The Hidden Consequences of Balanced Budget Requirements Part 2

3.5 Sample selection Our sample period covers the fiscal years 1998 through 2006. We use multiple public data sources to construct the sample. We obtain the state governments’ financial data and asset sale data from the Census Bureau’s Annual Survey of Governments. We collect the book value of the assets sold from the state’s Comprehensive Annual Financial Reports (CAFR). 18 We also collect states’ budget data from the NASBO fiscal surveys. The NASBO conducts a survey each year on all 50 state budget offices. These surveys focus on states’ general funds and obtain information on the actual and appropriated revenues and expenditures in the general fund balances. We obtain the state population from the U.S. Census Bureau and the state unemployment rate from the Bureau of Labor Statistics. We collect states’ GDP and personal income data from the Bureau of Economic Analysis. After requiring non-missing data, our final sample has 449 state-year observations. 4. Results 4.1 Descriptive statistics Table 1 presents descriptive information on variables used in estimating equation (1). We find that the average asset sale during the sample period is around $14 million or $7.5 per capita. The net revenues without considering the revenues from asset sales are on average $3.7 per capita. The positive net revenues are largely driven by the beginning (1998-2000) and latter sample periods (2005-2006) when most states were running surpluses. 18 We hand collect the book values of the assets sold for the asset proceeds test from schedules reconciling the Statement of Revenues and Expenditures for the Governmental Fund to the government-wide Statement of Activities in the CAFRs. This reconciliation schedule is required by the GASB Statement 34, which was implemented by most state governments for fiscal years beginning after June 15, 2001. While states are required to disclose the reconciliation, the level of detail is discretionary. Therefore, we only perform the asset proceeds test for the subset of states that report the book values of assets sold (117 state-year observations). . In figure 1 we plot the per capita state asset sales and the aggregate state general fund balances as a percentage of state spending from 1998 to 2006. The figure indicates that states face unusual fiscal difficulties between 2001 and 2004, coinciding with the recession. Nearly every state was in fiscal crisis during this period and had enacted spending cuts. The average spending cuts during this period is $152 million, which is larger than the average spending cuts of $6.6 million in the remainder of the sample period (p-value < 0.0001, untabulated). During 2001-2004, states are on average running deficits of $108 million or $18 per capita. Figure 1 also shows that during fiscal downturns, state asset sales increased dramatically. The per capita asset sales reached its highest in 2002 to $11. Overall, figure 1 shows that when states face financial difficulties, selling off assets is one of the fiscal adjustments they utilize to bring in revenues. Returning to Table 1, we find that 74% of our sample has strong balance budget restrictions that do not allow states to carry forward deficits. The average annual long-term debt issuance is $484 per capita and states on average receive $1200 per capita from the federal government. An average state’s population is about 5.6 million people. During our sample period the states’ GDP growth and personal income growth are both about 5% and the average unemployment rate is about 4.7%. Table 1 also shows that the average expenditures in the general fund are about $10 billion or $1,800 per capita. The average personal income per capita is about $31,000. During the sample period, states have an average surplus of $9.7 per capita. This average surplus is again mostly driven by the beginning and latter sample periods. During the recessionary period of 2001-2004, states on average were running a deficit of $10.5 per capita. For our measure of expense shifting, by construction EXP_RESID has a mean value of zero. LEADEXP_RESID has a mean value of -0.871, which is not statistically different from zero (p-value = 0.91). LEADEXP_RESID is our proxy for expenditure shifting. By requiring the lead value, we reduce our sample size by 50 (1 year of data). Table 2 presents Pearson correlations among main variables. We find that asset sales are negatively correlated with the net revenues in the general fund, suggesting that states are more likely to sell assets when they are in financial difficulties. Asset sales are positively related to a state’s dependence on credit markets, its revenue limitations, and revenue uncertainty and are negatively related to a state’s population size. Federal funding is negatively correlated with a state’s net revenues, suggesting that states are more likely to receive federal aid when they have fiscal trouble. 4.2 The relation between asset sales and balanced budget requirements Table 3 presents the regression results of equation (1). The model explains a reasonable portion of the variation in asset sales, with an adjusted R2 of 0.39. We find that the coefficient on NETREV is negative and significant (-0.091, t-statistic of -2.249) and that the coefficient on the interaction term NETREV*STRONGBBR is negative and significant (-0.145, t-statistic of 2.028). This result suggests that in general states facing deficits are more likely to sell public assets to increase revenues and that states who cannot carry deficits into the next fiscal cycle sell more public assets in response to deficits than states who do not face such strong balanced budget restrictions. For states with weak anti-deficit rules, a $100 per capita deficit induces an asset sale of $9.1. This amount is more than doubled in states with strong anti-deficit rules. For states that cannot carry the deficits forward, a $100 per capita deficit leads to an asset sale of $23.6. We find that the coefficient on STRONGBBR is not significant. This is not surprising, because balance budget restrictions are not binding unless the state incurs a deficit. Our other control variables are generally related to asset sales in the way we expect. Specifically, we find DISSUE to be positively associated with asset sales, suggesting that states relying more on the credit markets may need to maintain higher liquidity and, therefore, are more likely to use asset sales to bring in cash. We also find asset sales are related to a state’s revenue attributes. States with less diversified revenue streams and states with more volatile revenues are more likely to sell public assets to raise revenues. Finally we find that asset sales are more likely to happen when the state has higher unemployment rates. 19 Overall, Table 3 provides evidence that states sell assets when the net revenues in their general funds are low. The requirement to balance the budget intensifies states’ incentives to use asset sales to fill the budget holes. 20 4.3 Cross-sectional variation in the relation between asset sales and balanced budget requirements In Table 4, Panel A, we investigate whether the relationship between asset sales, NETREV, and STRNGBBR is influenced by limitations on a state’s ability to raise debt. We expect that for states facing limitations on the amount of debt they can issue, they are more likely to have incentives to seek other fiscal adjustments, such as asset sales, to solve financial difficulties. To determine whether a state has imposed legal limits on state debt issuances, we follow the budgetary report issued by NASBO in 2008 (NASBO 2008). We partition the sample 19 One concern with the finding is that our control variables do not capture the constraints on asset sales. Specifically, states that incur large asset sales in prior years may have fewer assets available for sale and therefore, are less likely to use asset sales to curb balanced budget restrictions. To investigate this possibility, we collect data on states’ capital assets in 2004 and 2005. We focus on these 2 years to reduce the data collection cost. We find that on average the amount of the assets sold constitutes only 2% of the total capital assets. Therefore, it is unlikely that states face constraints on selling assets because they do not have enough assets available for sale. 20 Given our small sample size, one concern with our regression analysis is that one state or year could influence our overall results. To address this concern, in all regressions we calculate the Cook’s Distance to identify any outlying observations. In untabulated analysis, we then run robust regressions, eliminating any outliers; our results and inferences remain unchanged. into two groups and run our regressions within each group. Table 1 shows that 62% of the sample has some form of limits on debt issuance. Table 4, Panel A reports the regression results. Column [1] reports the results for the subsample of states that have limits on debt issuances. For this sample, we find that NETREV is negative and significant (-0.088, t-value = -2.33), and so is the coefficient on the interaction term NETREV x STRONGBBR (-0.272, t-value = -2.95). The estimates suggest that for states with debt limits but with weak balanced budget requirements, a $100 per capita deficit in the general fund is associated with $8.80 per capita in asset sales. This result is magnified for states facing strong balanced budget provisions; a $100 per capita deficit in the general fund is associated with $36 per capita of asset sales. In contrast, the results in column [2] show that for the sub-sample of states without debt limits, NETREV and NETREV*STRONGBBR are insignificantly related to ASALE. Analysis of the difference in the coefficients across column [1] and column [2] reveals that the coefficients on NETREV and NETREV*STRONGBBR are statistically different for the subsample of states with debt limits versus the states without debt limits. Taken together, the results in Table 4, Panel A suggest that states that have limits on raising debt are more likely to sell off assets in a downturn than states without debt limits. This result is further magnified for states that have strict budget requirements. In Panel B of Table 4, we investigate the relation between asset sales and balanced budget restrictions by partitioning the sample into states with rainy day funds that are sufficient to cover a revenue shortfall and states that do not have sufficient rainy day funds. Rainy day funds are mechanisms used by various states to save surpluses in prosperous times to cover shortfalls in economic downturns. We expect that the propensity to sell assets to meet shortfalls will be lower in states that have rainy day funds that are sufficiently large enough to fill revenue shortfalls. The results in column [1] show that states with rainy day funds that are not sufficient to fill a revenue shortfall are more likely to sell assets; the effect is stronger for states that have strong budget requirements. A $100 per capita revenue shortfall results in a $12.40 per capita sale of assets for states with insufficient rainy day funding. In states with strong BBRs, the impact of a $100 per capita revenue shortfall increases the magnitude of asset sales to $39.20 per capita. The results in column [2] suggest that when states have sufficient rainy day funds, there is no relation between a revenue shortfall and the propensity to sell assets. For this subsample, the coefficients on NETREV and NETREV*STRONGBBR are both insignificant. A t-test for the difference in coefficients across the two subsamples confirms that NETREV and NETREV*STRONGBBR are statistically different from each other. Overall, the results suggest that the propensity to sell assets to meet revenue shortfalls is mitigated when states have a rainy day fund to fill the revenue gap. Finally, in Panel C of Table 4, we investigate the relation between revenue shocks, asset sales, and balanced budget requirements for subsamples of states that have negative shocks and positive shocks. Specifically, we partition the sample into states that have negative NETREV and states that have positive NETREV. Based on our previous discussion, we would expect the asset sale behavior only for states that have revenue shortfalls. Column [1] of Panel C shows that states with negative revenue shocks and strong balanced budget requirements are more likely to sell assets. In contrast, column [2] shows that states with positive net revenues do not sell assets, regardless of the strictness of the balanced budget requirements. These results validate our hypothesis that states sell assets when their revenues fall short of their expenditures and are more likely to do so when there are strict balanced budget requirements in place. Overall, Table 4 provides consistent evidence that states tend to sell public assets when facing financial difficulties and this behavior is intensified when they also face fiscal rules limiting their ability to borrow. The fact that we find consistent evidence across subsamples where we expect states to be more likely to sell assets mitigates the concerns that our main results are driven by omitted correlated variables. 4.4 The relation between gains and losses from asset sales and balanced budget restrictions Table 5 presents the results from our regression model of the asset sales price as a percentage of the book value of the assets at the time of the sale. Column [1] shows that NETREV is positive and significant (0.002, t-statistic= 2.06) and the interaction between NETREV and STRONGBBR is positive and marginally significant (0.001, t-statistic=1.87). These results suggest that states in poor financial health are more likely to sell assets at a loss, compared to states not facing financial difficulties. Further, the coefficient on the interaction term is consistent with our hypothesis that states facing strong anti-deficit rules are more likely to sell capital assets in fire sales when their financial condition deteriorates in order to meet the budget requirements. These results are economically significant. For states with weak BBR restrictions, a $100 per capita deficit results in a 20% reduction in the ratio of the proceeds from asset sales to the book value of the assets sold. For states with strong BBRs, a $100 per capita deficit results in a 30% reduction in this ratio. Thus, the impact of strong BBRs represents a 50% reduction in the asset sale proceeds, relative to states with weak BBR rules. We also find that states with more uncertain revenues (REVUNCERTAIN) are more likely to sell assets at a loss than states with more stable revenues. This suggests that states might sell assets as a source of revenue when alternative sources are less certain. While we do not find that other control variables relate to the asset sale relative to book value, overall our model explains a reasonable portion of the variation in our dependent variable, with an adjusted R-square of 0.257. Overall, the results in Table 5 suggest that states are more likely to sell assets at a loss when they have negative financial shocks and such effect is intensified when the states also face strong balanced budget provisions. The evidence is consistent with the idea that states may engage in sub-optimal fire sales to meet their strict balanced budget requirements. 4.5 The relation between expenditure shifting and balanced budget restrictions Table 6 reports the regression results of equation (3). The coefficient on LAGEXP is 0.990, suggesting that expenditures are sticky, tracking previous year’s values closely. The positive coefficient on LAGSURPLUS (0.822, t-statistic of 6.336) suggests that state governments increase expenditures following surpluses and reduce them following deficits. The magnitude of the coefficient suggests that a $100 surplus from last year is associated with $82 expenditures in the current year. The positive coefficient on state personal income (0.005, tstatistic of 2.230) and the negative coefficient on state unemployment rate (-14.633, t-statistic of -3.906) suggest that general fund expenditures are procyclical. States tend to increase expenditures when the general economic conditions are good and reduce expenditures when the economic conditions are bad. Our expenditure model explains a significant portion of the variation in EXP, with an adjusted R2 of 0.95. The strength of the results is reassuring in that our independent variables control for the factors determining the normal level of expenditures and that the residuals are reasonable estimates for the discretionary portion of the expenditures, which may be resulted from shifting. We expect that state governments are more likely to shift expenditures when they face financial difficulties. Similar to the analyses on asset sales, we measure state governments’ financial conditions using the net revenues in their general funds. We define the net revenues (NETREV_NOSHIFT) as the per capita total general fund revenues minus the predicted EXP from equation (3). Since the residual from equation (3) does not enter the calculations of net revenues, NETREV_NOSHIFT is free from the influence of expenditure shifting and our regression does not capture a mechanical relationship. Table 7 reports the regression results of balanced budget restrictions and expenditure shifting. We find that the coefficient on NETREV_NOSHIFT is not statistically significant (tstatistic of 0.100). However, the sum of the coefficients on NETREV_NOSHIFT and NETREV_NOSHIFT*STRONGBBR is negative and significant (p-value = 0.01). This result suggests that states that have weak balanced budget rules do not shift expenditures when their net revenues in the general funds are low. In contrast, states that have strong balanced budget rules tend to defer payments when facing low net revenues. Overall, Table 7 suggests that when a state cannot carry deficits forward, they tend to shift payments to the future when its net revenues in the general fund are low. We do not find similar shifting behavior for states that can carry the deficits to the next fiscal cycle. One concern with the findings in Table 7 is that we do not differentiate between expenditure shifting and expenditure deferring. When states face financial difficulties, they may defer some discretionary spending to future years and our expenditure model will capture such behavior as expenditure shifting. To investigate this concern, we exclude deferrable expenditures from our analysis. Clemens (2012) finds that non-construction capital expenditures and expenditures related to utilities are more likely to be deferred when states enact midyear spending cuts. These expenditures largely correspond to maintenance and equipment purchase related to gas and water supply systems, electric power systems, and public mass transit systems. In untabulate analysis we exclude these expenditures and rerun the test. We find that our results are not sensitive to excluding deferrable expenditures. Specifically, the coefficient on the interaction term NETREV_NOSHIFT*STRONGBBR is negative and significant (t-statistic of 1.80) and the sum of the coefficients on NETREV_NOSHIFT and NETREV_NOSHIFT*STRONGBBR is negative and significant (p-value = 0.013). This result suggests that our findings are not simply driven by states deferring expenditures to future fiscal years. We conduct additional robustness test using debt limits as an alternative proxy for budget restrictions and we find similar results. Specifically, we find that NETREV_NOSHIFT is not significant, but the sum of the coefficients on NETREV_NOSHIFT*DEBTLIMIT is negative and significant (p-value = 0.04). These results suggest that when facing financial difficulties, states tend to delay their payments to the next fiscal cycle if they have legal limitations on debt issuances. We do not find similar evidence on expenditure shifting for states that do not have such debt limits. 4.6 The tradeoff between asset sales and expenditure shifting In Table 8 we report the results of our tests examining the tradeoff between a state’s decision to shift expenses into the future and the decision to sell assets to meet strong balanced budget requirements. We partition our sample based on whether our expense-shifting model classifies the firm as an expense shifter in year t+1 (i.e. the residual of the model is positive) or a non-shifter. We then examine the relationship between asset sales and balanced budget restrictions both within and across this partition. In the first column of Table 8, we report the results of our asset sales model for states that are unlikely to have shifted expenses. Consistent with the results we report in Table 3, we find that in states with strong BBRs, asset sales are more likely (larger) when there is a deficit shock. The coefficient on NETREV*STRONGBBR is negative and significant (-0.199, t-statistic of 1.71). In the second column, we tabulate the results for states that were likely to shift expenses and also find that in states with strong BBRs, asset sales are more likely (larger) when there is a deficit shock. However, the magnitude of the coefficient on NETREV*STRONGBBR decreases (-0.094, t-statistic of -1.96). We then test whether the coefficients on NETREV*STRONGBBR are different across our partition, and find that the effects of balanced budget restrictions on asset sale behavior are weaker in states that are likely to have shifted expenses into the future. Thus, these results imply that accounting discretion is a substitute for asset sale behavior. 5. Conclusions This paper investigates the consequences of balanced budget requirements. Almost all states in the U.S. have some form of balanced budget provisions to curb deficit spending by state governments. The widespread use of these provisions suggests that balanced budget restrictions are perceived to be net beneficial as a control mechanism. Previous research finds that strong balanced budget requirements result in greater reductions in expenditures in periods in which a state experiences a fiscal shock. We add to this literature, providing evidence that when states face fiscal problems they tend to sell off public assets, these sales generate relatively lower proceeds, and states tend to shift expenditures to the future to meet the balanced budget provisions. We also find that there is a trade-off between asset sales and the use of accounting discretion; on average, states sell fewer assets when they use accounting discretion to avoid balanced budget requirements. These results are robust to different measures of the states’ fiscal condition and an alternative proxy for budgetary constraints. The paper contributes to the accounting literature along a number of dimensions. First, given the heightened interests in the accountability of state and local governments, our paper is timely in that we provide evidence that state governments sell assets and shift expenditures to meet budget constraints. Our paper also adds to the literature on the consequences of using budget as a control mechanism and the literature on real earnings management, both of which have largely focused on the for-profit sector. We extend these literatures by identifying the specific actions (asset sales and expenditure shifting) governmental entities engage in when they face fiscal problems and by providing evidence that such behaviors are more prominent when they also face strong balanced budget requirements. We also find results which suggest that there is a trade-off in these activities, whereby states use accounting discretion to avoid balanced budget restrictions to avoid having to sell off assets. The results of the paper should be of interest to policymakers, governmental entities, and more broadly, organizations that have enforced balanced budget provisions. They should consider whether behaviors such as asset sales and expenditure shifting are optimal, especially when asset sales may be done at fire sale prices. They should also consider whether they need additional mechanisms constraining these behaviors when enacting balanced budget requirements. References Alt, James E., and Robert C. Lowry. "Dividend government, fiscal institutions, and budget deficits: Evidence from the states." American Political Science Review 88, no. 4 (1994): 811-828. Bartov, E. (1993). The Timing of Asset Sales and Earnings Manipulation. The Accounting Review , 68 (4), 840-855. Bouwens, J., Kroos, P., 2011. Target ratcheting and effort reduction. Journal of Accounting and Economics 51, 171-185. Chaney, B.A., P.A. Copely, and M. Stone. (2002) The Effects of Fiscal Stress and Balanced Budget Requirements on the Funding and Measurement of State Pension Obligations. Journal of Accouting and Public Policy 21 (4,5): 287-313. Clemens, J. (2012) “State Fiscal Adjustment During Times of Stress: Possible Causes of the Severity and Composition of Budget Cuts.” Working Paper, University of California at San Diego Copeland, Ronald M., and Robert Ingram. "The association between municipal accounting information and bond rating changes." Journal of Accounting Research 20 (1982): 275289. Core, John, Wayne R. Guay, and Rodrigo S. Verdi. "Agency problems of excess endowment holdings in non-for-profit firms." Journal of Accounting and Economics 41 (2006): 307333. Gold, Steven D. "Recent developments in state finances." National Tax Journal 36 (1983): 1-29. Gore, Angela K. "Why Do Cities Hoard Cash? Determinants and Implications of Municipal Cash Holdings." The Accounting Review 84, no. 1 (2009): 183-207. Herrmann, D., Inoue, T., & Thomas, W. B. (2003). The Sale of Assets to Manage Earnings in Japan. Journal of Accounting Research , 41 (1), 89-108. Hou, Yilin, and Daniel L. Smith. "A framework for understanding state balanced budget requirement systems: Reexamining distinctive features and an operational definition." Public Budgeting & Finance 26 (2006): 22-45. Leone, A. J. and S. Rock. (2002). Empirical Tests of Budget Ratcheting and its effect on Manager;s Disretionary Accrual Choices. Journal of Accounting and Economics 33 (1): 43-67. Moody's. "General Obligation Bonds Issued by U.S. Local Governments." 2009. NASBO. "Budget processes in the states." Washington, D. C., 2008. NCSL. "NCSL Fiscal Brief: State Balanced Budget Provisions." Washington, D.C. , 2010. Perry, S., & Grinaker, R. (1994). Earnings Expectations and Discretionary Research and Development Spending. Accounting Horizons , 8 (4), 43-51. Petersen, Mitchell A. "Estimating standard errors in finance panel data sets: Comparing approaches." Review of Financial Studies 22 (2009): 435-480. Poterba, James M. "Capital Budgets, Borrowing Rules, and State Capital Spending." Journal of Public Economics 56 (January, 1995): 165-87. Poterba, James M, 1996. "Budget Institutions and Fiscal Policy in the U.S. States," American Economic Review, American Economic Association, vol. 86(2), pages 395-400, May. Rogers, William. "Regression standard errors in clustered samples." Stata Technical Bulletin 13 (1993): 19-22. Roychowdhury, S. (2006). Earnings management through real activities manipulation. Journal of Accounting and Economics , 42, 335-370. U.S. General Accounting Office. "Budget issues: State balanced budget practices." Washington, D.C., 1985. U.S. General Accounting Office. Balanced Budget Requirements: State Experiences and Implications for the Federal Government. Washington, D.C.: GAO, 1993. Zimmerman, J. (2001). Conjectures Regarding Empirical Managerial Accounting Research. Journal of Accounting and Economics 32: 411-427.