Professional Documents
Culture Documents
www.elsevier.com/locate/jaccpubpol
Abstract
This study examines the effects of disclosure regulation on municipal managersÕ
incentives to disclose financial report information to the bond market. I compare dis-
closure levels of municipal governments in Michigan, which requires GAAP, with those
in Pennsylvania, which has unregulated disclosure. In the absence of disclosure regu-
lation I find that managers have bond market-induced incentives to disclose informa-
tion. Controlling for other incentives to disclose, the evidence implies that regulation
induces additional disclosures for low-debt governments, and is not binding for high-
debt governments.
Ó 2003 Elsevier Inc. All rights reserved.
1. Introduction
Although financial reporting and disclosure have been regulated for over 60
years, relatively few studies have evaluated the effects of these regulations. The
lack of research critically evaluating the effects of disclosure regulation is
surprising given the impact that this regulation has had on the accounting
*
Tel.: +1-541-346-3329; fax: +1-541-346-3341.
E-mail address: agore@oregon.uoregon.edu (A.K. Gore).
0278-4254/$ - see front matter Ó 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.jaccpubpol.2003.11.002
24 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
1
For summaries of the arguments for and against disclosure regulation, see Easterbrook and
Fischel (1991), Watts and Zimmerman (1986), and Leftwich (1980).
2
I define ‘‘regulated’’ disclosure as requiring GAAP regulations, and ‘‘unregulated’’ disclosure
as requiring neither GAAP nor other, state-mandated disclosure regulations.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 25
Second, I find that GAAP disclosure levels are significantly higher in the
regulated state than in the unregulated state. The difference in disclosure levels
is significantly smaller among those governments with higher debt, however,
which implies two things. First, this evidence suggests that regulation is not
binding for governments with high debt levels. 3 In other words, regulated
governments with high debt levels are required to disclose GAAP information
that they would have voluntarily disclosed in the absence of regulation. Sec-
ond, it implies that regulation induces the production of information, or is
binding, for governments with low debt levels, indicating that these govern-
ments may incur costs by being forced to disclose more information than the
market requires. These results are interesting given that prior governmental
research (e.g., Ingram and DeJong, 1987; Copley, 1991) does not find a sig-
nificant difference in disclosure level between states that require GAAP and
states with unregulated disclosure.
The evidence presented in this paper is relevant for evaluating the impli-
cations of the recent trend of increasing GAAP disclosure requirements (Barth
and Murphy, 1994). For example, the GASB recently enacted a new disclosure
standard for governments that expands the level of required disclosures
(GASB Statement No. 34). Opposition has been strong, due to the costs of
implementing and monitoring this standard, and due to questions about
whether there is demand for the information (Copley et al., 1997). My results
suggest there is an economic basis for governmentsÕ choice to voluntarily
provide financial information. Mandating more information imposes costs on
governments with lower bond market interaction, and ultimately the taxpay-
ers, while conferring benefits to public accountants by increasing monitoring
costs.
One caveat needs to be considered when interpreting my results, however.
My analysis focuses on disclosure incentives in the context of bond markets.
Therefore, I do not consider alternative perspectives addressed in prior studies,
such as disclosure incentives derived from political markets (Baber, 1990).
However, as discussed later in the paper, there is a growing body of evidence
that indicates capital market participants are the primary users of the financial
reports, while voters rarely use them. In addition, while I do not focus on
political incentives to disclose, I do use control variables to represent them in
the empirical analysis.
The remainder of the paper is organized as follows. In Section 2, I discuss
the governmental accounting environment, and in Section 3, the theory and
hypothesis development. Section 4 describes the research design, Section 5
discusses the results, while Section 6 concludes.
3
The term ‘‘binding’’ refers to a binding constraint, applying the standard definition found
in most microeconomic theory texts.
26 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
In this paper, the term ‘‘GAAP’’ refers to standards issued by the Gov-
ernmental Accounting Standards Board (GASB), the governmental equivalent
of the Financial Accounting Standards Board (FASB). The GASB prescribes
the minimum GAAP requirement for financial reports as a set of general
purpose financial statements, which are comprised of combined financial
statements and footnote disclosures. Some governments choose to prepare a
more comprehensive report known as the comprehensive annual financial re-
port (CAFR), however. In addition, I use the term ‘‘non-GAAP’’ disclosure to
refer to financial report disclosures that are not required under GAAP or state-
specific regulations. 4
Although the GASB issues GAAP standards, the level of disclosure required
of local governments is regulated by the states in which they are located. States
may require that local governmentsÕ financial report disclosures follow GAAP,
state-specific regulations, or may be unregulated. Approximately 27 states re-
quire that local governments follow GAAP, 14 have state-specific disclosure
regulations, and the remaining nine have unregulated disclosure (NASACT,
1996). 5
4
The traditional term for disclosure outside of the set of GAAP disclosures is ‘‘voluntary
disclosure’’ (see Core, 2001). However, because all disclosure is voluntary in Pennsylvania, it
renders the term ‘‘voluntary’’ confusing in my study.
5
The State of Michigan adopted GAAP in 1968. An examination of news articles from this time
period did not reveal the reasons why State officials decided to adopt GAAP. However, discussions
with State of Michigan Treasury Department officials indicate that GAAP was most likely adopted
because Michigan had recently re-convened the state constitutional convention, and at that time
enacted mandatory audit requirements. When new GAAP and auditing regulations came out in
1968, they were automatically adopted because it was in the state constitution.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 27
In the GASB focus group sessions, user groups representing voters and
regulators indicate that the two primary sources of financial information are
the budget; and conversations with individuals employed by the government.
The majority of these user groups do not use the annual financial reports.
Evidence in the literature also supports the proposition that bond market
participants commonly use the annual financial reports, while voters and
regulators do not. Empirical research, discussed in detail in Ingram and
Robbins (1987), shows that bond market representatives use governmental
financial reports to assess default risk. Other studies such as Zimmerman
(1977), Gaffney (1986), Jones et al. (1985), and Copley et al. (1997) suggest that
citizens rarely use governmental reports.
Although a growing body of evidence indicates that bond market partici-
pants use financial reports, the literature has had inconsistent results about the
relation between debt proxies and disclosure levels. Some studies such as
Robbins and Austin (1986) and Evans and Patton (1987) find positive rela-
tions, while others such as Evans and Patton (1983), Ingram and DeJong
(1987) and Copley (1991) find either an insignificant relation or report mixed
results. However, these studies confine their analyses to large municipalities,
do not distinguish between GAAP and non-GAAP disclosures, and most
do not consider differences in statesÕ regulatory environments.
6
For example, FSA, a municipal bond insurer, states on their website that revenue bonds
typically require more analysis than do general obligation bonds.
28 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
information because they are more likely to access the bond market. Therefore,
the finding of no difference in disclosure between large cities in GAAP-regu-
lated and unregulated disclosure states is not surprising.
In another study, Copley (1991) uses a dummy variable to control for
whether municipalities are in GAAP-regulated states, and finds this vari-
able is insignificant. However, CopleyÕs measure of disclosure includes both
GAAP and non-GAAP disclosures. The inclusion of non-GAAP disclosures
could cause the lack of significance because municipalities can voluntarily
disclose information regardless of whether the state requires GAAP. In addi-
tion, their analysis is again restricted to large cities, with a mean population
of 177,000.
In contrast to these studies, I employ a research design that restricts the
sample to two states in order to minimize the potentially confounding effects of
state-specific regulations, and also include smaller municipalities. In addition,
my disclosure measures distinguish between GAAP and non-GAAP disclo-
sures.
H1: Ceteris paribus, within a state that has unregulated disclosure, there is a
positive association between disclosure level and debt.
7
One may conjecture that investors merely need to examine bond ratings in order to assess risk.
In the case of municipal bonds, however, many issues are not rated by ratings agencies, and for
these unrated issues, public disclosure is especially important to help resolve information
asymmetry. Empirical research such as Fairchild and Koch (1998) finds the decrease in debt costs
from public disclosure is greater for unrated issues. Even when bonds are rated, however, the
ratings alone are not sufficient for investors to forecast risk. Bond ratings only represent a range for
the probability of default, not the probability of recovery, and the ratings agencies do not assess
the liquidity of the underlying assets.
8
The term ‘‘optimal’’ refers to optimal disclosure for the municipal government, rather than
financial statement users. As such, I implicitly assume that the level of state regulation is exogenous.
I acknowledge that there may be additional cost–benefit concerns involved with statesÕ decisions
to mandate GAAP, however, I consider them outside the scope of this study.
30 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
Fig. 1. Hypothesized relation between total disclosure and bond market interaction.
Note: This figure shows the relation between disclosure and debt, and does not consider other
incentives for disclosure. ‘‘GAAP’’ represents the level of disclosure required under GAAP regu-
lations. ‘‘Regulated GovernmentsÕ Disclosure’’ represents the relation between debt and disclosure
for governments required to comply with GAAP. Note that Regulated Governments in Region 1
(low debt) disclose at levels required by GAAP; while Regulated Governments in Region 2 (high
debt) voluntarily disclose above the levels required under GAAP. ‘‘Unregulated GovernmentsÕ
Disclosure’’ represents the relation between debt and disclosure for governments that do not have
to comply with disclosure regulations.
Still other governments have equilibrium levels of disclosure above the re-
quired level (see Region 2 of Fig. 1). In this case, the mandated disclosure level
is below that which is optimal for governments, and is not binding.
Assuming municipalities are faced with varying demands for and quantities
of debt capital, then the benefits of information production will vary across
governments. Absent regulation, governments with little bond market inter-
action lack incentives to disclose information. In a GAAP-regulated state,
however, governments with low bond market interaction are required to dis-
close more information than they would in the absence of regulation. There-
fore, consistent with Fig. 1, for the subset of governments with low debt, the
level of GAAP disclosures will be higher in the GAAP-regulated state than it is
in the unregulated disclosure state. In other words, regulation will be binding
for low-debt governments, which leads to the following hypothesis:
H2A: Ceteris paribus, for the subset of governments with low debt, the level
of GAAP disclosures is higher in a state that requires GAAP than it is
for governments in a state that has unregulated disclosure.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 31
H2B: Ceteris paribus, for the subset of governments with high debt, the level
of GAAP disclosures is the same in a state that requires GAAP as it is
for governments in a state that has unregulated disclosure.
The combination of H2A and H2B represents a joint test of the effects of
regulation and bond market interaction on information production. Theoret-
ically, there should be a large difference in GAAP disclosure between regulated
and unregulated disclosure states for low-debt governments, and a relatively
smaller, or no difference between the two states for the high-debt governments,
consistent with Fig. 1. In addition, note that I confine my analysis of H2 to
GAAP disclosure because a priori, it is unknown how GAAP regulation affects
non-GAAP disclosure.
9
An alternative research design would encompass the examination of governments in one state,
before and after the enactment of GAAP disclosure requirements. This is not feasible in part due to
the lack of data availability.
10
Discussions with State of Michigan regulators reveal that they routinely test GAAP
compliance by examining a sample of municipal financial reports in detail. If they find that a
financial report is not in compliance with GAAP, then the municipality must re-submit revised
reports within a specified time frame.
32 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
However, because there are only nine states that have unregulated disclosure, I
control for the differences that are most likely to affect the level of financial
report disclosure.
Michigan and Pennsylvania are similar in the following respects. Both states
have city and township governments that do not have overlapping boundaries,
to ensure that the financial reporting relation is clear. This criteria ensures that
municipalities only record transactions (such as debt) for that specific munic-
ipality, rather than for other entities as well (for example, school districts). 11
The states are also similar with regards to the types of bonded debt allowed,
both general obligation and revenue bonds, as well as the restrictions on the
amount of debt, in that both allow cities to issue debt directly. I use these
requirements because regulations affecting incentives to issue debt can indi-
rectly cause variation in disclosure. 12 Finally, both states require local gov-
ernments to be audited, however, Pennsylvania allows the choice of either an
external or an internal auditor. 13 I use this last requirement because mandated
audits potentially influence the level of GAAP disclosures.
Table 1, Panel A describes the selection criteria for the Michigan sample.
Similar to Zimmerman (1977), I select those governments with populations
greater than 10,000, for a total of 166 observations. Note that the use of this
restriction effectively biases against finding results because smaller governments
are less likely to have interaction with the capital markets (Gore et al., 2004).
From these, I collected financial reports for 88 municipalities from the Mich-
igan Department of Treasury for 1995. 14 I limit the analysis to one year be-
cause disclosure policies appear to remain relatively constant over time
(Botosan, 1997), and due to the availability of data.
Financial report data for the Pennsylvania municipalities are obtained in a
similar manner. I randomly selected 92 financial reports from those govern-
ments with populations greater than 10,000, roughly equivalent to the number
gathered from Michigan (see Table 1, Panel B). I collected a total of 87 reports,
11
Note that this does not refer to component units, but rather, organizations that are not
normally considered component units, such as school districts.
12
I hypothesize that there is a positive relation between debt and disclosure. If a state restricts
the amount of debt issued, then it is possible that I could find no relation between debt and
disclosure, which effectively biases against finding results.
13
Internal auditors are defined as auditors that agree to perform one municipal audit per year,
and may or may not be Certified Public Accountants. The audits performed by internal auditors are
in accordance with GAAP standards. However, I am unable to assess the impact the use of internal
auditors may have on audit quality.
14
Local governments are required to submit two copies of their financial reports to the
Michigan Department of Treasury on an annual basis. The Department of Treasury contributed
the ‘‘second copy’’ of the financial reports to this study for approximately one half of the cities and
townships with populations in excess of 10,000. This sample cannot be construed as random,
however, there is no known bias in the selection of individual financial reports.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 33
Table 1
Summary of financial report selection procedures
Panel A: Michigan sample
Total local governments in Michigan 1,800
Less governments with population <10,000 and 1;634
governments other than cities and townships
Total Michigan cities and townships with populations 166
>10,000
Financial reports randomly eliminated 78
Final sample, collected from Michigan Department 88
of Treasury
Panel B: Pennsylvania sample
Total local governments in Pennsylvania 2,498
Less governments with population <10,000 and 2;288
governments other than cities and townships
Total Pennsylvania cities and townships with 201
populations >10,000
Randomly selected 92
Financial reports collected directly from municipalities 65
Financial reports collected from Pennsylvania Depart- 22
ment of Community and Economic Development
Financial reports not available 05
Total financial reports 90
Final sample 87
Michigan Pennsylvania Total
Panel C: Total sample
Cities 44 21 65
Townships 44 66 110
Total 88 87 175
4.2.1.1. Disclosure indices. I measure disclosure with two disclosure indices, one
measuring the extent of GAAP disclosure, and the other measuring the extent
of non-GAAP disclosure. The GAAP index is comprised of GAAP disclosures
identified on a checklist provided by the Michigan Department of Treasury, the
entity that oversees local governments in Michigan. The checklist, which only
contains GAAP disclosures, is used to determine whether municipalitiesÕ
financial reports are in compliance with GAAP. When compiling the GAAP
index, I retain all disclosures on the checklist that apply to all municipalities. 15
Otherwise, the index could proxy for complexity, as larger and more complex
governments would tend to score higher. 16 The non-GAAP disclosure index
includes those disclosures identified through the prior literature and through
Standard & PoorÕs (1986) as useful to bond market representatives.
I compile the two indices from disclosures contained within the annual
financial reports. The GAAP disclosures are drawn from elements contained
within the basic financial statements, as well as from the footnotes. Non-GAAP
disclosures are from the basic financial statements, footnotes, and supplementary
statistical information. Each disclosure is counted as part of the disclosure index
if it is present, and for the GAAP index, appears to be substantially in compli-
15
For example, commitments and contingencies are required to be disclosed under GAAP, and
this disclosure is listed on the checklist. However, many municipalities do not have contingencies.
I therefore exclude this disclosure, since it is likely to proxy for complexity.
16
Prior studies vary as to disclosure index construction. Some studies such as Robbins and
Austin (1986) use indices comprised of disclosures found useful to the bond market. The use of such
an index would be biased against finding evidence consistent with my hypotheses if some of the
disclosures required by GAAP are not useful to creditors. Rather than using one of these indices, I
use the checklist because it is a broad measure of GAAP, is somewhat more objective than those
used in prior studies, and also because it references recent disclosures. For example, studies such as
Copley (1991), Ingram and DeJong (1987), and Ingram (1984) use disclosures available in the mid-
1980s, and so do not include recent GASB pronouncements such as the investment disclosure.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 35
4.2.1.2. Debt. The debt variable represents current and future incentives to re-
duce the costs of issuing debt. For current debt, the disclosure of information
reduces the enforcement and monitoring costs of existing contracts. It also re-
duces the costs of future forays into the debt market. The bond market incentives
should increase as the amount of debt, and therefore, the level of interaction with
the bond market, increases, and as such is expected to be positively associated
with disclosure. I use the total debt per capita as the proxy for the debt level. 18
Total debt is comprised of all debt, including general obligation bonds, revenue
bonds, and other debt such as bank notes payable. I use this ratio to be consistent
with prior studies such as Ingram and DeJong (1987) and Copley (1991).
17
The use of an equally-weighted index is supported by Robbins and Austin (1986).
18
In Michigan, debt issued on behalf of local governments at the county or state level, called
overlapping debt, is recognized on the local governmentsÕ financial statements. In Pennsylvania,
local governments are required to include overlapping debt in calculations of the borrowing base,
which makes it likely that they include it in their financial reports. It is also likely to be included
because their financial reports are audited. However, because disclosure is not regulated in
Pennsylvania, it is also possible that they do not include it. Therefore, the debt variable may
understate the bond market incentives in Pennsylvania, which effectively biases against finding
results for my first hypothesis.
19
An additional variable commonly used in disclosure studies is a proxy for funding sources, the
ratio of intergovernmental revenues to total revenues. In this study, part of the sample is drawn
from the state of Michigan, where most of the intergovernmental revenue is comprised of state-
shared revenue allocated based upon population. I therefore do not include this variable because it
does not adequately represent a governmentÕs reliance on non-capital market sources of revenue.
36 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
prior studies such as Robbins and Austin (1986), Evans and Patton (1983), and
Copley (1991) find it positively associated with disclosure.
Government type is included as a control because Gore et al. (2004) finds
that for a sample of Michigan governments, the township government type is
negatively associated with the level of GAAP disclosure compliance. 20 The
negative relation is potentially attributable to restrictions in debt issuance
imposed by the state of Michigan, or because townships in general have fewer
incentives to comply with GAAP. 21 Alternatively, this could be due to the
more fundamental issue of why a government chooses to operate as a city
versus a township. Because the effect of government type on financial report
disclosure is unclear, this variable is included as a control. A dummy variable is
used to indicate the city government type.
With the exception of the audit quality and population variables, all data are
obtained directly from the financial reports. Data for the audit quality vari-
ables are hand-collected from the Michigan Department of Treasury and the
Pennsylvania Department of Community and Economic Development. I verify
the accuracy of the lists by comparing auditors on the lists with those in a
random sample of financial reports. Population data are from the 1990 census.
20
It is also common for disclosure studies to include a dummy variable indicating the city
manager form of government. Prior studies show that the presence of the city manager form of
government is positively associated with citiesÕ disclosure level. Studies such as Zimmerman (1977)
commonly describe this variable in terms of reducing agency costs that exist between city councils
and appointed city managers. However, in this study, both townships and cities are included in the
sample, and a control variable is included to measure differences in disclosure due to government
type. Because this variable is highly correlated with the city manager dummy variable, the latter
is not included in the regression. When I include the city manager variable, however, the results
remain consistent with those presented.
21
The State of Michigan does not allow townships to directly issue debt for bond issues greater
than $1,000,000. Instead, the county in which the township resides issues the debt, which the
township in turn must repay to the C ounty. This may reduce townshipsÕ incentives to issue debt
relative to cities. However, anecdotal evidence reveals that townships respond by issuing several
small issues, each under $1,000,000, in order to circumvent this regulation.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 37
5. Results
Table 2 describes the total sample (Panel A), the subset of Pennsylvania and
Michigan governments (Panel B), and partitions based upon the level of debt
(Panel C). I assess the reasonableness of comparing Michigan and Pennsyl-
vania municipalities through non-parametric statistical tests in Table 2, Panel
B. These comparisons indicate there are no significant differences in population
or debt, however, differences between CITY and AUDITS are statistically
significant at the 0.01 level. Half of the Michigan sample is comprised of cities,
while only 24% of the Pennsylvania sample consists of cities. Also, Michigan
auditors perform a significantly higher number of governmental audits on
average than do Pennsylvania auditors (median 57 for Michigan versus median
4 for Pennsylvania). The difference in the number of audits performed is
probably due to PennsylvaniaÕs use of internal auditors, who are restricted by
law to one municipal audit per year. I do not consider this a problem that will
impact the hypothesis tests because the two variables are control variables in
the multivariate specifications.
Panel B of Table 2 reveals that Michigan municipalitiesÕ median score for
GAAP is 17 out of a possible 18, indicating that there is not full compliance
with GAAP. This effectively biases against finding results for H2 , however. In
addition, the non-parametric tests in Panel B show a significant difference
38 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
Table 2
Descriptive statistics and tests of significance from non-parametric tests of differences between
Michigan and Pennsylvania
Variable Mean Median Std. dev.
Panel A: Total sample (n ¼ 175)
POP 3.065 2.90 0.701
CITY 0.371 0.00 0.485
AUDITS 30.503 9.00 38.700
DEBT 578.979 393.68 757.501
STATE 0.503 1.00 0.501
GAAP 14.006 15.00 4.390
NONGAAP 2.935 1.00 3.888
Pennsylvania (n ¼ 87) Michigan (n ¼ 88) Mann–
Whitney
Mean Median Std. dev. Mean Median Std. dev.
p-values*
Panel B: Subset of Pennsylvania and Michigan governments
POP 3.095 2.90 0.772 3.035 2.895 0.627 0.542
CITY 0.241 0.00 0.430 0.500 0.000 0.503 0.003
AUDITS 7.149 4.00 6.681 53.591 57.000 43.205 0.000
DEBT 584.337 324.54 715.237 573.683 408.158 801.170 0.833
GAAP 11.552 13.00 4.810 16.432 17.000 1.923 0.000
NONGAAP 2.756 2.00 3.226 3.102 1.000 4.428 0.238
Low debt (n ¼ 58)a High debt (n ¼ 58)b
Mean Median Std. dev. Mean Median Std. dev.
Panel C: Subset of governments partitioned by debt
POP 2.858 2.73 0.450 3.258 2.924 0.909
CITY 0.259 0.00 0.442 0.517 1.000 0.504
AUDITS 25.086 8.50 35.968 33.707 9.000 41.267
DEBT 108.469 118.48 66.159 1252.437 956.247 1003.721
STATE 0.517 1.00 0.504 0.500 0.500 0.504
GAAP 12.328 14.00 4.673 15.793 17.000 3.259
NONGAAP 1.618 1.00 2.423 4.193 2.000 4.756
*Indicates the two-sided statistical significance levels for a test of differences between Pennsylvania
and Michigan.
This table presents descriptive statistics for the total sample (Panel A), and the subset of Michigan
and Pennsylvania governments (Panel B). Panel C shows the total sample partitioned into three
groups based on the total debt per capita, with the top and bottom third of the total sample
designated as high debt and low debt. Variable descriptions are as follows: POPit is the log of
population in year t for government i; CITYit is a dummy variable where 1 indicates the city form
of government and 0 indicates the township form of government; AUDITSit is the number of
municipal audits performed in year t by government iÕs auditor; DEBTit is the total debt per capita
in year t for government i; STATE is a dummy variable where 1 indicates Michigan (GAAP
disclosure state) and 0 indicates Pennsylvania (unregulated disclosure state); GAAPit is an index of
GAAP disclosures in year t for government i; and NONGAAPit is an index of voluntary disclosures
in year t for government i.
a
n ¼ 30 Michigan, 28 Pennsylvania.
b
n ¼ 29 Michigan, 29 Pennsylvania.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 39
between the two states for GAAP disclosure, but no significant difference in
non-GAAP disclosure. The latter indicates that it is appropriate to compare
the two states, because in the absence of disclosure regulation, municipalities
voluntarily disclose information at similar levels.
Table 3 reports correlation matrices for the total sample (Panel A), the
subset of Pennsylvania governments (Panel B), and for partitions based on the
level of debt (Panels C and D). For the total sample and some of the partitions,
the correlation between STATE and variables such as AUDITS is high (0.54). I
therefore test for multicollinearity by computing condition indices following
the procedure outlined in Belsley et al. (1980). In each case the condition index
is below 10, which indicates that multicollinearity is not severe.
Table 3
Correlation matrices
POP CITY AUDITS DEBT STATE
Panel A: Total sample (n ¼ 175)
POP 1.000
CITY 0.238* 1.000
AUDITS )0.009 0.154* 1.000
DEBT 0.212** 0.290** 0.057 1.000
STATE )0.046 0.270** 0.540** 0.016 1.000
Panel B: Subset of Pennsylvania governments (n ¼ 87)
POP 1.000
CITY 0.242* 1.000
AUDITS )0.037 )0.158 1.000
DEBT 0.347* 0.287** )0.090 1.000
22
The Wald test is used because it does not rely on the assumption of constant variances across
samples, as does the Chow test. Preliminary analysis of the data indicated that the variances of the
cities could be different from the variances of the townships. The test statistic used is found in
Greene (1993), and is as follows:
where h1 , and h2 are two normally distributed estimators of a parameter based upon independent
samples (cities and townships), with variances V1 and V2 . The variance estimates are obtained from
separate OLS regressions of the cities and townships. The Wald statistic has a chi-squared distri-
bution with K degrees of freedom, and tests whether the differences between the parameters are
significantly close to zero.
23
Interpretations of my results need to allow for the possibility that fundamental differences
between the states could lead to differences in disclosure regulation. For example, there could be
differences in the underlying incentives to borrow that lead to why Michigan requires GAAP, while
Pennsylvania does not. However, analyses in Poterba and Rueben (1997) show that the fiscal
institutions (e.g. the level of debt restrictions, balanced budget and spending limits) of the two
states are very similar. I consider any remaining differences outside the scope of this study.
A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52 41
Table 4
OLS results for a test of the relation between disclosure level and debt in Pennsylvania, the
unregulated disclosure state (H1 )
Model: DISCLOSUREit ¼ ai þ b1i POPit þ b2i CITYit þ b3i AUDITSit þ b4i DEBTit þ eit
Variable Expected Dependent variable:
sign
GAAPa (n ¼ 87) Non-GAAPa Totala (n ¼ 87)
(n ¼ 87)
Coeffi- t-statistic Coeffi- t-statistic Coeffi- t-statistic
cient cient cient
Intercept 3.194 1.46 )0.850 )0.55 1.832 0.60
POP + 1.841 2.37** 0.712 1.31 2.740 2.53**
CITY + 1.011 0.90 )1.199 )1.55 )0.145 0.09
AUDITS + 0.198 3.00** 0.060 1.25 0.247 2.68**
DEBT + 0.002 2.01* 0.000 3.53** 0.004 3.11**
Adjusted 0.30 0.27 0.36
R2
F -statistic 10.20 8.45 13.2
(p-value) (<0.0001) (<0.0001) (<0.0001)
*Significant at p < 0:05; based on two-tailed tests.
**Significant at p < 0:01; based on two-tailed tests.
Note: Variable descriptions are as follows: POPit is the log of population in year t for government i;
CITYit is a dummy variable where 1 indicates the city form of government and 0 indicates the
township form of government; AUDITSit is the number of municipal audits performed in year t by
government iÕs auditor; DEBTit is the total debt per capita in year t for government i; GAAPit is an
index of GAAP disclosures in year t for government i, described in Appendix A; and Non-GAAPit
is an index of voluntary disclosures in year t for government i, described in Appendix B; and Total
is the sum of the GAAP and Non-GAAP disclosure indices.
a
WhiteÕs t-statistics. All regression results have been checked for the presence of influential data
points using CookÕs D statistic; no influential data points were detected. All models have also been
estimated using multinomial logit with the results generally consistent with those presented here.
Table 6 reports results for my second hypotheses (H2A and H2B ). Results for
H2A show that the difference in disclosure between low-debt governments in
Michigan and Pennsylvania is significant (p < 0:01), represented by the vari-
able LOWDEBT,MI. Consistent with H2A and Fig. 1, for the subset of low-
debt governments, the level of GAAP disclosures is significantly higher in the
regulated state. This result is consistent with regulation inducing the disclosure
of information for governments with little bond market interaction.
H2B tests whether the difference in disclosure between high-debt govern-
ments in Michigan and Pennsylvania are not significantly different from zero,
or ðb4i b5i Þ is not significant. I find that high-debt municipalities in Michigan
disclose significantly more information, inconsistent with H2B . However, this
result appears to be sensitive to my selection of the debt proxy, because when I
use alternate debt proxies (see sensitivity analysis below), I consistently find
no significant difference in disclosure among the high-debt governments.
42 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
Table 5
OLS results for a test of the relation between the GAAP disclosure level and the state
GAAPit ¼ ai þ b1i POPit þ b2i CITYit þ b3i AUDITSit þ b4i DEBTit þ b5i DEBTit STATEit
þ b6i STATEit þ eit
Variable Expected sign Total sample (n ¼ 175)
Coefficient t-statistica
Intercept 6.714 5.63**
POP + 1.072 2.72**
CITY + 1.018 3.06**
AUDITS + 0.014 2.42*
DEBT + 0.002 2.89**
DEBT STATE ) )0.002 )2.21*
STATE + 4.957 7.38**
Adjusted R2 0.47
*Significant at p < 0:05; based on two-tailed tests, two-tailed otherwise.
**Significant at p < 0:01; based on two-tailed tests, two-tailed otherwise.
Note: Variable descriptions are as follows: POPit is the log of population in year t for government i;
CITYit is a dummy variable where 1 indicates the city form of government and 0 indicates the
township form of government; AUDITSit is the number of municipal audits performed in year t by
government iÕs auditor; DEBTit is the total debt per capita in year t for government i; STATE is a
dummy variable where 1 indicates Michigan (GAAP disclosure state) and 0 indicates Pennsylvania
(unregulated disclosure state); and GAAPit is an index of GAAP disclosures in year t for gov-
ernment i, described in Appendix A.
a
WhiteÕs t-statistics. All regression results have been checked for the presence of influential
data points using CookÕs D statistic; no influential data points were detected. All models have also
been estimated using multinomial logit with the results generally consistent with those presented
here.
I also test whether the difference in disclosure between the low-debt gov-
ernments is significantly greater than the difference in disclosure between high-
debt governments, or b8i ðb4i b5i Þ > 0, consistent with Fig. 1. The results of
this test, reported in Table 6, show that the difference is significant (p ¼ 0:003).
The results of this test are consistent with regulation being binding for low-debt
governments, and not binding for high-debt governments.
Table 6
OLS results for tests of the relation between regulation and disclosure (H2 )a
GAAPit ¼ ai þ b1i POPit þ b2i CITYit þ b3i AUDITSit þ b4i HIGHDEBT; MIit
þ b5i HIGHDEBT; PAit þ b6i MEDDEBT; MIit þ b7i MEDDEBT; PAit
þ b8i LOWDEBT; MIit þ eit
Variable Expected Total Sample (n ¼ 175)
sign
Coefficient t-statisticb
Intercept 4.764 4.11**
POP + 1.357 4.53**
CITY + 1.007 2.40*
AUDITS + 0.013 2.76**
HIGHDEBT,MI + 6.909 7.74**
HIGHDEBT,PA + 4.406 4.22**
MEDDEBT,MI + 6.184 6.78**
MEDDEBT,PA + 2.278 2.09*
LOWDEBT,MI + 6.003 7.22**
Adjusted R2 0.51
Test of equality of coefficients
F -statistic for difference in coefficients:
HIGHDEBT,MI ) HIGHDEBT,PAc 7.04
(p-value) (0.0088)
LOWDEBT,MI ) (HIGHDEBT,MI ) 9.00
HIGHDEBT,PA)
(p-value) (0.0031)
*Significant at p < 0:05; based on two-tailed tests, two-tailed otherwise.
**Significant at p < 0:01; based on two-tailed tests, two-tailed otherwise.
Note: See Table 2 for variable definitions.
a
The total sample is ranked by the total debt per capita, with the bottom third designated as low
debt (LOWDEBT) governments, the top third designated as high debt (HIGHDEBT) govern-
ments, and the middle third as medium debt (MEDDEBT) governments. MI designates Michigan
governments, and PA denotes Pennsylvania governments.
b
WhiteÕs t-statistics. All regression results have been checked for the presence of influential data
points using CookÕs D statistic; no influential data points were detected. The model has also been
estimated using multinomial logit with the results generally consistent with those presented here.
c
The p-value of this test is insignificant (p ¼ 0:19) when multinomial logit is used.
interaction. Therefore, I also use the log of total debt, the ratio of total debt/
total revenue, and the number of new bond issues in the last five years, to
measure debt. The latter variable represents the frequency of recent forays
into the bond market. The results from these analyses are consistent with my
tests of H1 . However, for my test of H2B , I now find that for the alternate
measures of debt, there is no significant difference in disclosure level among
high-debt governments, or ðb4i b5i Þ is not significant, consistent with H2B .
Because my original specification with debt per capita as the debt proxy
includes population as a deflator and again as a separate control variable, it
is possible that this caused an econometric problem. This is plausible because
44 A.K. Gore / Journal of Accounting and Public Policy 23 (2004) 23–52
when total revenue is used as an alternate size deflator (in the ratio of total
debt/total revenue), I find no significant difference in disclosure between high-
debt municipalities in Pennsylvania and Michigan. In addition, when I use
multinomial logit for the specification that includes debt per capita, I again
find that there is no significant difference in disclosure between high-debt
governments.
Third, I assess the sensitivity of the results to the disclosure index measure
by using two alternative disclosure indices, the first based on Copley (1991) and
the second on Ingram and DeJong (1987). I chose these two indices because
both purport to represent measures of GAAP disclosure. A comparison of the
content of these alternate indices with the current studyÕs is provided in
Appendices A and B. Although I separate the GAAP and non-GAAP com-
ponents into two separate indices in the appendices to facilitate comparison,
for the purpose of this sensitivity test, I combine the indices in the same manner
as that used by the two aforementioned studies. The results of this specification
do not differ significantly from those reported, and as such, my results are not
likely due to index design.
Fourth, I assess the effect of the point of sample partition in the test of H2B ,
which uses the upper and lower third of the total sample to represent high and
low debt, by instead using the upper and lower quarter of the total sample. The
results from this analysis are again consistent with those presented. In partic-
ular, when I use this alternative partition point, the coefficient estimates for the
low-debt subset are slightly larger, and for the high-debt subset slightly smaller,
than the partition that divides the sample into thirdÕs.
6. Conclusion
debt levels. In other words, regulated governments with high debt levels are
required to disclose GAAP information that they would have voluntarily
disclosed in the absence of regulation. Second, it implies that regulation in-
duces the production of information, or is binding, for governments with low
debt levels. The latter result suggests that mandating GAAP information im-
poses costs on governments with lower bond market interaction, and ulti-
mately the taxpayers, while conferring benefits to public accountants through
monitoring costs.
My study examines disclosure levels for the governmental sector for one
year. This focus enables a more precise analysis of some of the consequences of
accounting regulation. However, it also means the results may not be gen-
eralized to the corporate sector and/or other time periods. This issue could be
addressed in future studies by applying a similar methodology to research areas
where corporate disclosure is differentially regulated, for example, in interna-
tional research.
Another limitation is that because some of the tests examine whether the
difference in disclosure between the two states is not significantly different from
zero, it is possible that the insignificance is due to model misspecification. A
final limitation of this study is that it necessarily restricts the examination of
disclosure to a bond market perspective. However, as discussed previously,
both focus groups commissioned by the GASB as well as prior academic re-
search indicate that bond market participants are the primary users of muni-
cipal financial reports. An additional benefit of the bond market focus is that it
allows analysis of non-political incentives to disclose, an area of the govern-
mental literature not extensively explored.
Acknowledgements
46
This table presents the GAAP disclosure index used in regressions. The index is compiled directly from disclosures
contained within the governmentsÕ annual financial reports, and is drawn from both the basic financial statements and
the footnotes. The GAAP index contains 18 disclosure elements that are required by GAAP. Thus, governments can
47
fixed assets
48
Appendix A (continued)
This table presents the non-GAAP disclosure index, comprised of seven disclosure elements that are not required by
GAAP. Thus, governments can attain a non-GAAP index score of between 0 and 7. The non-GAAP disclosure elements
are drawn from the basic financial statements, footnotes, and supplementary statistical information. The frequency of
49
50
Appendix B (continued)
References
Ingram, R., Robbins, W., 1987. Financial reporting practices of local governments. Governmental
Accounting Standards Board, Stamford, CT.
Jones, D., Scott, R., Kimbro, I., Ingram, R., 1985. The needs of users of governmental financial
reports. Governmental Accounting Standards Board, Stamford, CT.
Leftwich, R., 1980. Market failure fallacies and accounting information. Journal of Accounting
and Economics 2 (3), 3–30.
Myers, S., Majluf, N., 1984. Corporate financing and investment decisions when firms have
information that investors do not have. Journal of Financial Economics 13, 187–222.
National Association of State Auditors, Comptrollers, and Treasurers (NASACT), 1996. State
comptrollers: Technical activities and functions. NASACT, Lexington, KY.
OÕKeefe, T., King, R., Gaver, K., 1994. Audit fees, industry specialization, and compliance with
GAAS reporting standards. Auditing: A Journal of Practice and Theory 13 (2), 41–54.
Poterba, J., Rueben, K., 1997. State fiscal institutions and the U.S. municipal bond market.
Working paper, National Bureau of Economic Research, Cambridge, MA.
Robbins, W., Austin, K., 1986. Disclosure quality in governmental financial reports: An assessment
of the appropriateness of a compound measure. Journal of Accounting Research 24 (2), 412–
421.
Sengupta, P., 1998. Corporate disclosure quality and the cost of debt. The Accounting Review 73
(4), 459–474.
Standard & PoorÕs, 1986. Debt Ratings Criteria––Municipal Overview. Standard & PoorÕs, New
York, NY.
Watts, R., Zimmerman, J., 1986. Positive Accounting Theory. Prentice Hall, Englewood Cliffs, NJ.
Zimmerman, J., 1977. The municipal accounting maze: An analysis of political incentives. Journal
of Accounting Research 15 (Suppl.), 107–144.