Fiscal Stabilization in the United States: Lessons for Monetary Unions

Abstract

The debate about the use of fiscal instruments for macroeconomic stabilization has regained prominence in the aftermath of the Great Recession, and its relevance has suddenly increased further, after the recent Covid-19 shock. The analysis of fiscal stabilization in the United States, a monetary union equipped with a common fiscal capacity, has often informed the literature on the European EMU and could serve as a reference for its possible future reforms. This paper expands that literature in three ways: first, by measuring stabilization not only as inter-state risk-sharing of asymmetric shocks, but also as intertemporal stabilization of common shocks; second, by doing this for specific items in the US federal budget, both on the revenue and on the expenditure side; and third, by also measuring the impact of the federal system of unemployment benefits and of its extension as a response to the Great Recession. Corporate and personal income tax, on the revenue side, and social security benefits and federal grants, on the spending side, are the most effective items. The US federal system of unemployment insurance provides great stabilization in the event of a large shock, in particular when enhanced by the discretionary program of extended benefits. These findings imply that a proper design of the budget can maximize its stabilization effect, when it helps bridging the gap between higher mobility of capital and lower mobility of labor, by collecting revenues based on the income of the most mobile factor (corporate income tax) and providing support to the income of the least mobile factor (social security).

Introduction

The interaction between monetary, fiscal, and structural policies determines the way the economy grows and responds to cyclical fluctuations and shocks; an appropriate macroeconomic policy mix is crucial to ensure growth and stability. In general, the coordination of macroeconomic policies is the task of central institutions in most federal systems, where stabilization and redistribution typically operate at a federal level, while allocation is often partly decentralized (Musgrave 1959; Escolano et al. 2014).

Beyond the important role of monetary and structural policies, the focus of this work is on fiscal stabilization, in particular it tries to disentangle the role of fiscal policy conducted at the federal level in stabilizing the economy, by studying the case of the United States. In the US, monetary policy is conducted at the federal level by the Federal Reserve; structural policies are determined to some extent at the federal level, but also at the state and local levels; and fiscal policies are conducted at federal, state, and local levels. Nevertheless, the largest capacity for conducting fiscal policy to stabilize the economy remains at the federal level.

The literature on fiscal stabilization and risk sharing in the US has traditionally provided a reference for the establishment of the European Economic and Monetary Union (EMU). For example in the early 1990s Sachs and Sala-i-Martin (1992); Von Hagen (1992); and Bayoumi and Masson (1995) proposed different approaches to quantify the role of fiscal transfers for redistribution and risk sharing in the US and to draw lessons for the forthcoming EMU. More recently, Bilbiie et al. (2021) use the US experience to discuss the option of a common European Unemployment Insurance Scheme.

Yet, it is worth highlighting that due to its specific features the EMU does not represent in itself a single country or a political union1 and the option of fiscal transfers is necessarily constrained. The EMU has a different setup from the US, without common fiscal instruments for macroeconomic stabilization (Nikolov 2016; Bibow 2019). While monetary policy is fully centralized at the common level, the responsibility for fiscal policies remains entirely national, without a common fiscal capacity, in an unprecedented divergence between the main monetary and fiscal authorities (Goodhart 1998).

The analysis of fiscal stabilization in the US is especially relevant because of this specific element of the EMU architecture: centralized monetary policy vs. de-centralized fiscal policy. The budget of the European Union (EU) is small in comparison to the sum of the national budgets; it accounts for roughly 1 percent of the EU’s GDP, and it mainly performs an allocative function, so far. On average, 80 percent of the budget returns to the member states, and recent estimates show that its net redistributive and stabilization impact is much lower than in the US (Pasimeni and Riso 2018).

Proposals2 for the future of the EMU contain provisions for a euro area fiscal stabilization function to be developed over the longer term. The recent response provided by the EU to counteract the effects of the pandemic includes for the first time a common borrowing capacity to finance the “Next Generation EU” instrument, that some see as a possible embryo of a stable fiscal stabilization capacity (Guttenberg et al. 2021). It seems therefore useful to study how fiscal stabilization works in the United States – as a pertinent example of a well-established monetary union in a federal country, comprising 50 states.

This paper enhances our knowledge about the actual macroeconomic stabilization performed by fiscal instruments in the US, in three ways: first, it measures stabilization not only as inter-state risk-sharing of asymmetric shocks, but also as intertemporal stabilization of common shocks; second, it provides a detailed breakdown of specific items in the US federal budget, both on the revenue and on the expenditure side; and finally it also measures the impact of the federal system of unemployment benefits and of its extension as a response to the Great Recession.

The analysis helps draw some insights from the experience of a monetary union equipped with fiscal shock absorbers, which can guide the reflections about the future of the EMU without prejudice to the pace of its further deepening. In particular, we investigate the relative importance and stabilization impact of the US federal system of unemployment benefits and of its extension as a response to the Great Recession. The insights will be particularly useful in view of the new, deeper and more abrupt shocks caused by the COVID-19 pandemic. In drawing lessons from the US experience during the Great Recession, we always acknowledge the institutional differences between a federal state, such as the US, and a supranational entity, such as the EMU.

The paper is structured as follows: the next section discusses the problem of macroeconomic stabilization in supranational entities and reviews the literature on this topic; the following section briefly illustrates how fiscal stabilization works in the US and in the EMU. Section 4 explains the methodology and the data used in our empirical strategy. Section 5 presents the results of the estimations of the role of budgetary items for fiscal stabilization in the US, by distinguishing the role of intertemporal and interstate stabilization and by applying both a static and a panel vector autoregression (panel VAR) methodology. Section 6 elaborates further on the role of the emergency unemployment compensation enacted in response to the Great Recession. Section 7 discusses the results obtained and the last section, finally, concludes by suggesting some implications for the EMU architecture.

Macroeconomic Stabilization in Monetary Unions

The mitigation of the impact of macroeconomic shocks in supranational economic systems refers to two different functions: intertemporal and interregional stabilization. The first can be described as stabilization of symmetric shocks or common fluctuations, and the second as insurance (or risk sharing) against idiosyncratic shocks or, to be more precise, shocks having asymmetric consequences, regardless of their original nature. The first is stabilization over time; the second is stabilization across space. In studying fiscal stabilization, we refer to the capacity of fiscal policies to insure disposable income against major economic shocks; this can be measured as the capacity to cushion the effect of a GDP shock on consumption. In monetary unions, where exchange rate flexibility is not available as an automatic stabilizer, the need for macroeconomic stabilization is even greater; all the more so, if financial markets are incomplete (Fahri and Werning 2017). Such need is actually inversely proportional to the degree of business cycle synchronization among participating countries (Feldstein, 1997; Bofinger and Mayer 2007); as Afonso and Furceri (2008) show in terms of unsmoothed macroeconomic shocks to GDP.

Market mechanisms are often called to play a key stabilizing role through improved mobility of factors: capital and labor (Mundell 1973; Eichengreen 1992).3 On top of them, specific structural and fiscal policies at the national or state level can further enhance the capacity of the system to absorb macroeconomic shocks. Nevertheless, in federal states there is always a public channel providing macroeconomic stabilization through a common fiscal capacity.

Market mechanisms allowing for greater mobility of capital consist of the so-called capital market channel, the credit market channel, and the cross-border labor compensation channel, and they can provide sufficient stabilization to the extent they are stable and efficient in the allocation of resources. There is indeed evidence that in the US, for instance, they provide a great degree of stabilization (Nikolov 2016), which is nevertheless supported by public mechanisms for stabilization, such as the federal budget. As Fahri and Werning (2017) explained, even if financial markets are complete, privately optimal risk sharing is constrained inefficient. In this work, we apply a similar framework to the one Nikolov (2016) used to disentangle the relative importance of market vs public mechanisms in providing stabilization, but we apply it specifically to the different fiscal items within the US federal budget.

The experience of the Great Recession proved that the amount of risk sharing provided by markets remains generally inadequate (Berger, Dell’Ariccia, Obstfeld ). In the US, exceptional fiscal measures were necessary to complement market mechanisms and stabilize the economy. In the EMU, markets contracted and exacerbated the procyclical trend, and the ability to smooth the shock was rather reduced, because private risk sharing did not play a stabilizing role. Ferrari and Rogantini-Picco (2016) even found “a decrease in risk sharing over the period following the introduction of the Euro.” Furceri and Zdzienicka (2015) found that the degree of risk sharing in the EMU falls sharply in severe downturns; just when it is needed most, the increased inability to smooth output shocks is driven by the lack of consumption smoothing provided by private saving via the credit channel, and this is particularly true for severe downturns that are persistent and unanticipated.

Labor mobility is another market mechanism that improves the stabilization capacity of the system. Several studies found similar rates of mobility between the US and EU (Molloy et al. 2011), with a significant increase in the EMU in recent years (Dao et al. 2014). Recent research suggests that it is unlikely that cross-country migration flows will become a key driver of labor market adjustment after large shocks in the EMU (Draghi 2014), given that the regional adjustment process in Europe is already not that different from the one in the US, once controlling for country-specific factors (Beyer and Smets 2015).

Other policy instruments that can perform stabilization in a monetary union are a common monetary policy and national or state-level fiscal and structural policies. A common monetary policy can provide a first response to stabilize the economy in the event of shocks affecting the whole area, acting through the interest rate. Problems arise when the interest rate is close to the zero lower bound. In those cases, further reductions in nominal interest rates to reach equilibrium between aggregate demand and supply may be difficult; so-called “unconventional” tools are needed, but the more they are used, the lower the returns they provide (Blanchard et al. 2015), as the recent experience shows. Moreover, a common monetary policy cannot react to individual country shocks (Berger et al. 2018).4

Structural reforms help correct the structural reasons for the asymmetries in a monetary union. In particular, flexibility in product and labor markets play an important role for adjustment. However, they cannot replace the effectiveness and the speed of the exchange rate mechanism in absorbing idiosyncratic shocks and reducing asymmetries (Friedman 1953; Meade 1957). Moreover they have important short-term costs (Eggertsson et al. 2014), particularly when implemented during negative cyclical conditions (OECD 2015) when monetary policy is already constrained (Vogel 2014), posing a drag on aggregate demand (Duval and Furceri 2017).

National fiscal policies do also play a stabilization role and, given the size of national budgets compared with the common one in the EU, they are particularly important. Recent analyses (European Commission 2017a) show that the direct stabilization effects are relatively sizeable in the EU, with roughly one-third of the income absorbed by the national tax and benefit system following a shock to market income. Nevertheless, if shocks are large enough, decentralized national fiscal policies suffer from a systematic pro-cyclical bias in a monetary union, when the monetary policy is close to the effective lower bound and also when it is not (Landmann 2018), and this limits their capacity to stabilize.5

The case for a common fiscal instrument for macroeconomic stabilization in the EMU has long been discussed (Kenen 1969; European Commission 1975, 1977, 1989; Eichengreen et al. 1990; Forni and Reichlin 2001), and its relevance highlighted for both the case of asymmetric and common shocks (De Grauwe 2013; Bibow 2019). The economic rationale for common fiscal instruments for macroeconomic stabilization in a monetary union stems from the limits to market mechanisms and other instruments. These limits apply both in cases of intertemporal stabilization of common shocks and interregional stabilization of asymmetric shocks.

The reason why the two objectives of intertemporal and interregional stabilization are both important and need to be considered together is that we can think of a trade-off in the use of instruments to achieve each objective or, to be more precise, there is a trade-off in the “non-use” of a fiscal instrument for these two types of stabilization.

The less a monetary union relies on a fiscal capacity for risk sharing and insurance against idiosyncratic shocks and the more it relies on improving the adjustment capacity at the national level through structural reforms and prudent fiscal policies, the stronger the deflationary pressure that develops on the area (OECD 2015; Duval and Furceri 2017), thus the stronger the pressure on monetary policy towards its effective lower bound when counteracting such deflationary pressure.6 This leads to a greater need for a fiscal instrument for intertemporal stabilization to relieve the system from the deflationary pressure (Corsetti et al. 2019). And vice versa. The more the system pushes monetary policy toward its limits to achieve intertemporal stabilization without active support by fiscal policy, the lower the capacity to sustain all countries and free them from a deflationary pressure (Corsetti et al. 2019). This results in higher short-term costs of structural reforms and fiscal consolidation (Eggertsson et al. 2014; OECD 2015), lower effectiveness, and a greater need to compensate through a fiscal instrument for interregional stabilization. In other words, common instruments for fiscal stabilization cannot be ruled out in both functions, and the less we use them for intertemporal stabilization, the more we will have to use them for interregional stabilization.

A single fiscal instrument could also address both issues and perform both functions, but it should then include two legs: a basic arrangement for cross-country risk sharing, and a debt-issuing possibility for intertemporal stabilization. The US federal budget operates in this way by addressing both objectives, and the US system of unemployment insurance (UI) is an example of an instrument that operates on both fronts, with its mixed system of states’ responsibility in normal times and extended and emergency benefits provided by the federal system (financed through borrowing) in times of crisis.

One of the main contributions of this work is precisely to measure the fiscal stabilization capacity of the US federal budget along the two dimensions, of interstate risk-sharing of asymmetric shocks and of intertemporal stabilization of common shocks. The literature on fiscal stabilization has traditionally focused on the first dimension only, we try instead to expand it on the other dimension too. We also assess the specific stabilization impact of the US system of unemployment insurance.

Fiscal Stabilization in the EMU and in the US

Fiscal stabilization in the EMU is so far entrusted to the individual member states, with non-negligible effects (European Commission 2017a); nevertheless, stabilizing large shocks for the whole of the EMU through an appropriate aggregate fiscal stance requires a high degree of coordination, which has so far proved difficult. Leaving the formation of the aggregate fiscal stance as a sum of national fiscal policies may lead to a suboptimal aggregate stance (Hamada 1985), with a high probability of recreating imbalances. The fact that supportive monetary policy makes some fiscal space at the national level does not solve this coordination problem, and may instead lead to further distortions or misallocation.

The economic governance in the EMU has been revised considerably in recent years, with a view to achieving better coordination; however, the coordinated fiscal stabilization challenge has proved remarkable. The reason can be found in the key feature that differentiates the EU finances from those of other federations: the “reverse vertical fiscal imbalance.” The reverse vertical fiscal imbalance means that most of the central budget is so far dependent on upward transfers from the member-state level toward the top level, contrary to what usually happens in federations (Escolano et al. 2014). This provides an unstable framework for any form of common budgetary capacity.

In other words, it is extremely difficult for a central budgetary authority, which has only subsidiary budgetary powers (within the limits set by the lower level, the national one), to credibly enforce limits on the budgetary powers of that lower level. The necessary maneuvering room for fiscal policies cannot be credibly removed from the national level if the supranational one is not equipped to take it over as it is currently the case in the EMU. A fiscal counterpart to the common monetary authority could play the role of providing a coordinated and targeted fiscal impulse, minimizing distortions and maximizing impacts. There is recent evidence that the business cycles of euro area countries are increasingly correlated (Campos et al. 2017; Martínez-Martin et al. 2018), while the amplitude of the cycles differs. This implies that the need for stabilization of common shocks is becoming more pressing than the need for stabilization of asymmetric ones.

The US federal system, instead, has a considerable federal budget, which represents, on average, 20 percent of GDP, and is the main source of public expenditures in the multilevel governance of the federation (Fig. 1). The EMU has no specific budget; it is a subset of the EU, whose budget, as a matter of comparison, represents only 1 percent of GDP. The US federal government, unlike the EU, has the possibility to run deficits and borrow. Another important characteristic that differentiates the US system from the EU is that it allows for a higher degree of “cross-border” flows between states, particularly during large recessions.7 These two characteristics—common borrowing capacity and cross-border transfers—strongly determine the stabilization capacity of the US system.

Fig. 1.

Government Expenditures per Level of Government in the US (percent of GDP, 1985–2017)

Overall, the stabilization capacity of the federal budget in the US is much larger than in the EU. Parsley and Popper (2021) document that, despite increasing asynchronous business cycles and politically divide across US states, risk-sharing in the US increases and is definitely higher than among separate countries. Feyrer and Sacerdote (2013) found that, on average, between 1996 and 2011, a one-dollar shock to state income in the US is offset by a $0.20 fiscal response at the federal level; this response occurs entirely through the tax system. Pasimeni and Riso (2018) found that the same effect in the EU is thirty times smaller, given the limited size and the rigidity of the EU budget. This paper joins this literature, and in the following sections we measure in a more detailed way the specific stabilization capacity of each item in the US federal budget.

The US federal government has the power to collect taxes directly, something the EU cannot do8; it also gives direct transfers to states and individuals under several programs—from personal and corporate income taxes, to Social Security, and grants to and taxes from states.

We assess the net stabilization impact of these federal transfers in order to understand their relevance as fiscal stabilizers. In particular, we will try to estimate the relative importance of the federal system of unemployment insurance, and of its extension implemented to counteract the Great Recession. To do so, we analyze empirically the stabilization capacity of federal fiscal policies in the US. In particular, we estimate the relative importance of the different federal-to-state-government risk sharing channels that jointly contribute to the capacity of the US federal system to smooth large shocks. First, we apply a standard income decomposition approach, and then test its robustness to endogeneity issues, by means of a panel VAR approach.

Empirical Analysis of the US Fiscal Risk Sharing Channels

Methodology

Our methodological approach builds on the well-known framework for measuring risk-sharing proposed by Asdrubali et al. (1996), and complements it with a further breakdown of the data available from Nikolov (2016).9 Our estimation begins with the general Asdrubali et al. (1996) specification. In essence, this is a way to decompose the cross-sectional variance in gross state product into fractions that are smoothed over by various mechanisms or left unsmoothed. Asdrubali et al. (1996) propose a series of regressions of three balancing items to estimate the relative importance of several risk sharing channels: gross state product (GSP); gross state income (GSI); gross state disposable income (GSDI); and state consumption (SC), both private and public.

The identity $GSP=\frac{\mathit{GSP}}{\mathit{GSI}}\ast \frac{\mathit{GSI}}{\mathit{GSDI}}\ast \frac{\mathit{GSDI}}{\mathit{SC}}\ast SC$ shows that a relationship $1={\beta }_{\mathit{fi}}+{\beta }_{\mathit{tr}}+{\beta }_s+{\beta }_u$ exists10 where the beta terms are the estimates of the panel regression coefficients in:

$$\mathrm{\Delta }log{\mathit{GSP}}_t^i-\mathrm{\Delta }log{\mathit{GSI}}_t^i={\mu }_{fi,t}+{\beta }_{\mathit{fi}}\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+u_{fi,t}^i$$ 1

$$\mathrm{\Delta }log{\mathit{GSI}}_t^i-\mathrm{\Delta }log{\mathit{GSDI}}_t^i={\mu }_{tr,t}+{\beta }_{\mathit{tr}}\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+u_{tr,t}^i$$ 2

$$\mathrm{\Delta }log{\mathit{GSDI}}_t^i-\mathrm{\Delta }log{\mathit{SC}}_t^i={\mu }_{s,t}+{\beta }_s\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+u_{s,t}^i$$ 3

$$\mathrm{\Delta }log{\mathit{SC}}_t^i={\mu }_{u,t}+{\beta }_u\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+u_{u,t}^i$$ 4

The difference operator represents annual change; thus the degree of risk sharing is measured in terms of a change of each variable from the previous period.11 All variables are in constant prices and in log per capita terms. The right hand side captures changes in gross state product and similarly to Asdrubali et al. (1996) is treated as exogenous. Later on we relax this assumption by applying a panel VAR to treat all variables as endogenous.

The difference in the balancing items in Eq. (1) is due to the elements that represent smoothing of shocks to state product as it is transformed into state income, i.e., net factor income from abroad, such as dividends, rents, and wages earned abroad but spent at home. The difference in the balancing items in Eq. (2) is due to the elements that represent smoothing of shocks to income as it is transformed into disposable income, i.e., different fiscal elements such as income taxes and social support. The difference between disposable income and consumption that appears in Eq. (3) is savings or borrowing. All these elements on the left side of Eqs. (1) to (3) are regressed on changes of output. Finally, in Eq. (4), the change in consumption is regressed on the change in output to measure the part of the output shock that is directly passed on to consumption and thus not smoothed.

The cross-sectional dimension of the panels in the regressions described above represents the 50 US states (indexed by i); the beta terms are interpreted as the relative weights of cross-border risk sharing due to net factor income, fiscal transfers, savings, and borrowings on credit markets respectively; and u represents the error terms. The betas are not restricted to sum up to one, thus unsmoothing by a particular channel is allowed.

When the time fixed effects (μ) are excluded, the beta coefficients measure the amount of smoothing of both asymmetric shocks and shocks that are common to all 50 US states simultaneously. That is to say that the time fixed effects that are part of the original Asdrubali et al. (1996) methodology pick up all changes in the variables that are common to all 50 states at the same time in a given period. When these time fixed effects are excluded, such common responses are picked up by the beta coefficients.12

Next, when we include time fixed effects (μ), the beta coefficients show the amount of an asymmetric shock (i.e., the response in each state that is distinct to it and not shared with all other 49 states) that is being smoothed by each channel. This detail is important, because we can calculate the difference in the coefficients between the regressions with and without time fixed effects in order to estimate the capacity to stabilize common shocks.13

In addition to regression (2), through which we can estimate the overall risk sharing and stabilization impact of federal transfers on consumption, we use data that allow us to further detail the different channels of fiscal stabilization in the US federal budget by estimating the following set of regressions for different Xs:

$$\mathrm{\Delta }log{\mathit{GSI}}_t^i-\mathrm{\Delta }log{({\mathit{GSI}}_t^i\pm X}_t^i)={\mu }_{x,t}+{\beta }_x\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+u_{x,t}^i$$ 5

where X represents a different federal-to-state revenue or an expenditure item: a Social Security tax paid by state residents to the federal Social Security administration enters with a negative sign, while Social Security receipts received by state residents from the federal government enter with a positive sign. Note that if all revenue and expenditure items are added and subtracted from gross state income, the balance represents gross state disposable income and the sum of the set of Eq. (5) for each X becomes equivalent to Eq. (2). This stems from the nature of the construction and calculation of the national account items used, shown below, which together create the difference between GSI and GSDI and not by an explicit constraint.

The set of regressions in Eq. (5) evaluates the stabilizing impact of the following items separately, so in each of the regressions in Eq. (5) X represents one of the following:

• Federal personal income taxes paid;

• Federal corporate income taxes paid;

• Social Security contributions paid;

• Social Security benefits received;

• Federal grants to states;

• Medical benefits from the federal government;

• Supplementary income from the federal government;

• Federal excise taxes paid;

• Other federal transfers received (this includes a multitude of items as detailed below); and

• All other taxes and transfers including federal unemployment benefits received.

Federal grants to states include a variety of items, such as medical assistance, and housing and educational programs, as well as money distributed by the Federal Highway Trust Fund that funds road constructions. It is worth noting that the primary objective of these federal-to-state aid programs is not the short-term stabilization of income and consumption, but longer-term convergence goals, yet these programs may also have a stabilization role. In accordance with Asdrubali et al. (1996), medical benefits do not include Medicaid, which is administered by the states.

Supplementary income includes Supplemental Security Income (SSI) benefits to low-income people who are either aged 65 or older, visually impaired, or disabled. It also includes the Supplemental Nutrition Assistance Program (SNAP), also known as food stamps, and income maintenance benefits such as Temporary Assistance for Needy Families (TANF) and others.

“Other federal transfers received” include a multitude of diverse government support programs.14 One of them, which is of particular interest, is the Federal Additional Compensation for Unemployment; this extended benefit unemployment program was, as an exception, funded at 100 percent by the federal government, according to the provisions of the American Recovery and Reinvestment Act (ARRA) of 2009. Unfortunately, personal transfer receipts that resulted from it are not available separately and are grouped together with other items in the category “all other taxes and transfers,” which includes unemployment compensation for federal employees. Taken together the items outlined above comprise almost all taxes and transfers from the federal government to the states in the US, except for the common service that the federal government provides to the states, such as conducting foreign policy and defense.

The time period covered by the data is between 1998 and 2014. The estimation is with the Prais-Winsten procedure, which is a form of feasible generalized least squares (FGLS) with panel-corrected standard errors (PCSE). This estimation method assumes that the disturbances are heteroskedastic and there is first-order autocorrelation within panels with a common autocorrelation coefficient. There is a strong evidence for this type of panel correction based on tests performed.15 Such estimation is well suited for panels with relatively large cross-sections and relative short time periods, as discussed in Hepp and von Hagen (2013). Our Appendix 1 provides details on the tests performed in order to detect heteroskedasticity and autocorrelation among errors and motivates the choice of the estimation technique.

Data

The data on gross state product and consumption at the state level are available from the US Bureau of Economic Analysis (BEA). On the other hand, we have to construct the data for income and disposable income (i.e., income after all receipts and outlays vis-à-vis the federal government) in each state. We attribute to the state level the same share of those revenue and expenditure items that are only available at the US level.16

The observations of the main cross-state risk sharing balancing items—which are GSP, GSI, and GSDI, plus SC (both private and public)—for the 50 US states come from various sources and are calculated in the following way.

The BEA publishes data for the annual gross domestic product and personal consumption expenditure by state. In order to calculate the public sector consumption by state, we use data for state government expenditure (published as statistics on government finances by the US Census Bureau) minus the state transfers directed by state governments to individuals (which are already measured in personal consumption expenditure).

The calculation of GSI and GSDI—which is the former minus federal-to-state net transfers in the form of taxes, subsidies, or other types of benefits and contributions—closely follows the approach taken in Asdrubali et al. (1996). Gross income by state is defined as residents’ earnings (such as wages and rents), plus distributed corporate profits, plus corporate taxes. This is equal to the income approach to GDP for a particular state, i.e., all labor income (such as pretax wages, rents, etc.), all non-retained corporate income (such as earnings before interest, taxes, depreciation, and amortization [EBITDA]), and net factor income generated from across the state border. The calculation of gross state income involves using data on personal income from the BEA, and tax data from the federal government and the US Census Bureau.

GSDI is then GSI plus the net federal-to-state transfers including taxes, federal grants to states, benefits, and contributions measured in personal current transfer receipts. Federal grants to states are available from the US Census Bureau, and all personal taxes, contributions, and transfers are available from the BEA.

Intertemporal and Interstate Stabilization Through the Federal Budget

Estimations with the Static Decomposition Approach

While in the previous section we have presented the full set of risk sharing Eqs. (1) to (4),17 our focus is now on Eq. (2). The regression based on that formula measures the stabilization achieved through cross-state fiscal transfers, i.e. the stabilization properties of the federal budget. The other three equations are interpreted by Asdrubali et al. (1996) as risk sharing through cross-state factor income, such as: wages, dividend and rental income earned across state borders Eq. (1) risk sharing by saving and borrowing with the help of financial intermediaries Eq. (3) and part of the shock to output that is not smoothed and is thus directly transferred to consumption Eq. (4).18

Therefore, our focus is on the results obtained by running regression (2) and its decomposition in (5). We first estimate the stabilization effect of each channel in response to both kinds of shocks: common and asymmetric. To do so, we run regressions (2) and (5) without time fixed effects.

Table 1 shows the results. About 21 percent of shocks are smoothed through fiscal stabilization (column 1 in Table 1), both in terms of interstate risk sharing and of intertemporal stabilization, through the operation of the federal budget. This compares with 28 percent found by Poghosyan et al. (2016: Table 4). Alcidi and Thirion (2017), on the other hand, found that for the period 1995–2013 around 18 percent of symmetric and asymmetric shocks are absorbed by the federal budget. It is important to note that the main purpose of the federal budget is not to provide macroeconomic stabilization, given that it is designed to perform many other functions, for example it has a redistributive role. However, the structure of its revenues and expenditures also allows for a significant degree of macroeconomic stabilization.

Table 1.

Estimated Results without Time Fixed Effects: Asymmetric and Common Shocks

Table 4.

Estimated Results with Time Fixed Effects: Asymmetric Shocks only, % of GSP shock smoothed

It is interesting to note that federal corporate income taxes, Social Security benefits, and federal grants are the items in the federal budget with the highest stabilization potential; see Table 1, columns 8, 3, and 2, respectively. These are not the largest items in the federal budget; in particular the corporate income tax is the sixth-largest item, representing only 1.7 percent of GDP, but provides the largest stabilization effect, of about 5 percent.

Figure 2 shows the estimated coefficients of regressions (2) and (5) without time fixed effects, as well as their 95 percent confidence intervals. It is clear from the figure, and from the Wald tests reported in Table 1, that three items do not contribute to risk sharing when both common and asymmetric shocks are considered. These are other income from the federal government, federal personal income taxes, and federal excise taxes.

Fig. 2.

Estimated Coefficients without Time Fixed Effects: Asymmetric and Common Shocks. %

We can now estimate the specific risk sharing capacity of these items in the federal budget in the event of asymmetric shocks only, by adding the time fixed effects in the regressions. The results, reported in Table 2, indicate that around 10 percent of a shock19 to the GSP is smoothed through the fiscal channel (column 1, Table 2). This result is similar to previous results found by Nikolov (2016) and Alcidi and Thirion (2017), who find about 11 percent of asymmetric shocks smoothed by the federal budget for the period 1980–2013, as well as in Poghosyan et al. (2016), who find that risk sharing through fiscal means smooths about 12 percent of an asymmetric shock. Parsley and Popper (2021) find a somewhat higher coefficient, of about 16%, using annual data from 1997 through 2015.

Table 2.

Estimated Results with Time Fixed Effects: Asymmetric Shocks Only

We then estimate the stabilization effect of each fiscal revenue and expenditure item; these are shown in the remaining columns of Table 2 and Fig. 3. They have a varying degree of stabilization capacity. For example, Social Security and medical benefits (except Medicare, which is partially state administered, and its federal portion is included in the category “federal grants to states”) smooth between 2 and 3 percent of an income shock, despite having not been primarily designed for stabilization purposes. Interestingly, the different degree of stabilization effect is not correlated with the size of the item (in terms of percentage of GDP).

Fig. 3.

Estimated Coefficients with Time Fixed Effects: Asymmetric Shocks Only, %

The main items that contribute to year-to-year consumption smoothing after asymmetric shocks to income are Social Security benefits received, medical benefits received, and personal income taxes paid, each smoothing between 2 and 3 percent of the output shock. With the exception of personal income taxes, these are not among the largest items in the federal-to-state tax and transfer realm. The combined amount of the risk sharing roles of Social Security benefits received at 2.8 percent and Social Security benefits paid at -1.0 percent is positive, meaning that the Social Security system in the US has a mild positive risk sharing role of 1.8 percent.

The only item that does not contribute to risk sharing after asymmetric shocks is “other transfers received”, which refers to other income from the federal government” (see above for a description of its components). It is noteworthy to mention that the role of the different fiscal items for risk sharing after asymmetric shocks is quantitatively similar to the one found in table II of Asdrubali et al. (1996), even though the time period covered in their paper is between the 1960s and 1990. This points to a relatively stable role of the budgetary items for cross-state stabilization after asymmetric shocks.20

The literature has mainly focused on the measurement of interstate risk sharing only, therefore including time fixed effects. Our methodology also follows this approach, but in addition we try to estimate intertemporal stabilization by subtracting from the overall coefficient derived above. This methodology has been used by Alcidi and Thirion (2017), see Eq. (4) on page 3. Regarding the treatment of time fixed effects, we follow Poghosyan et al. (2016).21

We can now calculate the difference between the estimation, including both common and asymmetric shocks, and one that includes only the smoothing of asymmetric shocks; the results give us the stabilization effect against shocks common to all US states at the same time for each channel. We can then plot the results for the specific stabilization effect of each item in the US federal budget in the case of common and in the case of asymmetric shocks, adding also a third dimension, which is the size of the item in the federal budget (Fig. 4).

Fig. 4.

Intertemporal and Interstate Stabilization through Fiscal Channels in the US

On the horizontal axis, Fig. 4 shows the intertemporal stabilization effect of each channel against common shocks; the vertical axis shows the amount of interstate risk shared against asymmetric shocks; finally, the size of the points represents the size of each item in the federal budget (as a share of GDP).

We can see how the different items in the US federal budget have different effects in terms of stabilization, and these effects are measured along two dimensions: cross-country stabilization of asymmetric shocks (on the vertical axis) and stabilization of common shocks over time (on the horizontal axis).

The first thing we observe is that the stabilization capacity of each item is not directly related to its size, meaning that even small items can have relevant stabilization effects. Some items stand out for their effect, which is statistically significant along one of the two dimensions. On the one side, Social Security benefits together with federal personal income taxes are the most effective items in the federal budget for providing interstate risk sharing, i.e., stabilization against asymmetric shocks. On the other side, federal corporate income taxes, although quite small in terms of overall size, are the most effective item in the federal budget for providing intertemporal stabilization against common shocks; their small size implies they are also one of the most efficient ways to provide stabilization. Section 7 discusses further these findings.

Estimations with the Panel VARs Approach

In order to dispel the doubt that our estimations obtained through a standard static approach in the previous section may be subject to an endogeneity bias, we replicate the analysis by applying a panel VAR approach. Specifically, Eqs. (1)-(4) may lead to the doubt that they do not constitute a well-defined causal experiment, but report a correlation analysis that is subject to endogeneity. For example, using the underlying variables in Eq. (2), a shock to output will cause a corresponding response by the fiscal authorities in providing fiscal stabilization through various fiscal items, which however could also simultaneously cause output to change.

This problem has been addressed in the literature by Asdrubali and Kim (2004), by using panel VARs, in order to make assumptions about what causes the output process to change and capture the dynamic interactions between the different channels. The usefulness of a VAR model in our case is that it treats all the variables in the system, including output, as endogenous, and allows for dynamic feedbacks among those variables. It can address the undetermined endogeneity problem, by assuming contemporaneous or lagged influence of one variable on another, based on theoretical insights and empirical evidence.

In particular, we want to test the assumption that budgetary items that provide stabilization respond contemporaneously to output shocks, i.e. within the year as our time profile is annual changes, while the impact of their changes on output variations comes only with a lag. This is based on the intuition that the immediate reaction of the fiscal authorities, i.e. within the year of the shock, takes some time to be reflected in output, i.e. in the following year, as consumer spending recovers from the initial output shock with the help of fiscal risk sharing.

We group the variables in (1)-(4) into an endogenous vector $X_t^i$ where:

$$X_t^i=\left[\begin{array}{c}\mathrm{\Delta }log{\mathit{GSP}}_t^i\\ \mathrm{\Delta }log{\mathit{GSP}}_t^i-\mathrm{\Delta }log{\mathit{GSI}}_t^i\\ \mathrm{\Delta }log{\mathit{GSI}}_t^i-\mathrm{\Delta }log{\mathit{GSDI}}_t^i\\ \mathrm{\Delta }log{\mathit{GSDI}}_t^i-\mathrm{\Delta }log{\mathit{SC}}_t^i\end{array}\right]$$

The ordering in the vector $X_t^i$ guarantees the assumption of the contemporaneous nature of state output vis-à-vis the other endogenous variables and their lagged effect on output. This is used in an identification strategy of the impulse response functions (IRFs) that presents the individual responses of each risk sharing item to a shock in output. In order to construct the IRFs we need a moving average representation of the following structural VAR22:

$$G_0{X}_t^i=d_t+G\left(L\right)X_{t-1}^i+e_t^i$$ 6

where the structural parameters are estimated from the estimation of a reduced form VAR representation23 as in Eq. (2). $G(L)$ is a matrix polynomial in the lag operator L, d is a column of time fixed effects and the disturbance terms $e_t$ are serially uncorrelated and $var\left(e_t^i\right)=\mathrm{\Lambda }$. The variance–covariance matrix $\mathrm{\Lambda }$ is assumed to be diagonal, so that there is no mutual correlation among the structural disturbances. The contemporaneous structural parameter matrix $G_0$ serves to identify the impulse responses and is a lower triangular matrix with 1 s in the principal diagonal.

The overall fiscal risk sharing channel is small compared to the other two principal channels: capital and credit markets, as shown by Fig. 5, which is also broadly in line with the general findings of Asdrubali and Kim (2004).

Fig. 5.

Impulse response from the standard risk sharing channel

Our strategy now is to modify the endogenous vector $X_t^i$ in order to separate its third element (the fiscal risk sharing channel) into two distinct elements: each fiscal risk sharing item separately and all other fiscal risk sharing items together. Specifically the endogenous vector $X_t^i$ looks like this:

$$X_t^i=\left[\begin{array}{c}\mathrm{\Delta }log{\mathit{GSP}}_t^i\\ \mathrm{\Delta }log{\mathit{GSP}}_t^i-\mathrm{\Delta }log{\mathit{GSI}}_t^i\\ \mathrm{\Delta }log{\mathit{GSI}}_t^i-\mathrm{\Delta }log{({\mathit{GSI}}_t^i\pm X}_t^i)\\ \mathrm{\Delta }log{({\mathit{GSI}}_t^i\pm X}_t^i)-\mathrm{\Delta }log{\mathit{GSDI}}_t^i\\ \mathrm{\Delta }log{\mathit{GSDI}}_t^i-\mathrm{\Delta }log{\mathit{SC}}_t^i\end{array}\right]$$

Here the strategy is to identify the IRFs in the same way by ordering the elements in the endogenous vector in such a way as to ensure a contemporaneous effect of the output disturbance on both fiscal items included (each individual item separately and all remaining items together), while not allowing the opposite. Equation (6) has to be run separately for each fiscal item, with the modified endogenous vector, where the third element captures the effect of each of the fiscal items individually and the fourth element captures the effect of all the remaining fiscal items together.

Figures 6 and 7 present the impulse response of the various individual fiscal items to a shock in GSP. The IRFs are scaled to match a cumulative shock in GSP after 5 years that equals 100% so that they can be directly compared to the static fiscal risk sharing estimates obtained with Eq. (5).

Fig. 6.

IRFs from panel VARs without Time Fixed Effects: Asymmetric and Common Shocks

Fig. 7.

IRFs from panel VARs with Time Fixed Effects: Asymmetric Shocks only

Similar to the results presented in Table 1 and Figs. 2 and 6 shows that corporate income taxes, federal grants, social security benefits, medical benefits and all other taxes and transfers provide the largest share of stabilization against common and asymmetric shocks. Federal personal income taxes, which were not statistically significant in the static estimation, provide instead a positive and statistically significant stabilization effect in the first year after the shock. Social security contributions, which in the static estimation were found to be a dis-smoothing channel, have the same property in the first year for the dynamic estimation, but become a positive risk sharing item in the second and third year after the shock. At the same time social security benefits turn negative, so the overall impact of the social security system (benefits minus payments) has a positive risk sharing property until the fourth year after the shock.

As in the case of common and asymmetric shocks, the results of panel VARs with time fixed effects (measuring stabilization against asymmetric shocks only—see Fig. 7) leads to similar results to those produced by the static approach (see Table 2 and Fig. 3). Social security benefits, medical benefits and personal income taxes provide the largest contribution to stabilization after asymmetric shocks. The former two are statistically significant until the third year after the shock to GSP, while the latter is statistically significant only immediately after the shock. Federal grants to states is not statistically significant, unlike in the static estimation, while all other taxes and transfers (rest in Fig. 7) provides statistically significant risk sharing until the end of the first year after the shock.

Table 3 shows that the panel VAR IRF estimates of the size of the fiscal stabilization channels in response to common and asymmetric shocks are very similar to the estimates obtained through the static approach. For example, the static estimate of corporate income taxes after a common and asymmetric GSP shock is 5% while the panel VAR estimates the IRF of corporate income taxes to a common and asymmetric shock in GSP is found to be 4% immediately, 3% one year after the shock and not significant thereafter. The other fiscal items have similar orders of magnitude across the static and panel VAR estimation.

Table 3.

Estimated Results without Time Fixed Effects: Common and Asymmetric Shocks, % of GSP shock smoothed

Also in the case of stabilization of asymmetric shocks only, the results for the various channels after estimated with a panel VAR remain similar (Table 4). For example, social security benefits are estimated to smooth about 3% of the GSP shock in the static regression based on Eq. (5) and 1.2% immediately, 0.8% after one year and 1% after two years, in the VAR approach.

The panel VAR approach allows us to conclude that the static estimates do not suffer from a significant bias, i.e. from the fact that GSP could be affected contemporaneously by the change of the fiscal items that its shock has caused. This conclusion is based on the similarity of the results obtained through the two approaches, the static one and the structural panel VARs. The latter explicitly tests that the fiscal risk sharing items do not affect GSP contemporaneously, thus making it strictly exogenous within the model. The estimates of the static approach and the panel VARs may not be exactly equal, but the relative importance of the various risk sharing channels remains the same.

The Role of Emergency Unemployment Insurance During the Great Recession

Unemployment Insurance in the United States

The US system of unemployment insurance is a joint federal-state program that provides direct support to eligible workers to sustain their income during a spell of unemployment. The overall objectives of the program are: to provide workers who lose their jobs with partial wage replacement, to help maintain purchasing power and provide macroeconomic stabilization, and to prevent dispersal of the trained labor force by promoting reemployment. The only condition that the states have to fulfil is to have an unemployment benefit scheme in place, but large differences exist in terms of coverage, replacement rates, and generosity of the benefits (Fischer 2017).

Under the Federal Unemployment Tax Act (FUTA), the tax rate on employers is 6 percent on the first $7,000 of each worker’s annual wage. However, states that are compliant with all federal rules can lower this rate to a minimum of 0.6 percent, which finances administrative costs and the federal share of the extended benefit program (Whittaker and Isaacs 2016). The extended benefit program is triggered under specific conditions that are linked to an increase in the unemployment rate above certain thresholds, which provides 13 additional weeks of benefits on top of the standard 26 weeks. There are several layers of extended benefits, each triggered by a higher threshold; some of these extensions are mandatory, others are voluntary for the state.

Interestingly, while in principle the extended benefit program is jointly paid for at the state and the federal level, in practice during the deepest recessions the federal level contributes more and the system leads to permanent transfers. As an illustration of this fact, O’Leary (2013) shows that the federal share of the total unemployment benefit cost increases enormously during the deepest recessions. This happens because under the extended benefit program, if a state unemployment benefit scheme is underfunded and cannot afford the full coverage, the state can borrow from the federal level; the borrowing then should be paid back in two years, otherwise the compulsory federal tax rate of 0.6 percent under the FUTA can be increased by 0.03 percent.

As Fischer (2017) notes, this incentive is extremely weak to prevent moral hazard, so that states have a clear preference for keeping the unemployment scheme underfunded by maintaining a low tax rate so as to avoid relocations of companies to other states. This actually leads to the key rationale for having an unemployment scheme at the highest level of government in a federation: higher mobility of capital than labor implies the likelihood of a race to the bottom among states on corporate tax rates, and either lower standards for protection of the unemployed or structural underfunding.

Vroman (2010) analyzed the stabilization effect of UI in the US during 2007–10 and found that both regular and extended benefits had a multiplier effect of 2.0. The stabilizing effect of the regular UI program estimated by Vroman was about one-tenth of the real GDP shortfall caused by the 2007 recession. For the three separate components of UI, the proportional gap-closing effects of the program during 2008Q3–2010Q2 were as follows: increased regular UI benefits = 0.105; extended benefits = 0.085; and increased UI taxes = –0.007. On average, the UI program closed 0.183 of the gap in real GDP caused by the recession. For the Great Recession, the UI program has provided stronger stabilization of real output than in many past recessions.

Moreover, during the Great Recession, the US federal government launched an additional program in July 2008, Emergency Unemployment Compensation (EUC08), which ended in December 2013 (Whittaker and Isaacs 2016). This is one of the most relevant discretionary actions taken by the US federal government to counteract the effect of the shock caused by the Great Recession. As Cashin et al. (2018) note, the impulse provided by fiscal policy in the US during the Great Recession was significantly stronger than in previous contractions. We will therefore try to assess the net stabilization effect of this program, both in terms of interstate risk sharing and of common intertemporal stabilization.

The Role of the Emergency Unemployment Compensation (EUC08)

The conventional channels used to measure the degree of risk sharing through federal support to states and then through personal transfer receipts do not separate out the role of ad hoc measures that were enacted as a response to the Great Recession. In particular, one of the most important policy actions taken by the US federal government to counteract the big shock caused by the Great Recession was the ARRA of 2009. The ARRA mandated full federal support to the extended benefit unemployment program (which is available when a state is experiencing a sharp rise in unemployment); unfortunately data on personal receipts for this particular program are not available.

The US Department of Labor, however, publishes data related to the EUC08 program, which was adopted in July 2008 and expired in December 2013. The EUC08 was a federally funded response to the common crisis shock across all states and as such represents an important instrument for macroeconomic stabilization through fiscal means.

In order to test the impact of the EUC08, we adapt regression (2) in the following way:

$$\mathrm{\Delta }log{\mathit{GSI}}_t^i-\mathrm{\Delta }log{\mathit{GSDI}}_t^i={\beta }_{\mathit{tr}}\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i+{\beta }_{\mathit{tri}}\ast \mathrm{\Delta }log{\mathit{GSP}}_t^i\ast \mathrm{\Delta }log{\mathit{Cl}}_t^i+u_{tr,t}^i$$ 7

where Cl represents the number of initial claims made from each state to the EUC08 program each year between 2008 and 2013, thus the change in log claims, which appears in (7) approximates the percentage change in claims, when the value of the latter is small. In this way, we measure the marginal impact of the various cross-state fiscal stabilization items in interaction with the increase of unemployed people who were eligible to benefit from the EUC08 program, which is proxied by the change of successful initial claims per state. The rationale of this approach is to condition the amount of fiscal risk sharing on the degree of hardship that each state suffered as a result of the common shock. Note that regression (7) does not contain time fixed effects and thus measures the impact of common and asymmetric shocks together.24

Table 5 shows the estimated coefficients of regression (7) and, in columns 5 and 6, the difference between the coefficients of two regressions—when time fixed effects are included and excluded, so as to measure the fiscal stabilization effect against the common shock. As shown by the results, the total amount of risk sharing through federal-to-state revenue and expenditure items is influenced by the inclusion of an interactive term that captures the need for using the EUC08 program in each state. That interactive term is positive and statistically significant (0.43 in column 4, Table 5) while the coefficient of fiscal risk sharing drops by close to 6 percentage points (when the number of initial EUC08 claims is evaluated at its average across 50 states between 2009 and 2013; see column 7, Table 5). Note that the marginal effect of federal-to-state net transfers has to be evaluated at a particular level of the number of EUC08 claims that enter regression (7) in interaction with GSP. Therefore, a simple subtraction of the coefficient in column 2 from column 4 will not give the result reported in column 7.

Table 5.

Estimated Stabilization Effect of the Emergency Unemployment Compensation

The inclusion of the interactive term picks up the information contained in the number of initial claims to the EUC08 program (a proxy for state needs for additional support). Therefore this suggests that during the years that it was enacted by all fiscal channels, the impact of the emergency unemployment compensation was to change the smoothing of the common and idiosyncratic income shock for the average state by around 6 percentage points, i.e., this is the difference in the marginal impact of the change in GSP on the change of the left-hand side variable in Eq. (7) with and without interaction. This means that when the role of this crisis-induced policy measure is explicitly taken into account, the contribution of total combined fiscal channels of cross-state risk sharing falls by 6 percentage points, pinpointing a value for the stabilization role of the EUC08 program.

It is also noteworthy that dropping the time fixed effects, which in effect represents measuring total stabilization in response to common shocks and asymmetric shocks together, makes the coefficient change much less when the interactive term is included—5 percentage points versus almost 12 percentage points without the interactive term (see columns 5 and 6, Table 5). This is a result of the EUC08 program being oriented toward common shocks, thus the interactive term picks up the response to common shocks and becomes positive and statistically significant (compare the coefficient of the interactive term in columns 3 and 4, Table 5).

Figure 8 presents an alternative way to look at the marginal effect of the EUC08 program on fiscal risk sharing. It shows the marginal effect on the beta coefficient that represents the relationship between shock to state output and the difference between state income and state disposable income (the fiscal risk sharing channel), together with its estimated 95% confidence interval, against the range of changes in EUC08 claims in each state during the program’s duration.

Fig. 8.

Marginal Effect on the Fiscal Risk Sharing Coefficient during EUC08, Common Shocks, 95% Confidence Intervals

The depicted marginal effect corresponds to the results on fiscal risk sharing after common shocks shown in column 5 (which is the difference between columns 2 and 1) of Table 5. The estimated fiscal risk sharing is not statistically different from zero when interacted with the change in EUC08 claims when the latter is negative.25 However, when the number of unemployed who benefitted from this additional government program for income support rises, the estimated fiscal risk sharing coefficient becomes statistically significant. Moreover, there is a positive relationship between the increase in EUC08 claims and the intensity of the fiscal risk sharing channel. As designed, the EUC08 program provided more consumption smoothing in states with more unemployed and thus eligible workers. It also only provided this service in the case when shocks to output were common across the 50 states and thus the private risk sharing channel of inter-state labor mobility could not have provided the desired stabilization.

These results prove the effectiveness of an ad hoc, contingent fiscal measure adopted by the US federal government in stabilizing the large common shock of the Great Recession among all 50 US states. The joint federal-state program of unemployment insurance, in spite of being permanently underfinanced in its state-level component and being prone to moral hazard at the state level, was able to provide very significant intertemporal stabilization during the Great Recession, mainly for its capacity to address common shocks. That is in fact the main rationale for having such a program, which can be scaled up in the event of large macroeconomic shocks, as it was the case then.

Discussion

With an average size of about 20 percent of GDP, over the period considered, the US federal budget is able to stabilize on average more than 21 percent of macroeconomic shocks through its system of federal-to-state net transfers, including interstate stabilization of asymmetric shocks (about 10 percent) and intertemporal stabilization of common shocks (about 11 percent). Different items in the federal budget have different stabilization properties, independent of their size; for example, the corporate income tax represents a small item in the budget (1.7 percent of GDP), but provides the largest stabilization effect (5 percent of shock smoothed in the static approach, while in the panel VAR it is 4% immediately and 3% after one year).

While Social Security benefits, federal personal income taxes, and medical benefits from the federal government are the most effective items for interstate risk sharing, i.e., stabilizing against asymmetric shocks; federal corporate income taxes are the most effective item for providing intertemporal stabilization against common shocks, and their small size implies they are also one of the most efficient ways to provide stabilization. This finding is consistent with similar results in the case of Germany, where Buettner and Fuest (2010) found that about 8 per cent of an initial shock to gross revenues is smoothed on average by corporate income tax, and that this effect tends to increase during cyclical downturns.

Corporate income taxes are generally collected with longer lags compared with other taxes, as it is often observed in several jurisdictions. This fact is indeed consistent with the finding that this item in the federal budget provides sensible stabilization over time, while it is not particularly relevant for inter-state risk sharing. Both methods applied in the empirical analysis confirm such result. The other item which provides relevant stabilization, on the revenue side, is personal income tax, but its effect materializes mainly in the case of asymmetric shocks. This is consistent with the predominantly redistributive nature of this tax.

On the expenditure side, we find that the net stabilization effect of the Social Security system is positive, when asymmetric shocks are concerned. This can be seen by adding the positive effect of Social Security benefits received, which smooths close to 3 percent of idiosyncratic shocks, and the negative effect of Social Security taxes paid, which has a small dis-smoothing role, adding an additional 1 percent of the output shock to consumption. This is not surprising, as workers might be tempted to consider early retirement after a negative localized shock to income, or to postpone retirement plans in the opposite case. At the same time, Social Security taxes are usually proportional or a fixed sum for a given income bracket above the wage base, so they may not have a stabilization role.

These findings are robust to different methodologies, in particular testing for a possible endogeneity bias. In a structural panel VAR, we obtain the identification by imposing restrictions on the impact of GSP on the fiscal items and vice versa. These restrictions force a simultaneous effect of GSP on the fiscal items, but not vice versa. The dynamic approach finds results of the stabilization effects that are very similar to those found by the static one.

The fact that the two items in the budget with the highest stabilization effect are corporate income taxes, on the revenue side, and social security benefits, on the expenditure side, suggests a reflection on the possible mechanisms at play. Corporate income is the most mobile tax base, the one that moves easily and quickly across states. Social security, instead, is the type of expenditure that tries to mitigate the adverse effects of a shock on the income of those who cannot relocate that easily. The combination of these two opposite fiscal flows, i.e. the collection of revenues from the most mobile factor and the payment of benefits to the least mobile one, may represent the key ingredient that allows the federal budget to stabilize consumption, in the event of a large shock.

The sizeable stabilization capacity of ad-hoc transfers, to top up the state-based US unemployment insurance after a common shock, prove the need and the effectiveness of a federal-state transfer system. Such transfers can fill in gaps occurring in the state unemployment schemes during bad times for the Union as a whole. This serves to keep the integrity of the single market for labor in the US, avoiding scars caused by long-term unemployment. We find that such supplementary ad-hoc federally-funded system works mainly after common shocks and does not provide significant stabilization after asymmetric shocks. This is because in a fully-fledged monetary union, such as the US, market mechanisms, such as cross-border labor mobility, work more effectively to adjust to state-specific shocks. Moreover, to the extent that cross-border labor mobility cannot fully clear the shock, social security benefits paid by the federal budget guarantee further income smoothing in the affected states.

On top of this, the fact that corporate income taxes are partly collected at federal level, makes the revenue side of the budget less volatile and less sensitive to the high mobility of this tax base, therefore increasing its stabilization capacity. On the revenue side, the mechanism at play has to do with tax elasticities over the business cycle, but also with the mobility of the tax base across states (in the US) or countries (in the EU). Tax elasticities over the business cycle tend to increase in the short run, but to be rather stable over the long run (Boschi and d’Addona 2019). Corporate income tax has quite a high long-term elasticity to the business cycle (Mourre and Princen 2019), however, what matters even more in the setting of a multilevel governance system (such as a federation like the US, or a monetary union like the EMU) is the heterogeneity of tax elasticities across states or countries.26 Such heterogeneity implies that the mobility of the tax base could destabilize the revenue side of the state/country budget, so collecting those taxes at the federal/common level shields the budget from the volatility induced by the cross-border mobility of that tax base.

The key point which makes a difference in terms of stabilization in a multilevel governance system is the difference between the mobility of the tax base in a monetary union with a central fiscal authority, as in United States, and the mobility of the tax base in a union comprised of separate national fiscal authorities, as in the euro area.

Conclusions

This paper has shown that the US federal budget allows for several channels of fiscal stabilization, given its flexible structure and its capacity to borrow. In spite of not being designed to primarily perform macroeconomic stabilization, the combined structure of its revenue and expenditure sides allows for a significant degree of stabilization. We contribute to the literature of fiscal stabilization in the US in three ways: we measure stabilization not only as inter-state risk-sharing of asymmetric shocks, but also as intertemporal stabilization of common shocks; we do this for specific items in the US federal budget, both on the revenue and on the expenditure side; and we also measure the impact of the federal system of unemployment benefits and of its extension as a response to the Great Recession.

The results obtained in the study of fiscal stabilization in the US may offer a reference for the EMU, but we have to acknowledge that the latter is not a full political union and a federation like the former. This implies that in the EMU the option of fiscal transfers is inherently constrained. Nevertheless, to the extent that the experience of an economic and monetary union that is also a full federation and political union (such as the US) can be a reference, the findings of this paper suggest that channels of fiscal stabilization through the federal budget are relevant. In particular, there is a case for addressing both common and asymmetric shocks, but the instruments chosen have different impacts on the capacity to address these distinct stabilization needs.

In the case of the EMU, the efficiency of the design of such a common budget becomes absolutely crucial, given the limited size of a common budget and the strong constraints in terms of borrowing capacity. This work has shown that a careful design of the revenue and expenditure sides can maximize the stabilization capacity. The structure of a federal (or common) budget, and in particular the composition of its revenue and expenditure sides, determines a great deal of its stabilization capacity, beyond the obvious relevance of its size. On the revenue side, corporate income taxes collected at the federal level are the single most effective and also most efficient item for providing stabilization, given that even with a smaller size than other items they can provide more important effects, mainly against common shocks. On the expenditure side, the most effective item for achieving stabilization against asymmetric shocks is Social Security benefits.

Even a small common budget can maximize its stabilization potential by collecting corporate income taxes at the federal level and then paying benefits to individuals in the form of an unemployment benefit. The key is to bridge the gap between higher mobility of capital and lower mobility of labor, by collecting revenues based on the income of the most mobile factor (corporate income tax) and by providing support to the income of the least mobile factor (social security). If instead a specific and contingent stabilization function is considered, the discretionary program of extended unemployment benefits, mainly funded by the US federal budget and supported by the borrowing capacity of the federal government, proves a powerful example of a timely and effective stabilization instrument.

In practical terms, these findings highlight the relevance, the potential, and the complementarity of two specific projects: the work on a European common consolidated corporate tax base and on a common European unemployment insurance scheme. A common scheme of unemployment insurance could be financed by the regular collection of a share of the corporate income taxes collected by the Member States, and it could pay unemployment benefits. It would not need to substitute the national unemployment benefit schemes, but it could rather complement them, in the event of large shocks. The automatic stabilizers associated with these schemes would act more strongly at the European level, than what they do at the national level, because the revenues of a European scheme would be less affected by cross-border mobility of the tax base and the benefits could be paid to the unemployed regardless of their cross-border mobility.

The scheme could be activated in coherence with monetary policy conditions and once a given threshold is reached. Such threshold can refer to the unemployment rate or to a combination of other indicators (i.e. GDP, inflation, policy interest rate). In the event of a common shock large enough to exhaust the fund’s endowment, a borrowing capacity would allow it to provide intertemporal stabilization. Further research on this topic could focus on the possible design of such a common fiscal capacity, trying to estimate its efficiency and effectiveness in providing macroeconomic stabilization.

Appendix 1 Testing the Panel Data Error Structure for Serial Correlation within Panels and Heteroskedasticity Across Panels

We run a test for serial correlation in the idiosyncratic errors of the model discussed by Wooldridge (2002). Serial correlation in the disturbances can bias the standard errors and decrease efficiency. Under the null hypothesis of no serial correlation of the errors, the residuals from the regression of the first-differenced variables should have an autocorrelation of –0.5. This implies that the coefficient on the lagged residuals in a regression of the lagged residuals on the current residuals should be –0.5. We perform a Wald test on this hypothesis. The results are shown below.

As is evident from Table 6, there five cases where fiscal breakdown regressions could be run without correction of serial correlation of errors within panels: federal grants, Social Security benefits, medical benefits, supplementary income, and corporate income taxes. In the nine other cases the null of no serial correlation within panels cannot be rejected. In view of harmonization across specifications we decided to use the Prais-Winsten PCSE estimator with a correction for serial correlation within panels in all specifications. As a robustness check we have run the regressions for these five items also assuming no serial correlation of error within panels. Estimated coefficients do not differ substantially as shown below in Table 8.

Table 6.

Results of Tests for Serial Correlation of the Errors Following Wooldridge (2002)

Table 8.

Estimated Results with Time Fixed Effects: Asymmetric Shocks Only, With and Without Correction of Serial Correlation among Errors

In addition, we also run a test of error variance that is specific for each cross-sectional unit (in our case, each state). There is relatively strong evidence of some form of heteroskedasticity among panels, as the 50 states differ widely in their geographic and socioeconomic characteristics. This is confirmed in the following table, which shows results from a modified Wald statistic for group-wise heteroskedasticity in the residuals. This is the reason why we decided to use the Prais-Winsten PCSE estimator, which is an alternative to the feasible generalized least squares estimator used in cases when disturbances are assumed to be heteroskedastic and contemporaneously correlated across panels (Table 7).

Table 7.

Results of Tests for Heteroskedasticity of the Errors Following Greene (2000)

In response to the performed test, Table 8 shows results from estimations of those regressions in Eq. (5) that have shown no serial correlation, alternatively assuming serial correlation and no serial correlation among errors in order to compare the results.

Standard errors do not differ considerably and, as a result, we decided to proceed with all regressions by correcting for autocorrelation of errors.

Funding

Not applicable.

Availability of Data and Material

The authors state that in this paper they use publicly available data, downloaded from official datasets, and the sources are mentioned in the text.

Code Availability

Not applicable.

Declarations

Conflicts of Interest

The authors state that there are no conflicts of interest.