIMITED_REVENUE

=

Herfindahl type index of revenue diversification calculated as (T01/ (T01 + T09 + T40) ×
(T09/ (T01 + T09 + T40) × (T40/ (T01 + T09 + T40).

BIG4

=

An Indicator variable that equals 1 if the auditor is classified as one of the Big 4 auditors;
otherwise 0.

=

An Indicator variable that equals 1 if the auditor is classified as a government audit specialist
(by being a member of the Government Audit Quality Center of the AICPA) that is not one
of the Big 4 auditors, large auditors, or state auditors; otherwise 0.

MUNICIPAL_SPECIALIST

STATE_AUDITOR

MayorCouncil

=

An Indicator variable that equals 1 if the auditor is classified as a State Government Auditor;
otherwise 0.

=

An Indicator variable that equals 1 if the form of government is Mayor–Council and 0 if it is
Council–Manager.

Source: Form SF-FAC from the Federal Audit Clearinghouse, the 2011 Government Audit Quality Center membership list, and the
U.S. Census Bureau

45

Table 3
Sample Description
Panel A: Sample Selection
Full Sample
ICMA Form of Government Survey 2001

4,244

ICMA 2001 Data; population over 10,000

1,738

Cities in Federal Audit Clearinghouse Database 2001–2010
Census Annual Survey of Governments 2001–2005 (Counties, Cities, and
Townships

16,400
34,553

Panel B: Cities by Year
ICMA
2001
2002
2003
2004
2005

4,244

Census
1,855
27,954
1,644
5,890
5,902

FAC
3,366
3,438
3,609
2,951
3,036

Census FAC
Merge
168
204
159
224
225

Census/ FAC/
ICMA
166
200
157
222
221

Panel C: Cities by Year with population >= 50,000
ICMA
2001
2002
2003
2004
2005

360

46

Table 4
Descriptive Statistics of No Reportable Condition Firm-Year versus Reportable Condition
Firm-Years
Panel A:

RC_FinStat = 0

Variables

mean

p25

p50

p75

STRUCTURE1
STRUCTURE2
STRUCTURE4
COMPLEXITY
SIZE
DEBT_PER_CAPITA
LOW_RISK
LIMITED_REVENUE
BIG4
MUNICIPAL_SPECIALIST
STATE_AUDITOR
Observations

0.136
0.081
0.056
71.830
11.573
1.650
0.796
0.713
0.136
0.502
0.074

0.000
0.006
0.000
69.000
11.113
0.774
1.000
0.518
0.000
0.000
0.000

0.000
0.006
0.000
72.000
11.414
1.196
1.000
0.629
0.000
1.000
0.000

0.000
0.009
0.001
75.000
11.811
2.089
1.000
1.000
0.000
1.000
0.000

648

(continued on next page)

47

Table 4 (continued)
Descriptive Statistics of No Reportable Condition Firm-Year versus Reportable Condition
Firm-Years
Panel B:

RC_Finstat = 1
mean

STRUCTURE1
STRUCTURE2
STRUCTURE4
COMPLEXITY
SIZE
DEBT_PER_CAPITA
LOW_RISK
LIMITED_REVENUE
BIG4
MUNICIPAL_SPECIALIST
STATE_AUDITOR
Observations

p25

p50

p75

p

0.414
0.099
0.077
72.283
11.768
1.964
0.414
0.754
0.125
0.480
0.138

0.000
0.000
0.000
69.000
11.108
0.914
0.000
0.531
0.000
0.000
0.000

0.057
0.006
0.007
73.000
11.433
1.549
0.000
0.698
0.000
0.000
0.000

0.970
0.035
0.012
76.000
12.056
2.207
1.000
1.000
0.000
1.000
0.000

0.000
0.332
0.236
0.265
0.002
0.012
0.000
0.032
0.725
0.637
0.011

152

All variables are defined in Table 1. Each of the continuous variables is winsorized at 1% and 99% to mitigate outliers.

48

Table 5
Correlation Matrix

RC_FinStat

RC_FinStat
b/p
1.000

STRUCTURE1
b/p

0.294
(0.00)
0.034
(0.33)
0.042
(0.24)
0.039
(0.27)
0.110
(0.00)
0.089
(0.01)
-0.335
(0.00)
0.076
(0.03)
-0.012
(0.73)
-0.017
(0.64)
0.090
(0.01)

1.000

STRUCTURE1
STRUCTURE2
STRUCTURE4
COMPLEXITY
SIZE
DEBT_PER_CAPITA
LowRisk
LIMITED_REVENUE
BIG4
MUNICIPAL_SPECIALIST
STATE_AUDITOR
Observations

-0.078
(0.03)
-0.070
(0.05)
-0.042
(0.23)
0.148
(0.00)
0.033
(0.36)
-0.355
(0.00)
0.260
(0.00)
0.007
(0.83)
-0.132
(0.00)
0.367
(0.00)

800
49

STRUCTURE2
b/p

STRUCTURE4
b/p

1.000
-0.009
(0.81)
-0.043
(0.22)
-0.004
(0.90)
0.040
(0.26)
0.023
(0.52)
0.034
(0.34)
0.110
(0.00)
-0.062
(0.08)
0.061
(0.08)
(continued on next page)

1.000
-0.029
(0.41)
-0.028
(0.43)
-0.043
(0.22)
0.035
(0.32)
0.112
(0.00)
0.005
(0.90)
0.011
(0.75)
-0.048
(0.17)

Table 5 continued
Correlation Matrix
COMPLEXITY
b/p

SIZE
b/p

DEBT_PER_
CAPITA
b/p

LowRisk
b/p

RC_FinStat
STRUCTURE1
STRUCTURE2
STRUCTURE4
COMPLEXITY

1.000

SIZE

-0.476
(0.00)
-0.409
(0.00)
0.010
(0.77)
0.133
(0.00)
-0.261
(0.00)
0.111
(0.00)
-0.060
(0.09)

DEBT_PER_CAPITA
LowRisk
LIMITED_REVENUE
BIG4
MUNICIPAL_SPECIALIST
STATE_AUDITOR
Observations

1.000
0.423
(0.00)
-0.111
(0.00)
-0.145
(0.00)
0.478
(0.00)
-0.150
(0.00)
-0.015
(0.66)

800
50

1.000
0.021
(0.55)
-0.139
(0.00)
0.273
(0.00)
-0.064
(0.07)
-0.041
(0.25)
(continued on next page)

1.000
-0.149
(0.00)
-0.028
(0.42)
0.117
(0.00)
-0.149
(0.00)

Table 5 continued
Correlation Matrix
LIMITED_
REVENUE
b/p

BIG4
b/p

MUNICIPAL_
SPECIALIST
b/p

STATE_
AUDITOR
b/p

RC_FinStat
STRUCTURE1
STRUCTURE2
STRUCTURE4
COMPLEXITY
SIZE
DEBT_PER_CAPITA
LowRisk
LIMITED_REVENUE

1.000

BIG4

-0.003
(0.94)
0.035
(0.32)
-0.050
(0.15)

MUNICIPAL_SPECIALIST
STATE_AUDITOR

1.000
-0.391
(0.00)
-0.121
(0.00)

1.000
-0.306
(0.00)

Observations
800
All variables are defined in Table 1. Each of the continuous variables is winsorized at 1% and 99% to mitigate outliers.

51

1.000

Table 6
Determinants of Internal Control Deficiencies and Tests of the Effect of Governance Structure
Dependent Variable: Financial Statement Reportable Condition

Variable

Pred.
Sign

Intercept
STRUCTURE1

+

STRUCTURE2

?

STRUCTURE4

?

COMPLEXITY

+

SIZE

?

DEBT_PER_CAPITA

+

LowRisk (d)

-

LIMITED_REVENUE

-

Base Model
b/t
-10.889**
(-2.033)

Base Model
Marginal
Effect
b/t

0.082*
(1.819)
0.645**
(2.373)
0.221**
(1.992)
-1.482***
(-5.112)
-2.432**
(-2.046)

52

0.010*
(1.799)
0.082**
(2.302)
0.028**
(2.035)
-0.231***
(-4.470)
-0.308**
(-1.997)

Full Model
Marginal
Effect
b/t

Full Model
b/t
-12.689**
(-2.279)
1.454**
0.177**
(2.295)
(2.309)
1.712**
0.209***
(2.573)
(2.683)
1.234
0.151
(1.533)
(1.487)
0.109**
0.013**
(2.306)
(2.274)
0.522*
0.064*
(1.799)
(1.771)
0.226*
0.028*
(1.936)
(1.953)
-1.444***
-0.217***
(-4.714)
(-3.997)
-2.415**
-0.295**
(-2.092)
(-2.044)
(continued on next page)

Table 6 continued
Determinants of Internal Control Deficiencies and Tests of the Effect of Governance Structure

Variable

Pred.
Sign

Base
Model

Base
Model
Marginal
Effect

BIG4 (d)

?

MUNICIPAL_SPECIALIST (d)

?

STATE_AUDITOR (d)

?

-0.746
(-1.277)
0.550
(1.336)
2.698**
(2.370)
Yes
Yes
119.847
32
0.2748
680

-0.078
(-1.515)
0.071
(1.372)
0.547**
(2.305)
Yes
Yes
119.847
32
0.2748
680

Year Indicators
State Indicators
Chi-square
Degrees of Freedom
Pseudo r-square
Number of Observations
Marginal effects
(d) for discrete change of dummy variable from 0 to 1
* p<0.10 ** p<0.05 *** p<0.01

Full
Model

Full
Model
Marginal
Effect

-0.795
(-1.232)
0.578
(1.385)
1.935*
(1.716)
Yes
Yes
132.607
35
0.2959
680

-0.079
(-1.490)
0.071
(1.423)
0.365
(1.392)
Yes
Yes
132.607
35
0.2959
680

All variables are defined in Table 1. Each of the continuous variables is winsorized at 1% and 99% to mitigate outliers.

53

Table 7
Determinants of Internal Control Deficiencies Using Form of Government Measure
Dependent Variable: Financial Statement Reportable Condition

Variable

Pred.
Sign

Intercept
MayorCouncil (d)

+

COMPLEXITY

+

SIZE

?

DEBT_PER_CAPITA

+

LowRisk (d)

-

LIMITED_REVENUE

-

BIG4 (d)

?

MUNICIPAL_SPECIALIST (d)

?

Base Model
b/t
-10.889**
(-2.033)

Base Model
Marginal
Effect
b/t

0.082*
(1.819)
0.645**
(2.373)
0.221**
(1.992)
-1.482***
(-5.112)
-2.432**
(-2.046)
-0.746
(-1.277)
0.550
(1.336)

0.010*
(1.799)
0.082**
(2.302)
0.028**
(2.035)
-0.231***
(-4.470)
-0.308**
(-1.997)
-0.078
(-1.515)
0.071
(1.372)

Full Model
b/t
-12.256**
(-2.161)
1.422***
(2.899)
0.107**
(2.165)
0.485*
(1.690)
0.253**
(2.272)
-1.397***
(-4.658)
-2.525**
(-2.191)
-0.633
(-1.022)
0.664
(1.643)

Full Model
Marginal
Effect
b/t

0.212**
(2.480)
0.013**
(2.168)
0.059*
(1.668)
0.031**
(2.284)
-0.208***
(-3.947)
-0.307**
(-2.168)
-0.065
(-1.193)
0.082*
(1.668)

(continued on next page)
54

Table 7 continued
Determinants of Internal Control Deficiencies Using Form of Government Measure

Variable
STATE_AUDITOR (d)

Pred.
Sign
?

Year Indicators
State Indicators
Chi-square
Degrees of Freedom
Pseudo r-square
Number of Observations
Marginal effects
(d) for discrete change of dummy variable from 0 to 1
* p<0.10 ** p<0.05 *** p<0.01

Base
Model
2.698**
(2.370)
Yes
Yes
119.847
32
0.2748
680

Base
Model
Marginal
Effect
0.547**
(2.305)

119.847
32
0.2748
680

Full
Model
2.089*
(1.787)
Yes
Yes
128.336
33
0.2989
679

All variables are defined in Table 1. Each of the continuous variables is winsorized at 1% and 99% to mitigate outliers.

55

Full
Model
Marginal
Effect
0.401
(1.474)

128.336
33
0.2989
679

Table 8
Comparison of Governance Measures

Pred.
Sign
Intercept
MayorCouncil

+

STRUCTURE2

?

STRUCTURE4

?

COMPLEXITY

+

SIZE

?

DEBT_PER_CAPITA

+

LowRisk

-

LIMITED_REVENUE

-

Governance
Structure
Measure
b/t
-12.689**
(-2.28)

1.449***
(3.61)
1.230*
(1.79)
1.317**
(2.36)
0.137***
(3.59)
0.564**
(2.21)
0.256**
(2.50)
-1.522***
(-5.70)
-0.495
(-0.72)

1.454**
(2.29)
1.712**
(2.57)
1.234
(1.53)
0.109**
(2.31)
0.522*
(1.80)
0.226*
(1.94)
-1.444***
(-4.71)
-2.415**
(-2.09)

+

STRUCTURE1

Governance
Structure
Measure
b/t
-17.569***
(-3.68)

56

Form of
Government
Measure
b/t
-16.786***
(-3.52)
1.404***
(4.42)

Form of
Government
Measure
b/t
-12.256**
(-2.16)
1.422***
(2.90)

0.135***
(3.55)
0.498**
(1.96)
0.279***
(2.76)
-1.563***
(-5.93)
-0.477
(-0.69)

0.107**
(2.17)
0.485*
(1.69)
0.253**
(2.27)
-1.397***
(-4.66)
-2.525**
(-2.19)

Both
Both
Measures
Measures
b/t
b/t
-16.949***
-12.625**
(-3.55)
(-2.25)
1.473*
1.562
(1.83)
(1.41)
-0.028
-0.169
(-0.03)
(-0.14)
0.431
0.761
(0.58)
(0.77)
-0.356
-0.641
(-0.34)
(-0.48)
0.136***
0.110**
(3.57)
(2.29)
0.507**
0.498*
(1.97)
(1.73)
0.275***
0.264**
(2.72)
(2.32)
-1.559***
-1.429***
(-5.79)
(-4.65)
-0.497
-2.239**
(-0.72)
(-1.99)
(continued on next page)

Table 8 continued
Comparison of Governance Measures

Pred.
Sign

Governance
Structure
Measure

Governance
Structure
Measure

Form of
Government
Measure

Form of
Government
Measure

Both
Measures

Both
Measures

BIG4

?

-0.800
(-1.53)

-0.795
(-1.23)

-0.613
(-1.20)

-0.633
(-1.02)

-0.619
(-1.19)

-0.632
(-0.99)

MUNICIPAL_
SPECIALIST

?

STATE_AUDITOR

?

0.121
(0.38)
-0.211
(-0.35)
Yes
No
78.591
14
0.1888
799

0.578
(1.38)
1.935*
(1.72)
Yes
Yes
132.607
35
0.2959
680

0.235
(0.74)
-0.133
(-0.23)
Yes
No
80.703
12
0.1947
798

0.664
(1.64)
2.089*
(1.79)
Yes
Yes
128.336
33
0.2989
679

0.249
(0.78)
-0.175
(-0.30)
Yes
No
84.085
15
0.1964
798

0.676*
(1.67)
2.009*
(1.82)
Yes
Yes
130.692
36
0.3018
679

Year Indicators
State Indicators
Chi-square
Degrees of Freedom
Pseudo r-square
Number of Observations

* p<0.10 ** p<0.05 *** p<0.01

57

Table 9
Incidence of At Least One Reportable Condition over Time
Panel A: Cumulative Reportable
Conditions (2002–2005)
Number of Reportable
Conditions

Cumulative
Frequency

0
1
2
3
4

Cumulative
Percent

63
10
6
6
7

Frequency

68.48
10.87
6.52
6.52
7.61

63
73
79
85
92

Percent
68.48
79.35
85.87
92.39
100

Panel B: Reportable Conditions by Year
2002
Reportable Condition
N
Y
Total

2003

Audit Year
2004

2005

Total

79
13

75
17

70
22

76
16

300
68

92

92

92

92

368

(continued on next page)

58

Table 9 (continued)
Incidence of At Least One Reportable Condition over Time
Panel C: Frequency by Population
0

Cumulative Incidence of Reportable Conditions
1
2
3

4

Total

Population
Groups
under
100,000

Frequency
Percent of Total
Percent of Population Group
Percent of Incidence Group

27
29.35
65.85
42.86

5
5.43
12.20
50.00

3
3.26
7.32
50.00

2
2.17
4.88
33.33

4
4.35
9.76
57.14

41
44.57

Between
100,000
and
150,000

Frequency
Percent of Total
Percent of Population Group
Percent of Incidence Group
Frequency
Percent of Total
Percent of Population Group
Percent of Incidence Group

20
21.74
86.96
31.75
16
17.39
57.14
25.40

2
2.17
8.70
20.00
3
3.26
10.71
30.00

1
1.09
4.35
16.67
2
2.17
7.14
33.33

0
0.00
0.00
0.00
4
4.35
14.29
66.67

0
0.00
0.00
0.00
3
3.26
10.71
42.86

23
25.00

Total
Percent Total

63
68.48

10
10.87

6
6.52

6
6.52

7
7.61

92
100.000

Over
150,000

59

28
30.43

Table 10
Poisson Regression of Incidence of Reportable Conditions over a Four-Year Period
Pred. Sign

STRUCTURE1

+

SIZE_mean

?

DEBT_PER_CAPITA_mean

+

LOWRISK_mean

-

LIMITED_REVENUE_mean

-

BIG4_mean

?

-31.850***
(-13.38)

0.153***
(3.35)
0.447**
(1.98)
0.272***
(2.79)
-1.973***
(-5.48)
1.699**
(2.08)
0.171
(0.36)

0.104**
(2.23)
0.739***
(3.44)
0.209
(1.30)
-1.271***
(-4.37)
-0.552
(-0.45)
-0.491
(-0.82)

-17.547***
(-3.61)
0.721*
(1.94)
1.601**
(2.02)
1.391***
(3.30)
0.178***
(3.74)
0.352
(1.60)
0.297***
(2.76)
-2.044***
(-5.06)
0.686
(0.69)
-0.013
(-0.03)

-30.599***
(-4.28)
0.056
(0.10)
0.408
(0.41)
0.515
(1.05)
0.103*
(1.85)
0.662***
(2.76)
0.207
(1.20)
-1.456***
(-3.38)
-0.541
(-0.42)
-0.457
(-0.75)

?

COMPLEXITY_mean

Full Model
b/t

?

STRUCTURE4

Full Model
b/t

+

STRUCTURE2

Base Model
b/t

-17.363***
(-3.30)

Intercept

Base Model
b/t

(continued on next page)

60

Table 10 continued
Poisson Regression of Incidence of Reportable Conditions over a Four-Year Period
Pred. Sign

Base Model

Base Model

Full Model

Full Model

MUNICIPAL_SPECIALIST_mean

?

STATE_AUDITOR_mean

?

-0.085
(-0.20)
0.252
(0.56)
No
61.911
8
0.2851
92

-0.389
(-0.69)
1.213*
(1.67)
Yes
.
5
0.4313
92

-0.243
(-0.63)
-0.202
(-0.46)
No
60.232
11
0.3214
92

-0.402
(-0.70)
1.058
(1.27)
Yes
.
13
0.4335
92

State Indicators
Chi-square
Degrees of Freedom
Pseudo r-square
Number of Observations
* p<0.10 ** p<0.05 *** p<0.01

61

APPENDICES

62

APPENDIX A: MUNICIPAL GOVERNANCE MECHANISMS
In this appendix, I include an excerpt from Marshall (2010) that discusses various governance
mechanisms and how they affect individual incentives and organizational goals. I also include a
figure from Marshall (2010).
Marshall (2010) Excerpt:
“I examine the following governance mechanisms: form of government, Chief Elected
Official (CEO) power, direct democracy, and legislative terms and independence. A summary of
my predictions of how municipal governance mechanisms affect incentives and monitoring
based on prior literature and my own conjecture are found in Table 1. I examine four broad
governance mechanisms which are proxied by eight different survey items. Each of these
governance mechanisms is associated with different individual incentives and organizational
objectives. I expect that these mechanisms act together to fulfill organizational missions and
mitigate agency issues by varying incentives and monitoring. Therefore, I expect that if some
mechanisms increase certain incentives that they will exist with other mechanisms that decrease
other incentives. How monitoring, incentives, and organizational goals are associated with these
measures is described in the following sections.
Form of government
In the first two rows of table one are two governance mechanisms, mayor-council versus
council manager form of government and whether there is a Chief Appointed Officer. These two
mechanisms fall under the category of form of government. As described in previous sections,
the mayor-council form of government is characterized by a separation of powers between the
executive and legislative branches of government whereas the council-manager form of
63

government is similar to governance we see in the for-profit sector where an elected board hires
and supervises an executive/manager. In the mayor-council form of government, having two
branches of government reduces monitoring since the costs and benefits of agents’ actions are
less concentrated (Zimmerman 1977) (Table 1 row 1, column1). The career incentives are
different compared to council-manager form of governments since the mayor is looking to
further his political career compared to a career in the private sector. This reduces career
incentives (Table 1 row 1, column 2) and increases political incentives (Table 1 row 1, column
3). At an organizational level, having a mayor-council form of government increases the
organizational goal of reflecting voter preferences (Table 1 row 1, column 4) because this form
of government is more political and thus stewardship of resources is not as important (Table 1
row 1, column 5).
The Chief Appointed Officer was created to lessen the demands on mayors in mayorcouncil form of government cities (Frederickson, Logan, and Wood 2003). In the councilmanager form of government the CAO is the city manager. Frederickson, Logan, and
Wood(2003) document that the presence of CAO in mayor-council cities has increased to over
50 percent and this increase is due to the trend of having greater administrative effectiveness.
The presence of a Chief Appointed Officer (Table 1 row 2, column 1) increases monitoring since
the costs and benefits of his actions are concentrated in the elected officials that monitor him as
well as municipal residents. Since the CAO is a professional and is only hand-picked by the
mayor is few cases (11%, Svara 1999), he has greater career incentives (Table 1 row 2, column
2) and less political incentives (Table 1 row 2, column 3). As the CAO does not directly answer
to voters and is non-political by design, his presence decreases meeting the organizational goal

64

of fulfilling voter preferences (Table 1 row 2, column 4) and better aligns the organization with
the mission of being a good steward of resources (Table 1 row 2, column 5).
Chief elected official (CEO) power
I include two items for Chief Elected official power, whether the CEO sets the budget by
himself, and whether department heads are elected or appointed, based on their emphasis in prior
literature (Renner and DeSantis 1994; Frederickson, Logan, and Wood 2003; Feiock, Keong, and
Kim 2003). Having the power to set the budget gives municipal officials the ability to repay
political favors if they are politically motivated. For municipal officials that have low-powered
political incentives budgetary authority may increase career incentives through greater job
fulfillment and less turnover (Renner and DeSantis 1994). Budgetary authority is usually
delegated to the mayor with approval by council in mayor-council cities and is delegated to the
city manager in council-manager cities but that has changed in recent years (Renner and
DeSantis 1994, Frederickson, Logan, and Wood 2003). With the introduction of the chief
appointed officer position, budgetary authority has been increasingly delegated to the CAO in
mayor-council cities making them more similar to council manager cities (Frederickson, Logan,
and Wood 2003) thus form of government no longer automatically determines budget setting
authority. If the budget is set by the chief elected official, i.e. the mayor, then monitoring is
decreased (Table 1 row 3, column 1) since the mayor is only accountable to voters and not the
council. Career incentives are decreased for the chief appointed official if the chief elected
official does not share budget setting authority with him and increases the political incentives for
the CEO (Table 1 row 3, column 3) since he can now use budgetary authority to give political
favors (Table 1 row, 3 column 2). At an organizational level voter preferences should be better
met (Table 1 row 3, column 4) since unreformed municipal structures are more responsive to
65

their constituencies (Lineberry and Fowler 1967 and DeSantis and Renner 2002) and unreformed
municipal structures are characterized by mayor-council form of government where the mayor
has budgetary authority and council is elected by district. The goal of being a good steward of
resources is diminished if the CEO sets the budget (Table 1 row 3, column 5) since the position
has high-powered political incentives and thus short-term projects that will gain favor for
reelection will be pursued over longer-term projects that may be more prudent (Frant 1996).
If a municipality’s department heads are elected then those department heads are
accountable to the electorate and not directly to the city manager or CAO. This decreases
monitoring (Table 1 row 4, column 1) since the actions of the agent (the department heads) are
no longer as concentrated (Zimmerman 1977). Also by having department heads elected career
incentives for the department heads are decreased (Table 1 row 4, column 2) and political
incentives increase (Table 1 row 4, column 3) since the position is no longer a professional civil
service position but is now political. By making department heads more political through being
elected, voter preferences are more likely to be met (Table 1 row 4, column 4) and stewardship
of resources is less likely (Table 1 row 4, column 5).
Direct democracy
Direct democracy refers to the ability of the electorate to participate directly in their
government. Maser (1998) refers to the provisions of initiative, referendum, and recall as direct
democracy and explains that these provisions are meant to provide a threat to municipal officials
that may not abide by the interests of their constituents. Maser (1998) argues that these
provisions reduce the cost of monitoring and that the provision for recall provides incentives for
elected officials ‘to identify the medians on multiple issues, to avoid deviating from them.’ This
leads to the following predictions, monitoring is higher (Table 1 row 5, column 1), career
66

incentives are lower (Table 1 row 5, column 2), and political incentives are higher (Table 1 row
5, column 3) when provisions for referendum, initiative, and/or recall are present. Career
incentives are lower when these provisions are present because they have the potential to
destabilize government (Maser 1998) and when there is political instability more turnover is
present for city managers (Renner and DeSantis 1994). Because provisions for direct democracy
help to align elected officials with median voter preferences then the organizational goal of
fulfilling voter preferences becomes more important (Table 1 row 5, column 4). I conjecture that
the goal of stewardship of resources is also more important when there are direct democracy
provisions because of the use of referenda in fiscal issues such as the issuance of bonds (Maser
1998).
Legislative characteristics
I use three measures of legislative characteristics: whether the chief elected official is on the
council, whether the chief elected official votes on the council, and whether the council members
are elected concurrently or in staggered terms. Because of separation of powers the chief elected
official is not on the council in the mayor-council form of government but can be in the councilmanager form of government. As a member of council, the chief elected official or mayor can
have the authority to vote in all matters, only vote in the case of breaking a tie, or not vote at all.
For the chief appointed officer (CAO) or city manager career incentives are higher (Table 1 row
6, column 2) when the CEO is on the council since this means that the CEO has less executive
power. Because the CEO is not a separate executive when he is on the council political
incentives are lower (Table 1 row 6, column 3). Having the CEO on the council is a
characteristic of reformed governments which are less responsive to their constituencies
(DeSantis and Renner 2002 thus voter preferences are lower (Table 1 row 6, column 4).
67

Staggered terms are associated with more political stability which leads to less city
manager/CAO turnover (Renner and DeSantis 1994), therefore; monitoring of the CAO is lower
(Table 1 row 8, column 1) and career incentives are higher (table 1 row 8, column 2). The
governance mechanism of having the CEO vote on the council increases the political incentives
of the CEO (table 1 row 7, column 3) since the CEO becomes more involved in the legislative
process. Voter preferences are higher when the CEO votes on the council (table 1 row 7, column
4) and stewardship of resources is lower (table 1 row 7, column 5) since political incentives for
the CEO are higher when they are able to vote and higher political incentives lead to greater
allocative efficiency (Frant 1996).When council elections are staggered fewer council members
run at a time thus each council member has more scrutiny. This causes higher political incentives
on behalf of council (table 1 row 8, column 3). Staggered terms also bring stability into the
political process slowing political change thus lowering the goal of voter preferences (table 1
row 8, column 4). Gore (2009) shows that councils with staggered terms have higher agency
costs as proxied by administrative expenses thus I predict that staggered council terms lead to
lower stewardship of resources (table 1 row 8, column 5). ”

68

APPENDIX B: LATENT CLASS ANALYSIS
In this appendix I include an excerpt from Marshall (2010) that describes latent class analysis. I
then compare latent class analysis to factor analysis and cluster analysis.
From Marshall (2010):
“I use latent class analysis to group municipalities by governance types based on answers to the
2001 ICMA Form of Government Survey. Latent class analysis (LCA) determines the
probability of being in a particular class and further determines the probabilities of each item
response given membership in a certain class. From these probabilities, posterior probabilities of
class membership are determined. The analysis works by analyzing a contingency table using
maximum likelihood to determine the parameter estimates. The contingency table is made up of
cells of every combination of answer to the items included in the analysis. For example, if I
created a contingency table from three items that had two possible answers each the table would
be made of eight cells (2 × 2 × 2). One issue with using LCA is that when more items are added
the contingency table that is used for the maximum likelihood estimation becomes very large.
This becomes a problem for model specification since there needs to be enough observations in
the different cells of the contingency table. This problem of the average expected cell count
being small is called sparseness and is a function of the sample size (N) divided by the size of the
contingency table (W) (Collins and Lanza 2010). For example, the ICMA 2001 Form of
Government Survey has thirty-one questions with some questions having multiple parts. If I used
the 21 items with dichotomous answers, that would create a contingency table with 2,097,152
(221) cells and my sample size is only 1738. I do not use all 31 questions since I do not have a
large enough sample for my model to be specified, thus I choose the eight most important

69

questions based on the prior literature. Since there are eight dichotomous indicators there are 256
possible combinations of answers. Because the latent governance construct mitigates agency
issues while fulfilling organizational missions certain combinations of answers should be more
prevalent than others and latent class analysis can determine what combinations hang together.”
Latent class analysis is used for governance structure instead of factor analysis for a few
reasons. The first is that measurement of governance mechanisms is discrete instead of
continuous. Factor analysis assumes that measures are continuous; so, it is not well suited for
discrete measures. There are some ways to correct for this, such as tetrachoric correlation;
however, there is another reason that factor analysis is not used. Factor analysis also assumes
that the underlying latent variables are continuous. I do not view governance structure as a
continuum ranging from good to bad but as discrete types developed to deal with various issues
in a particular situation. Because of my view of governance, factor analysis would not be
appropriate since it would capture a latent variable that does not have the properties to deal with
discrete governance types.
Latent Class Analysis shares some similarities to cluster analysis. Both analyses
determine groupings of observations based on multiple observed characteristics. In both
analyses, these groups are not directly observed. Latent class analysis differs from cluster
analysis in that it is based on a conditional probability model and cluster analysis is not based on
any statistical model. There are statistical methods that can be used for different types of cluster
analysis but determining the probability of an observation being in a cluster based on its
characteristics is not possible with cluster analysis. I use these probabilities of class membership
as a measure of governance structure and I would not have this information using cluster
analysis.
70

FIGURES

71

Figure 1: Governance Mechanisms and Influence on Incentives and Organizational Missions
Reproduced from Marshall (2010)

Incentives
Governance
Mechanisms
Form of
Government
CEO Power

Direct
Democracy
Legislative
Characteristics:
Terms and
Independence

Items
(1) Form MayorCouncil
(2) CAO
(3) Budget
setting CEO
only
(4) Department
H. elected
(5) Provisions
(6) CEO on
council
(7) CEO votes
on council
(8) Staggered
Terms

−

Individual
Career
Incentives
(2)
−

Political
Incentives
(3)
+

+
−

+
+ if CEO

−
+

−
+

+
−

−

− If CAO
−

+

+

−

+

−

+

+

+

No prediction

+

−

−

+

No prediction

No Prediction

+

−

−

+

−

−

Monitoring
(1)

72

+

Organizational
Voter
Stewardship of
Preferences
Resources
(4)
(5)
+
−

REFERENCES

73

REFERENCES
Armstrong, C., W. Guay, and J. Weber. 2010. The role of information and financial reporting in
corporate governance and debt contracting. Journal of Accounting and Economics 50 (23): 179–234.
Ashbaugh-Skaife, H., D. Collins, and W. Kinney. 2007. The discovery and reporting of internal
control deficiencies prior to SOX-mandated audits. Journal of Accounting and
Economics 44 (1–2): 166–192.
Baber, W., A. Gore, K. Rich, J. Zhang. 2010. An Empirical Investigation of Accounting
Restatements and Governance in the Municipal Context. Working Paper. Georgetown
University.
Collins, L. and S. Lanza. 2010. Latent Class and Latent Transition Analysis with Applications in
the Social, Behavioral, and Health Sciences. Hoboken, NJ: John Wiley & Sons.
Committee of Sponsoring Organizations (the Treadway Commission). 1992. Internal Control –
Integrated Framework.
Copeland, R. and R. Ingram. 1982. Municipal Market Measures and Reporting Practices: An
Extension. Journal of Accounting Research 20(2):766-772.
Dey, A.2008. Corporate Governance and Agency Conflicts. Journal of Accounting Research
46(5):1143-1181.
DeSantis, V. and T. Renner 2002. City Government Structures: An Attempt at Clarification.
State and Local Government Review 34 (2): 95–104.
Doyle, J., W. Ge, and S. McVay. 2007. Determinants of weaknesses in internal control over
financial reporting. Journal of Accounting and Economics 44 (1–2): 193–223.
Dye, T. 1973. Politics in States and Communities. Englewood Cliffs, NJ: Prentice-Hall.
Evans III, J. H. and J. M. Patton. 1987. Signaling and monitoring in public-sector
accounting. Journal of Accounting Research 25: 130–158.
Feiock, R., M. Jeong, and J. Kim. 2003. Credible Commitment and Council-Manager
Government: Implications for Policy Instrument Choices. Public Administration Review.
63(5): 616-625.
Frant, H.1996. High-powered and low-powered incentives in the public sector. Journal of Public
Administration Research and Theory: J-PART 6(3): 365–381.
Frederickson, H., B. Logan, and C. Wood. 2003. Municipal reform in mayor-council cities: A
well-kept secret. State & Local Government Review 35 (1): 7–14.
74

Frederickson, H., G. Johnson, and C. Wood. 2004. The Adapted City: Institutional Dynamics and
Structural Change. Armonk, NY: ME Sharpe.
Garrison, J., A. Sewell, and R. Winton. 2011. FBI probes spending in Montebello; FBI, U.S.
attorney are looking into potential misuse of $1.3 million in development funds. Los
Angeles Times (July 1)
Ge, W., and S. McVay. 2005. The disclosure of material weaknesses in internal control after the
Sarbanes-Oxley Act. Accounting Horizons 19, 137–158.
Giroux, G., and A. McLelland. 2003. Governance structures and accounting at large
municipalities. Journal of Accounting and Public Policy 22: 203–230.
Gore, A. 2009. Why Do Cities Hoard Cash? Determinants and implications of municipal cash
holdings. The Accounting Review 84(1):183–207.
Greene, W. 2002. The Behaviour of the Maximum Likelihood Estimator of Limited Dependent
Variable Models in the Presence of Fixed Effects. Econometrics Journal 7:98-119.
Hoitash, R., U. Hoitash, and J. C. Bedard. 2009. Corporate governance and internal control over
financial reporting: A comparison of regulatory regimes. The Accounting Review 84 (3):
839–867.
Keating, E.K., M. Fischer, T. P. Gordon, and J. Greenlee. 2005. The Single Audit Act: How
compliant are nonprofit organizations? Journal of Public Budgeting, Accounting and
Financial Management 17 (3), 285–309.
Krishnan, J., 2005. Audit committee quality and internal control: An empirical analysis. The
Accounting Review 80, 649–675.
Krishnan, G. V., and G. Visvanathan. 2007. Reporting internal control deficiencies in the postSarbanes-Oxley Era: The role of auditors and corporate governance. International
Journal of Auditing 11: 73–90.
Larcker, D., S. Richardson, and I. Tuna. 2007. Corporate governance, accounting outcomes, and
organizational performance. The Accounting Review 82(4): 963–1008.
Lineberry, R. and E. Fowler. 1967. Reformism and Public Policies in American Cities. The
American Political Science Review 61(3):701-716.
Lopez, D., and G. Peters. 2010. Internal control reporting differences among public and
governmental auditors: The case of city and county Circular A-133 audits. Journal of
Accounting and Public Policy 29 (5): 481–502.
MacDonald C. 2011. Ousted supervisor sues Detroit; says she tried to halt spending. The Detroit
News (August 17)
75

Marshall, D. 2010. A Classification of Municipal Governance. Working paper, Michigan State
University.
Maser, S. 1998. Constitutions as Relational Contracts: Explaining Procedural Safeguards in
Municipal Charters. Journal of Public Administration Research and Theory. 8(4): 527 –
564.
Newland, C. 1995. The Future of Council-Manager Government. In Ideal and Practice in
Council-Manager Government, edited by H. George Frederickson. Washington, D.C.:
International City/County Management Association.
Ogneva, M., K. Subramanyam, and K. Raghunandan. 2007. Internal control weakness and cost
of equity: Evidence from SOX section 404 disclosures. The Accounting Review 82 (5):
1255–1297.
Petrovits, C., C. Shakespeare, and A. Shih. 2011. The causes and consequences of internal
control problems in nonprofit organizations. The Accounting Review 86 (1): 325–357.
Raman, K. K., and E. R. Wilson.1992. An empirical investigation of the market for ‘‘single
audit” services. Journal of Accounting and Public Policy 11 (4): 271–295.
Renner, T., and V. DeSantis. 1993. Contemporary patterns and trends in municipal government
structures. In The municipal yearbook 1993, 57–69. Washington, DC: International
City/County Management Association
Renner, T., and V. DeSantis. 1994. City manager turnover: The impact of formal authority and
electoral change. State and Local Government Review 26 (2): 104–111.
Schneider, A., A. Gramling, D. Hermanson, and Z. Ye. 2009. A Review of academic literature
on internal control reporting under SOX. Journal of Accounting Literature 28: 1–46.
Svara, J. 2001. Council Roles, Performance, and Form of Government. Paper presented at the
2001 American Political Science Association Conference.
Svara J., and D. Watson. 2010. More than mayor or manager: Campaigns to change form of
government in America’s large cities. Washington, D.C.: Georgetown University Press.
United States Government Accountability Office (GAO). 2007. Government Auditing Standards.
July 2007 Revision.
U.S. Census Bureau. 2010. Federal Aid to States for Fiscal Year 2009. Washington, D.C. U.S.
Department of Commerce.
U.S. Congress, 1984. Public Law 98-502: The Single Audit Act (SAA), 19 October, 1984, 98
Stat. 2327. US Congress, Washington, DC.
76

Vijayakumar, J. 1995. An empirical analysis of the factors influencing call decisions of
local government bonds. Journal of Accounting and Public Policy 14: 203–231.
Wallace, W. 1981. Internal control reporting practices in the municipal sector. The Accounting
Review 56(3): 666–689.
Ward, D., R. Elder, and S. Kattelus. 1994. Further evidence on the determinants of municipal
audit fees. The Accounting Review 69 (2): 399–411.
Zhang, Y., J. Zhou, and N. Zhou. 2007. Audit committee quality, auditor independence, and
internal control weaknesses. Journal of Accounting and Public Policy 26 (3): 300–327.
Zimmerman, J. 1977. The municipal accounting maze: An analysis of political incentives.
Journal of Accounting Research 15: 107–144.

77