An Empirical Analysis of Theories on Factors Influencing State

An Empirical Analysis of Theories on
Factors Influencing State Government
Accounting Disclosure
Rita Hartung Cheng
This study develops a politico-economic
model based on the theoretical and empirical
work in public choice and political science to explain state government accounting
disclosure choice. Measures of the theoretical constructs hypothesized to influence
accounting disclosure choice are selected from the literature. An updated 1986 practice
index, based on Ingram’s (1984) 12-practice categories, is used as the indicator of
accounting disclosure choice. The model is then tested for other indicators of accounting disclosure choice and a reanalysis is performed for 1978. Because of the complexity
of the political context, LISREL methodology was used to test the model. The evidence
supports the implication that state government accounting disclosure choice is dependent on factors in the political environment and on institutional forces. The model is
robust over time and for different measures of accounting disclosure choice.
1. Introduction
Numerous accounting studies have addressed the question of why accounting
choices are made. The private sector has developed a body of literature and
empirical findings into a positive theory of accounting (Watts and Zimmerman
1986, p. 14). The governmental
accounting researchers have also addressed
accounting choice and quality of financial reporting questions; however, few
empirical studies have been conducted which focus on the unique aspects of the
governmental
institutional
environment.
Government
accounting researchers
(Zimmerman
1977, p. 133; Baber 1983, p. 221; Baber and Sen 1984, pp.
102-103; Evans and Patton 1983, pp. 168-173; Evans and Patton 1987, p.
148; Ingram 1984, p. 139; Magann 1983, pp. 30-40; Robbins and Austin
1986, p. 418; Banker et al. 1989, p. 37; Giroux 1989, p. 211) have recognized
Address
School
reprint requests to
of Business
Administration,
Professor
P.O.
I I,
Rita
Box
Journal of Accounting and Public Policy,
I-42
0 1992 Ekvier Science Publishing Co., Inc.
Hartung
742,
Cheng,
Milwaukee,
University
of
Wisconsin-Milwaukee,
WI 53201
1
(I992)
0278-4254/92/.$5.00
2
Rita Hartung Cheng
the intercorrelation
of several economic and political measures with accounting
choice. These researchers used various measures of accounting disclosure
choice, making it difficult to compare the research findings. In addition, due to
multicollinearity
among the independent
variables,
data reduction methods
resulted in different measures being used for similar constructs in each
analysis. As a result of the differences in the selection of measures for both the
dependent and independent constructs, identifying specific variables that are
most important for explaining accounting choice has been problematic.
A
summary of governmental
accounting disclosure choice research is presented
in Table 1. A review of this table confirms the use of different variables for
similar constructs and the resulting lack of comparability of findings.
The purpose of my study is to integrate the findings of prior accounting
research with both the theoretical and empirical work in political science,
public choice, and public administration.
State accounting policy choices and
decisions to report financial information
are posited in my study to be
influenced by a number of factors in the political environment.’
The political
science literature has introduced a wide array of social, economic, cultural,
political, and institutional
variables as potential factors influencing
public
policy in different political systems. In addition, the literature on public choice
provides an analysis of the complex political environment and is important to
my study for its explanation of why voters, interest groups, politicians, and
bureaucrats should be viewed as dominant actors in government decisions to
adopt particular accounting practices.’ The complementary theories in political
science and public choice literature provide a necessary basis on which to
model voter/politician,
voter/interest-group,
interest-group/politician/bureaucrat, and politician/bureaucrat
relationships and their effect on the outcome of
accounting disclosure choices, This body of research also helps to explain why
assumptions
in the models of prior accounting
studies may have led to
conflicting results.3 Integration of the complementary
theories of the political
environment with accounting research to date should provide additional insight
into state government financial reporting choices.
’ McCormick and Tollison (1981, pp. I- 12) provide a discussion of the unique characteristics
of the
political (market) environment,
the complexity of the government
agency relationships,
and differences
that exist in the constraints that are imposed on self-interested
agents in the political setting from those in
th5 capital markets.
See Mueller (1979, 1989) for a comprehensive
review of the vast public choice literature. See Dye
and Gray (1980) for a comprehensive
review of the development
of determinant
research in political
science.
3 Ingram (1984, pp. 138%l39), Baber (1983, p. 22l), and Baber and Sen (1983, p. 105) recognized
the
intercorrelation
of several economic and political measures in their research. Data reduction techniques
and ad hoc selection of measures for similar constructs were employed. As a result, it is difficult to
interpret their findings and identify the specific variables
that are most important
for explaining
accounting practice choices. My study employs a path diagram to sort out these relationships.
Institutional measures are included to interpret the link that management control of the accounting system may
have to accounting practices and the mediating effect of the bureaucracy on accounting policy decisions.
Length of
report,
Audited and
by whom
Form of
Political
government*-Factors
mayor YS.
-Form
of
manager
government
Independent
Variables
-Audttor
-Financial
ability
-Managerial
ability
Capacity
Cay’s
Internal
-
I
-
Officials
wages
Controlled
for:
PaflY *
--Voter
tUrnoUt
-..
agent
_
_
-
renumeration
PolitIcal
PaflY*
--Yoter
turnout
held by
minority
--Seats
--seats
held by
minorIt>
Polittcal
competmon
Participation
fund
-
-
_
_
-Profe\\lonally
acttve off&al*
-PrlW
participation*
Quahty of
Management
Form of
government*mayor “I.
manager
CCP
Ctty
Evans and
Patton (1983)
GAAFR
State
Babcr and
Sen (1984)
Choice Research
Political
competitlon
state
Audit
Budget
state
Baber
(1983)
Dtsclosure
CPA
mayor YS.
manager
Audit Opinion,
MFOA Certlf.
* pages,
# exhibits,
Timeliness,
city
City
Magatltl
(1983)
Accountmg
Dependent
Variable(s)
Zimmerman
(1977)
of Government
system
Political
Study
Table 1. Summary
-
_
--
Management
_
-Accountant/
audttor selection*.
Administrator
Select10n
-Appomtivc
power*
COalltloll
of Voters
-Polltlcal
competition*
-Urbanization*
-Income*
-Educatwn
Index
Number of
pGKt,ces
state
Ingram
(1984)
^
\--
Income
Form of
government*mayor v\.
manager
Compound
Index and
Ingram (1984)
Index
Ctty
Robbtns and
Austtn (1986)
-
^
_-
-
CFO \alary*
CFO education
Profc\\lonally
active CFO
Prmr CCP
participation*
^
^-
Revemwwdent
Enrollment
mayor v.5
manager
ma>or “5.
milnager
PolItIcal
competition
Form of
government*-
-=~
budget
and statlstlC\
indice
index. and a 24.
practice Index
Income
Average
tax*
competttion
POlltlCd
manager
Form of
governmentmayor “\.
PensIon &
benefit,
city
Ciiroux
(1989)
Ingram tndcx.
Essential practice
School dtstnct
Banker. Bunch,
and Strau\\ (1989)
Form 01
government*--
CCP
Partlclpatlon
oty
Elan\ and
Patton (1987)
system
Polittcal
Study
Size
Controlled
for-
city
Ztmmerman
(1977)
Table 1. Continued
-Size*
-Financial
data
-Demographic
Population
size
City’s
Internal
Needs
-scope
-Complexity
Legislature
New debt
Federal
funds
State
Baber
(1983)
-Legislature
regulation*
-state
External
Constraints
-contracts
City
Ma&Q””
(1983)
Turnover*
statutory
barriers
Debt
state
Baber and
Se” (1984)
Population*
Debt*
Ctty
Evans and
Patton (1983)
Alternative
Information
-Newspaper
circulation’
-Population
reYe”“e
-Debt
--IntergLl”er”“le”t
state
Ingram
(1984)
firm we*
Population
Audit
Debt*
Intergovernment
revenue*
City
Robbins and
Austin (1986)
Population*
State GAAP*
Debt*
Tax revenue
complexity
city
Evans and
Patton (1987)
State GAAP*
Independent
Debt*
IntergO”e~““X”t
rwenuc
School district
Banker, Bunch.
and Strauss (1989)
CPA*
State GAAP
audit opmion
Tax revenue
complexfity
city
Glroux
(1989)
5
State Government Accounting Disclosure
A politico-economic
model, including measures for factors hypothesized to
influence accounting disclosure, is developed and tested in my study. As many
of the relationships and concepts are not new, this is an attempt to integrate
these relationships
into a broader analysis. Due to the complexity of the
political context, analysis of covariance
structures (causal modeling),
frequently called LISREL (Linear Structural RELations),
is selected for this
study. The advantage of this approach to data analysis is that it is not necessary
to assume perfectly measured variables. A distinction is made between theoretical variables and their indicators. LISREL can be described as a combination
of multiple regression and factor analysis. It combines two statistical traditions:
the structural model from econometrics and the measurement, or factor, model
from psychometrics.
The strength of LISREL is its ability to generate measures of theoretical
constructs by using several nonperfect indicators for each of these inherently
complex constructs. The method assumes that there is a causal structure among
a set of theoretical,
or latent, variables.
The latent variables appear as
underlying causes of the observed variables, i.e., indicators (Joreskog and
Sorbom 1986, p. 3). The model specifies the structural relationships among the
latent variables. In addition, the model provides for the estimation of the
measurement relations between the latent variables and their respective empirical indicators.
The use of LISREL for my study is important in order to expand upon prior
research which has relied on single measures for factors hypothesized to
influence accounting choice. This technique was developed especially to deal
with the complex measurement and structural modeling problems of the social
sciences, particularly psychology, economics, and sociology. LISREL is highly
regarded in other disciplines.
Its adoption could also serve to improve our
understanding of accounting choice in state governments.
My paper is organized as follows. A model to explain accounting disclosure
choice is developed from prior research in the next section. The research
method is explained in Section 3. Results of the statistical analysis are included
in Section 4. Conclusions
are made based on the results of the study,
with appropriate discussion of limitations and plans for future extensions in
Section 5.
2. A Causal Model: Theoretical Constructs
Operational Indicators
and
A review of the literature suggests that accounting policy choices are not
simply a function of economic or political factors, but are a result of legislative/governor/bureaucratic
decisions shaped by voter preferences,
interestgroup pressures, party competition, institutional forces, external demands and
constraints, and the financial condition of the states. In the following discus-
Rita Hartung Cheng
6
sion,
these factors are divided into four broad categories:
1. Socioeconomic
factors
2. Political system factors
3. Characteristics of the bureaucracy
4. Factors that represent other external
demands
and constraints
Factors within these broad categories are called latent constructs. Each is
inherently complex and no single measure is likely to fully characterize each
latent construct. Thus, sets of potentially important observable measures are
suggested from the literature.
Accounting Disclosure Choice
The problem of selecting measures of accounting disclosure choice has been
addressed by various researchers in governmental accounting. Early studies of
state government
relied on measures, such as length of financial report,
whether the financial statements were audited and by whom; and size of the
state audit budget, to surrogate extent and quality of disclosure. In more recent
years, different indices have been developed to represent the extent of disclosure (Ingram 1984, pp. 131- 134); extent and perceived importance of disclosures (Robbins and Austin 1986, pp. 414-417);
perception of disclosure
quality (Governmental
Finance Officers Association
(GFOA) Certificate of
Achievement for Excellence Program), and self-reported conformance to generally accepted accounting principles (NASACT 1986).
Ingram’s (1984, pp. 131- 134) 12-practice index of accounting disclosure
choice is included in my study for several reasons. First, Ingram developed an
index of practices based on the Council of State Governments’
Inventory of
Current State Government Accounting and Reporting Practices (CSG
1980). This index has been found to be robust when compared to weighted
measures and has been used in prior research to proxy both quantity and
quality of accounting disclosure (Ingram 1984, p. 127; Robbins and Austin
1986, pp. 413-417;
Banker, Bunch, and Strauss 1989, p. 30). Second, this
index was used to analyze disclosure practices of the states in 1978 and no
update has since appeared in the literature on state government practices. In
my study, the 1986 financial reports of all fifty states in the United States were
examined to compile an updated 1986 practice index using Ingram’s categories. A description of Ingram’s index and a comparison of 1986 practices
with those reported by Ingram is provided in Table 2.
The politico-economic
model developed below will be analyzed for its
ability to explain differences in the 1986 practice index and will also provide a
reanalysis of factors that may explain differences in Ingram’s 1978 index. In
addition, the model will be verified using other measures of accounting
State Government
Accounting
Table 2. Comparison
Accounting
Practice
2
3
4
5
6
7
8
9
10
11
12
Disclosure
of 1986 Practices
7
with
1978 Practices
Number of States
Adopting Practice
Description
1978
1986
General fund balance sheet
Statements of revenue and expensesenterprise funds
Basis of accounting disclosed
Comparison of general fund to budget
Accounts payable recorded
Fixed assets reported
Statement of long-term debt
GAAP-based funds
Taxes receivable reported
Accrued vacation leave reported
Short-term borrowing reported
Leased assets reported
35
34
44
32
29
18
18
17
13
8
8
7
4
44
38
32
21
36
25
22
25
12
12
39
A comparison of the number of states following each of the above major accounting practices is
indicated for the years 1978 and 1986. The 1978 practice data was reported in the Council of State
Governments
(1980) survey and summarized by Ingram (1984, Tables 2 and 3, pp. 132, 135). The 1986
data for the same twelve practices Ingram used in his 1984 study was developed by this author from an
analysis of the 1986 Comprehensive
Annual Financial Report (CAFR) of each of the 50 states. The
accounting disclosure choice indices used in my study are calculated from the above information.
Ingram
(1984, pp. 131- 134) calculated an index of disclosure choice based on the extent to which the practices
were adopted by each state. Each state was given a number representing the number of practices (from
the twelve listed) that were adopted (Ingram 1984, p. 134). I developed a 1986 practice index for my
study based on the same twelve practice categories as found in Ingram (1984, pp. 134- 135). The states’
1986 annual financial reports were examined to determine the extent to which each state adopted each of
the above practices.
This information
was used to compile the 1986 practice index of accounting
disclosure choice. The accounting practice number and description were the same used by Ingram (1984,
p. 135) in his Table 3.
disclosure as suggested from the literature. Results of this analysis should lead
to a better understanding of the political environment and the complex linkages
among social, political, and economic factors, and disclosure practices. It will
also provide a comparison of factors that may help to explain disclosure
practices in 1986 with those that explain 1978 disclosure practices.
Socioeconomic Factors and Environmental Conditions which
Influence the Demand for Accounting Information
The theoretical basis formalized in political science and public choice literature
is that factors in the environment
indirectly and directly influence policy
decisions of government bodies. Beginning with the work of Downs (1957),
the focus of much of the public choice literature has been on the citizen/voter.
In public choice research, socioeconomic variables are used as surrogates for
the median voter. The importance of including the relationships between the
voter and politician, and between voter and bureaucrat in the study of public
8
Rita Hartung Cheng
choice is recognized by these economists. My study also attempts to address
the impact of these relationships.
A basis for the use of socioeconomic
variables is found in the work of
Milbrath (1965). Milbrath (p. 119) documents higher participation and voting
rates as the population of a state develops into a higher socioeconomic status.
Downs (1957, p. 147) and Zimmerman (1977, p. 136) later conclude that the
voter lacks incentives to directly acquire information
about state policies.
Evidence has shown, however, that as society increases in population, urbanization, and economic and social differentiation,
diverse organizations develop
to represent these segmented interests (Dahl and Tufte 1973, p. 33). Research
findings, however, do not support the effects of socioeconomic
status on
interest-group
strength as many of these organizations’
interests have been
found to offset one another (Becker 1983, pp. 388-394).
Governmental
accounting choice literature has also used various indicators
of socioeconomic development. Ingram (1984, p. 139) found significant relationships among urbanization,
income, education, and political competition,
confirming the political science literature, and then created a construct, Coalition of Voters, to represent external monitoring. Baber (1983, p. 218)
controlled for state population in his study of the association of political
competition and the supply of auditing in state government. Evans and Patton
(1983, p. 163) found population to be a significant variable in the Municipal
Finance Officers Association
Certificate of Conformance
Program (CCP)
participation.
Robbins and Austin (1986, p. 418) included per capita income
and population in their study and found a significant relationship between the
former variable and the quality of financial reporting. Banker et al (1989, p.
34) selected enrollment as a proxy for coalition formation.
Consistent with public choice and political science literature, the effects of
socioeconomic
variables on financial reporting choice will be tested in my
study. As economic development and social diversity increase, political competition is expected to increase. This increased political competition will, in turn,
put pressures on the political system to effect accounting disclosure.
The
impact of socioeconomic development on interest-group numbers is thought to
be positive; however, the socioeconomic development effect on interest-group
strength is expected to be negative consistent with Becker’s (1983, pp.
372-373) theory of competition among pressure groups.
To extend prior research, multiple indicators for the construct socioeconomic development
are selected for the analysis. Potential indicators are
population, urbanization,
industrialization,
per capita income, and education
level of the citizenry based on support found in prior research.
The Political System
A very important relationship,
recognized and examined at great length in
public choice studies and political science literature, is that of the voter/politi-
State Government Accounting Disclosure
9
cian. The political system, comprised of voter, interest groups, and elected
politicians, is an important domain for the study of public policy decisions and
may also contribute to the study of accounting choice. The complex agency
relationships found in the political arena have been discussed by Downs (1957,
pp. 138-141); McCormick and Tollison (1981, pp. I-12); and Bendor and
Moe (1985, p. 757).
Measures for political competition, interest-group strength, and measures for
two key political actors, i.e., the governor and legislature
(that have a
significant role in the decision making of state government) are selected. As
discussed above, both interest groups and political parties organize individuals
to make claims upon government (Moorehouse 1981, pp. lOO- 101). However,
these two forms of political organization differ. The political party has a basic
function to organize a majority of citizens for the purpose of governing and is
less concerned with policy issues (Downs 1957, p. 137). Interest groups seek
to influence specific policies of government and give expression to the interests
of minority groups (McCormick
and Tollison 1981, pp. I- 12). Previous
accounting studies (e.g., Baber 1983, p. 217; Baber and Sen 1984, p. 96;
Ingram 1984, p. 131) have looked at the effects of interest groups or political
competition on accounting choice, but no study has used both forms of political
organization.
The governor and legislature must respond to the demands of
individual voters and interest groups. Prior research (e.g., Abney and Lauth
1986, p. 64; Brudney and Hebert 1987, p. 199; Moorehouse
1981, pp.
203-305;
Schlesinger 1971, pp. 220-234;
Stigler 1976, p. 31; McCormick
and Tollison 1981, pp. 61-77, pp. 113-121) suggests that characteristics of
these political actors, discussed later, will influence public policy outcomes and
may also help to explain accounting disclosure practices.
Political Competition
The general political environment of a state is defined by Baber (1983, p. 215)
as “the strength of opposition that a political entrepreneur expects to encounter
in future elections. ” It is assumed in political science literature that strong
party competition and the accompanying
prospect of close partisan elections
will provide an incentive for the governor and legislators to exercise influence
over the bureaucracy (Dye and Robey 1980, p. 7; Schlesinger 1971, p. 227).
My study depicts political competition
as positively related to financial
disclosure because of incentives political participants
have to monitor the
behavior of the opposition in order to maximize the number of votes in an
election (see Downs 1957, p. 138). The impact of political competition will
manifest in pressures placed on the political structures to disclose accounting
information.
Many of the prior studies in political science and accounting have used the
degree of interparty competition (in political science, Dawson and Robinson
10
Rita Harhmg Cheng
1963, p. 276; Dye 1966, p. 296; Plotnick and Winters 1985, p. 463; and in
accounting, Ingram 1984, p. 137; Baber 1983, p. 217; Baber and Sen 1984, p.
96), partisan control of state government (Ranney 1976, p. 61; Klass 1980, p.
146, Baber and Sen 1984, p. 96), and the level of voter turnout (Dye 1966, p.
258; Baber 1983, p. 217; Baber and Sen 1384, p. 96) as typical characteristics
of the political system which influence public policy. Consistent with this
research, indicators chosen for my study are: 1) an index of interparty
competition
developed by Ranney (1976, pp. 51-60) and recalculated by
Bibby et al. (1983, p. 66); 2) percentage of seats held by minority party in the
legislature; and 3) percentage vote for the winning party in the last gubernatorial election. A proxy for intraparty competition,
voter turnout in the most
recent gubernatorial primary, is also included in the model.
Interest-Group Activity
The economic interest-group theory asserts that voters use interest groups to
reduce the vast quantities of information required to make informed decisions
in elections. The early work of Downs (1957, pp. 147-149); Olson (1965, pp.
22-23);
Buchanan and Tullock (1962, pp. 213-214);
Bartlett (1973, pp.
55-58);
and Stigler (1971, p. 12) focused on why rational voters would
delegate their information-processing
and decision-making
responsibilities
to
interest groups in order to reduce the high costs of monitoring government.
More recent elaborations of economic interest-group theory have assumed that
interest groups in turn exert much influence on politicians, voters, and bureaucrats (Peltzman 1976, pp. 221-222; Becker 1983, p. 372).
Interest-group
theory suggests that a principal-agent
relationship exists between government
officials and various interest groups (McCormick
and
Tollison 1981, p. 5). Interest-group theory views legislators, the governor, and
bureau administrators
as economic agents who respond to their institutional
environment.
Interest groups are the principals monitoring and lobbying for
political influence (McCormick and Tollison 1981, p. 5). Interest groups have
been linked with legislature influence (Brudney and Hebert 1987, p. 198),
legislative decision making (Weingast 1984, pp. 149- 151), governor monitoring of states’ policies (Crain and Tollison 1979, p. 165; McCormick and
Tollison 1981, p. 114), and as an important party in the legislator/bureaucrat
relationship (Bendor and Moe 1984, pp. 757-761).
Interest-group
strength is expected, a priori, to be positively related to the
monitoring of politicians’ behavior and the demand for accounting information.
The construct interest-group strength, however, is difficult to measure. Consistent with prior research, Moorehouse’s (1981, pp. 108- 112) impressionistic
classification of the states according to pressure-group
strength is selected for
my study. Abney and Lauth’s (1986, p. 101) index of the level of interaction is
used as another indicator of interest-group strength. In addition, the number of
Political Action Committees
(PACS) registered with the Federal Election
State Government Accounting Disclosure
11
Committee is employed as a crude measure of interest-group
strength. This
measure was selected in lieu of state PACS because of data constraints on state
lobbying groups. This proxy, however, is thought to create offsetting lobbying
behavior as interest groups compete for political influence (Becker, 1985, p.
342).
Power of the Governor
It is widely argued that formal powers of the governor result in a more
responsive bureaucracy (Schlesinger 1971, p. 217; Abney and Lauth 1986, p.
64; and Brudney and Hebert 1987, p. 199). Measures of governor power found
in political science research are tenure potential (Moorehouse 1981, p. 205;
and Schlesinger 1971, p. 225), power of appointment (Moorehouse 1981, p.
228; and Schlesinger 1971, p. 227), salary (Schlesinger 1971, p. 23.5), size of
staff (Abney and Lauth 1986, p. 5), and Schlesinger’s (1971, pp. 220-234)
general index comprised of the general tenure provisions, appointive powers,
responsibilities
for budget preparation, and the power to veto bills passed in
the legislature. It is argued (Schlesinger 1971, p. 225) that the length of term
and re-election probability impact on the incentives of the governor to monitor
a state’s activities and on the control a governor has over personnel who may
outlast him/her.
The signaling literature (Spence 1973, pp. 356-357; Ross
1977, pp. 24-25) also supports the governor incentives for an outward show of
quality of financial reporting.
Power of appointment
is the most widely
appreciated means of controlling bureaucratic officials (Moorehouse 1981, p.
228).
The formal power of the governor is expected in my study to be positively
related to the monitoring of appointed officials and the demand for information
among individuals in government. One index used in my analysis is based on
the governor’s powers of appointment in sixteen major functions and offices as
developed by Schlesinger (1971, p. 227). Salary may also proxy for governor
status and/or power. High salary is thought to be indicative of independence
from external influence and may reflect more devotion to internal goals
(Schlesinger 1971, p. 235). Size of staff is another proxy for governor power
(Abney and Lauth 1986, p. 5). Finally, the combined index developed by
Schlesinger (1971, pp. 220-234) is tested.
Legislative
Power
One of the most important premises of government agency theory is that, in the
absence of capital-market mechanisms, the legislature is the primary monitor
of bureaucratic behavior (Fama 1980, p. 295; Miller and Moe 1983, p. 311;
Weingast 1984, p. 148; Spencer 1982, p. 198; Shepsle 1986, p. 136; Banks
1989, p. 672). Legislators are viewed as attempting to maximize their chances
for re-election by providing
a monitoring
function on state bureaucratic
12
Rita Hartung Cheng
behavior (Bendor, Taylor and Van Gaalen 1987, p. 815). Where legislative
power is strong, active administrative
lobbying has been documented (Abney
and Lauth 1986, p. 69) A major strategy in lobbying is to provide neutral
legislators and those with influence with information. Legislative size (Stigler
1976, p. 31); appropriate authority (Schlesinger 1971, p. 227); appointment
power (Moorehouse
1981, p. 228); professionalism
(e.g., wages, length of
session, number of committees)
(Moorehouse
1981, p. 288); and tenure
(Patterson 1983, p. 155) have all been used in prior research as indicators of
legislature power and are included in this analysis. These factors are thought to
be related to monitoring incentives which may result in an increase in the
quantity of financial disclosure. Due to the lobbying reaction of the bureaucratic administrative departments and the resulting increase in political actions
of these units discussed in Rowley and Elgin (1985, p. 43), however, the
quality of financial reporting may not be significantly
related to legislative
strength.
The Bureaucracy: Internal Needs for Information,
Ability to Provide Quality Accounting Disclosure
Incentives, and
Political science and public choice researchers have emphasized the importance
of characteristics
of the bureaucracy
for public policy decisions (Niskanen
1971, pp. 24-35; Migue and Belanger 1974, p. 28; Bendor, Taylor, and Van
Gaalen 1985, p. 1044; Rowley and Elgin 1985, p. 48; Abney and Lauth 1986,
p. 5). The theory of institutions, particularly in the bureaucratic realm, argues
that “dimensions of the bureaucracy and bureaucratic behavior are responsible
for variations in policy outputs” (Downs 1976, p. 11). Niskanen (1971, pp.
45-52) provided the first economic-utility
maximization
model of the public
bureau. Niskanen’s model led the way for a rich body of literature in which the
bureaucracy is analyzed on the basis of universal self-seeking assumptions,
discarding the public-interest Weberian (Weber 1947) model of elected government. This institutionalism
considers the relative autonomy of political institutions and the importance of bureau interaction with the environment.
Abney
and Lauth (1986, p. 222) refer to the importance of neutral competence in
their study of state and municipal governments.
Fama’s (1980, p. 289) notion
about outside managerial market monitoring and the literature on fiscal illusion
(Pommerehne and Schneider 1978, pp. 384-385; Wagner 1976, p. 51; West
and Winer 1980, p. 617) also supports the inclusion of variables which
characterize management ability and incentives in my study.
In previous governmental
accounting research Ingram (1984, p. 137) used
salaries, CPA status, and selection variables to surrogate management ability.
Evans and Patton (1983, p. 161; 1987, p. 145) discussed quality of management as important and used bond ratings, education, and salary as surrogates
for management quality. Baber (1983, p. 218), and Baber and Sen (1984, p.
94) included political agent renumeration
as a proxy for quality. Consistent
13
State Government Accounting Disclosure
with prior research (e.g., Downs 1976, p. 11; Ingram 1984, p. 137; Baber
1983, p. 218; and Baber and Sen 1984, p. 94), the following have been
selected as potential indicators of bureaucratic ability and quality of management: salary and professional
certification of the auditor general and chief
accountant; size of the auditing and accounting departments; number of CPAs;
whether these positions are appointed versus elected; mean wage of public
employees; and percentage of unionized positions.
Magann (1983, p. 23) and Ingram (1984, p. 139) have also argued that
bureaucratic complexity and financial ability to provide information may
impact on the amount and quality of accounting information; however, neither
used measures for complexity from the political science literature. In my study
an attempt is made to provide measures of this construct. Observable measures
to proxy for extent/complexity
of bureaucracy
include total expenditures,
number of full-time equivalent employees, and number of governmental units.
The financial ability of the government to provide information demanded of it
may also be an important determinant of accounting disclosure since the costs
of complying with generally accepted accounting practices must be weighed
against the benefits of reduced costs that result from contracting with interested
parties in the political market. Banker et al. (1989, p. 36) selected revenueper-student as a measure of fiscal ability. Ingram (1984, p. 139) found own
revenue as a percentage of total revenue to be significantly related to financial
accounting disclosure. This measure of financial ability is included in my study
and is expected to positively affect the quantity and quality of financial
reporting.
External
Demands
and Constraints
Political science literature and accounting studies have recognized other external influences on state policy decisions. Other agency relationships have also
been analyzed in public choice research. Four additional external forces
discussed below are: 1) contracting agreements in the debt market; 2) the
federal government; 3) outside audit firms; and 4) the press.
Contracting Agreements. Despite the lack of a well-defined theory of
municipal bond valuation, there is preliminary evidence that state and municipal accounting information have an effect on bond ratings and interest costs.
Conclusions are that the market reacts to the perception of good accounting
practices in general, although individual practices may not be reflected in debt
characteristics.4 The relationships among accounting information, bond ratings,
4 See Ingram et al. (1987)
market research.
for a comprehensive
review of the development
of governmental
capital
14
Rita Hartung Cheng
and bond yields (Wallace 1981, p. 511; Ingram and Copeland 1984, pp.
33-36; Wilson and Howard 1985, p. 222) suggest that there may be incentives
on the part of state officials to improve the quantity and quality of financial
reporting when there is outstanding debt. A review of government
finance
literature also reveals conceptual arguments and empirical findings which
suggest that the information included in municipal and state financial reports
may be relevant for the analysis of debt issues (Petersen 1974, p. 76;
Rabinowitz 1969, p. 136). In addition, Standard and Poors (S & P 1982) has
indicated that the quality of accounting disclosure will impact on their bond
rating decisions.
Baber and Sen (1984, p. 103) and Ingram (1984, p. 139) found an
insignificant relationship between debt and their measures of quality of financial disclosure for state government.
Evans and Patton (1987, p. 149) and
Robbins and Austin (1986, p. 418), however, found debt to be a positive and
significant explanatory variable for municipal disclosure quality. Banker et al.
(1989, p. 44) also found debt to be significant for school districts. These
findings suggest that debt may be more of a factor at the local level of
government than at the state level. Outstanding debt per capita is incorporated
in this study as positively influencing the quantity and quality of disclosure due
to incentives governments have to minimize the cost of debt. Further incentives
are expected if the state has a significant proportion of its bonds rated by the
rating agencies (Moody’s or Standard and Poors) or if net interest costs are
high.
Federal Government. Public organizations researchers have concluded that
although the presence of a substantial proportion of a state’s funding by the
federal government may serve to increase federal influence and monitoring of
the state’s disclosure, such funding may also serve to insulate a state from
influence-attempts
by the legislature and governor, and, perhaps, by interest
groups (Wright 1982, p. 199; Wamsley and Zald 1973, p. 42). This may
explain the mixed results to date in the accounting literature. Baber (1983, p.
218) and Ingram (1984, p. 137) used surrogates for federal funds in their
research. Ingram’s proxy, intergovernmental
revenue/total
revenue, was not
significant. Baber (1983, p. 221) did find population to be a significant control
variable, but questions arise as to the appropriateness
of population as a proxy
for federal influence.
In my study, the percentage of total state revenue from the federal government is selected as a proxy for federal government influence. The direction of
the relationship between federal government influence and accounting disclosure choice is not stated a priori.
Outside Audit. In my study the existence of an outside auditing firm is also
posited to affect state financial reporting. Rubin (1987, p. 17) associates the
amount of audit activity to improvements
in quality of accounting information
State Government
Accounting
Disclosure
15
argument
for the demand for audits. Baber (1983, p. 215)
selected state audit budgets as a surrogate for the amount of monitoring of state
bureaucracy. Magann (1983, p. 58) and Banker et al. (1989, p. 40) found state
auditing requirements
to be a significant determinant of financial disclosure
quality. Robbins and Austin (1986, p. 418) specifically looked at size of audit
firm as an important determinant of quality in municipal financial reporting.
Banker et al. (1989, p. 41) also found independent external auditors to be a
major influence in the level of financial disclosure and conformance to generally accepted accounting principles (GAAP).
No study to date has looked at the influence of outside audit on state
government. Two variables from municipal research are selected in my study
as potential indicators of the impact of external audit on the decisions to report
financial information: 1) the existence of an independent private auditor, and 2)
the size of the state audit budget.
in his theoretical
The Press. Zimmerman (1977, p. 121) argued effectively that the press
plays an important role in monitoring the activities of public officials and,
therefore, a strong press will increase the incentives for public officials to
disclose financial information.
Downs’ (1957, p. 146) analysis of voting
behavior suggests that the press may play a significant role in voting decisions
by reducing the costs of information.
Alternatively,
as discussed in Zimmerman (1977, p. 121) the kind of information demand primarily facing the press
is an important factor in the role of the press in the agency relationship
between voters and politicians. Ingram (1984, p. 141) found a press proxy,
newspaper circulation per capita, to have a significant, but negative, relationship to accounting disclosure. Ingram speculated that the press may be a cost
effective substitute for disclosing accounting information, or that a strong press
may provide incentives for public officials to disclose less information
to
protect themselves from negative reports. Another explanation,
discussed in
Zimmerman
(1977, p. 121), may be voters demand for entertainment,
as
opposed to information, from the press.
In my study, the existence of a strong press is assumed to facilitate voting
decisions, as well as interest-group
formation,
and to result in increased
accounting disclosure consistent with the literature. Two indicators of a strong
press selected for this analysis are newspaper circulation per capita and the
number of newspapers per capita in each state.
General Hypothesis
Figure 1 is a graphical summary of the links among theoretical
the political environment
that may help to explain accounting
model posits eleven unobserved theoretical variables that directly
may affect the decisions to provide accounting information by
ments. Although prior accounting research has found proxies of
constructs in
choice. This
or indirectly
state governsome of these
16
Rita Hattung Cheng
Figure 1. The politico-economic
model of accounting disclosure choice. + / cates predicted sign of relationship. Arrow indicates predicted causal path.
indi-
theoretical constructs to be related to accounting choice, no accounting study
has appropriately
addressed the complex interrelationships
of the political
environment.
An analysis of the relationships among the theoretical constructs
is of primary interest in my paper. The reliability and validity of the indicators
for each theoretical construct will also be tested. Table 3 lists the observable
indicators for each theoretical construct suggested from the literature and
discussed earlier.5
Finally, the explanatory power of the overall model (how well the model fits
the data) will be tested. The general null hypothesis testing the overall model
is:
The politico-economic
model developed from a priori information and theory
provides an explanation for the financial reporting status of state governments.
5 Data necessary to test the model and hypotheses
was obtained from the U.S. Bureau of the
Census(l978a,
1978b, 1986a. 198b): the Council of State Governments
(1980); the National Association
of State Auditors, Comptrollers
and Treasurers (NASACT,
1986); Federal Election Commission;
data on
state bond issues that I purchased from Public Securities Association;
Moody’s Municipal Government
Manual (1978, 1986), and other published sources. In the absence of theoretical guidelines for specification of the time requirement
for a state government
to react to changes in its political environment,
a
two-year difference in the explanatory and dependent variables was selected. Data for 1984 was used to
develop measures to explain 1986 accounting practice choice and data for 1976 was used to estimate the
model explaining
1978 practices. This temporal lag was selected due to data availability
of accurate
census information and is consistent with Ingram (1984, p. 137.)
State Government
Table
3. Theoretical
Construct/Indicator
Accounting
Constructs
17
Disclosure
and Observable
Indicators
Variable Definition/Measure
QUAL12
Accounting disclosure choice
Ingram’s (1984) practice index recalculated
DIV
URBAN
INDUST
PINCOME
EDUC
POP
Socioeconomic development and diversity
percent population in standard metropolitian areas
percent employed in manufacturing
per capita personal income
percent population completing four or more years college
population
PC
MINOR
WINN
RANNEY
Political competition
percent legislative seats held by minority party
percent vote for winning party in last gubernatorial election
Ranney (1976) index of partisan control of state government
(governor, senate, and house)
voter turnout for last gubernatorial primary
QUA
TURNOUT
IGS
PACS
ACTIVITY
INTERACT
for 1986
Interest-group strength
number of groups/capita
registered with the Federal election commission
Moorehouse (1981) classification of states according to level
of interest-group strength
state interaction index (average deviation from mean)
GOV
GAPPT
GTENURE
GSALARY
GENINDEX
Power of governor
degree governor has sole power over 46 functions or offices
5-point scale of governor’s tenure potential
governor’s salary
Schlesinger (1971) formal index, 23 tenure potential,
appointive powers, budget powers, organization powers,
and veto powers
LEG
LSIZE
APPRAUTH
LWAGE
SESSION
COMMIT
TURNOVER
Legislative power
number of seats in house and senate
number of bills passed/number
of bills introduced
mean legislative wage
number of days in regular session
number of legislative committees
number of membership changes/total
number of members
BIA
WAGES
AUDREQ
AUDSAL
ACCTSAL
CPA
SIZEACAU
APPTELECT
UNION
EXPEND
FTES
Bureaucracy needs and abilities
average earnings, non-education employees
1 if CPA required for state auditor position; 0 otherwise
state auditor salary
chief accountant salary
number of CPA’s on staff
total positions in accounting and auditing departments
1 if audit agency head appointed; 0 otherwise
percent of state employees organized
total government expenditures/capita
number of state full-time equivalent employees/capita
18
Rita Hartung
Table
Cheng
3. Continued
Construct/Indicator
Variable Definition/Measure
GOVUNITS
OWNREV
number of government units in state
total own revenue/capita
DC
LTDEBT
NIC
Contracting
long-term debt/capita
average net-interest cost over three-year period prior to
financial statement
current Moody’s bond rating (Moody’s Municipal and
Government Manual 1986)
BONDRATE
FED
FEDFUNDS
Federal influence
intergovernmental
revenue from federal government/total
AD
AUDIT
ABUDGET
Outside audit
1 if use outside auditor; 0 otherwise
audit agency budget (1,000,000’s)
PR
CPRCIR
CPRNUM
Press
newspaper circulation/capita
number of newspapers/capita
revenue
3. Research Method
The statistical procedure used to estimate the model developed in the previous
section is an application of the LISREL model developed by K. Joreskog
(1973, pp. 86-87)(j. The LISREL model consists of two parts, the measurement model and the structural model, which are estimated simultaneously.
LISREL allows the researcher to posit multiple observable indicators for the
underlying unobservable variable, or latent construct, and through the use of
factor analysis, to propose and test a measurement model of the construct and
its indicators (Joreskog and Sorbom 1986, p. 3). The measurement
model
specifies how each imperfect real-world measurement is related to the underlying latent construct and is used to describe the measurement properties, i.e.,
‘Over the years, statistical techniques have been developed
for dealing with situations in which
multiple variables, some unobserved,
are involved and where measured variables only rarely correspond
on a one-to-one basis with the unobserved constructs of interest to the researcher. Various names have
been used to refer to an extremely general technique for analyzing data. such as: “covariance
structure
of linear structural
relations;”
“structural
“analysis
of covariance
structures; ” “analysis
model;”
equation modeling;”
“causal modeling;”
and “analysis of moment
equation modeling; ” “simultaneous
structures”
(for a review of the development
of this methodology
see Long, 1983, pp. 7-13). Used
carefully, these names do not refer to the same thing, but have become common terms that refer to the
method implemented
in such computer packages as LISREL (Joreskog and Sorbom 1986). COSAN
(McDonald
1978), and EQS (Bentler 1985). Joreskog’s (1973, pp. X6-87) general Linear Structural
RELations
(LISREL) model is the oldest and most widely used of these programs and has become
synonymous with the methodology.
19
State Government Accounting Disclosure
validities and reliabilities (Joreskog and Sorbom 1986, p. 3). The structural
model tests the causal relationships among the latent constructs (Joreskog and
Sorbom 1986, p. 3).’ In this analysis, accounting disclosure choice is specified
as a function of the latent constructs defined in the measurement model.
The general LISREL model (Joreskog and Sorbom 1986, p. 6) is defined by
three equations:
Structural equation model:
r] = BV + r[
+ [
(1)
Measurement
model for y : y = A,7 + E
(4
Measurement
model for x : x = A,[
(3)
+ 6
where,
. . >7,) : a random vector of latent dependent
71’ = (?l,,?l2,.
E’ = (E,, Ez,. . . > E,): a random vector of latent independent
B(m x m) and I?(m x n): coefficient
C’ = (.C,,&..,s’,):
constructs
constructs
matrices
a random vector of residuals
(disturbance
terms)
Equations 2 and 3 state that although the political and economic constructs, q
and .$ that are thought to affect accounting choice cannot be observed, a
number of other variables denoted as indicators y’ = (y,, y,, . . . , y,) and
x’ = (x,, x2,. . .) x,), that are imperfect measures of the political and
economic constructs, are observable (Joreskog and Sorbom 1986, pp. 5-6),
where,
y’: the observed dependent
x’: the observed independent
A,(p
x m): regression
A,(q x n): regression
variables
variables
(indicators)
(indicators)
(factor) matrix of y on v
(factor) matrix of x on C;
E and 6: vectors of errors of measurement
in y and x respectively
The estimation procedure of LISREL is to fit the variance-covariance
matrix
implied by the model C, to the observed variance-covariance
structure of the
’ The use of structural equation (causal) models requires statistical tools which are based upon, but
which go well beyond, conventional
regression analysis and analysis of variance. A full mathematical
discussion of this method can be found in Hayduk (1987, pp. 87-138);
Lwhlin
(1987, p. 49-53);
Joreskog (1973, pp. 107-112); Carmines and McIver (1983, pp. 52-66). and Long (1983, pp. 13-28).
While the multiple-indicator
methodology
is mathematically
complex, its logic is relatively straightforward. The estimation of indicator reliabilities
is analogous to factor analysis’ estimation of indicator
correlations with the underlying factor. The estimation of the construct-to-construct
link follows the logic
of path analysis, where the correlation between indicators equals the product of the paths connecting
them. In actuality,
the indicator reliabilities
and construct links are estimated by a single set of
simultaneous
equations,
rather than two distinct operations
(Joreskog and Sorban
1986, p. 2). An
appendix which includes a mathematical
discussion of the analysis of covariance structures is available
from the author of this paper to the reader upon request.
Rita Hartung Cheng
20
data, S (Joreskog and Sorbom 1986, p. 27). Maximum-likelihood
estimates,
the most frequently used estimation procedure in LISREL analyses, were
employed for this study to simultaneously
estimate parameter estimates for
both the measurement and structural sections of the model. This is done by
minimizing the function:
F=logJCJ
+tr(SC-‘)
-1og)Sl
-
(p+q)
(4
where tr(SX-‘)
is the sum of the diagonal elements, 1C ( is the determinant of
C, p and q are the number of observed y and x variables (Joreskog and Sorbom
1986, p. 28)
Table 4 presents the correlation matrix for pairs of indicators. This correlation matrix consists of three types of correlation
coefficients:
1) product
moment (Pearson) where both variables are continuous; 2) polychoric where
both variables are ordinal; and 3) polyserial where one variable is continuous
and the other is ordinal.8 This correlation matrix served as initial input for the
LISREL analysis. Significant correlations among several of the indicators are
evident from the correlation matrix. As discussed before, LISREL requires
that all the indicators for a given latent theoretical construct be correlated.
However, high collinearity between indicators causes interpretational problems
in LISREL similar to those in regression equations with proxy variables (Jagal
1982, p. 432). The principal advantage of this estimation procedure over
standard regression is that sources of multicollinearity
can be identified and
overcome (Hayduk 1987, pp. 175-176).
4. Analysis of Results
The statistical problem is this analysis is not one of testing a given hypothesis,
but rather one of fitting the model to the data and deciding whether the fit is
adequate. In addition to specifying a model and the mathematical fitting of this
model using the maximum likelihood technique, steps in the assessment of a
LISREL model include statistical evaluation of the fit of the model; criticism of
the reliability, validity, and areas of lack of fit; and proposed reformulation of
the model (Carmines and McIver 1983, p. 53). The last two steps are unique to
LISREL and allow for several nested, or alternative, models to be compared in
this analysis. Evaluation of several plausible models rather than a single
hypothesis facilitates testing, evaluation, and interpretation of a final solution.’
a When observed variables are of mixed-scale type (ordinal and interval), the use of ordinary product
moment correlations
is not recommended
(Olsson et al. 1982, p. 338). Polychoric
and polyserial
correlations have been found (Olsson et al. 1982, p. 347) to be unbiased, efficient correlations,
in the
sense of being closest to the true p. The LISREL program was employed to produce a correlation matrix
consisting of all three types of correlations,
where each correlation was estimated separately (Joreskog
an$ Sorbom 1986, p. 43).
In evaluating competing models, Carmines and McIver (1983, pp. 63-65) suggest that differences in
x2 values can be examined. If the drop in x2 is large compared to the difference in degrees of freedom
(df), there is an indication that the change made in the model represents a real improvement.
If, on the
other hand, the drop in x2 is close to the difference in number of degrees of freedom, this is an indication
that the improvement
in fit is obtained by capitalizing
on chance,
and the added parameters may not
have any real significance or meaning (Carmines and McIver 1983, p.64).
State Government Accounting Disclosure
21
Several nested models, i.e., model MO can be obtained by constraining one or
more parameters of model M, , are compared by examining differences in their
relative x2( x2 /df) and other goodness-of-fit
criteria (Carmines and McIver
1983, p. 63). Procedures for relaxing and otherwise modifying the model were
followed according to Joreskog and Sorbom (1986).
Results of an iterative process of simultaneously
estimating the parameters
of the measurement model and estimating the structural coefficients for 1986
are shown in Figure 2. The final parsimonious model is also applied to 1978
data (Figure 3).
The Measurement Component
The initial choice of observable indicators was based on content validity or
face validity as determined by the review of the theoretical literature and prior
research findings. Next, convergent validity, the extent to which the observable variables appear to be measuring the same construct, is addressed in
LISREL.”
Finally, to satisfy the constraints of LISREL, indicators for each
latent construct must be correlated (convergent
validity), but must not be
highly correlated with many other indicators (discriminant
validity) (Howell,
1987, p. 121). The discriminant
validity of the measures is tested by
observing the correlation of the observed measures. For example, if observed
indicators of two proposed latent constructs are highly correlated, they may
measure only one latent variable instead of the two proposed, and the validity
of the two constructs is nil.
Standardized regression weights (factor loadings) and the corresponding
t
values, given for each regression equation of the measurement model, are used
to assess the reliability and validity of the proposed measures, or indicators. As
expected, not all of the observed indicators from the literature meet the
reliability and validity criteria of LISREL, which suggests that some proxy
variables used in prior research or suggested from the literature are not
appropriate measures for the theoretical constructs in the political arena. In
addition, conflicting research findings in prior studies can be explained by the
researchers’ use of different measures and the high degree of multicollinearity
among their measures. ’ ’
A measurement
model, which shows how indicators that meet the reliability and validity tests are related to the theoretical constructs, is presented in
Table 5.
“Some judgment
is involved in this step because no specific correlation
standards for convergent
validity have been established.
Nunnally (1978, p. 95) has suggested that the correlation should exceed
0.5.
” For example,
Ingram (1984, p. 139) addressed
multicollinearity
of the data by reducing the
independent variables into four broad factors, making the interpretation
of individual results difficult. In
addition, Baber’s (1983, p. 222) multivariate
tests of political competition
showed different results
depending on the measure of political competition used.
URBAN
POP
EDUC
PINCOME
INDUST
ACTIVITY
INTERACT
RANNEY
MINOR
WINN
TURNOUT
LSIZE
APPRAUTH
LWAGE
LSESSION
COMMIT
TURNOVER
GAPPT
GTENURE
GENINDEX
GSALARY
LTDEBT
NIC
BONDRATE
FEDFUNDS
AUDIT
1.000
.565**
.348**
.547**
.244*
- .378**
- ,198
.321**
.08.5
- ,043
-.121
.075
- .373**
.514**
,234
,137
- .416**
.190
.228
.270*
.375**
- .136
.04.5
.103
- .474*
- .310**
URBAN
1.000
,042
.279**
,199
-.147
-.154
,173
.158
- .216
- ,101
,191
- .312**
.589**
,164
.489**
- .387**
.253*
.353**
,119
.350**
- .250*
- ,038
,049
- .171
- ,148
POP
1.000
.654**
- .255*
- .367**
- .247**
- .251*
,221
- ,077
-.145
- ,065
- ,017
.232
,076
- ,215
.024
.060
.301**
.449**
,097
.332**
,003
-.103
- .451**
- ,099
EDUC
1.000
.047
.529**
,124
,095
,406
- .187**
- ,165
,013
- ,229
.588**
,151
- ,054
- ,182
,067
.285**
.374**
.336**
.394**
- ,069
,181
- .701**
-.189
-
PINCOME
,211
- ,074
,073
- .246*
.454**
- ,229
,068
,052
.325**
- ,094
,175
- .237*
- .161
,064
- ,223
,074
-.159
.245*
- ,068
,127
1.000
-.162
INDUST
Table 4. Product Moment (Pearson) Polyserial and Polychoric Correlation
1.000
.067
.322**
- .331**
.021
.416**
- ,034
,123
- .328**
- .206
- .008
.056
- .313**
- .455**
- .561**
-.136
,008
,241
- ,053
.164
-.159
ACTIVITY
,071
,079
- .236*
-.175
.cy2
,036
,037
,223
.072
-.167
,138
1.000
- ,056
- ,144
-.183
- .022
,062
- ,024
- ,180
- .338**
INTERACT
1.000
- .477**
,090
.414**
.Oll
-.198
.056
,078
,182
- .070
.004
- .284**
- .247*
,227
-.142
.279*
.290*
- .014
-.139
RANNEY
Matrix of Observed Variables
1.000
- .314**
- .353**
-.154
-.174
.341**
.206
- ,051
- ,228
.051
.292**
,164
,110
,099
,164
- ,226
- .336**
,048
MINOR
,199
.024
.146
1.000
- ,234
,178
.301**
- .272**
-.lll
-.170
,022
- .284**
-.113
- ,066
-.102
- .252*
- ,054
WINN
1.000
- ,121
- .196
,057
- ,012
,084
.056
,029
- ,139
- .145
,105
.186
,208
- .105
- ,039
,040
TURNOUT
1.000
- .120
- ,004
-.127
.392**
.002
- ,075
-.160
-.133
,195
- ,212
p.163
- .023
.257*
-.181
LSIZE
.334**
.550**
,181
.359**
CPAS
SIZEACAU
APPTELECT
UNION
FTES
- .250*
COMMIT
,,
- .044
LTDEBT
.248*
.472**
- .479**
GSALARY
-.004
.095
.003
.449**
- .286**
GENINDEX
-.130
.356**
-.131
GTENURE
.219
- .132
,014
- .309**
1.000
.274*
LSESSION
.304**
-.125
,186
.398**
.328**
- ,133
.293**
.367**
.540**
,037
.114
- ,110
,031
.328**
,051
.657**
- ,003
EDUC
.353**
-
.374**
- .224
GAPPT
,153
- .079
LSESSION
TURNOVER
- .523**
1.000
LWAGE
APPRAUTH
1.000
.242*
- .443**
.273*
.I08
-.150
.618**
.355**
LWAGE
APPRAUTH
QUAL12
- .726**
.332**
CPRNUM
.321**
CPRCIR
- .098
.199
PACS
OWNREV
GOVUNITS
.050
,041
.911**
.475**
.349**
- .493**
.254*
ACCTSAL
.630**
- ,083
- .282**
.561**
AUDSAL
-.176
,160
AUDREQ
.274**
.607**
POP
- ,138
.426**
WAGES
EXPEND
.320**
ABUDGET
URBAN
Table 4. Continued
- .252*
.541**
- .191
.056
,046
- ,179
1.000
COMMIT
,205
-.159
.485**
.462**
.475**
,158
,156
.447**
.570**
,154
.332**
,013
.165
.562**
.127
.743**
.163
PINCOME
.482**
-.186
- ,096
- ,087
-.153
1.000
TURNOVER
,144
- .343*
,097
- ,073
- .306*
,128
- .389**
- .379**
,135
- .025
,170
,060
- ,087
,009
-.193
- .255*
.118
INDUST
-
,174
- ,064
.584**
,080
1.000
GAPPT
- ,105
.142
- .520**
- ,144
,068
- .217
.308**
.03 1
- .565**
- a40
- ,203
.149
- ,036
- .272*
-.144
- .379**
-.195
ACTIVITY
,068
- ,048
‘224
.519**
1.000
GTENURE
- .239*
.288*
- ,044
.014
.021
- .038
,136
- ,030
- .236*
,203
- ,069
- .034
- ,038
- .061
,169
,045
1.000
GEINDEX
.257*
- .431**
- .238*
,070
1.000
GSALARY
.014
.106
.439**
,191
,219
- .078
- .022
.396**
- ,056
,204
.296**
- ,209
- .003
- .087
- .066
-.169
,135
,088
.057
,169
- ,057
,126
-.116
.489**
.138
MlNOR
,149
.205
,151
-.106
-.164
- ,031
,064
- .309
RANNEY
INTERACT
1.000
LTDEBT
-.121
- ,026
- .217
- ,208
- .279**
- .280**
- .249*
NIC
,109
.099
- .240*
-.189
.300**
- ,220
.306**
.305**
-.103
- .266*
-.159
,184
-.107
.074
- ,030
- .045
- .268*
,051
- ,057
TURNOUT
- ,192
- .236*
- .002
,106
- ,088
,158
- .356**
- ,146
WINN
BONDRATE
,207
-.119
,066
-.lll
- .287**
.196
- .348**
- .299**
.116
,046
.217
,106
.068
.151
- .349**
- .241*
,222
LSIZE
,092
SIZEACAU
APPTELECT
QUAL12
CPRNUM
CPRCIR
.294**
- .320**
.288**
- ,101
.442**
- .262*
,190
.403**
.089
- ,141
OWNREV
PACS
.415**
- .017
-.180
.005
.379**
.399**
-.005
.620**
.271*
- .036
GOVUNITS
FTES
- .291**
- .383**
CPAS
- .122
- ,188
ACCTSAL
EXPEND
.355**
-.104
AUDSAL
UNION
.610**
- .282*
AUDREQ
.606**
- .207
.278**
WAGES
,519
.176
.065
- .357**
AUDIT
ABUDGET
- .475**
.191
FEDFUNDS
.308*
- ,188
,113
- .201
LWAGE
BONDRATE
NIC
APPRAUTH
Table 4. Continued
.215
- ,077
,069
- .023
- .163
.114
- .284**
- .340**
- .287
.144
.230
.221
.4&a**
- .292**
.OOO
.119
.066
.279**
- ,148
.282**
- ,039
.063
.096
.164
- ,097
,212
-.190
-.143
.109
- .356**
- .142
,179
-.004
- .189
.382**
.184
- .297**
.078
-.124
.049
,137
- .089
-.121
.235*
.057
- .198
,077
-.104
.260*
- .131**
,214
.248*
- ,140
.381**
- .006
- .193
,033
.425**
- .376*
,130
,008
.118
TURNOVER
COMMIT
- .362**
LSESSION
.096
,219
- .163
,168
.051
- .152
.164
- ,080
- .146
.326**
.151
.358**
- .OOl
,023
.012
- .283**
.037
.266*
- .090
.080
- .023
-
GAPPT
.021
-.141
.283**
,202
.013
.378**
- .244*
,041
.237*
,017
.364**
,130
.114
.327**
.449**
- .022
,164
,100
- .268*
,069
- .178
GTENURE
,111
- .184
.266*
,163
,156
.095
,093
,200
.567
- .053
.235
- ,046
.175*
,179
-.008
.400**
.237*
.233
-.113
- .149
- ,056
GINDEX
.223
- .258*
,179
,231
.207
.115
- ,077
,171
.109
- ,031
.406**
.281**
.565**
.545**
-.046
.239*
.368**
- .453*
- .240*
.143
,078
GSALARY
.673**
.818**
.359**
.097
,158
- .021
.209
,103
,072
.828**
- .278*
-
- .086
- .274*
.163
,143
.526**
- .083
,222
- .315**
- ,056
- .033
LTDEBT
,065
- .232
- .446**
.192
- .155
- .017
,111
- ,063
- .153
.261*
-.121
-.128
.OlO
- .OlO
.220
- ,201
- .352**
- .122
- ,014
.142
1.000
NIC
.002
-.004
- .013
,117
.202
.168
,130
,186
- .069
- ,011
,053
.136
.lll
.051
.060
,094
- .066
,141
- .262*
1.000
BONDRATE
FTES
1.000
- .493**
.711**
.032
- .087
.264*
- .051
**significant
at ,015
* significant at 10
FTES
GOVUNITS
OWNREV
PACS
CPRCIR
CPRNUM
QUAL12
FEDFUNDS
FEDFUNDS
Table 4. Continued
.148
1.OOO
-.199
.032
.296**
.008
GOVUNITS
AUDIT
1.000
.127
,100
.246*
,011
OWNREV
ABUDGET
1.000
.Oll
- .247*
.211
PACS
WAGES
1.000
.024
.085
CPRCIR
AUDREQ
l.ooO
- .404**
CPRNUM
AUDSAL
1.000
QUAL12
ACCTSAL
CPA.9
SIZEACAU
APPTELECT
.472**
- .266*
.392**
.105
UNION
1.000
.743**
- .225
.974**
.089
,096
.245*
.023
EXPEND
Rita Hartung Cheng
Factor loadings appear below each indicator name. Indicators assigned as unit of
measurement have loadings fixed at 1.0 and, thus, no standard errors. Asymptotic t
statistics for other loadings are in parentheses.
Beta coefficients for paths in the
structural model are given with t statistics in parentheses. *p < . lO,**p < .05,***p <
.Ol
The Structural Model in Equation Form:
IGS = .764***DIV
(3.366)
GOV =
- .444* PC + .839** IGS
(- 1.500)
(1.852)
ABIL = .630***IGS
(2.617)
QUA =
(-
.269** ABIL (1.715)
- ( _tl;:s,
Coefficient of determination
Asymptotic
t statistics are
equations are derived from
deviations from their means.
,026 GOV
,165)
(-
,133 GOV ,995)
.354***
(-2.623)
PR
DC + (::09;) BIA
for the structural equations = .716. R2 for QUA = .192.
in parentheses.
No constant terms appear because the
a covariance matrix in which variables are measured as
*p < .lO,**p < .05,***p < .Ol
Figure 2. Results of LISREL analysis for 1986 accounting disclosure choice. x2 =
193.61 with 102 df; Goodness-of-fit
index = ,721; Adjusted goodness-of-fit = ,582;
Root mean square residual = .136.
State Government
Accounting
Disclosure
27
Factor loadings appear below each indicator name. Indicators assigned as unit of
measurement have loadings fixed at 1.0 and, thus no standard errors. Asymptotic t
statistics for other loadings are in parentheses.
Beta coefficients for paths in the
structural model are given with t statistics in parentheses. *p < .lO, **p < .05, ***p
< .Ol. An explanation of the Ingram (1984) Index is in the text.
The Structural Model in Equation form:
IGS = .541*** DIV
(2.186)
GOV = - .434***PC
(-2.375)
ABIL = .630***IGS
(2.617)
QUA = (:;;:)
-
.342** IGS
(2.425)
(-
.269 GOV
,885)
ABIL + (;4E12)
***GOV
.422***
(-3.942)
Coefficient of determination
Asymptotic
t statistics are
equations are derived from
deviations from their means.
+
-
.310***
(-2.734)
PR
DC + .384***BIA
(3.149)
for the structural equations =
in parentheses.
No constant
a covariance matrix in which
*p < .lO, **p < .05, ***p <
,455. R2 for QUA = ,505.
terms appear because the
variables are measured as
.Ol.
Figure 3. Results of LISREL Analysis for 1978 accounting disclosure choice. x2 =
164.63 with 87 df. Goodness-of-fit
index = .733; Adjusted goodness-of-lit = ,583;
Root mean square residual = ,158
28
Table 5. Parameter
Rita Hartung
Estimates
of the Measurement
Cheng
Model
Estimate
Disclosure Choice (QUA) Construct
QUAL12
Socioeconomic Development and Diversity (DIV) Construct
EDUC
PINCOME
FEDFUNDS
PACS
Interest-Group Strength (IGS) Construct
ACTIVITY
UNION
CPRCIR
Political Competition (PC) Construct
RANNEY
MINOR
TURNOUT
Governor Power (GOV)
GAPPT
GENINDEX
Bureaucratic Ability/Legislative
Strength (ABIL)
LWAGE
AUDSAL
ACCTSAL
SIZEACAU
CPAS
Press Strength (PR)
CPRNUM
Debt Covenants (DC)
NIC
Bureaucratic Financial Ability (BIA)
LTDEBT
EXPEND
OWNREV
t Value
l.BQO’
1.otxl
1.027
- .678
,442
(5.512)
(-4.455)
(3.079)
l.ooll
- ,841
- ,554
(-4.789)
(- 3.446)
l.GQO
- .391
.307
(-2.108)
(1.875)
1.000
,482
(2.404)
l.ooa
.803
.519
,710
.373
(5.894)
(2.596)
(5.633)
(2.525)
l.ooa
1.ocKl
1.000
.978
,997
(10.025)
(10.294)
’ The above results are for a measurement model with all paths of the structural model constrained to
zero. Indicators assigned as unit of measurement have loadings fixed at 1 .O and, thus, no standard errors.
Measures of goodness of fit for the whole model: x2 = 355.28 with 197 degrees of freedom; Goodnessof-fit index = .660; Adjusted goodness-of-fit = ,523; and Root mean square residual = .132.
The indicators for socioeconomic development and diversity that meet the
convergent validity, discriminant validity, and reliability criteria of LISREL
are education level of the citizenry (EDUC); personal income per capita
(PINCOME); federal funds flowing into the state as a percent of total state
budget (FEDFUNDS),
and the number of registered political action committees (PACS). The first two indicators are as predicted; however, the latter two
indicators are discussed in the literature as indicators of interest-group strength
and federal government influence, respectively. EDUC, PINCOME, and PACS
load positively on the latent construct, socioeconomic development and diver-
State Government Accounting Disclosure
29
sity. FEDFUNDS loads negatively. The findings suggest that these are indicators of state development consistent with Ingram (1984). Ingram (1984, p. 139)
found personal income and urbanization
to be positively correlated with
intergovernmental
revenue, but did not term these variables as proxies for
socioeconomic
development.
He (1984, p. 128) designated this group of
variables as indicators of coalition formation. Other indicators that did not
meet the convergent and discriminant
validity were urbanization
(URBAN),
industrialization
(INDUST), and state population (POP). Although urbanization and population were found to be significant by Ingram (1984, p. 141) and
Baber (1983, p. 221), respectively,
these measures may be proxies for
something other than socioeconomic development.
Significant measures of interest-group strength are Moorehouse’s (198 1, pp.
108- 112) classification
of interest-group
strength (ACTIVITY),
unionism
(UNION), and press circulation
per capita (CPRCIR).
ACTIVITY
loads
positively on IGS; the coefficients for UNION and CPRCIR are negative. The
relationship of the press to interest-group
strength is not surprising.
This
finding is consistent with Downs (1957, pp. 146- 148), who considered interest
groups and the press as information specialists. The negative loading of the
circulation per capita measure suggests that when newspaper circulation is low,
information
costs are high, which causes elected officials to rely on and
respond to interest groups since their political careers depend on their ability to
assess and fulfill the desires of their constituency.
Unions also are another
source of information for the citizen/voter
and politician. When the percentage
of state employees covered by a collective bargaining unit is low, interest
groups provide the mechanism for information exchange and policy influence.
This finding could also explain Ingram’s (1984, p. 139) negative correlation
between press and extent/quality
of financial reporting. Two new measures did
not load in the LISREL analysis. Abney and Lauth’s (1986, p. 101) state
interaction index (INTERACT) did not converge with any other indicators and
PAC, was highly correlated with the socioeconomic
indicators discussed
earlier.
Measures of political competition in the LISREL model are percent-minority
party in legislature (MINOR), voter turnout in last gubernatorial
election
(TURNOUT), and the Ranney index of political competition (RANNEY). The
measure percentage vote for winning party in last gubernatorial
election
(WINN) does not show significant factor loadings in any of the models. The
signs of the factor loadings are not all positive. Baber and Sen (1984, pp.
102-103) and Baber (1983, pp. 221-223) also found the above measures of
political competition to have different signs. One explanation is that RANNEY
is a comprehensive
index, and along with MINOR, measures interparty
competition. TURNOUT is a measure of intraparty competition and was also
found to be negative by Baber (1983, p. 221) in regression results when other
political competition measures were positive.
30
Rita Hartung Cheng
Significant indicators for strength of the governor are appointment power
(GAPPT) and Schlesinger’s (1971, p. 227) formal power index (GENINDEX).
Other indicators,
governor
salary
(GSALARY)
and governor
tenure
(GTENURE), do not have the expected relationship. In his 1984 study, Ingram
(p. 139) did not find salary of the governor to be significantly
related to
appointive power of the governor consistent with the current findings. No other
accounting study has used these measures.
The interrelationships
between indicators of political competition and indicators of legislative influence, and between legislative influence and indicators of
audit/accounting
ability of the bureaucracy are very complex and the model is
not able to isolate the latent construct legislative power (LEG). Other attempts
to measure legislative influence have also been unable to isolate the construct.
Ingram (1984, p. 139) included legislative wage and other salary data in his
coalition formation variable. Baber (1983, p. 221) found an insignificant
negative coefficient for legislative size and a significant correlation of this
measure with political competition measures in his model explaining quality of
financial reporting. Indicators selected for this construct did not meet the
discriminant
validity test of LISREL. As a result of this finding, the LEG
construct was not included in the models. Results of the measurement analysis
also support dividing the bureaucracy construct into an ability construct and an
incentive construct. One of the indicators selected to measure legislative
influence, mean legislative wages (LWAGE), converges with indicators of
bureaucratic ability; understanding this collinearity is important when interpreting the results of the analysis.
Other measures of the accounting and auditing ability of bureaucracy
to
provide quality financial information
(ABIL) that load with LWAGE are
average salaries of accounting personnel (ACCTSAL);
number of certified
public accountants in the accounting staff (CPAS); size of the accounting and
audit staff (SIZEACAU), and average salary of the audit personnel (AUDSAL).
Salary information, along with size of accounting and audit staff, and number
of certified employees gives an indication of professionalism,
and accounting
and audit ability to provide information demanded by forces in the political
environment.
Other indicators suggested from the literature mean wage of
public employees (WAGES) whether audit agency head is appointed or elected,
(APPTELECT),
and if CPA is required for state audit position (AUDREQ) do
not pass the discriminant validity test.
Indicators of the internal incentives and financial ability of the bureaucracy
to provide quality financial information
(BIA) which have significant factor
loadings are total expenditures (EXPEND) and percentage of own revenue to
total revenue (OWNREV). Long-term debt per capita (LTDEBT) also shows a
highly significant factor loading on BIA. The relationship of long-term debt per
capita to size and complexity of government, and financial ability of government, is consistent with prior studies (Ingram 1984, p. 139), and may explain
31
State Government Accounting Disclosure
why prior accounting studies have not found level of state long-term debt to be
a significant explanatory
variable in understanding
the incentives of state
government to provide accounting information.
Results of the measurement model for debt market influence indicate a need
for better indicators of this ~eoretic~ly
important variable. Long-term debt
per capita (LTDEBT) and net interest costs (NIC) are used, but their factor
loadings are insignificant
(although in the expected direction). Bond rating
(BONDRATE) does not enter the model. Since LTDEBT is highly correlated
with EXPEND and OWNREV, NIC was used as the sole measure of debtmarket influence on quality of financial disclosure when the bureaucracy
measures were included in the model.
The two suggested measures for the audit variable, the existence of an
outside auditor (AUDIT) and size of the audit budget (ABUDGET),
do not
show significant factor loadings. In addition, ABUDGET is highly correlated
with other indicators and does not meet the dis~riminant
validity for the
LISREL model.
per capita
The two indicators
for the press, number of newspapers
(~PR~UM)
and newspaper circulation per capita (CPRCIR), are not highly
correlated. CPRCIR loaded well on interest group strength, while CPRNUM,
a measure not tested in prior accounting research, was retained as the sole
indicator of the strength of the press.
The final paths of the structural component in Figure 2 were obtained by
evaluating the overall explanatory value and statistical fit of the estimated
likelihood function of several alternative nested models (Carmines and McIver
1983, p. 63), starting with a test of the paths as suggested in Figure 1. Results
of these alternative specifications help to establish the reasonableness
of the
findings. ‘*
Measures of the overall fit of the model to the data are provided in LISREL.
One test, the x2, is a measure of how much evidence there is against the model
(Joreskog and Sorbom 1986, pp. 38-41).
The x2 is sensitive to model
I2 Comparison
of Nested Models:
Model
Measurementmodel-no structurai relations
Politico-economic
model-Figure I
Model in Figure 2-1986 practice index
Model in Figure 3-1978 practice index(Ingram
1984)
Beta coefficients for each path in the alternative
coefficients are available from the author.
X2
df
X’/df
GF’I
AGF
RMR
355.28
592.31
193.61
164.63
197
215
102
87
1.80
2.16
1.90
1.90
,660
.559
.721
,733
,523
,433
S82
583
.I32
.413
.136
.158
models,
and reported
t statistics
for these
32
Rita Hartung Cheng
complexity, sample size, and severe departures from normality and is not used
as a statistical test, but rather as a measure of relative goodness of fit.13 Other
measures of model fit include the goodness-of-fit index (GFI), a measure of the
relative amount of variances and covariances jointly accounted for by the
model (GFI is independent of sample size and has a value between 0 and 1; its
statistical distribution is unknown and no standard is available with which to
compare it); the adjusted goodness-of-fit index (AGFI), which adjusts the GFI
for degrees of freedom; and the root mean square residual (RMR), a measure
of the average size of the estimated residuals, which can be interpreted in
relation to the sizes of the observed variances and covariances in the data
(Joreskog and Sorbom 1986, pp. 40-41).
Standardized regression weights (beta weights) and corresponding
t values
are also given for each regression equation for latent constructs in the
structural model. In addition, squared multiple correlations, R's,are given for
each endogenous latent construct of the structural model as a measure of the
strength of relationship between two constructs. Finally, the coefficient of
determination
is a measure of the strength of the relationships in all of the
structural equations of the model. A large value is associated with a high
explanatory power (Joreskog and Sorbom 1986, p. 37).
The theoretical constructs identified in the structural analysis as directly or
indirectly affecting accounting disclosure choices are: 1) socioeconomic development and diversity; 2) interest-group
strength; 3) political competition; 4)
strength of the governor; 5) debt market influences; 6) press strength, and 7)
characteristics of the bureaucracy. The nested models do not include the latent
constructs, legislative strength, outside audit, or federal government influence,
because only the observable variables that meet the reliability and validity tests
of LISREL are included in the final model. In addition, several measures that
meet the convergent validity tests, and are discussed in the measurement model
section, but continue to violate the discriminant
validity of LISREL, are
constrained to zero in order to present and test a parsimonious model. Even
though the final model does not include these indicators, e.g., RANNEY,
FEDFUNDS, PACS, UNION, other indicators which measure the same latent
constructs, but are not highly correlated with many other variables are included. These findings improve our understanding
of the appropriateness
of
indicators for the theoretical constructs discussed in the political science and
public choice literature, and how multicollinearity
can confound results.
The model in Figure 2 constrains insignificant paths, socioeconomic development and diversity (DIV) to political competition (PC), and interest group
strength (IGS) to accounting disclosure (QUAL12), to zero in order to improve
I3 The probability level of x2 is the probability of obtaining a x2 value
obtained given that the model in correct. The objective is to develop and
small x2 relative to the degrees of freedom (df). Wheaton et al. (1977, p.
compute a relative x’/df. Carmines and McIver (1983, p. 64) suggests that
to 1, or 3 to 1 are indicative of an acceptable fit between the hypothetical
larger than the value actually
test a model that produces a
99) instruct the researcher to
x*/df ratios in the range of 2
model and the sample data.
State Government Accounting Disclosure
33
the model fit. These constraints
and adaptations are consistent with the
theoretical literature, e.g., individual voters influence public policy through
interest groups and interest groups interact with appointed and elected officials.
Results of this iterative effort are a x2 of 193.61 with 102 df. Goodness-of-fit
indices and the average size of residuals (RMR) indicate an adequate model
that is more parsimonious
and theoretically
reasonable.
The coefficient of
determination of the model is .716.
Examination of the results suggests that socioeconomic development (DIV)
has a strong negative effect on interest-group strength (IGS). The results are in
the predicted direction and suggest that highly developed states will not have
strong interest groups. Rather, interest-group
strength rises when education
level of the citizenry and personal income per capita are low, and ties to the
federal government are high. This relationship is also suggested in the correlation matrix and is consistent with Becker (1983, p. 380). Although a causal
relationship between socioeconomic development and political competition was
posited from the literature, this path did not hold in the LISREL analysis. In
the final model, this path was constrained to zero.
Interest-group strength (IGS) is found to have a significant negative effect on
bureaucratic
accounting and auditing ability (ABIL) variable. This finding
suggests that interest-group
strength has a causal effect on the ability of the
bureaucracy to produce quality financial reports. These findings are consistent
with Bendor and Moe’s (1985, p. 771) public-policy research. Political competition (PC) has a weak positive effect on governor strength (GOV). The lack of
significant findings suggests that the relationship between political competition
and gubernatorial power is still open to further investigation. The relationship
between power of the governor and bureaucratic accounting ability (ABIL) is
also insignificant. The path from IGS to GOV is significant at the .lO level of
significance.
A construct that is significantly related to accounting disclosure choice is
press strength (PR). This finding is consistent with Ingram (1984, p. 141). The
number of newspapers per capita has a significant negative effect on financial
reporting choice in most models. This result suggests that newspapers may be a
cost-effective substitute for accounting
disclosure as discussed by Ingram
(1984, p. 143). Another possibility is that when the number of newspapers is
low, the few large papers have a powerful effect on public opinion and
government policy decisions, including accounting disclosure choice.
Bureaucratic accounting/audit
ability (ABIL) also has a significant positive
effect on extent and quality of financial reporting. Measures of this construct
correlate highly with overall size measures; therefore, in addition to measuring
accounting ability this construct also brings to the model the ability of large
states to have quality financial disclosure. Other causal effects, not significant,
but of the expected sign, are bureaucracy
complexity and financial ability
(BIA), and debt covenants (DC). The LISREL results are consistent with a
weak positive causal relationship between debt market and the extent and
34
Rita Hartung Cheng
quality of financial reporting. The findings suggest that low average net-interest
cost three years prior to disclosure, and amount of debt positively affect the
extent and quality of financial reporting.
Interest-group
strength (IGS) and governor (GOV) also have insignificant
direct effects on accounting
disclosure choice. These results suggest that
interest groups may not be interested in accounting disclosure nor in funding
the costly financial information
systems necessary for quality disclosure.
Carpenter (1987, p. 106) found similar results and suggests that the incentives
of interest groups to support or oppose funding requests by governments must
be incorporated into a theory of government accounting information production. The insignificant path between governor strength and accounting disclosure suggests that a strong governor will not determine the basic quantity of
accounting disclosure. Results of prior accounting studies, however, support
the contention that a strong governor may affect the outward show of quality,
e.g., awards for excellence in financial reporting.
The LISREL model for 1978 (Figure 3) supports the findings of 1986. The
model in Figure 3 is a test of the model developed above using Ingram’s (1984)
1978 practice index as the sole indicator of accounting disclosure choice. The
analysis yields a x2 of 164.63 with 87 df. Other goodness-of-fit measures are
also indicative of a good fitting model. This analysis was performed to confirm
the robustness of the model and also to compare results with the 1986 analysis
(reported in Figure 2) and Ingram (1984, pp. 139-142). The results suggest
that the power of the governor;
debt market interest costs; bureaucratic
financial ability and incentives; bureaucratic accounting and audit ability, and a
strong press are consistent causal factors in the decision to provide quality
financial reports. An interesting finding is the strength of the positive causal
relationship of power of the governor and the relationship of high net-interest
costs three years prior to disclosure on the quality and extent of financial
reporting. These findings are different from the 1986 models. One explanation
for these results is that original forces for change in quality of financial
reporting came from the debt market (Standard & Poors, 1982). Initial
demands for change were very controversial and required the intervention of
the governor. By 1986 most states had made some improvement
to their
financial reporting, and the debt market costs and power of the governor had
less influence on accounting choice.
Robustness
An examination of the correlations between the indicators included in the final
model provides some insights into the robustness of the results. Correlations
that were problematic
in prior regression studies were effectively handled
through the stringent reliability and validity tests of the measurement model.
Many significant correlations that remain among indicators are reflected in the
beta weights of the paths among the theoretical constructs of the structural
model. The highest correlation between any two indicators included in the
State Government Accounting Disclosure
35
parsimonious model, not connected with a structural path, is .447. In addition,
the LISREL output of the normalized
residuals and modification
indices
suggests that most, but not all, of the correlation among the indicators have
To the extent unexplained multicollinearity
been explained in the modelI
continues to exist among the indicators, parameter estimates of the measurement and structural model must be interpreted with caution.
To provide further evidence about the statistical significance of this model,
an analysis was performed on another measure of accounting disclosure
choice, self-reported conformance to generally accepted accounting principles
(GAAP) as reported by the National Association
of State Comptrollers
(NASACT, 1986). This self-reported GAAP measure was the only one suggested in the literature that met the convergent and discriminant validity tests
of LISREL and was found to be significantly correlated (.916) with Ingram’s
(1984, p. 134) practice index.15 Two analyses were performed, one substituting self-reported GAAP as the sole indicator of accounting disclosure choice in
place of the 1986 practice index (QUAL12), and a second analysis using
QUAL12 and self-reported GAAP as joint indicators of accounting disclosure
choice. Results of these analyses were consistent with the 1986 results shown
in Figure 2. The relative x * = 1.67 and x2 = 1.60, respectively,
were
consistent with the results of the 1986 model. Significant beta weights in the
measurement and structural equations remained stable. I6 Further, coefficients
I4 The problem of multicollinearity
in structural equation models with latent variables is not resolved.
Some researchers suggest that problems are similar for structural equation models to those found in other
econometric models (Jagpal 1982, p. 432). Judge et al. (1980, p. 459) suggest a correlation of 0.8 as
indicative of a serious collinearity
problem. Although LISREL allows for correlated error terms, and
therefore allows for multicollinearity
between indicators of different latent constructs,
interpretational
confounding may occur, and such modifications
to the model must be attempted with caution (Hayduk
1987, p. 188). My study assumes uncorrelated errors across all equations in the model and does not test a
mygel with correlated error terms because of these interpretational
problems.
In 1986, 26 states self-reported
financial statements conforming to GAAP and 24 states reported
non-GAAP
financial reporting.
Although
the dichotomous
variable to represent the Governmental
Finance Officers Association (GFOA) Certificate of Achievement for Excellence award is also correlated
(.641) with the 1986 practice index, only seven states were awarded the certificate from the GFOA in
1986. Further analysis indicated that the GFOA variable did not meet the discriminant
validity test of
LISREL, i.e., the dichotomous
variable was significantly
correlated with many of the indicators, and
models tested with the variable did not converge. I also developed an index similar to Robbins and
Austin’s (1986, p. 417) compound index by using five of the Ingram (1984, p. 132) disclosure-practice
categories and two audit-activity
indices. This index was not found to bc correlated with the 1986 practice
index. ‘The insignificant correlation (. 156) suggests that the addition of the audit variables has resulted in
a variable that does not measure the same underlying
theoretical
construct at the state level of
government.
Other measures suggested from prior studies, the size of the audit budget (Baber 1983, p.
215) and existence of an external audit firm (Robbins and Austin 1986, pp. 418-419;
Banker et al. 1989,
p. 40), were not significantly
correlated
with the other measures of accounting
disclosure
choice.
Co;;elations
are (.130) and (.017), respectively.
Results of the model with self-reported GAAP as an indicator of disclosure choice:
Model
Self-rewrted
Self-reported
GAAP
GAAP & 1986 practice index
X2
df
170.68
188.34
102
117
r’/df
1.67
1.60
GFI
AGFI
.731
.728
,597
.608
RMR
,142
,139
Beta coefficients for each path in the alternative models and reported I statistics for these coefficients are
consistent with the model presented in Figure 2, which has the 1986 practice index as the sole indicator of
disclosure choice.
36
Rita Hartung Cheng
of determination of .646 and .659 indicate that the relationships measured in
the structural equations in the model have explanatory power.
The evidence supports the implication that state government
disclosure
choice is dependent on factors in the environment
and institutional forces.
Findings of the lack of collinearity of other measures of government accounting disclosure choice help to explain contradictory findings among other prior
studies. Further study is necessary to determine what factors influence a state’s
decision to apply for the GFOA Certificate of Achievement for Excellence
Award and what factors influence audit choice.
5. Conclusions, Limitations, and Future Extensions
The political-economic
model developed in this study from existing economic
and political theories provides a plausible explanation for state government
accounting disclosure choices based on the standard goodness-of-fit criteria for
structural equation models with latent variables. Relationships are consistent
across models and years for the causal effect of socioeconomic development
(DIV) on interest-group
strength (IGS); the causal effect of interest-group
strength (IGS) on governor strength (GOV) and bureaucratic accounting/auditing ability (ABIL); the causal effect of political competition (PC) on governor
strength (GOV); and the causal effect of bureaucratic
accounting/auditing
ability (ABIL) and the press (PR) on state government financial reporting
choice (QUAL12). The model developed to explain 1986 financial disclosure
practices is also significant for 1978. In 1978, governor strength (GOV), debt
market influence (DC), and bureaucracy size/complexity
and financial ability
(BIA) are also significantly related to extent and quality of financial disclosure.
Perhaps the most significant feature of this study is that it applies the study
of political markets to accounting choice. Our understanding
of the incentives
“nonmarket
decision
makers”
(Mueller
1979,
p.
3)
in
an accounting
of
context is limited. A more complete analysis of the political environment
is
accomplished
in my study through the use of the LISREL simultaneous
estimation of a system of structural equations. Applying such an analysis,
similar to that applied in other public-choice
studies, is important to our
increased understanding
of this political process. Important principal agent
relationships in the political setting not previously addressed in an accounting
context are identified as well.
In addition to exploring the political process in more depth than prior
accounting studies, my paper has built on Ingram’s (1984, p. 134) work by
updating an inventory of accounting practices by state government and by
sorting out the multicollinearity
of the political and economic variables. Many
of the LISREL findings are consistent with regression results of prior accounting research. The models are also able to represent the complex relationships
between the theoretical constructs and indicators in much more detail than
previous research. The additional relationships studied will contribute toward
our understanding of state government accounting choice.
State Government
Accounting
Disclosure
37
Several
limitations
of the study should be noted. Although this paper
includes additional public choice and political science theories, in the past
thought to be competing, but now viewed as complimentary,
observed relations
can be misleading to the extent unspecified factors affect accounting choice. In
addition, although careful testing of content, convergent,
and discriminant
validity criteria was done throughout the paper, to the extent that the measurement model is incorrect, structural parameter estimates of the relationships
between the latent constructs are biased. Finally, the absence of theoretical
guidelines precludes accurate specification
of the time required for state
governments to react to changes in the political environment.
The results do not prove the model, merely that it is a plausible explanation.
The x2 goodness-of-fit
test may be quite sensitive to sample size, model
complexity, and severe departures from normality. Other goodness-of-fit measures are not sample dependent and were included to compensate for the x2.
Significance tests must also be interpreted cautiously since the observations do
not represent a random sample and, as such, they serve only as relative
measures of importance of the associations. Finally, even with the advancements in multivariate analysis, this study is cross-sectional;
therefore, we must
remain circumspect about drawing causal inferences. Further study is required
before the question of causality can be fully addressed.
Future research should investigate whether the effects found in this study can
be replicated by studying individual states more closely. A case-study approach
comparing several states may be helpful in sorting out the complexities of the
state government environment and the effect of the relationships in the political
market on accounting decisions. Replications must also be performed for other
years and for other government accounting policy choices.
This paper is based on my doctoral dissertation completed at Temple University in 1988. I wish to
thank Mary Anne Gaffney for her guidance and support throughout the entire research project.
This paper also benefited from the helpful suggestions and comments of Ruth Ann McEwen, Harry
F. Bailey, Jr., James Arbuckle, and three anonymous reviewers.
References
Abney, G. and Lauth, T. 1986. The Politics of State and City Administration.
Albany: State University of New York Press.
Baber, W. December 1983. Toward understanding the role of auditing in the public
sector. Journal of Accounting and Economics 5(3):213-221.
Baber, W. and Sen, P. Summer 1984. The role of generally accepted reporting methods
in the public sector: An empirical test. Journal of Accounting and Public Policy
3(2):91- 106.
Banker, R., Bunch,
financial reporting
counting,
pp. 27-56.
B. and Strauss, R. 1989. Factors influencing school district
practices. In Research in Governmental and Nonprofit AcVolume 5, (J. Chan and J. Patton, eds.) Greenwich, CT: JAI Press Inc.,
38
Rita Hartung Cheng
Banks, J. August 1989. Agency budgets,
Journal
costs information,
and auditing.
American
State Government
Accounting
Downs, G. 1976. Bureaucracy,
Lexington
Politics,
American States.
Innovation,
Lexington,
Books, pp.
Signalling and
and Economics
E. April
Agency
Political
151- 175.
problems
Fall 1989.
interests
Public Policy
1987. Structural
Baltimore: Johns
R. February
1984.
Copeland,
and credit
Schwartz, P.
19-40.
incentives
Joreskog, K.
In
0. Duncan,
K.
Judge, G.,
Econometrics.
choice of
Governmental
government
CT: JAI
for estimating
W., Hill,
research
Inc., pp.
l-126.
models
the Social
linear structural
system.
(A. Goldberger
Press, pp.
LISREL-VI
D.
and Lee,
John Wiley
User’s
1980. The
Sons.
of policy
of Public
politics in
(T. Dye
4th Edition.
and Practice
american states,
V. Gray,
pp. 141-
Latent Variable
An Introduction
Factor, Path,
Structural Analysis.
NJ: Lawrence
Associates.
Loehlin, J.
with
Marketing Research
Models
Sorbom,
IN: Scientific
capital
and Nonprofit
in structural
Journal
New York:
1980. The
In The
Toronto: Lexington
and
Chan, ed.)
A general
New
and measurement
Research in
1982.
Equation
Essentials and
Research 22(l):
Wilson, E.
Part B
November
variables.
with
R.
The association
municipal accounting
and return.
Advances in
Volume 1,
A. Wright,
Greenwich CT:
Press Inc.,
Raman, K.,
A review.
ing, Volume
disclosure.
Press.
structure
126.
Journal of
R.
accounting
199-217.
Modeling
Research 24(l):
Ingram,
Spring
accounting
Klass,
of
the firm.
and
University
Covariance
Journal of
Joreskog,
the theory
of Accounting
Hayduk,
Jagpal,
accounting.
88(2):288-307.
Giroux,
Ingram,
in
in public
130- 158.
J. April
An economic
of participation
the
officers association
of conformance
Journal
of
Ingram,
in
of the
(T. Dye
of Public
Accounting Research
J. and
municipal
Outcomes
versus economics:
In The
MA:
Patton, J.
Journal
the Public:
MA:
Rand McNally.
J. 1980.
Evans, J.
Public Policy.
Economics,
Dye, T.
T. and
on policy
V. Gray,
39
Disclosure
40
Rita Hartung Cheng
Long, J. 1983. Covariance Structure Models: An Introduction
Hills. CA: Sage Publications.
to LISREL.
Research
Magann, J. 1983. Municipal financial disclosure: An empirical investigation.
for Business Decisions, Number 58. Ann Arbor: UMI Research Press.
McCormick,
R. and Tollison,
An Inquiry
Beverly
R. 1981. Politicians,
into the Interest-group
Legislation, and the Economy:
Theory of Government. Boston:Martinus
Nijhoff.
McDonald, R. May 1978. A simple comprehensive model for the analysis of covariante structures.
British Journal of Mathematical and Statistical Psychology
31(2):59-72.
Migue, J. and Belanger, G. Spring 1974. Toward
discretion. Public Choice 17(1):27-43.
Milbrath,
L. 196.5. Political Participation.
a general
theory
of managerial
Chicago: Rand McNally & Company
Miller, G. and Moe. T. June 1983. Bureaucrats, legislators, and the size of government. American Political Science Review 77(2):297-322.
Moody’s Municipal and Government ManuaI. 1978. New York: Moody’s Investor
Service.
Moody’s Municipal and Government Manual. 1986. New York: Moody’s Investor
Service.
Moorehouse, S. 1981. State Politics, Parties, and Policy. New York: Holt, Rinehart
& Winston.
Mueller, D. 1979. Public Choice. Cambridge,
UK: Cambridge
Mueller, D. 1989. Public Choice ZZ. Cambridge,
National Association
of State Auditors, Comptrollers,
State Comptrollers-Technical
Activities
University
UK: Cambridge
Press.
University
Press.
and Treasurers (NASACT) 1986.
Lexington, KY: NAS-
and Functions.
ACT.
Niskanen, W. 1971.
Aldine-Atherton
Bureaucracy
Nunnally, J. 1978. Psychometric
and Representative
Government.
Chicago:
Theory. New York: McGraw Hill.
Olson, M. 1965. The Logic of Collective Action.
Press.
Cambridge:
Harvard
University
Olsson, U., Drasgow, F., and Dornas, N. September
coefficient. Psychometrika 47(3):337-347.
1982. The polyserial
correlation
S. 1983. Legislators and legislatures in the american states. In Politics in
the American States, 4th edition. (V. Gray, H. Jacob, and K. Vines, eds.) Boston:
Patterson,
Little, Brown and Company,
Peltzman,
S. October
pp. 135-179.
1980. The growth
of government.
Journal
of Law and
Economics 23(2):209-280.
Petersen,
J. 1974. The Rating Game. Twentieth Century Fund.
Plotnick, R. and Winters, R. June 1985. A politico-economic
theory
redistribution.
American Political Science Review 79(2):458-473.
of income
State Government
Accounting
41
Disclosure
Pommerehne, W. and Schneider, F. 1978. Fiscal illusion, political institutions and local
public spending. KykIos 31(3):381-407.
Rabinowitz, A. 1969. Municipal
Wiley & Sons, Inc.
Bond Finance and Administration.
New York:
Ranney, A. 1976. Parties in state politics. In Politics in the American States
(H. Jacob and K. Vines, eds.) Boston: Little, Brown and Company, pp. 51-92.
Robbins, W. and Austin, K. Autumn 1986. Determinants
municipal financial reports: Some additional evidence.
of disclosure
quality in
Journal of Accounting
Research 24(2):412-421.
Ross, S. Spring 1977. The determination of financial structure: The incentive signalling
approach. BeN Journal of Economics 8(1):23-40.
behavior. In Public
Choice, Public Finance and Public Policy, Essays in Honour of Alan Peacock
Rowley, C. and Elgin, R. 1985. Towards a theory of bureaucratic
(D. Greenaway
Rubin,
and G. Shaw, eds.) Oxford,
M. 1987. A theory of demand
Research in Governmental
Chan, ed.) Greenwich,
UK: Basil Blackwell,
for municipal
and Nonprofit
Ltd., pp. 31-50.
audits and audit contracts.
Accounting,
In
Volume 3, Part A (J.
CT: JAI Press Inc., pp. 3-33.
J. 1971. The politics of the executive. In Politics in the American States,
2nd edition (H. Jacob and K. Vines, eds.) Boston: Little, Brown and Company, pp.
210-237.
Schlesinger,
Shepsle, K. 1986. The positive theory of legislative institutions: An enrichment
social choice and spatial models. Public Choice 50(2): 135- 183.
Spence,
A. August
1973. Job market signaling.
of
Quarterly Journal of Economics
87(3):355-379.
Spencer, B. 1982. Asymmetric
information and excessive budgets in government
bureaucracies: A principal and agent approach. Journal of Economic Behavior and
Organization 3(2-3):197-224.
Standard and Poors. July 26, 1982. Credit Comment.
Stigler,
G. Spring
1971. The theory of economic
Belt Journal of Eco-
regulation.
nomics and Management Science 2( 1):3-21.
Stigler,
G. January
1976. The sizes
of legislatures.
Journal
of Legal Studies
5(l): 17-34.
U.S. Bureau of Census 1978a. State Government
States Government Printing Office.
Finances. Washington
DC: United
U.S. Bureau of Census 1978b. Statistical Abstract of the United States. Washington,
DC: United States Government Printing Office.
U.S. Bureau of Census 1986a. State Government Finances. Washington,
States Government Printing Office.
DC: United
U.S. Bureau of Census 1986b. Statistical Abstract of the United States. Washington,
DC: United States Government Printing Office.
Wagner,
R. Spring
1976. Revenues
structure,
fiscal illusion,
and budgetary
choice.
Public Choice 25(1):45-61.
Wallace, W. Autumn
selected
financial
19(2):502-520.
1981. The association
reporting
practices.
between municipal
Journal
of
market measures
Accounting
and
Research
42
Rita Hartung Cheng
Wamsley, G. and Zald, M. 1973. The Political Economy
Bloomington, IN: Indiana University Press.
Watts, R. and Zimmerman,
NJ: Prentice Hall.
J. 1986. Positive Accounting
of Public Organizations.
Theory. Englewood
Cliffs,
Weber, M. 1947. Bureaucracy. In From Max Weber: Essays in Sociology (H. Gerth
and C. Wright Mills, eds.) New York: Oxford University Press, pp. 196-244.
Weingast, B. 1984. The congressional-bureaucratic
tive. Public Choice 44(1):147-191.
West, E. and Winer,
S. 1980. Optimal
system: A principal-agent
perspec-
fiscal illusion and the size of government.
Public Choice 33(5):607-622.
Wheaton, B., Muthen, B., Alwin, D., and Summers G. 1977. Assessing the reliability
and stability in panel models. In Sociological Methodology (D. Heise, ed.) San
Francisco: Jossey-Bass, pp. 84- 136.
Wilson, E. and Howard, T. 198.5. Information for municipal bond investment decisions: Synthesis of prior research, and extension and policy implications. In Research in Governmental and Nonprofit Accounting,
Volume I (J. Chan, ed.)
Greenwich, CT: JAI Press, Inc., pp. 213-263.
Zimmerman, J. 1977. The municipal accounting maze: An analysis of political incentives. Journal of Accounting Research 15(supplement): 107- 144.