Our study provides evidence of the association between accrual-based financial information provided by the GASB Statement No. 34 reporting model and new issue municipal bond borrowing costs. Given the controversy that surrounded the implementation of the GASB Statement No. 34 reporting model, questions that continue to be raised about the usefulness of the model, and recent calls for research on Statement No. 34 by the GASB, our study into the informativeness of the GASB Statement No. 34 model is timely and helpful in informing the GASB decision process. The findings of our study indicate that the Statement No. 34 model provides incremental information to the new issue market beyond that provided by the pre-Statement No. 34 model. Our results indicate that the Statement No. 34 model may allow creditors to better assess issuers' longer-term performance.

In 1999, the Governmental Accounting Standards Board (GASB 1999) issued Statement No. 34, Basic Financial Statements—and Management's Discussion and Analysis—for State and Local Governments. For the first time, state and local governments are required to provide entity-wide governmental financial accounting information on an accrual basis. The standard was heralded as the most comprehensive and significant change to occur to local government financial reporting. As such, implementation was phased in over a number of years to allow local governments to address the numerous reporting changes wrought by the new standard. At the time of implementation, and subsequently, the value of the standard has been questioned, with little empirical research to support whether the new reporting model achieves the objective of providing more useful information to users than was provided previously.

Our study informs the discussion by examining whether the new government-wide accrual and modified accrual information added by GASB Statement No. 34 is incrementally useful for assessing municipal bond default risk. Prior to the implementation of Statement No. 34, the primary operations of government were reported using only a modified accrual basis of reporting. This basis of accounting focuses on current sources and uses of funds and, as such, allows users to assess short-term performance and fiscal compliance. However, under the prior model it was not possible for users to assess the longer-term effects of government's decisions on financial position and condition (GASB 1999, para. 219). To address this shortcoming, along with what were perceived to be other shortcomings, Statement No. 34 introduced accrual accounting-based financial reporting in addition to the modified accrual basis of accounting. One of the results of reporting on the accrual basis has been the recognition of long-term operating liabilities that, under the modified accrual basis, were recognized using the “pay as you go” method. Recognizing longer-term operating liabilities, such as pensions and retiree health care, has significantly impacted the financial statements; however, it is unclear how reporting these items has affected user decisions.

One of the other extensive changes resulting from Statement No. 34, for which there is little empirical data, relates to the usefulness of the requirement that aggregated modified accrual financial information be provided. In the past, information was presented for individual components (funds) of government, but there was no “consolidation” to provide a comprehensive entity-wide perspective. Under Statement No. 34, both the modified accrual information provided by the individual governmental funds and the accrual information are aggregated on a government-wide basis.

Although Statement No. 34 provides considerable additional information that potentially benefits users, it also adds to reporting complexity. Concerns were expressed about the cost of implementing a system that requires two bases of accounting, which some believed had no demonstrated value (Foltin 2008). Additionally, concerns were expressed about whether the presentation of statements using two bases of accounting would confuse users. Finally, some expressed the sentiment that the statement represented an attempt to “be all things to all users” (GASB 1999). It was suggested that the GASB narrow the range of user needs and concentrate on just one basis of accounting for presenting financial information (GASB 1999).

Research on the GASB Statement No. 34 financial reports is of considerable interest to the GASB, as evidenced by its latest call for research (GASB 2013b) on the usefulness of Statement No. 34. Research on the usefulness of Statement No. 34 will assist the GASB with its recently scheduled examination of whether Statement No. 34 is meeting the GASB's reporting objectives (GASB 2013c). Another project currently on the GASB agenda impacting Statement No. 34 is the development of a concept statement related to recognition under the modified accrual basis of accounting (GASB 2013a). There have been calls for the GASB to delay the concept statement project until an examination of the impact and usefulness of the models required under Statement No. 34 is complete (GASB 2013a).

Our study also adds to the discussion on the usefulness of government financial reports to a primary user group of government financial information, the municipal bond market. Disclosed inequities in the rating agencies' application of bond rating criteria to corporate and municipal bonds (Bullock 2010; Seymour 2010), coupled with the financial instability of bond insurers (Lauricella and Richardson 2007; Neumann 2010), increases the importance of financial information for assessing default risk. Additionally, several provisions of the Dodd-Frank Wall Street Reform and Consumer Protection Act (U.S. House of Representatives 2010), signed into law on July 21, 2010, are intended to decrease the importance of credit ratings for assessing default risk. As such, the Act could elevate the usefulness of financial reports in assessing a municipal bond's default risk.

One way to assess the usefulness of the Statement No. 34 reporting model is to determine whether information added by the model is correlated with the cost of debt. Given that the presence of accrual information for use in assessing default risk on general obligation debt is relatively new, there is not a large body of research that has examined the impact of the statement. The two published studies (Plummer, Hutchison, and Patton 2007; Reck, Wilson, and Schiffel 2009) that have looked at the informativeness of accrual information in a government context provide mixed results. Thus, whether government's cost of debt has been affected by the addition of Statement No. 34 information remains unresolved. Adding to the research on the impact of Statement No. 34 on this large user group is important as the GASB considers whether Statement No. 34 is meeting its objectives.

We examine two aspects of the Statement No. 34 reporting model that the GASB asserts increase the usefulness of government financial reports to users. The first is the addition of accrual information to help assess the longer-term impact of management decisions on performance. Second is whether aggregated information increases the usefulness of financial reports.

The results of our study indicate that, as hypothesized, the government-wide accrual financial information provided by GASB Statement No. 34 is significantly associated with bond interest costs, above and beyond the contribution of pre-Statement No. 34 general fund financial information. However, when the government-wide accrual information is compared to a model that contains aggregated modified accrual information, the government-wide accrual model does not appear to provide additional explanatory power. Further, the aggregate modified accrual information required by GASB Statement No. 34 does not appear to add significantly to the explanatory power of a model with only general fund information.

Our findings for aggregated information may not be surprising, as governmental funds, other than the general fund (i.e., capital projects and debt service funds), mainly report capital and debt-related financial flows rather than operating flows. The informativeness of financial statement information about these flows may be diminished by the extensive information about capital flows and debt that is reported in the notes to the financial statements and the body of bond offering statements. Moreover, GASB standards commonly require the reporting of financing inflows in earlier periods than the related capital and debt outflows. This may significantly distort single-year aggregated modified accrual fund balance and operating measures, such as those calculated for this study. As a result, the inherent noise associated with these measures may limit the power of tests that examine aggregated governmental fund financial information.

The paper proceeds by presenting the hypotheses in the next section, followed by the research design. The final two sections of the paper are the results, and discussion and summary.

Government Reporting Model

Government financial statements provide users with information about government activities that are supported primarily by nonexchange revenues, such as taxes and grants, and other sources generated by governmental activities. They also provide information about business-type activities, supported primarily by user charges, and fiduciary activities. Our study focuses on the revenues and expenditures or expenses related to governmental activities since it is these activities, rather than business-type or fiduciary activities, that are of primary interest to general obligation municipal bond investors.

Before Statement No. 34, government financial statements reported these governmental activities on a modified accrual basis of accounting and separated the activities by fund type (referred to as governmental funds). Modified accrual basis accounting provides information useful for assessing short-term performance by focusing on sources and uses of funds. Additionally, information provided on the modified accrual basis was presented in separate fund types: the primary operating fund, known as the general fund; and other self-balancing sets of accounts earmarked for specific operating purposes, interest and debt principal payments, and capital projects. Since the fund types were presented in separate columns of the financial statements, users were prevented from obtaining an overall or government-wide perspective on the government's financial performance and position. The inability to obtain an overall perspective was due to activity among funds, which made it difficult, if not impossible, for the user to aggregate the fund information provided in the columns to meaningfully assess government-wide performance.

The Statement No. 34 model retains the fund-type modified accrual information provided under the previous reporting model. However, unlike the previous model, the Statement No. 34 model also aggregates the fund types providing a government-wide modified accrual perspective. Because the government-wide perspective under modified accrual is obtained through aggregation of the individual fund type information, we refer to this government-wide perspective as aggregate modified accrual throughout the paper. In addition, the Statement No. 34 model expands available information, providing creditors with accrual information on governmental activities on an aggregate government-wide level, but not at the fund level. In the paper we refer to the presentation of aggregate governmental activities information on an accrual basis as government-wide accrual. To assist users with understanding these differences in reporting, Table 1 is provided.

TABLE 1

Differences between Pre-GASB Statement No. 34 and GASB Statement No. 34 Reporting Models as Used in the Paper

Differences between Pre-GASB Statement No. 34 and GASB Statement No. 34 Reporting Models as Used in the Paper
Differences between Pre-GASB Statement No. 34 and GASB Statement No. 34 Reporting Models as Used in the Paper

Usefulness of Accrual Financial Accounting Information

The importance of accrual accounting information has been studied in the corporate and municipal debt markets. Unlike the equity markets, the debt markets are not concerned with the residual returns to shareholders. Rather, the debt markets are concerned with the entity's ability to generate sufficient cash to meet its debt obligations (default risk). As with the equity markets, debt markets use financial accounting information to assess firm performance. Through the use of analytic models, Fischer and Verrecchia (1997) show the relationship between firm performance and debt markets, indicating that debt yields (prices) decrease (increase) with increasing earnings. Researchers (Das, Hanouna, and Sarin 2009; Khurana and Raman 2003) find support for the Fischer and Verrecchia model. Khurana and Raman (2003) find a significant negative association between accounting information and yields on new debt issues. Additionally, Das et al. (2009) find that accounting-based models explain about two-thirds of the spread for credit default swaps. The usefulness of accounting information to the bond market is also supported by the practitioner literature (Howe 2001; Strumeyer 2005), which indicates that financial accounting information is useful for assessing the financial health of an entity and its ability to meet future obligations. Both practitioners (e.g., Van Horne 1990) and researchers (e.g., Ederington, Yawitz, and Roberts 1987; Khurana and Raman 2003) indicate that financial accounting information is important incrementally, since bond rating information is not fully informative of the financial performance or default risk of entities. The finding that bond ratings are not fully informative to the corporate bond market is reinforced by research in the municipal bond market (Liu and Seyyed 1991; Ingram and Wilson 1999; Reck, Wilson, Gotlob, and Lawrence 2004).

The accrual information provided to corporate markets allows creditors to not only assess the cash or short-term position of the borrower, but also the longer-term effects of management decisions on financial performance. Under the Statement No. 34 reporting model, modified accrual accounting information continues to allow creditors to assess short-term financial performance. In addition, the model provides, for the first time, accrual-based information to help municipal creditors assess the longer-term effects of management decisions on government financial performance.

Although very little research exists on the impact of accrual information in assessing municipal default risk, prior research (see reviews by Ingram, Raman, and Wilson [1987] and Reck et al. [2004]) finds that information provided on the modified accrual basis is useful to municipal creditors for assessing default risk. Of interest, then, is whether the addition of accrual financial information concerning the longer-term effects of management decisions is associated with default risk assessment. The results of the cited research in the corporate markets, and one recent study (Reck et al. 2009) in the municipal markets, suggest that accrual information may be incrementally relevant.

Using pre-Statement No. 34 data, Reck et al. (2009) find that crude proxies of accrual operating performance, constructed from information provided in the modified accrual statements, are associated with new issue bond costs. However, such crude proxies measure accruals imperfectly. Moreover, capital investment and borrowing-decision policies cannot be determined solely from modified accrual statements.

Plummer et al. (2007) provide preliminary evidence on the relevance of Statement No. 34 information. More specifically, their study finds that some accrual measures appear to be useful for assessing municipal bond default risk. In their study, measures constructed from the balance sheet are significantly associated with bond ratings. However, none of the accrual operating measures is significantly associated with bond ratings. Given that default risk is a measure of the entity's ability to generate sufficient cash to pay its debt, and the findings in the corporate bond market that operating measures are associated with bond costs, the lack of significance for operating measures found by Plummer et al. (2007) is unexpected.

Initially we focus on the general fund, a modified accrual basis fund, in assessing default risk. The reason for this focus is that the general fund is the primary operating fund of government, and depends heavily on nonexchange tax revenues for its operation. Due to its size and the source of its resources, the operating performance of the general fund has been found by previous researchers (Ingram et al. 1987; Wilson 1990; Reck et al. 2004) to be useful for assessing the default risk on general obligation bonds. Consequently, we test whether the additional information required by GASB Statement No. 34 adds explanatory power in assessing default risk beyond a model that contains only general fund financial measures.

Our hypothesis is based on research conducted in the corporate markets, pre-Statement No. 34 and preliminary Statement No. 34 markets. However, unlike the corporate markets, two bases of accounting are provided to creditors in municipal markets. As a result, while short- and longer-term performance is evaluated using only accrual information in the corporate markets, short-term performance and budgetary and fiscal constraints can be analyzed in the municipal markets using modified accrual information, while the analysis of longer-term performance requires the use of accrual information. Based on the pre-Statement No. 34 research findings, we expect information provided on the modified accrual basis to be associated with measures of default risk; that is, stronger financial information is associated with lower borrowing costs.

Additionally, and in accordance with the preponderance of corporate literature and the limited Statement No. 34 research, we expect the government-wide accrual operating and financial position information to be significantly correlated with bond interest costs. However, we are reluctant to predict a directional effect of such information on borrowing cost, since longer-term financial information can convey either a positive or negative signal, independent of the signal conveyed by short-term general fund information. For example, as discussed more fully in the “Results” section, long-term operating information is more likely to be driven by expense accruals for pensions, compensated absences, and court judgments that are reported on the government-wide accrual operating statement, but not the governmental fund modified accrual statement.

H1: 

In the presence of general fund modified accrual basis financial information, government-wide accrual accounting information required by Statement No. 34 provides additional information that is associated with new issue bond interest costs.

While we expect that accrual-based financial position and performance will be associated with bond interest costs, a lack of support for H1 could be observed because of the unique operating cycle of governments. For municipal governments, which basically operate from one budget cycle to the next, creditors could be focused mainly on the short-term effects of management decisions. Given the year-to-year nature of most government budgets, creditors may feel that the impact of management decisions affecting longer-term cash flow performance can be undone or altered with the next budget. If so, the value of accrual accounting information is diminished.

Usefulness of Aggregated Financial Information

Prior to Statement No. 34, only disaggregated modified accrual fund information was presented in the state and local government financial reports. As indicated earlier, under the pre-Statement No. 34 model, information was provided by fund type; however, information was typically not provided on government-wide performance. The GASB believed that the disaggregated nature of the pre-Statement No. 34 reporting model made it difficult for users to obtain a clear picture of a government's overall performance (GASB 1999, para. 248). Additionally, governments had considerable flexibility identifying and defining the funds used, making it difficult to compare information cross-sectionally or longitudinally. As a result, the Statement No. 34 model requires the aggregation of modified accrual information, providing a government-wide modified accrual perspective on performance. In support of the GASB's assertion that aggregate information is useful, preliminary work by Reck et al. (2009) finds that aggregate financial information is negatively associated with bond interest costs. However, a survey of municipal bond analysts (Wilson 1990) indicates that while aggregate information might be helpful in assessing government performance, the disaggregated information provided by funds, in particular the government's primary operating fund, the general fund, may be more important to the decision process. The survey results raise the question of whether aggregate information has incremental value relative to the disaggregate information, particularly the information provided by the general fund. Unfortunately, the work by Reck et al. (2009) does not address whether aggregate information has incremental value relative to the disaggregate information.

Given the expectation by the GASB concerning the usefulness of government-wide information and the preliminary research work done in this area, it is important to empirically test whether favorable aggregate modified accrual information has incremental value in assessing default risk and, therefore, will lower bond interest costs. Prior research (see reviews by Ingram et al. [1987] and Reck et al. [2004]) reports associations between modified accrual information and new issue bond interest costs. Therefore, we hypothesize that:

H2: 

In the presence of general fund modified accrual basis financial information, aggregate modified accrual basis funds financial information required by Statement No. 34 is associated with new issue bond interest costs.

The GASB Statement No. 34 reporting model raises an additional question: Does government-wide accrual financial information affect bond borrowing costs differently than aggregate modified accrual information? As discussed previously, the government-wide accrual information incorporates the effects of longer-term financial decisions, whereas the modified accrual fund information reflects current period spending decisions. The following hypothesis posits that accrual financial information provides incrementally useful information about longer-term financial decisions that is relevant to assessing borrowing costs on municipal bonds.

H3: 

In the presence of aggregate modified accrual financial information, government-wide accrual financial information required by Statement No. 34 provides additional information that is associated with new issue bond interest costs.

To test our hypotheses, we employ regression analysis that includes financial measures that are associated with bond interest costs in prior research. While this is admittedly an ad hoc process, a limitation of all bond research is a lack of theory concerning which financial accounting variables should be included in models testing for an association with bond interest costs/yields.

Sample Description

Bond issue and related market data were obtained from the Thomson Municipal Market Monitor (TM3) primary market database, a proprietary service of Thomson Reuters. Required financial data were obtained from comprehensive annual financial reports (CAFRs) provided by governments on their websites. As the Thomson market database essentially contains the full population of municipal bond issues made in the national market, we were able to obtain complete bond issue and market data for 1,822 bond issues. However, because of the difficulty of obtaining financial information for municipalities, our sample was restricted to only those cities that (1) had bond issues and published comprehensive annual financial reports for the relevant years on their websites, and (2) had fully complied with GASB Statement No. 34. Both constraints resulted in excluding numerous cities. Subject to these constraints, we hand collected financial information for 234 bond issues. Forty-nine observations were subsequently excluded from our sample because of incomplete financial data for some test variables. Our final sample consists of 185 general obligation bond issues made by 125 cities during the years 2002–2006. These cities are well distributed geographically across 30 states and are diversified by municipality size, ranging from a population of 6,334 to a population of 3,807,400, with an average population of 222,763.

Our sample addresses some of the data limitations from the prior research (Plummer et al. 2007; Reck et al. 2009) about the usefulness of Statement No. 34 information. We use a sample that covers a five-year period of time to allow for increased market familiarity with the information offered by Statement No. 34 financial reports, and the significance of environmental events that could impact a single year result. Our broad sample of municipalities from 30 states adds to the generalizability of the results, since state economic conditions can impact default risk (Fairchild and Koch 1998; Baber and Gore 2008). We use a continuous dependent variable, interest costs, rather than bond ratings. Interest cost is better able to capture the impact of accounting information disclosure on debt costs than is bond ratings, since bond ratings tend to change only at long intervals and, as a result, are not necessarily able to capture the association between current period operating activity and default risk. We use actual Statement No. 34 data and construct our ratios using data that were not available under the pre-Statement No. 34 model.

Regression Models

The dependent variable in our regression models is IntCost, the aggregate true or net interest cost reported by The Bond Buyer for each serial bond issue. Because most municipal bonds are issued with serial maturities, they typically have different coupon rates for different maturities, affecting the IntCost of the bond issue. For decades, net interest cost was the principal measure used in the primary market for municipal bonds when evaluating and awarding bids from competing underwriters.1 As a result of technological developments, true interest cost has largely replaced net interest cost as the primary measure of borrowing cost in the municipal bond market, as evidenced by the 87.7 percent of issues in our final sample that reported true interest cost. To control for potential differences between net interest cost and true interest cost, we include a dummy variable (NICdum) indicating issues that are reported at net interest cost.

As described more fully in the next section, hypotheses are tested by adding selected financial test variables to a base model that controls for a number of bond issue attributes, market factors, and issuer socioeconomic and fiscal factors (see Table 2 for descriptions of all variables). The specific models used for our tests take the following form:

formula

where Finvar is a vector of financial variables selected to address the hypotheses, and the other variables are control variables that constitute the base model.

TABLE 2

Description of All Variables Used in Regression Tests

Description of All Variables Used in Regression Tests
Description of All Variables Used in Regression Tests

To control for changes in market interest rates for municipal bonds over time, we include the Bond Buyer 20-Year Bond Average Yield Index (BBind) for the same week as each new bond issue. We control for other important bond issue or issuer attributes by including the average maturity of each serial issue (Avgmat), an indicator variable for whether the bond issue is insured (Insured), an indicator variable indicating whether the issue reported net interest cost (NICdum) rather than true interest cost, and the number of competitive underwriter bids received for the issue (NbrBids). Most bond yield or interest models also include one or more control variables for bond ratings. Given the relatively strong economy that prevailed during the years included in our study (2002–2006), only 26 of 185 bond issues had ratings below Moody's Aa category. Accordingly, we add only a single indicator variable, LoRating, indicating whether an issue was rated below Moody's Aa category.

Consistent with Plummer et al. (2007) we include controls for issuer-specific factors: the amount of outstanding general obligation (i.e., tax-supported) debt (Debt); two socioeconomic factors, the gross state product for each state (GspPCap) and the city's income per capita (IncPCap); and the proportion of revenues from own sources (OwnRev).2 These factors capture, respectively, the debt burden on taxpayers, the economic climate in each state, the ability of taxpayers to pay principal and interest on debt, and the degree of control city officials have over the revenue base.

We also include dummy variables for the years included in the sample to control for macroeconomic or market structural factors that could influence bond interest costs. Due to the small number of observations in years 2002 and 2003, these two years are combined and assigned a value of 0, effectively leaving dummy controls for 2004, 2005, and 2006.

The rationale for including the selected control variables and their expected associations with interest cost are explained as follows. General obligation bonds predominantly reflect market interest rates, adjusted for differences in default and interest rate, among other factors. Thus, the bond issues in our sample are expected to be positively and strongly associated with the Bond Buyer Index (BBind) for the same week as the issue. The longer the average term to maturity (Avgmat), the more susceptible are bond issues to interest rate and default risk. As a result, longer-maturity bonds generally provide higher yield.

Bond insurance was common in the municipal bond market during the sample period, as governments sought to enhance the credit quality of their bonds and reduce borrowing costs. Theoretically, the market should price all insured bonds as triple-A rated. However, researchers (Hsueh and Chandy 1989; Peng and Brucato 2004) indicate that the price of insured bonds is affected by the fact that the creditor could still incur losses should the insurer default. Creditors must therefore consider the joint probability that the insurer and the issuer will default (Insured). Bonds sold in competitive offerings generally sell at a lower interest cost (yield) than bonds sold in a negotiated offering. In general, a larger number of competing bids (NbrBids) and higher bond ratings reduce borrowing costs.

Finally, issuer-specific factors, such as those for which control variables are included, may affect an issuer's borrowing costs. High levels of debt outstanding (Debt) relative to capacity to service the debt may increase an issuer's borrowing cost on new debt issuances. Interest and principal on general obligation debt is paid with tax revenues, and such debt is backed primarily by the promise of the government for repayment. Critical to supporting higher debt burdens is the economic health of the state, the ability of taxpayers to pay taxes, and the ability of local officials to control the revenue base. Consequently, higher levels of economic activity (GspPCap), higher disposable income (IncPCap), and the ability to generate own-source revenues (OwnRev) are generally associated with lower default risk.

Hypothesis Testing

We test our hypotheses using models similar to those of Reck et al. (2004), Reck and Wilson (2006), Plummer et al. (2007), and Reck et al. (2009). For each hypothesis, we construct a null model consisting of the base model plus balance sheet and operating statement financial measures corresponding to the implied null hypothesis for each hypothesis. Each hypothesis is then tested by adding appropriate alternative balance sheet and financial performance test variables to the null model, as described below and depicted in Table 3.

TABLE 3

Hypotheses Test Models

Hypotheses Test Models
Hypotheses Test Models

To test H1, we compare the explanatory power of a null model that contains the base model plus modified accrual basis balance sheet (GFBal) and operating statement financial performance (GFPerf) variables, for the general fund only, to an alternate test model that adds accrual basis balance sheet (NetAssets) and financial performance (GAPerf) variables for government-wide governmental activities to the null model.

The same null model is used to test H2 but, rather than government-wide accrual basis information, the alternate model includes aggregate modified accrual basis balance sheet and operating statement variables, AllGovFB and AllGovPerf.

A different null model is used to test H3; the base model plus aggregate modified accrual basis balance sheet and operating statement variables, AllGovFB and AllGovPerf, for all governmental funds. The alternate test model for this hypothesis adds the government-wide accrual basis balance sheet and financial performance variables, NetAssets and GAPerf, to the null model.

To test each hypothesis, we employ the general linear test to obtain an F-statistic and related probability for a direct test of whether the alternate test model significantly reduces the error variance of interest costs.3

Prior research has found the general fund variables GFBal and GFPerf to be negatively associated with bond interest costs (e.g., see studies reviewed by Ingram et al. [1987] and Reck et al. [2004]). However, prior research findings (e.g., Wilson 1990; Reck et al. 2009) provide little evidence on whether aggregate modified accrual information for all governmental funds is incrementally useful in assessing default risk beyond the information provided by the general fund. If it is, we expect the aggregate balance sheet and financial performance test variables (AllGovFB and AllGovPerf) to be negatively associated with interest cost, similar to research findings relying on the general fund information. As we discussed earlier, we do not predict the direction of association between the accrual basis variables (NetAssets and GAPerf) and interest cost, as the longer-term effects of government-wide accrual information may be interpreted either favorably or unfavorably.

Descriptive Statistics

Table 4 provides descriptive statistics for all variables used in the regressions. The average interest cost of 4.043 percent for our sample is somewhat lower than the Bond Buyer 20-Year Bond Average Yield Index of 4.509 since the average maturity of 10.492 years for our sample is substantially shorter than the 20-year maturities of the bonds used to calculate the index. More than half (60 percent) of the bond issues in our sample are insured. These bonds are automatically assigned a Moody's Aaa bond rating, reflecting the credit enhancement of providing insurance. Consistent with the growing trend in the municipal market, most of the bond issues in our sample (85.4 percent) were awarded on the basis of lowest true interest cost. On average, the issues in our sample received more than six bids from competing underwriters. Only 14.6 percent of the 185 bond issues had a Moody's rating below the Aa category, indicating the overall positive creditworthiness of our sample.

TABLE 4

Descriptive Statistics for All Variablesa Used in Regression Tests Sample is 185 General Obligation Bonds Issues by Cities from 2002 through 2006

Descriptive Statistics for All Variablesa Used in Regression Tests Sample is 185 General Obligation Bonds Issues by Cities from 2002 through 2006
Descriptive Statistics for All Variablesa Used in Regression Tests Sample is 185 General Obligation Bonds Issues by Cities from 2002 through 2006
TABLE 5

Bivariate Spearman Correlations between All Variablesa Used in the Regression Models (n = 185)

Bivariate Spearman Correlations between All Variablesa Used in the Regression Models (n = 185)
Bivariate Spearman Correlations between All Variablesa Used in the Regression Models (n = 185)

Debt (the ratio of general obligation debt to total governmental revenues) averaged 77.2 percent of revenues and ranged from 0 to 266.4 percent of revenues. Overall, the cities in our sample raised 79.6 percent of their own revenues (OwnRev), which varied from 31.8 to 98.9 percent. Gross state product (GspPCap) averaged $40,665 per capita and ranged from $27,998 to $57,909. Income per capita (IncPCap) averaged $33,388 and was highly dispersed, ranging from a low of $14,137 to a high of $89,978 per person.

Means for the year dummies indicate the proportion of total bond issues made each year. Thus, the bond issues per year were 42 in 2004 (185 cities × 0.227), 70 in 2005, and 55 in 2006, leaving 18 in the excluded years of 2002 and 2003.

Both the modified accrual fund and accrual government-wide financial position variables reflect the generally favorable financial climate during most of the years covered by our study (2002–2006).4 The general fund balance (GFBal) of cities in our sample averages 23.9 percent of general fund revenues, more than four times the 5 percent level that credit analysts generally regard as a “red flag.” This represents an 87-day operating reserve for the average city. Aggregate modified accrual fund balance (AllGovFB) and accrual government-wide unreserved net assets (NetAssets) are even higher, averaging 38.3 and 27.1 percent of a year's revenue, respectively.

Operating statement variables also reflect relatively healthy financial performance. At the fund level, the general fund operating surplus (GFPerf) averaged 3.4 percent of annual revenues, although aggregate modified accrual funds (AllGovPerf) reported an average deficit of 18.7 percent. This can occur because funds servicing debt may report expenditures for principal paid from resources accumulated in prior years. Similarly, funds established for multiple-year capital projects may report construction expenditures in years subsequent to financing receipts. Government-wide accrual operating surplus, the excess of governmental activities revenues over expenses divided by governmental activities revenues (GAPerf), was similar to that for the general fund; 3.1 percent of annual revenues. As shown by the minimum values of the three operating performance measures (GFPerf, AllGovPerf, and GAPerf), some cities reported operating deficits, both at the modified accrual funds and government-wide accrual levels. Although not reported in Table 4, 60 of 185 cities (32.4 percent) reported general fund operating deficits on the modified accrual basis and 55 (29.7 percent) reported government-wide deficits on the accrual basis. Contrary to public perceptions, fund operating deficits do not necessarily reflect fiscal stress, since governments often intentionally budget for the use of beginning-of-year financial reserves (fund balance) to finance a portion of current year expenditures/expenses in order to reduce the reserve level to a specified target.

Table 5 presents Spearman correlations for all continuous variables used in the regression. Correlations are discussed with regression diagnostic tests in the next section.

Regression Results

We construct five regression models, as discussed previously in the “Research Design” section. Each model provides a test of the extent to which specified financial variables are associated with municipal bond borrowing cost after controlling for bond issue and market factors, and potential structural changes across years.

Consistent with most prior studies, the results displayed in Table 6 show that the borrowing cost on new municipal bond issues (IntCost) is driven primarily by the contemporary market interest rate (BBind) and the average maturity of the issue (Avgmat) (both p < 0.0001). Insured bonds were not associated with lower interest cost; however, this result may indicate that the insured bonds represent issuers that, without credit enhancement, have inherently higher default risk. As expected, bond issues measured using net interest cost (NICdum) tended to understate the true interest cost for new issues (p < 0.05). Also, as expected, competition for bond issues (NbrBids) lowers borrowing cost (p < 0.01). Unexpectedly, issues with a Moody's bond rating below Aa were associated with lower, rather than higher, interest cost. This may reflect the fact that 19 of the 26 issues that fall in the low rating category were insured, resulting in automatic upgrade to Aaa category.

TABLE 6

Regression Results Associations between Municipal Bond Interest Cost and Test Models Comparing General Fund and Aggregate Modified Accrual Fund-Based Financial Information and Government-Wide Accrual Financial Information (n = 185)

Regression Results Associations between Municipal Bond Interest Cost and Test Models Comparing General Fund and Aggregate Modified Accrual Fund-Based Financial Information and Government-Wide Accrual Financial Information (n = 185)
Regression Results Associations between Municipal Bond Interest Cost and Test Models Comparing General Fund and Aggregate Modified Accrual Fund-Based Financial Information and Government-Wide Accrual Financial Information (n = 185)

The issuer specific control variables, Debt, OwnRev, GspPCap, and IncPCap are generally insignificant in relation to the other control variables, except for models H03 and HA3, the models that examine only aggregate financial information. Controls for year of issue reflect a pattern of increasing municipal borrowing costs over those years, compared with the excluded years of 2002 and 2003.

Our primary focus is on the financial test variables, particularly the hypotheses tests. As shown in Table 6, the first set of regression results provides the results for the null model (denoted as H01 and H02) used to test both H1 and H2. This model consists of the control variables plus the modified accrual balance sheet and operating statement variables for the general fund only (see Table 3 for the summary of testing procedures). The general fund balance sheet variable (GFBal) and operating statement variable (GFPerf) are individually significant in the expected negative direction at the 0.05 and 0.10 levels, respectively. Although we did not formally hypothesize that these variables would be significant in the presence of the control variables, we did conduct a test comparing a full model that included the general fund financial variables to a reduced model that included only the control variables. We obtained an F-value of 4.25 that shows, consistent with most prior research, that the two general fund measures, as a pair, significantly reduce unexplained variance in interest costs (p < 0.05), compared with the base model only.

Model HA1 provides the alternate model for testing H1, that government-wide accrual information explains interest cost beyond general fund modified accrual information. Specifically, the error sum of squares of this model is compared to that of the null model using the general linear test procedure. The test result (F-value = 4.99) reported at the bottom of the table shows that the two government-wide accrual variables significantly add to the explanation of interest costs beyond the general fund variables (p < 0.01). Thus, the evidence supports H1.

Model HA1 suggests that government-wide accrual basis financial position (NetAssets) and financial performance (GAPerf) are positively associated with borrowing costs, although only GAPerf is significant (p < 0.05, two-tailed). As noted by Plummer et al. (2007, 213), and discussed more fully in the next section, the direction of association between government-wide accrual information and default risk is inherently ambiguous and cannot be predicted.

Results for model HA2 show that in the presence of general fund financial information, aggregated modified accrual information (AllGovFB and AllGovPerf) does not significantly add explanatory power (F = 1.57) and, therefore, does not support H2.

The results for the null model H03 indicate that the aggregate balance sheet (AllGovFB) and operating statement (AllGovPerf) variables are not significantly associated with interest cost. Our primary focus, however, is with H3. Model HA3 suggests that neither the aggregate modified accrual financial information nor the government-wide accrual information is significantly associated with interest cost. Moreover, the related F-value of 1.88 indicates that the government-wide accrual information does not incrementally explain interest costs in the presence of aggregate modified accrual information.

Sensitivity and Diagnostic Tests

To examine more directly the incremental influence of government-wide accrual-based operating information on interest cost, we conduct an additional test of H1 by first removing capital and debt service expenditures from general fund modified accrual expenditures. These components of general fund expenditures potentially introduce noise in the measure of operating variables relevant for H1. The remaining components of general fund expenditures are essentially short-term operating components, which are referred to as general fund operating expenditures. Further modification of general fund revenues is unnecessary since its components, for the main part, are inherently short-term operating components (taxes, intergovernmental, charges for services, fines and forfeitures, licenses and permits, etc.). We then construct a new operating statement performance measure, GFopPerf, defined as (general fund revenues − general fund capital and debt related expenditures)/general fund revenues. GFopPerf replaces GFPerf in the test of HA1 and permits a better test of the incremental explanatory power of longer-term operating accrual components beyond the short-term modified accrual components.5

These results (not tabulated) are similar to those reported in Table 6, with H1 still being supported at the 0.01 level. In addition, we evaluate H1 replacing the unrestricted government-wide accrual balance sheet measure, NetAssets, with total (rather than unreserved) governmental activities net assets divided by total governmental activities revenue. The results are slightly weaker than those reported in Table 6 with a slightly lower F-value of 4.27, which is still significant at the 0.02 level.

We also evaluate H2 using both a reformulated general fund modified accrual performance measure and a reformulated aggregate modified accrual performance measure. For both measures, capital and debt expenditures are removed, leaving primarily short-term operating expenditures. Testing in this manner is warranted because of the potentially distorting effect of large capital and debt expenditures in the governmental funds. At the aggregate level, the reformulation leaves mainly general fund plus all special revenue funds. Testing H2 with the reformulated variables produced positive associations for AllGovFB and AllGovopPerf in the alternate model, but the results are insignificant (F = 2.19; p = 0.12). This suggests that aggregated short-term modified accrual operating information does not provide additional information beyond that of the general fund alone in assessing default risk.

Overall, our results suggest that the market relies heavily on the short-term modified accrual information in the general fund's fund balance and its operating performance, but the longer-term operating statement accrual components provide additional information in assessing default risk.6

To test whether the hypothesized relations were affected by the credit enhancement (i.e., bond insurance) for 60 percent of our sample, we tested models for H1 where all four test variables were interacted with the dummy variable Insured. The incremental contribution of the two accrual basis government-wide test variables, NetAssets and GAPerf, relative to the two general fund modified accrual basis variables, GFBal and GFPerf, was even stronger than for the main test of H1. The F-value for the interacted test variables was 6.57 (p < 0.01) compared with 4.98 for the models reported in Table 6. This suggests that the explanatory power of the accrual basis test variables is stronger for issues with credit enhancement compared to issues without enhancement.

Multicollinearity does not appear to be a problem since variance inflation factors for all our test variables are well below 2.0. Analysis of the Spearman correlations (see Table 5) reveals no pattern of significant concern, with only moderate correlations observed among test variables. Additional tests indicate that observed associations are not attributable to influential observations. In particular, omitting nine observations with studentized residuals greater than two standard deviations from the mean yields F = 4.29 for tests of H1, compared with 4.98 reported in Table 6 for the full sample, although associations for individual test variables with interest costs are somewhat stronger than those reported for the full sample. Finally, White (1980) tests reveal no significant heteroscedasticity (nonconstant variance of residuals) problem.

Although we find, consistent with prior studies, that general fund modified accrual financial information is negatively associated with bond interest costs, the results for our sample suggest positive associations between government-wide accrual financial information and bond borrowing costs. Our study tests the joint contribution of the accrual basis variables to reducing interest cost error sums of squares rather than testing a direction of association of the individual government-wide test variables. Still, it is useful to understand why the observed positive associations may occur.

GFBal, the fund balance of the general fund, is a well-known financial policy variable in local government. Most well-managed governments have established a minimum fund balance policy, typically a financial reserve in the range of 10–25 percent of annual revenues or expenditures. GASB standards require governments to disclose this policy in the notes to financial statements. Fund balance is closely examined for both internal budgeting purposes and external credit analysis. Although the government-wide accrual balance sheet measure we use in our study, unrestricted net assets over revenues (NetAssets), captures some of the same working capital factors (e.g., cash, current receivables, and current liabilities) that define GFBal (Table 5 shows a correlation of 0.345), certain longer-term assets and liabilities impact NetAssets, but not GFBal. These include assets such as long-term receivables and deferred bond issue costs, and long-term operating liabilities such as compensated absences, unfunded pension liabilities, pollution liabilities, and court judgments. Moreover, given credit analysts' long experience with evaluating general fund financial information, it is possible that they do not yet place as much value on the newer government-wide accrual financial information.

Similarly, our government-wide accrual operating performance measure, the excess (deficiency) of revenues over (under) expenses divided by revenues (GAPerf), includes the operating components of the modified accrual basis revenues and expenditures that determine the general fund operating performance measure (GFPerf). In addition to the modified accrual components, however, GAPerf recognizes revenues receivable in the future that are not recognized in the general fund. GAPerf also recognizes expenses, such as accrued interest and increases in longer-term liabilities (e.g., compensated absences, pensions, pollution costs, and court judgments), that will not be paid until future periods. Further differences are generated by acquisitions of capital assets, which are reported as expenditures (the cost of the item) in the general fund, but as depreciation expense over the lives of the items at the government-wide accrual level. The potential impact of longer-term operating statement components is illustrated in our sample by the low 0.074 pairwise correlation between GFPerf and GAPerf.

The significant support for H1 reflects the incremental explanatory power of the longer-term accrual components over the short-term focused general fund measures. Although those components are useful in assessing default risk, their assessed effects may vary. Given that GAPerf reflects the longer-term operating decisions of the government, it could also be influenced by actions of some governments to boost short-term financial performance (i.e., GFPerf) by shifting the cost of current services to future periods. Governments have a history of deferring the cost of current services by offering more liberal vacation and sick leave policies, reducing pension contributions (Chaney, Copley, and Stone 2002; Apostolou, Reeve, and Giroux 1984), and deferring maintenance on capital plant. Under the pre-Statement No. 34 model, the deferral of the cost of current services was not as transparent as under the Statement No. 34 model, which provides longer-term accrual basis information.

Consistent with the notion that governments strategically manage costs by shifting the cost of current services to future periods, we observe that cities with the lowest values of general fund modified accrual balance sheet and general fund operating performance measures tend to have the highest values of the government-wide accrual basis balance sheet and operating performance measures. Conversely, cities that have the highest values of general fund modified accrual measures tend to have the lowest values of government-wide accrual financial measures.

As specific examples of how government-wide accrual financial measures tended to be opposite those for the general fund modified accrual measures alone, we examined whether cities with a general fund modified accrual operating deficit (i.e., expenditures greater than revenues) also had a government-wide accrual basis operating deficit (expenses greater than revenues). Of the 60 cities that reported a general fund deficit, only 24 also had a government-wide deficit. In addition, 73.3 percent of these 60 cities had government-wide accrual operating performance that was more positive than general fund modified accrual operating performance. Of the 125 cities that reported a general fund operating surplus, 31 reported a government-wide operating deficit, and government-wide operating performance was more positive than that of the general fund for only 42.4 percent of the cities. Thus, the inverse relationship between government-wide accrual and general fund modified accrual financial information is consistent with the argument that governments may manipulate long-term accrual information to improve overall results when short-term financial stress occurs. Nonetheless, the direction of the relationship does not detract from the overall usefulness of longer-term accruals in assessing default risk and determining borrowing cost.

In summary, our results suggest that it is the government-wide longer-term operating accruals, such as those just discussed, that explain the incremental explanatory power of the aggregate accrual basis information over general fund information only. Our results further suggest that the inherent noise in capital and debt related inflow and outflow measures reduce their usefulness in explaining bond interest costs. Our data did not permit examination of specific longer-term accrual components such as compensated absences, pension costs, pollution abatement costs, and judgments. Future research should focus on the impact of components such as these.

Apostolou
,
N. G
.,
J. M
.
Reeve
,
and
G. A
.
Giroux
.
1984
.
Accounting information and municipal bond net interest cost: An empirical evaluation
.
Journal of Accounting and Public Policy
3
:
9
28
.
Baber
,
W. R
.,
and
A. K
.
Gore
.
2008
.
Consequences of GAAP disclosure regulation: Evidence from municipal debt issues
.
The Accounting Review
83
(
3
):
565
591
.
Bullock
,
N
.
2010
.
Moody's to change U.S. muni ratings
.
Financial Times (London)
(
March 17
)
:
21
.
Chaney
,
B. A
.,
P. A
.
Copley
,
and
M. A
.
Stone
.
2002
.
The effect of fiscal stress and balanced budget requirements on the funding and measurement of state pension obligations
.
Journal of Accounting and Public Policy
21
:
287
313
.
Das
,
S. R
.,
P
.
Hanouna
,
and
A
.
Sarin
.
2009
.
Accounting-based versus market-based cross-sectional models of CDS spreads
.
Journal of Banking and Finance
33
(
4
):
719
730
.
Ederington
,
L. H
.,
J. B
.
Yawitz
,
and
B. E
.
Roberts
.
1987
.
The informational content of bond ratings
.
The Journal of Financial Research
10
(
3
):
211
226
.
Fairchild
,
L. M
.,
and
T. W
.
Koch
.
1998
.
The impact of state disclosure requirements on municipal yields
.
National Tax Journal
51
(
4
):
733
753
.
Fischer
,
P. E
.,
and
R. E
.
Verrecchia
.
1997
.
The effect of limited liability on the market response to disclosure
.
Contemporary Accounting Research
14
(
3
):
515
541
.
Foltin
,
C
.
2008
.
Unrest in government accounting
.
The CPA Journal
78
(
3
):
26
31
.
Governmental Accounting Standards Board (GASB)
.
1999
.
Basic Financial Statements—and Management's Discussion and Analysis—for State and Local Governments
.
Statement No. 34
.
Norwalk, CT
:
GASB
.
Governmental Accounting Standards Board (GASB)
.
2013
a
.
Conceptual Framework: Recognition Concepts for the Current Financial Resources Measurement Focus. Memorandum of July 22, 2013
.
Norwalk, CT
:
GASB
.
Governmental Accounting Standards Board (GASB)
.
2013
b
.
Request for Research: Gil Crain Memorial Research Grant, Announcement
.
Norwalk, CT
:
GASB
.
Governmental Accounting Standards Board (GASB)
.
2013
c
.
Technical Plan for the Final Third of 2013
.
Norwalk, CT
:
GASB
.
Howe
,
J. T
.
2001
.
Credit analysis for corporate bonds
.
In
The Handbook of Fixed Income Securities
,
edited by
F. J
.
Fabozzi
,
417
489
.
New York, NY
:
McGraw-Hill
.
Hsueh
,
L. P
.,
and
P. R
.
Chandy
.
1989
.
An examination of the yield spread between insured and uninsured debt
.
The Journal of Financial Research
12
(
3
):
235
244
.
Ingram
,
R
.,
and
E. R
.
Wilson
.
1999
.
Information intermediation and seasoned municipal bond yields
.
Research in Governmental and Nonprofit Accounting
10
:
1
31
.
Ingram
,
R
.,
K. K
.
Raman
,
and
E. R
.
Wilson
.
1987
.
Governmental capital markets accounting research in accounting: A review
.
Research in Governmental and Nonprofit Accounting
3
:
111
132
.
Khurana
,
I. K
.,
and
K. K
.
Raman
.
2003
.
Are fundamentals priced in the bond market?
Contemporary Accounting Research
20
(
3
):
465
494
.
Lauricella
,
T
.,
and
K
.
Richardson
.
2007
.
Credit pressure filters down to muni market; bond insurers' exposure to troubled mortgages ripples through system
.
Wall Street Journal (Eastern edition)
(
November 16
)
:
C.1
.
Liu
,
P
.,
and
F. J
.
Seyyed
.
1991
.
The impact of socioeconomic variables and credit ratings on municipal bond risk premia
.
Journal of Business Finance & Accounting
18
(
5
):
735
746
.
Neumann
,
T
.
2010
.
Bond insurers lose AAA-rated player—S&P downgrades assured guaranty units as outlook dims on the industry
.
Wall Street Journal (Eastern edition)
(
November 2
)
:
C.5
.
Peng
,
J
.,
and
P. F
.
Brucato
Jr.
2004
.
An empirical analysis of market and institutional mechanisms for alleviating information asymmetry in the municipal bond market
.
Journal of Economics and Finance
28
(
2
):
226
238
.
Plummer
,
E
.,
P. D
.
Hutchison
,
and
T. K
.
Patton
.
2007
.
GASB No. 34's governmental financial reporting model: Evidence on its information relevance
.
The Accounting Review
82
(
1
):
205
240
.
Reck
,
J. L
.,
and
E. R
.
Wilson
.
2006
.
Information transparency and pricing in the municipal bond secondary market
.
Journal of Accounting and Public Policy
25
(
1
):
1
31
.
Reck
,
J. L
.,
E. R
.
Wilson
,
D
.
Gotlob
,
and
C. M
.
Lawrence
.
2004
.
Governmental capital markets research in accounting: A review, extension, and directions for future research
.
Research in Governmental and Nonprofit Accounting
11
:
1
33
.
Reck
,
J. L
.,
E. R
.
Wilson
,
and
O
.
Schiffel
.
2009
.
Did bond prices reflect accrual basis approximations before GASB Statement No. 34?
Research in Governmental and Nonprofit Accounting
12
:
169
188
.
Seymour
,
D
.
2010
.
Recalibrations may have nudged spreads: Moody's, Fitch upgrades analyzed
.
Investment Dealers' Digest
76
(
18
):
9
.
Stevens
,
G
.
1999
.
Evaluation of underwriter proposals for negotiated municipal bond offerings
.
Public Administration & Management
4
(
4
):
435
468
.
Strumeyer
,
G
.
2005
.
Investing in Fixed Income Securities: Understanding the Bond Market
.
New York, NY
:
John Wiley & Sons
.
U.S. House of Representatives
.
2010
.
Dodd-Frank Wall Street Reform and Consumer Protection Act. Public Law 111-203 [H.R. 4173]
.
Washington, DC
:
GPO
.
Van Horne
,
J. C
.
1990
.
Financial Market Rates and Flows. 3rd edition
.
Englewood Cliffs, NJ
:
Prentice Hall
.
White
,
H
.
1980
.
A heteroscedasticity-consistent covariance matrix estimator and a direct test of heteroscedasticity
.
Econometrica
48
:
817
838
.
Wilson
,
E. R
.
1990
.
Financial Reporting by State and Local Governments: A Survey of Preferences among Alternative Formats
.
Norwalk, CT
:
GASB
.
1

Net interest cost is a single-value borrowing cost measure calculated as the weighted average interest cost across all maturities, adjusted for premiums or discounts. It, however, ignores the time value of money. Computer technology now facilitates the iterative process used to calculate true interest cost. True interest cost is the internal rate of return that discounts the irregular future interest and principal payments of a new bond issue to the underwriter's purchase price. Some municipal issuers still specify lowest net interest cost as the criterion for awarding competitive bond issues. Post-sale data reported by market data sources for net interest cost issues typically do not provide the true interest cost for these issues, nor the information needed to calculate it. Stevens (1999) provides comparative evaluations for negotiated offerings that indicate net interest cost is slightly lower than true interest cost because cash flows for all maturities are equally weighted using net interest cost; whereas, true interest cost discounts longer-term cash flows more than shorter-term flows.

2

Some prior bond studies (e.g., Plummer et al. 2007) control for entity size. We find no compelling reason to control for population in our study as our test measures are, by construction, size invariant. However, we also conducted tests that included the logarithm of population as a control variable. Since our results were unaffected by including a population control variable, we did not include it in our final tests.

3

The general linear test (also known as the general linear hypothesis) relies on an F-statistic that indicates whether an alternate (full) model significantly reduces the unexplained error variance of interest costs relative to a null (reduced) model. The F-statistic is calculated as: F-statistic = [(SSEnullSSEalt)/(dfnulldfalt)]/(SSEalt/dfalt), where SSE is the error sum of squares, null and alt denote the null and alternate models, respectively, and df is the degrees of freedom. Under the null hypothesis, the calculated F-statistic is distributed as F with numerator degrees of freedom of dfnulldfalt and denominator degrees of freedom of dfalt. This test assumes that least-squares estimators are statistically appropriate for these models. Evaluation of the error term distribution reveals no significant deviation from least-squares assumptions. Specification tests discussed in the “Results” section indicate that the error terms follow the normal distribution with an expected mean of 0, and the expected variance equal to the sample mean square error (MSE).

4

Reck and Wilson (2006) examined 595 new issues that occurred during the relatively weaker economy of the 1980s. The cities in their sample had average unreserved general fund balances of only 13.6 percent of annual revenues compared with the 23.4 percent for our sample. Similarly, Reck and Wilson (2006) reported general fund performance (ratio of the difference between general fund revenues and general fund expenditures to general fund revenues) of only 1.2 percent for their 1980s' sample compared with 3.3 percent for our sample.

5

While we would like to reconstruct our general fund balance sheet measure as well, the cumulative nature of fund balance and data limitations do not permit us to do so.

6

Alternatively, we retested H3 using an AllGovopPerf measure that also removed capital and debt service expenditures from total aggregate expenditures for all governmental funds. These results were not different from those reported in Table 6, suggesting that the test results for H3 were not affected by the large amounts of capital expenditures reported in the capital projects funds and debt service expenditures reported in the debt service funds.

Author notes

The authors thank the 2010 AAA Annual Meeting GNP reviewers and attendees and the faculty of the University of South Florida School of Accountancy. Professor Reck gratefully acknowledges the financial support received from the University of South Florida School of Accountancy.