The Financial Crisis, Fiscal Federalism, and the Creditworthiness of US State Governments

1 Introduction

The 2008–2009 financial crisis caused US states’ economies to downturn. During this time, state governments faced pressure to increase spending, while taking in less revenue (Gordon 2012). These forces put state finances in a perilous position in the short and long term (Kiewiet and McCubbins 2014). Despite the fiscal fallout from the Great Recession, we observe a curious contradiction: as the crisis unfolded, credit rating agencies offered optimistic assessments of some states’ financial health. Table 1 shows that during the years 2008–2009, Standard & Poor’s (S&P) upgraded 11 states’ credit ratings (Louisiana twice), as opposed to three downgrades. S&P’s ratings are assessments of a state’s “capacity” to avoid default (S&P Capital IQ 2014). Given the effects of the Great Recession on state budgets, a reasonable prior would expect widespread downgrades.

What explains S&P’s optimism in states’ creditworthiness in the face of unprecedented fiscal pressures? Why did this optimism vary across states and, as we will show, over time? While other research has examined the partisan and budgetary rules to explain variation in US state fiscal policies (Lowry and Alt 2001; Johnson and Kriz 2005; Crain 2009; Kelemen and Teo 2014), this study considers an alternative approach. Despite the counter-cyclical fiscal tendencies of central governments in OECD countries, sub-national governments are procyclical spenders, meaning that these governments spend more in times of growth and less during recessions (Abbott and Jones 2013). Building on this research, we offer a theory explaining the creditworthiness of US state governments. As a result of the procyclical nature of US state governments, economic well-being can actually have a negative effect on credit ratings. Procyclical states tend to implement policies during periods of economic growth that make it politically more difficult to pay debt later. This effect, however, is mitigated if a state is dependent on the counter-cyclical federal government. Federal government assistance lessens the demand for procylical policies and weakens the connection between a state’s own economy and credit evaluations.

We focus our analysis on US state creditworthiness because credit rating changes translate into political consequences. For example, Maryland’s state treasurer Nancy Kopp warned the Senate Budget and Tax Committee in 2011 that a downgrade in the state’s credit rating could cost the state millions of dollars as investors would be less likely to purchase low-interest bonds. In particular, Kopp warned the committee that because credit rating agencies were punishing Maryland for the federal government shutdown, it appeared as though Maryland’s credit-related fate was beyond its own control (Kowarski 2013). Despite the negative consequences of credit downgrades, the relationship between states’ procyclic behavior and creditworthiness is largely unexplored. While previous research has focused on alternative determinants of creditworthiness, such as partisanship and budgetary rules, we find that these factors do not help explain the credit increases during the financial crisis.

To test our theory, we gather S&P credit ratings of general obligation bonds issued by US state governments. We first model the Great Recession as an intervention, finding that states receiving less money from the federal government were more likely to experience increases in their credit scores following the crisis, while states with robust support from the federal government experienced little change in their credit ratings. We then test whether this pattern holds outside of the financial crisis. Using fixed effects error correction models for the years 1990–2006, we observe that increases in gross state product (GSP) were actually negatively correlated with credit ratings when there were little to no changes in federal dollars flowing into a state. As the federal government increased its assistance to a state, the negative effect of GSP was mitigated. The results are consistent with our argument that economic growth incentivizes leaders to undermine their states’ creditworthiness even when controlling for institutional or partisan variance.

2 Politics, Policies, and Credit Ratings

Creditworthiness represents a government’s credibility to repay debt, even in adverse political and economic conditions. States that can convince lenders of their credibility receive economic and political benefits. Credit alleviates states’ dependence on tax revenue in the short run, allowing state officials to avoid politically difficult decisions on taxation and spending cuts. Since taxation motivates citizens to monitor governments’ fiscal policy (Paler 2013), credit can isolate officials from their constituents’ scrutiny on fiscal matters. Consistent with this argument, there is comparative evidence that government officials who do not depend on their citizens for tax revenue have longer tenures in office (Morrison 2009; DiGiuseppe and Shea 2016, 2018). We expect that credit has the same potential political impact at the US state level.

Unlike most of types of debt, sovereign debt has little to no guarantees in place to ensure repayment. Once a government has borrowed money, investors have little recourse to compel governments to repay their debts.^[1] Consequently, governments have a credibility problem when they try to convince investors to lend money. While investors can observe a government’s economic fundamentals to determine a government’s ability to repay debt, investors have difficulty determining a state’s political willingness to repay.

Individual investors do not have incentives to monitor sovereigns, given the opportunities to free ride on the information acquisition of other investors (Hauswald and Marquez 2006). As an investor, it is more cost-effective to avoid monitoring activities and merely copy the investment decisions of those individuals who actually do monitor. This opens up market space for third-party credit rating agencies, such as Standard & Poor’s, Moody’s, and Fitch, which provide graded sovereign credit risk assessments. These agencies invest in information acquisition to attempt to reduce information asymmetry between governments and investors. As a result, credit rating agencies’ “grades” are generally considered accurate assessments of a government’s creditworthiness.

Credit rating agencies’ grades are assessments of default risk. While rating agencies do not completely reveal their rating methodology, they do provide a general view of their assessment strategies, claiming that grades are a function of both governments’ economic and political capacity to repay debt.^[2] All US states have economic options to raise taxes, cut spending, sell land, or implement some other policy that provides the capacity to repay debt obligations. In theory, these tools should ensure that states can always pay their debts. In practice, these policy options are fraught with political consequences for state officials, making credit evaluations more a function of political dynamics than economic fundamentals.

How do credit rating agencies assess US states, given their place in a federal political system? In some sense, these states are independent sovereigns that have the ability to raise and manage debt without interference from the federal government. Federal governments have demonstrated that they are willing to help sub-national government in fiscal distress, and given institutional incentives, have not credibly committed to ignore the fiscal problems of state governments (Rodden 2002). This latter point is often considered a threat to a state’s creditworthiness, rather than a benefit. Wibbels (2000) argues that a federal system gives sub-national governments incentives to avoid the political costs of fiscal adjustment.^[3] Additionally, the presence of federal institutions does not necessarily prevent the occurrence of extreme fiscal crises (von Hagen 1991). In general, these arguments are consistent with a moral hazard problem. That is, the expected bailout from the federal government provides state governments incentives to be more fiscally irresponsible. However, some have argued that the moral hazard does not apply to US states, as state politicians do not behave as if a federal bailout is on the horizon, if needed (Rodden 2012). If true, we would expect limited influence of federal assistance on states’ fiscal behavior and creditworthiness.

In the next section, we present a theory of US state creditworthiness that contends that the political hazards that threaten states’ creditworthiness originate from states’ procyclical incentives, not the moral hazard of federal assistance. The procyclical nature of state government finance prompts states experiencing economic growth to undermine their own creditworthiness. Procyclical incentives, however, can be mitigated by additional federal assistance.

3 A Theory of Creditworthiness

In the face of the worst economic crisis since the Great Depression, S&P increased credit ratings for several US states. This unusual tactic prompts two important questions. First, in the face of a historical fiscal crisis, why would credit ratings agencies increase credit ratings? Second, why did the ratings increase vary across states?

We begin our theory with a simple assumption about credit rating agencies: credit rating agencies are motivated to provide accurate assessments of a state’s creditworthiness. While mistakes occur, we take credit ratings to be rating agencies’ best attempts of evaluation of creditworthiness.^[4] When evaluating a state’s creditworthiness, credit rating agencies are likely to consider two critical factors: a state’s capacity to pay its current debts and its capacity to avoid new debts in the future. When gauging a state’s economic ability to pay its debts, credit rating agencies claim that they are concerned with the economic health of a state and a state’s revenue generating processes (S&P Capital IQ 2014). Most states’ primary means of generating revenue are through taxation. Although sales tax revenue is more volatile than income tax revenue (Dye and McGuire 1991), revenue from both sources is strongly tied to the economic welfare of a state. Thus, the economic ability of a state to pay its debts is tied to the economic health of that state. While many states have balanced budget requirements such that they cannot carry debts or surpluses forward through fiscal years, these rules only apply to general funds. As Bohn and Inman (1996) point out, this leaves many other funds from which states might run deficits or surpluses. Additionally, other types of tax and expenditure limits on states provide a number of loopholes that allow states to over or under-spend their revenue streams despite seemingly powerful institutional limitations.

While economic growth may increase a state’s economic ability to repay debt by increasing tax revenue, it may also undermine a state’s political willingness to repay debt. We assume that elected officials care about reelection, which is a function of their ability to effectively allocate fiscal benefits (Abbott and Jones 2013). When there are more fiscal resources available, officials will allocate more fiscal benefits to constituents. Regular electoral competition provides politicians incentives to discount future debt problems in favor of contemporary fiscal resources. Voters fail to punish politicians for exploiting sovereign debt because citizens fail to perfectly equate current debt with future taxation (Eslava 2011).^[5] If retaining office motivates politicians, and voters are retrospective, politicians have strong incentives to prioritize the next election over long-term goals (Nordhaus 1975). Because of these incentives, we expect that states will implement procyclical policies that undermine a state’s creditworthiness. Specifically we expect that states are more likely to allocate benefits to constituents in the form of lower taxes and/or higher spending during periods of growth. While the added revenue from growth might offset the lost revenue from the higher spending and/or lower taxes in the short run, in the long run it sets the state up for fiscal burdens when the economy eventually slows down.

Procyclical policies may prove politically difficult to undo when a state’s economy is growing. Credit rating agencies anticipate that officials will eventually face politically difficult fiscal choices because of their procylical policies, increasing the possibility of new debt or even default. During economic downturns, officials are more likely to implement policies that raise revenue through higher taxes and/or lower spending. Economic downturns do not always lead to repeals of giveaways, nor is the process automatic. However, economic crises give officials the political cover to reverse the policies that endangered a state’s creditworthiness in the first place. Therefore, repealing old giveaways is most likely to happen during a recession. In sum, governments will implement policies that decrease their capacity to repay debt obligations during periods of growth and will implement policies that increase their capacity to repay debt obligations during recessions. Credit rating agencies will change their assessment of states’ creditworthiness as a result. This leads to our first hypothesis:

Hypothesis 1 Credit ratings will decrease (increase) during states’ economic expansions (recessions).

Previous research supports the notion that state political economies are procyclic (Wagner and Weber 1977; Garand et al. 1991; Sørensen et al. 2001; Clemens and Miran 2012).^[6] In short, officials undo the fiscal policies that helped produce the extra income in the first place. Using this extra income in this way would not be as problematic if the added income was unexpected. However, growth projections are incorporated into many budget projections. In addition, budgets often do not account for the possibility of economic downturns, although credit agencies expect downturns to occur every 6–7 years (Petek 2014). In sum, any potential benefit of added revenue from growth is overshadowed by the perverse political incentives elected officials have during economic expansion to undo their fiscal discipline.

A state’s capacity to pay its current debts is not just tied to the political incentives created from economic performance. States also receive significant financial support from the US federal government. While federal support is often earmarked or guaranteed for specific programs, that support lessens the need for state governments to implement fiscally irresponsible policies in two important ways. First, increases in federal money help elected officials meet demands from constituents in terms of allocated fiscal resources. Federal assistance can be directed toward projects that the state otherwise would have paid for through procyclical policies. For example, federal highway spending has been shown to replace, not add to, state highway spending (Knight 2002). Thus, the added state income as a result of growth is less likely to be funneled towards procyclical projects if more federal support is in place to address some of its necessary expenses.

Second, federal assistance often contains requirements for matching funds or at least some state allocation towards a project. This requirement limits procyclical policies by increasing the political effectiveness of each state dollar spent (CBO 2013: p. 17). For example, if a state spends $1 on a project that results in the federal government contributing $4, the state receives $5 worth of funding. This contribution effect dampens the need for officials to engage in procylical behavior, given that state dollars will go further toward political ends when there is additional federal assistance.

Federal assistance offers state officials varying control over allocation, meaning that it should take large increases in federal support to greatly weaken the procyclic effects of state economic growth. The Congressional Budget Office (CBO) differentiates between block grants, which give states a high level of discretionary control, and categorical formula grants, which are far more specific on how funds can be used (CBO 2013). While most federal assistance to the states is in the form of formula grants, even formula grants provide some discretion to the states.^[7] For example, states have the authority to set the eligibility rules for Medicaid, as long as they adhere to the broader federal rules. In addition, states can apply for waivers to allow for greater discretionary control over federal funding.^[8] Nevertheless, the federal government’s provision of financial assistance to a state government helps ensure that at least some of the basic services provided by that state government can be paid for regardless of a state’s economic health.

There is reason to believe the procyclical effects of economic growth and the effects of federal assistance on state credit ratings are conditional on one another. New money from the federal government mitigates credit raters’ concerns about states’ procyclical behavior by ensuring that outside sources of revenue can help meet fiscal demands. This mitigating effect against the potential procyclical hazards associated with growth in state economies ought to result in a positive interaction effect in which the negative effects of state economic growth weaken as federal assistance to states increases. Growth in a state’s economy is more likely to be harmful to its credit ratings if procyclical incentives are not offset by federal government assistance. As federal fiscal assistance increases in a state, the effects of that state’s own economy on its credit rating ought to diminish.

Without the assistance from the federal government, when a state’s economy grows, its credit ratings will decline. Credit raters see economic growth in states as new avenues through which leaders can be fiscally irresponsible. As the federal government increases its assistance to a state, however, the state’s own economy will play a less prominent role in determining its credit rating. Growing federal revenue in a state assures credit raters that the influence of procylical behavior are diminished, and that any debts – old or new – will have a higher probability of being addressed by the state even in the face of a shrinking economy. Without increases in federal support to a state, economic growth in a state will more likely result in procyclical behavior. We argue that federal assistance can mitigate these potentially deleterious procyclical effects of economic growth on a state’s creditworthiness. This leads to our second hypothesis:

Hypothesis 2 Federal assistance decreases the effect that states’ economies have on credit ratings.

In sum, when a state receives very little new (or even decreased) federal financial support, increases in that state’s economic growth are likely to lead to downgrades in credit ratings. The strength of a state’s economy as a predictor of a state’s credit rating is conditional on changes in the amount of money the state receives from the federal government. As a state receives more money from the federal government, the influence of a state’s economy on its credit score is more likely to be mitigated. With enough additional financial support from the federal government, a state’s own economic performance should play little role in determining its credit rating.

4 Data and Methods

In order to test our hypotheses regarding the effects of state revenues on the credit ratings of state governments, we gathered general obligation bond credit ratings as provided by Standard and Poor’s Ratings Services, starting in 1990.^[9] With these data, we first systematically examine the effect of 2008–2009 financial crisis on state credit ratings. We then examine a panel time series of the changes in state credit ratings for years excluding the financial crisis.

Before turning to the multivariate analysis, we briefly describe our data. To put the ratings increases during the financial crisis into context, we pay particularly close attention to the S&P ratings just before the onset of the crisis.^[10] For our analyses, we convert the alphabetic credit rating scale of S&P’s to a numeric scale which runs from 9 (AAA status) to 1 (BBB status).^[11] Credit ratings of state governments tend to be quite sticky for individual states with some states having stable credit ratings across all 16 years of observation. For example, Maryland, Missouri, and Utah have each maintained the highest possible credit rating (AAA) from S&P’s since well before the beginning of our time series. Other states, however, experience a great deal of variance in their credit scores over time. The most obvious example is California, which had a perfect credit score from S&P in 1990, but by 2005 had seen its credit score fall to junk bond status. While the stability of Maryland, Missouri, and Utah’s credit ratings and the volatility of California’s credit rating are somewhat extreme examples, other states do experience changes in their credit ratings. Figure 1 plots the change in credit scores for nine states in the US from 1990 to 2006. The top left panel plots the credit score of California, while the remaining panels plot the credit ratings of New York, Rhode Island, Florida, Massachusetts, Louisiana, New Jersey, Washington, and Wisconsin. The plots demonstrate that Rhode Island’s credit score spiked in the mid-1990s before returning to AA status. Alternatively, New York and Rhode Island experienced increases in their credit ratings much later in the series.^[12]

Figure 1:

Credit rating history from 1990 to 2006 for nine US states.

(A) California’s credit rating history. (B) New York’s credit rating history. (C) Rhode Island’s credit rating history. (D) Florida’s credit rating history. (E) Massachusetts’ credit rating history. (F) Louisiana’s credit rating history. (G) New Jersey’s credit rating history. (H) Washington’s credit rating history. (I) Wisconsin’s credit rating history.

In addition to the purely temporal, or “within unit,” variation in credit ratings, there is also substantial cross-sectional variation in credit ratings. Figure 2 plots the credit ratings of all 50 US state governments in 2006. California and Louisiana both had the lowest credit rating in 2006 (A). Several states also have extremely high credit ratings at this time, including North Carolina, Virginia, Maryland, Georgia, Florida, Minnesota, and Utah. There does seem to be a regional pattern in credit ratings with Southern states receiving rather high credit scores. Since Southern states receive a great deal of financial support from the federal government, this is consistent with our expectation that federal financial support helps improve states’ credit ratings. The median credit rating of US state governments over the series is a AA (7) rating status.

Figure 2:

Credit ratings of US state governments in 2006.

Yearly credit ratings provide us with the necessary dependent variable to test our theory of risk aversion in credit rating assignments.^[13] In order to measure the concepts from our key hypotheses, we require measures of state economic growth and growth in the provision of federal funds to states. We measure the economic growth of a state, which confers on the state government added revenue, using the change in the gross state product (GSP) of a state in a given year relative to the prior year, measured in trillions of dollars.^[14] We measure the changing level of federal financial support for state governments using the change in total federal dollars a state receives.^[15]

We concentrate on changes in GSP and intake of federal dollars in states for two reasons. First, our theory focuses on how gaining access to new revenue will change credit raters’ views of states’ creditworthiness. This is necessarily a question about economic growth and growth in federal support rather than raw economic size or federal support size (which would be measured by the levels rather than differences). Additionally, over the 16 years of our study, both GSP and states’ intake of federal dollars have positive time trends for virtually all states. These temporal trends can create a host of problems in time series models (Granger and Newbold 1974; Enders 2004), but differencing trending data helps convert non-stationary time series into stationary time series. Thus, focusing on differences in GSP and federal dollars to states both more accurately reflects the theoretical concepts of interest and solves potential methodological problems.

Our first analysis concentrates on uncovering the patterns underlying credit ratings agencies’ responses to the 2008–2009 financial crisis. Our models of ratings agencies’ responses to the crisis are sparse, focusing only on the key covariates from our theory. We argue that the impacts of the crisis function as a natural experiment freeing us from worries about confounding (we defend this position more below). Our second set of models examines a longer time series, excluding the years from the financial crisis, in order to demonstrate the generality of our results. In our second set of models, we also control for potential confounders. First, we control for the population in states in case federal spending or credit ratings are operating differently in the largest US states (U.S. Census Data 2014).^[16] We control for unemployment levels across states to account for the possibility that credit rating agencies are not responding to states’ fiscal conditions, but rather only to the economic conditions of a state [data drawn from Kelly and Witko (2012)]. Finally, we control for the total debt of states as a proportion of their GSP to ensure that previous debt behavior is not confounding our results. [Data drawn from Klarner et al. (2012)].^[17] Summary statistics for these key independent variables appear in Table 2.

While we are primarily interested in the influence of economic growth and changes in federal assistance on state credit ratings, which requires us to difference state GSP and intake of federal dollars, this differencing is generally insufficient in accounting for the heterogeneity in economic growth across the US states. An increase in gross state product of $10 billion in Texas in 2006 means something very different than a $10 billion increase in GSP in Rhode Island in 2006 and a $10 billion increase in GSP in Texas in 1990. The baseline size of the Texas economy has changed enormously over time, and the Texas economy is always dramatically larger than the Rhode Island economy. This heterogeneity in economic size implies that our model requires both the level of GSP and the change in GSP on the right hand side of our formula. This would allow our coefficients on the difference in state economies to estimate the influence of economic growth on credit ratings controlling for the sizes of state economies. Fortunately, the error correction specification of the general dynamic model includes both the differenced and lagged values of independent variables in its linear predictor (De Boef and Keele 2008). Accordingly, we make use of panel error correction models with varying intercepts for states to estimate the effects of changing state economies and changing intake of federal dollars on state credit ratings, controlling for the historical sizes of state economies and intake of federal dollars through lagged versions of those variables.^[18]

To summarize our expectations, we expect that when states experience decreased financial support from the federal government (reflecting a decreased investment in a state by the federal government), the relationship between the growth in gross state product and a state’s credit rating will be negative. That is, credit raters will view economic growth in states not paired with increases in support from the federal government as problematic. We expect that as the amount of money a state receives from the federal government grows, the effect of gross state product will become increasingly positive. This implies that the gross state product-federal intake interaction term will be positive. In order to evaluate our conditional hypotheses, we use OLS regression controlling for unit effects across states using varying intercepts. Thus, our analysis concentrates on the within-unit variation as the fiscal health of the states change. We first focus on the financial crisis as an unexpected intervention to model credit rating changes. We then analyze credit rating from 1990 to 2006, purposely excluding the 2008–2009 financial crisis.

5 State Credit Ratings and the 2008–2009 Financial Crisis

We began our work by pointing out an unusual empirical pattern in credit ratings: that for many states, these ratings increased in the wake of the housing and banking crisis of 2008. Our theoretical model implies that in the face of this unexpected downward shock to state economies, some states will actually experience increases in their credit ratings. This would clearly reflect the negative relationships between economic growth and credit ratings we expect to see for state governments. Moreover, our theory suggests which states in particular should see increases in their credit ratings. States receiving less money from the federal government should be most likely to experience increases in their credit scores following the crisis, while states with robust support from the federal government should experience little change in their credit ratings.

The 2008–2009 financial crisis is a unique case to examine because a large portion of the economic sector was unable to predict its onset (Rose and Spiegel 2010). Indeed, much of the public outcry following the economic downturn focused on forecasters’ inability to see the crisis coming (Colander et al. 2009; Schneider and Kirchgässner 2009). This implies that the crisis is an unexpected, downward shock to state economic growth. In other words, the crisis could neither be caused by state credit ratings, nor could state governments have proactively responded to the crisis. Because the shock to state economies was largely unpredictable to the economic system, the possibilities for strategic behavior by state governments to influence our statistical analysis are effectively minimized.

To model the crisis intervention, we create a dummy variable that indicates when the US experienced the financial crisis. We identify the years 2008 and 2009 as financial crisis years, as these 2 years experienced sustained periods of economic stagnation or decline, consistent with the National Bureau of Economic Research’s assessment of the crisis timeline.^[19] To model the intervention effect of the financial crisis, we include the financial crisis dummy variable in an error correction model predicting changes in state credit ratings. To determine the financial crisis’ conditional effects on credit ratings, we interact the financial crisis dummy variable with our measure of federal financial support to states.

Before we examine the effect of the financial crisis on credit ratings, it is important that we evaluate the usefulness of the crisis as an instrument. While it presents the empirical puzzle at the heart of our work, the crisis is a useful circumstance in which to test our theory only if it produces a decrease in economic growth without simultaneously influencing federal financial investments in the states. In order to ensure this is the case, first, we run a simple difference of means test, comparing states’ growth in non-crisis years and crisis years. The results of this comparison are presented in the left hand panel of Figure 3. As expected, in non-crisis years states experience positive growth, whereas the crisis years produced economic stagnation. This demonstrates that the crisis did have real downward effects on states’ economies. Next, in the right hand panel of Figure 3, we examine whether the crisis affected states’ intake of federal dollars. We expect that the financial crisis should have a null effect on changes in federal intake. While states would have a higher demand for federal assistance because of economic stagnation, the federal government would be in a worse position to allocate money to states because of its own fiscal troubles. Our difference of means test comparing changes in federal intake during crisis and non-crisis years support our expectations. Thus, the 2008–2009 financial crisis represents an unpredictable, fully exogenous shock to gross state products while also having little to no influence on state dollars from the federal government. Thus, the only “path” through which the crisis could influence credit ratings in our model is through GSP, making it an ideal instrument for our analysis.^[20]

Figure 3:

2008–2009 financial crisis and changes in GSP and federal intake.

Having shown that the financial crisis affects states’ GSP, but not federal intake, we analyze the effect of financial crisis on states’ credit ratings in Table 3. We utilize an error correction set up to examine both the short-term and long-term effects of the financial crisis on credit ratings (Enns et al. 2016; Keele et al. 2016). The error correction specification requires that our model include both lagged and differenced versions of independent variables as covariates, along with a lagged dependent variable on the right-hand side of our model. This is also true for dummy variables like interventions. The equation for the first model is as follows:

The coefficient of Δ financial crisis dummy variable represents the short-term effects of the crisis, while the coefficient on the crisis intervention at t–1 represents the long-term effect of the crisis (Esarey 2016). We interact the intervention with our federal dollars covariates to capture the conditional effects of the exogenous depression of economic growth on credit ratings. As the table indicates the coefficient on the Δ crisis intervention indicator is positive and statistically significant, meaning that the financial crisis increased states’ credit rating when they received no change in federal financial support. The interaction coefficient shows that this relationship is negative, meaning that as federal assistance increases, the financial crisis had less of a positive effect on credit ratings.

To help clarify the inferences from this model, we graph the short-term and long-term marginal effects of the financial crisis over the range of changes in federal intake in Figure 4. As expected, the financial crisis has a positive short-run effect for the lower range of federal intake, but as federal intake increases, the short-term effects of the crisis transition to having a statistically insignificant relationship to credit ratings. This pattern is also true in the long-term effects of the crisis. The long-term effects of the crisis are positive at lower ends of the distribution of federal dollars, and that relationship also declines as federal dollars increase.

Figure 4:

The short-term and long-term marginal effect of the 2008–2009 financial crisis on state credit ratings (2004–2012).

The results of our intervention model show that in the face of the worst recession in the United States since the Great Depression, credit raters from S&P viewed many state governments as better bets to repay their debts than they had been just the year before. The financial crisis produced dramatic slowdowns in state economies and reduced state tax revenues by huge margins. In spite of these inarguably negative consequences for states’ economic well-being, credit ratings for many states climbed in the face of these struggles. Alaska, Indiana, Iowa, Wyoming, Oklahoma, Louisiana, Montana, and Wisconsin saw better credit evaluations in 2008 than they had in 2007, and North Dakota, Texas, and West Virginia saw credit rating increases in 2009.

We contrast our findings with an alternative explanation: moral hazard. This argument suggests that states are too-big-to-fail, thus states will get bailed out by the federal government if bankruptcy is a real possibility. As a result, states will engage in riskier behavior, knowing that they will be bailed out. While we do not rule out that a moral hazard dynamic exists between states and the federal government, we do not see how this dynamic can explain the results presented here. First, we expect that this dynamic is fairly constant, at least within states, and therefore would be ill-suited to explain variation overtime. Second, while the negative interaction effect is consistent with the moral hazard argument, the other components of the model are not. In the right panel of Figure 4, we observe the Great Recession has a positive effect on credit ratings in states with low intake of federal dollars. This is consistent with our argument that these states are more likely to reverse old giveaways. It is not clear why moral hazard would explain increases in credit ratings in the face of a recession. Similarly, Federal Intake_t−1 has a positive coefficient in the absence of a crisis. Under the moral hazard argument, federal money should signal to CRAs that state government will engage in riskier behavior, and thus have a negative relationship with credit ratings. By itself, federal spending does not negatively affect credit ratings. Instead, federal spending moderates the positive effects of recessions on credit ratings (and moderates the negative effects of economic expansion on credit ratings).

Most importantly for our arguments the states that were most likely to experience an increase in their credit rating in the wake of the crisis were the states that saw decreases in their fiscal support from the national government. That is, the states that decreased their reliance on the federal government at the time of the crisis saw their credit evaluations improve. In sum, our intervention analysis is consistent with our argument, which contends that economic stagnation can improve states’ creditworthiness. Economic stagnation prevents states from accruing new debts, thus improving their long-term credit outlooks, while also providing an impetus to state governments to get their fiscal houses in order. Specifically, economic downturns force states to raise taxes and decrease spending in contrast to their procyclic behavior during growth periods, putting the state in a better position to repay debt. However, this effect is conditional on the assistance states receive from the federal government. As states’ federal intake increases, this support from the federal government nullifies the effect of changes in GSP.

6 Credit Ratings in Non-Crisis Years

While our models of the financial crisis’ impact on credit ratings would seem to corroborate our account of the interactive effects of GSP and federal financial support, these effects may be a function of the unique nature of the crisis years. A more general test of our theory would examine the interactive effects of GSP and federal financial support on credit ratings in non-crisis years.

Table 4 presents the results of an OLS error correction model predicting the credit ratings assigned to state government by S&P’s Credit Rating Agency as a function of the economic growth of states and federal financial support of the states. We control for unit effects across states in the data set by incorporating varying intercepts for the states. We estimate two models. In the first, we incorporate only gross state product, dollars of federal support, and their relevant interaction terms as covariates.^[21] In the second model, we include our covariates of interest and a set of economic control variables. We take this “model building” approach to demonstrate that our inferences are robust to concerns about model dependency. Under a variety of model permutations, our results remain stable.^[22]

Recall that in the single-equation error correction model, the left-hand side dependent variable is differenced rather than held at its level. Thus, our model predicts changes (differences) in state credit ratings as a function of lagged levels and differences of our key covariates.^[23] Interpretation of error correction model results typically separate out the short-term effects of a covariate captured by the differenced version of the covariate, from the long-term effects of the covariate captured by the lagged version of the variable (De Boef and Keele 2008; Ramirez 2009; Kelly and Witko 2012).^[24] The equation for our full model is as follows:

Our theoretical expectations and our earlier empirical results suggest that when the federal dollars flowing to a state decreases, economic growth in that state will actually be negatively related to credit ratings. We also expect that when federal dollars to a state increase, the importance of that state’s economic growth in determining its credit rating will decline. These expectations are captured in the coefficients on Δ gross state product and the interaction of Δ gross state product and Δ dollars of federal intake. As Table 4 reports, across each of the model specifications, the effect of economic growth on credit rating is negative and significant. Because the model includes an interaction term, this coefficient represents the marginal effect of economic growth when there is no change in a state’s intake of federal dollars. The interaction term describes how the effect of economic growth changes when changes in federal intake move up or down. The interaction term suggests that when federal dollars to a state decline (implying a negative change), the marginal effect of economic growth becomes increasingly negative, matching our theoretical expectations. The effects of change in these variables is estimated controlling for historical levels of the covariates.

Because the direct interpretation of coefficients from interactive models can be difficult, Figure 5 presents the change in both the short-term and long-term marginal effect of gross state product as federal intake changes.^[25] The left hand panel of plot indicates that when the federal government maintains or decreases its financial support to a state indicated by Δ federal intake, growth in that state’s GSP is negatively and significantly associated with that state’s credit ratings. If the federal government’s financial support to a state grows, however, the influence of that state’s own economic growth on its credit rating becomes statistically insignificant. For example, when the financial support of the federal government to a state grows by 0.029 ($29 million) the effect of a state’s own economic growth is an insignificant predictor of its own credit rating. If federal support to the state changes by less than $29 million, then our analysis suggests that economic growth in a state’s economy is associated with a lower credit rating for that state. The right hand panel of the figure indicates that controlling for the change in the gross state product of a state, the size of the economy or level of gross state product has no independent influence on state credit ratings, no matter how large the level of investment from the federal government. These negative and conditional short-term effects match our expectations.

Figure 5:

The short-term and long-term marginal effects of state economic growth on a state’s credit rating as growth in federal financial support to that state changes. Dotted lines are 95% confidence intervals.

Moving beyond marginal effects plots, Figure 6 plots the predicted value of state credit scores for two hypothetical states as a function of their economic growth across the range of observed economic growth in our data set.^[26] The left hand panel plots the predicted change in credit score for a state receiving a large decrease in support from the federal government, while the right panel plots the predicted change in credit score for a state receiving a large increase in support from the federal government. When a state experiences a large decrease in support from the federal government, states with slow or shrinking economies receive significantly higher predicted credit scores than do states with robust, growing economies. However, when a state receives a large increase in support from the federal government, its own credit score is essentially unmoved by changes in its own economy.

Figure 6:

Predicted credit ratings for states as a function of changes in their GSP for two difference levels of federal financial support.

Thus, our empirical analysis of state credit ratings lends support to our theory that credit ratings are largely risk averse evaluations of the probability of state default while also corroborating our results on the effects of the financial crisis on credit ratings. When states’ economies grow and state governments have access to new revenue, credit raters will realize that this will not actually help states pay down their debt. Alternatively, when state economies shrink, the procyclic tendencies of state governments incentivize officials to cut spending or raise taxes, causing credit ratings to improve.^[27]

The procyclical incentives of state leaders are serious problems for the creditworthiness of states if those state governments have few alternative sources of financial support. Yet, if states have robust financial support from the federal government, their own economies are largely irrelevant in the eyes of credit raters and state economic growth plays little role in state credit ratings.

7 Discussion

Downgraded creditworthiness can have substantial political effects on states. Without credit, states need alternative revenue strategies that may bring political controversy and electoral consequences. Despite the political importance of credit, states’ creditworthiness is threatened by procyclical political incentives. Federal assistance, however, can dampen the demand for procyclical policies, decreasing the probability that governments will undo their fiscal discipline during times of growth. Our empirical models test these expectations and find consistent support for our arguments.

There are several implications of our findings. First, the results demonstrate that sovereign creditworthiness is largely a function of political willingness, not just economic ability. Growth decreases the credit markets’ confidence in a state’s capacity to repay debt because of the procyclical incentives for states to cut tax revenue and increase spending. These procyclical policies ultimately undermine a state’s ability to repay debt when a state faces an eventual economic downturn.

Second, our findings suggest that credit rating agencies are pessimistic that state institutions designed to curb procyclical policies actually work. While many states have balanced budget requirements and rainy-day funds, electoral incentives motivates politicians to find work ways around these policies. Consistent with this, Galle and Stark (2012) find that most states do not save enough during economic expansions. Several of the authors’ prescriptions for this problem involves more federal involvement in state fiscal politics. However, our research suggests that federal intervention creates a “damned-if-they-do, damned-if-they-don’t” effect in the sovereign credit market. Given that the US federal government cannot credibly commit to stay out of the financial affairs of state governments, the markets worry that individual states face moral hazard incentives, leaving the federal government responsible to cover states’ debt problems. Nevertheless, the federal government has also not been able to credibly commit to bailing out states given a lack of opportunity and constitutional concerns. We find that the federal government can indirectly help states through its assistance policies, which dampen the demand for procyclical policies. Alternatively, state governments need to establish more durable political coalitions to implement credible budgetary constraints, based on economic triggers rather than political incentives (Galle and Stark 2012).

Our research has similar implications for comparative sovereign credit markets, particularly for federal-like systems. For example, as the European Union becomes more economically and financially integrated, the general expectation in the sovereign credit market will be that EU countries will be bailed-out. Europeans countries attempted to prevent this expectation with a “no bail-out” provision in the Maastricht Treaty (von Hagen and Eichengreen 1996). However, as exemplified by Greece’s financial crisis, this provision was toothless in the face of economic turmoil. While it is unclear whether the Greece fiscal crisis was a result of a moral hazard created by the EU (a research project worthy of attention), we are confident from our research that Greece’s creditworthiness would benefit from clear signals of EU assistance. The financial crisis revealed that domestic political turmoil in European countries, particularly in Germany, created uncertainty around whether Greece would receive assistance from other European countries. Our findings suggest that Greece’s creditworthiness suffered from the EU’s inability to credibly commit ex ante to rescue Greece. Now that the bail-out is complete and the EU has signaled that it will commit to bailouts, Greece has been able to re-enter the sovereign credit market and borrow at rates comparable to pre-crisis rates (Edwards et al. 2014). Further economic integration into the EU will likely lead member nations to face similar moral hazards towards procyclic spending as the US states, but further economic integration will also insulate some member nations’ credit ratings from their own economic ups and downs. While obvious differences between the EU and the US states exist, our results do help us form a prediction for what the EU can expect from the credit market as its own economic integration increases.

Our research provides a novel explanation for an empirical puzzle: why did US state credit ratings rise in the wake of the 2008–2009 crisis? We suggest this comes from the procyclic tendencies of state governments, mitigated by the financial assistance states receive from Washington. While our account receives considerable empirical support, one might imagine future scholarship examining alternative mitigating circumstances. While we see little evidence that static institutions like tax and expenditure limits play a role in credit ratings (see our supplemental appendix), future work might consider the ability of state governments to forecast future revenues, or the professionalism of state governments as other covariates that might modify the relationship between state economies and credit ratings. We hope to see further exploration of these conditional relationships moving forward.

We argue and show that credit markets not only discount the benefits of economic growth, but are prone to downgrade growing states. US states have perverse incentives to limit their ability to repay debts because of the political attractiveness of tax cuts and increased spending during periods of growth. A policy prescription that can be inferred from our findings is that states should follow more counter-cyclical policies, such as limiting spending and raising taxes in times of growth. This would provide states with better fiscal fundamentals when an economic downturn eventually hits. However, this prescription would be politically unpopular, as state officials have electoral incentives to follow low tax, high spending policies until fiscal realities make these policies impossible. Therefore, the policy onus is put on the federal government. The federal government must be able to continue to increase federal assistance to states to lessen the demand for procyclical policies. In short, the federal government, the original reason why moral hazard exists for US states, is also the solution to nullifying the effects of procyclical policies in the states.