Does the Federal Budget Process Promote Fiscal Sustainability?

Does the Federal Budget Process Promote Fiscal Sustainability? By Denvil R. Duncan, John L. Mikesell, and Justin M. Ross School of Public & Environmental Affairs, Indiana University

WORKING DRAFT May 1, 2015 Circulated for Comment at IU-SPEA Public Finance Conference Do not cite without permission

Abstract This paper articulates a simple positive political economy theory of the American federal budget, namely that it employs processes that facilitate political exchanges for preference satisfaction. In this view, normative objectives like maintaining fiscal sustainability do not emerge directly from budgeting theory, and as a result theories of budget process cannot be judged as having failed to predict those fiscal outcomes. We put forward a comprehensive empirical model of budget process acts and features, and find that almost every process feature is irrelevant to outcomes or is very sensitive to specification. The only exception is the Budgetary Enforcement Act of 1990 that consistently reports an association with fiscal restraint. However, a synthetic control method suggests that the expiration of BEA90 actually reduced expenditure growth. We argue this pattern of findings is consistent with the view that a public budget process exists to facilitate political exchange only, and that budgeting process theory should be reconsidered for its ability to explain its role in adding value to political exchange. Keywords: Federal budget, budget process JEL Code: H6, H1 Acknowledgments: We would like to thank Phil Joyce, Bob Kravchuk, and participants of the 2015 Public Finance Conference on “Federal, State, and Local Budgets in Crisis” at Indiana University’s School of Public and Environmental Affairs.

1. Introduction The American federal budget process is generally regarded as a collective negotiation between the major political parties in both congressional chambers and the executive branch with an aim towards prioritizing individual public programs within aggregate fiscal constraints. To be successful, an institutional process like budgeting must be capable of coordinating diverse and changing actors in a way that is robust to environmental conditions in a manner that affects relevant outcomes. The long-term fiscal sustainability of America’s federal government will ultimately depend on some degree of fiscal restraint, and observers of the budget process have been critical of federal budgeting and budgeting theory in just about every conceivable respect. For decades scholars have raised concerns that the budget process, despite several reforms and adaptations, is increasingly deficient by way of systematic deficits, shrinking control of appropriate levers for correction, and increased political polarization. This long-running criticism is perhaps underscored most prominently by the widely decried fact that Congress has regularly failed to pass its own budget resolution, failed to pass appropriation laws before the beginning of fiscal years, and cannot avoid deficits even in a favorable economic climate. This paper reconsiders the federal budget process in terms of positive political economy. The normative idealization that a budget process should promote long-term fiscal sustainability does not seem likely to align with a positive budgeting theory that would arise when considering the budget as simply another arena for politics to serve the function of exchange between various stakeholders. In other words, processes in politics serve to assist politicians in promoting their own ends towards maximum value. From this perspective, the evolution of the budget process in terms of various activating and expiring acts that seek to promote fiscal sustainability is more likely to reflect the collective preferences of and balances of power between the two branches of government in employing fiscal powers in politics. This view has been articulated in the fiscal domain of federal government before, albeit in taxation rather than budgeting process. Most notably, Buchanan (1987) argued that the 1986 federal tax reform that eliminated numerous tax expenditures with base simplification could be understood in a model of public choice where the pre-reform complexity of the tax code had reduced the ability of political actors to extract gains from offering up additional exemptions, deductions, and credits. In other words, by simplifying

1

the base, Buchanan (1987) argued that the reform increased the value of what could be politically gained by amending the tax code. To consider this perspective, the emphasis of this paper will be on the Budget Enforcement Act of 1990 (BEA90), which was in effect during the fiscal years 1992 to 2002. This act required the executive and legislative branches to agree on a cap on discretionary spending growth and for only deficit-neutral or deficit-reducing rule changes to entitlements and taxes. The act was renewed in 1997 but permitted to expire in 2002. BEA90 is generally regarded as having aided in bringing about budget surpluses in fiscal 1998 through 2001, and its expiration permitted the passage of the Bush tax cuts and the Medicare Modernization Act of 2003, each of which (good policy or not) were significant contributors to the deficit. 1 Positive political economy theory provides reason to doubt the popular view of BEA90 by viewing it as endogenous, at least in terms of what the act might have encouraged in the post2002 expiration period. The BEA90 period provided a framework for political exchange among actors who preferred deficit reduction, in which case expenditure growth could have been unusually restrained even in the absence of the Act. Just as Tax-Expenditure Limits persist in states that generally favor lower levels of taxation and spending regardless of statutory code, federal budgetary processes that favor deficit reduction should reflect a political preference for low deficits that will be discarded when it no longer is convenient for the winning political coalition. In this public choice view, had the entitlement component of BEA90 not served as a barrier to tax cuts and Medicare reform, the budget process could have been capable of producing periods of exceptional discretionary growth by facilitating bigger bargains among those who preferred expenditure growth. The empirical examination begins with a time series analysis of fiscal outcomes as they relate to numerous budget processes. The only major budget process that is consistently associated with fiscal constraint is BEA90. Although a contribution by being more comprehensive than what has previously appeared in the literature, the time series analysis is largely descriptive rather than causal, so it is followed by a synthetic control method for

1

The Medicare Modernization Act of 2003 included Medicare Part D, which expanded entitlements in the form of subsidies to drug prescriptions.

2

identifying the effect of BEA90 on the deficit. To the best of our knowledge, this paper represents the first application of synthetic control methods in public budgeting. After introducing the federal budget process and major historical acts that have influenced it, the next section discusses how this paper ties together the disparate elements of budgeting theory and fiscal politics under the overarching public choice view of politics as exchange. In section 3, the aforementioned time series analysis is provided with the synthetic control analysis of BEA90 to follow in section 4. Section 5 summarizes how the major findings of the empirical analysis are consistent with the politics as exchange perspective in the context of BEA90 and how future research can advance politics as exchange theory in public budgeting. 2. The Federal Budget: Process and Theory Scholarship on the American federal budget process faces two problems which has directed empirical research away from consideration of fiscal sustainability; 1) the uniqueness of America and its federal budgeting system; 2) the limited predictive power of most budget theory - when it is applied to the American federal government in recent decades - as well as the difficulty of providing a broader theory of federal budgeting. To provide the necessary context, this section will proceed with an overview of the process and its major changes over time then proceed to the discussion of previous scholarship in budget theory. Overview of the Process and its Evolution The budget process provides the framework for the President and Congress to formulate, approve, and execute the expenditure programs of the federal government. While revenue issues are sometimes involved in the process, the primary focus of the process is on spending because major revenue proposals and their adoption operate through a different decision structure, i.e., different Congressional committees and not according to a regular cycle. The budget process establishes how proposals for expenditure will be developed and approved and verifies that spending has occurred according to the approved law. The process intends to provide a discipline that keeps overall spending within the bounds of available revenue (a fiscal sustainability concern), that allocates resources among possible uses to services most important to the nation, and that encourages operating units of government to manage resources they have been provided in the most efficient means possible (Mikesell 2014: 55 – 59). It requires the President and 3

Congress to perform certain defined budgetary tasks in a regular sequence if it is to function properly. The most basic element is in regard to control of expenditure. The federal Constitution states that “No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law…” (Article 1, Section 9), so expenditures must be approved in a law passed by Congress and signed by the President. The process for making these appropriations is generally the same as required for passage of any law: approval of an identical bill by both houses of Congress and signature of the bill into law by the President. This is the flow identified as starting in the appropriation committees in Figure 1. The appropriation process, however, represents approval of agency expenditure programs. It does not directly involve the development of fiscal plans that budgeting entails. Two important laws have created the current federal budget process that creates integrated fiscal plans: the Budget and Accounting Act of 1921 and the Congressional Budget and Impoundment Control Act of 1974. Although enactment of each law was driven by multiple intentions, one element in each involved a desire to get budgetary deficits under control. The Budget and Accounting Act of 1921 created the formal Presidential budget process. It required the President to present to Congress early in the calendar year an intended expenditure program for all operations of the federal government for the fiscal year that will begin several months in the future. No more would agencies bring their requests for appropriations directly to Congress when more resources were required. The President, assisted by a Bureau of the Budget created by that Act (now Office of Management and Budget), would present a single, government-wide request to Congress (a flow shown on Figure 1), developed according to his policy intentions and within the resources he believed to be available for that upcoming year. While the Presidential budget system was an element in an overall management agenda, it was also a product of concern about fiscal discipline. The expenses of World War I had caused deficits far in excess of experience in the early years of the twentieth century (deficits of 44 percent of outlays in 1917, 71 percent of outlays in 1918, and 72 percent of outlays in 1919, compared with surpluses in twelve of the prior years of the century) and there was much concern that there be a return to disciplined finances. An executive budget was seen as a device for re-establishing control – and the first nine years covered by the Presidential budget 4

requirement (1922 to 1930) showed a surplus in excess of twenty percent of total outlays. The Presidential budget, as a unified and organized proposition for federal outlays developed from constrained agency requests, continues as an entrenched element of the federal fiscal process. As an executive plan, the Presidential budget provided no actual spending authority to federal agencies. The process functioned with only a Presidential budget proposal for almost fifty years, through the fiscal stress coming from the Great Depression, World War II, and the Cold War. In non-shooting-war periods, there were years of surplus and years of deficit with deficits frequently in recession years. But from 1961, the pattern changed: from 1961 through 1973, there was only one surplus year, even though the federal government faced no extraordinary fiscal pressures in that time (the Vietnam war was not associated with the substantial run-up in defense spending traditionally associated with military action). This history was one of the major forces behind passage of the Congressional Budget and Impoundment Control Act of 1974 (Public Law 93 – 344). The Congressional Budget and Impoundment Control Act, among other things, required Congress to pass a Congressional Budget Resolution, developed in Budget Committees (shown in Figure 1), that establishes the fiscal program of Congress. 2 Key elements to be set forth in the resolution for the upcoming fiscal year included appropriate levels of outlays and new budget authority, an estimate of budget outlays and new budget authority for each major functional category of government operations, the surplus or deficit appropriate in light of economic conditions and other factors, recommended level of revenues and the aggregate amount by which revenue should increase or decrease, and the appropriate level of public debt and any appropriate change in the statutory debt limit. Hence, the Congressional Budget Resolution provided the Congressional guideline for the budget process in much the same way as the Presidential budget provided the executive branch view. The Resolution would be passed by Congress but would not be signed by the President, would guide the fiscal decisions by Congressional committees and the full Congress made on expenditures (and revenues) (shown in Figure 1), but would provide no obligation authority to agencies. Although Congress regularly passed its Budget Resolution 2

The Act also provided Congressional committees to consider the Budget Resolution, a Congressional Budget Office to provide non-partisan fiscal and economic analysis for Congress, and constrained the ability of the President to fail to spend all money provided in an appropriation act.

5

in the early years of the law, as Table 1 shows, a Congressional Budget has been passed for only eight fiscal years since 2000. The two budget process laws created Presidential and Congressional responsibility for development of fiscally constrained spending programs. However, the fiscal discipline record continued problematic. From fiscal 1977 (the first year with both Presidential and Congressional budgets) through the first half of the 1980s, the deficit was less than ten percent of total outlays in only one year and was over twenty percent in 1983, 1984, 1985, and 1986. At this point, lawmakers moved away from efforts to improve fiscal discipline through a strengthened budget process and proceeded to craft direct control mechanisms with triggers and consequences. Several control acts were passed to impose direct control over deficits. These are included as part of Table 1 and include the following: i)

Balanced Budget and Emergency Deficit Control Act of 1985 (Public Law 99 – 177) [Gramm-Rudman-Hollings]. The Act established deficit targets for future years, ultimately leading to budgetary balance, with a formula sequestration process if the appropriation process appeared to be producing a fiscal outcome that would violate the target for a year. The law was applicable for fiscal 1986 through 1990, although its application was sometimes suspended.

ii)

Budget Enforcement Act of 1990 (Public Law 101 – 508). The Act divided spending into discretionary (spending controlled by the annual appropriation process) and mandatory (spending controlled by a formula) and controlled each. It established firm ceilings on discretionary spending and a PAYGO requirement on mandatory spending and revenue provisions (PAYGO meaning that any legislative change that would increase mandatory spending or reduce revenue would have to be accompanied by a provision to pay for the change in the deficit). The law and its extensions applied to fiscal year 1992 – 2002.

iii)

Statutory Pay-As-You-Go Act of 2010 (Public Law 111-139) The Act requires that new legislation changing taxes, fees, or mandatory spending, taken together, may not increase the projected deficit. OMB is the scorekeeper and applies across-the-board (with exceptions) sequestration of mandatory spending if there is a violation.

6

According to the scoring rules of the law, no sequestration has yet been found to be necessary. The law applies to all legislation enacted after February 12, 2010. iv)

Budget Control Act of 2011 (Public Law 112 – 25) The Act establishes discretionary spending caps through fiscal 2021. Sequestration across-the-board applied by OMB if the caps are violated. The law applies to fiscal 2012 through fiscal 2021. In terms of added fiscal discipline, these extraordinary targets and controls do not show a

strong record of success. From fiscal 1986 through 2014, only four fiscal years – 1998, 1999, 2000, and 2001 -- have shown a surplus, with only one of those years with a surplus above ten percent of outlays. The rest of the years have shown deficits, twenty with deficits above ten percent of outlays. It is apparent in this record that it is easier to pass laws purported to control a budget deficit than it is to pass laws that actually reduce a deficit by increasing revenue or reducing spending. Previous Scholarship: Review and Critique Much of the scholarly literature in public budgeting discusses the federal budget process since the 1970’s as a departure from an incremental method. Incrementalism is the method of comparing proposed budgets to those of the previous period, with participants giving special attention to those components that are changing (Wildavsky, 1964; Wildavsky and Feno, 1966). Davis et. al. (1966), for instance, were able to show that from 1947 to 1963 congressional appropriations to various federal agencies were highly predictable from simple linear trend models. 3 Though it was never without its critics (e.g. Schultze 1968; Natchez and Bupp, 1973; Berry, 1990), incrementalism is still the dominant theory of budget evaluation and processing in the sense that alternative theories are compared for how they departure from it or are otherwise seeking to explain periods in which incrementalism is not followed. Like incrementalism, subsequent budgeting theories would emphasize the perceived new realities of the budget

3

A large literature testing the incrementalism hypothesis during this era exists, a critical methodological summary and review of which can be found in Dezhbakhsh et al. (2003).

7

process by ways of its features (e.g. Bozeman and Straussman, 1982; Caiden, 1984; Pitsvada and Draper, 1984; Joyce and Reischauer, 1992; Rubin, 2007). 4 By and large, these theories of budgeting are intended to describe the processes for the evaluation of programs. That is, these theories seek to provide observers with a means to understand the evaluation process as it progresses across political cycles, in the form of considering “program A” against “program B” whilst seemingly ignoring “program C”. Incrementalism, for instance, theorizes that programs will seldom be re-evaluated and political battles on programs will not be reopened. 5 Bozeman and Straussman (1982) addressed inadequacies of incrementalism by arguing that a theory of federal budgeting revised from incrementalism must incorporate the existence of two processes, one that is “top-down” that sets the parameters as well as a “bottom-up” process of negotiation. Rather than consider the end of incrementalism as failure of budgeting theory, the public choice perspective would be to instead regard these as a subtheory within a broader theory of political exchange. In other words, why was incrementalism useful for predicting the pattern of political exchange in certain eras and not others? A revised positive political economy view of this budgeting era might argue that political exchange was best facilitated when there existed a pre-commitment to not revise completed deals in the near future, which resulted in incremental budgeting processes. When the greatest gains in political exchanges would involve existing programs, either due to external pressures or ideological preferences, the observed process adapted to facilitate exchanges of this form. Efforts to create a budgeting theory that describe the process are almost certain to be limited in the time horizon of applicability, so a process oriented theory of budgeting can only be rightly judged within the era for which the observed process was designed. It is also important to recognize the uniqueness of the American federal budget system as a contributing factor to a relatively small number of studies that consider fiscal sustainability. 6 In terms of budgetary process, the federal government has a number of features that make budgetary and other fiscal comparisons with other countries difficult. On top of the usual 4

This shift in the literature’s attention is predicted by a literature review by Berry (1990: 167): “I argue that the most productive course would be to banish the term incrementalism from new scholarly literature, and instead, focus research n more specific characteristics of the budgetary process.” 5 This is a reasonable common-denominator view of what “incrementalism” means, but as Berry (1990) points out the term “incrementalism” has taken on many different theoretical meanings. Our attempt is not to define what it means 6 CITE THE ONES THAT DO HERE

8

challenges of geography and economic size of the country that plagues cross-country research, the structure of government reflected in the budget process represents an important point of departure: The politically and fiscally independent tiers of government associated with the federal system; The separation of powers between legislative and executive branches that distinguish the American system from that of parliamentary governments and make the executive budget only the starting point for fiscal debate; The existence of two legislative assemblies (Senate and House) that must approve appropriations rather than only one; legislative freedom to add expenditures to the budget presented by the executive; The single-year emphasis in both budgeting and appropriating; Absence of a distinct capital / development budget, Absence of over-arching deficit-ratio control rules (e.g. the 3% of GDP EU control), and the substantial amount of expenditure that goes unchecked by the appropriation structure via the operation of entitlement formulae or through the tax expenditure system. Gathering all these features together makes panel data on the national budget outcomes as controls for operations of the U. S. federal government a difficult challenge. Given the data challenges required for theory driven tests that relate the budget process to fiscal outcomes, fiscal sustainability of process has instead given way to a large literature of fiscal politics that offered more data and clearer path toward hypotheses testing. For instance, Albouy (2013) finds in a regression discontinuity design that districts represented by members of the majoritarian party with greater proposal power will receive a larger allocation of federal grants, and Berry et al. (2010) find similar evidence using data from the president’s proposed budget; Larcinese et al. 2006 provide evidence that states receive greater federal funds when they were heavy supporters of the incumbent president in the last election, and when the governor of their state is from the same political affiliation; Anderson and Woon (2014) finds that, consistent with bargaining theory, delays in the delivery of the presidential budget advantages in the form of greater concessions from congress to the executive. In addition to showing that political elections encourage deficit spending, Shi and Svensson (2006) show that the size of this effect depends on politicians rents of remaining in power and in the share of uninformed voters in the electorate. Internationally, Roubini and Sachs (1989) and Edin and Ohlsson (1991) find that majority coalition governments have greater public debt to GDP ratios, while Woo (2003) provides a number of pooled cross-sectional regressions between public debt and various political policy process variables. 9

Each represent hypothesis testing of the games within the budget process, but do not individually aim to paint a broader perspective of budget theory as it has been traditionally regarded, particularly with regards to macro outcomes like fiscal sustainability. In the public choice critique advanced here, however, these works in fact represent more direct contributions to a positive theory of political exchanges than do empirical tests of whether or not a specific budget process theory holds true. Tests of the influences of coalitions, how presidents can extract concessions from congress, etc. are in fact tests of how the terms of trade affect the distribution of gains in political exchange. If the process of budgeting is designed to maximize these gains from trade, then understanding the determination of payoffs is a crucial component. 3. Time Series Analysis of American Federal Budget Process Model Specifications and Variable Selections The previous section highlights the numerous changes to the federal budget process. This section provides a descriptive analysis in the form of a time-series regression to see how these processes correlate with the fiscal outcomes of the budgeting process: deficit as a percent of GDP (DEF t ), total outlays (OUT t ), and the year-over-year growth in outlays (ΔlnOUT t ). These three alternative dependent variables will be considered in regressions on two sets of explanatory variables. One set of regressors will relate to budget process features (Z t ) while another set control for various economic and political conditions (X t ). The three regressions will be specified as follows: (1) 𝐷𝐷𝐷𝑡 = 𝜌𝐷𝐷𝐷𝑡−1 + 𝛿𝑍𝑡 + 𝛽𝑋𝑡 + 𝜀𝑡

(2) 𝑂𝑂𝑂𝑡 = 𝜌𝑅𝑅𝑅𝑡−1 + 𝛿𝑍𝑡 + 𝛽𝑋𝑡 + 𝑣𝑡

(3) ∆ln(𝑂𝑂𝑂𝑡 ) = 𝜌∆ln(𝑅𝑅𝑅𝑡 ) + 𝛿𝑍𝑡 + 𝛽𝑋𝑡 + 𝑒𝑡

Each dependent variable represents an alternative perspective on how the long-run fiscal performance may be judged. Equation (1) is specified to be consistent with budget process that produces a deficit with a mean reverting process, motivating the inclusion of a lagged dependent variable. 7 Equation (2) considers the process as one that determines total outlays as a function of

7

Both the Augmented Dickey-Fuller and Phillips-Perron tests failed to reject the null hypothesis of a unit root without an autoregressive lag, but did support a stationary process under an AR(1) structure with p-values of 0.0268 and 0.084, respectively.

10

available receipts. Finally, equation (3) views the budget process as an incremental adjustment from the previous year. Equations (1) – (3) will each be estimated twice, one full specification as well as one restricted version that excludes the non-budget controls (i.e. β=0), which aids in judging the sensitivity of the budgetary variables to model specification. The budgetary variables included in vector Z include the major acts discussed in section 2 and listed in Table 1 that were in force for fiscal years 1962 to 2014. This era includes the Congressional Budget and Impoundment Control Act of 1974 (1975-current), the Balanced Budget and Emergency Deficit Control Act of 1985 (1986-1990), the Budget Enforcement Act of 1990 (1992-2002), and both the Statutory Pay-As-You-Go Act of 2010 and Budget Control Act of 2011 that both went into effect as of 2012. In addition to these major statutory acts, several annual indicators were collected on the progress of the budget process that actually occurred. Among these indicators is whether or not Congress passed a budget resolution, whether it had passed all budget appropriations by the start of the fiscal year, and whether or not the presidential budget was delivered on time. In all cases, these variables are coded such that their role in the budget process should be to promote fiscal sustainability, ceteris paribus, so that their expected sign in equations (1)-(3) would be negative in all cases. Of course, many of these budgetary rules coincide with numerous economic and political changes whose influence we attempt to mitigate with variables defined in vector X. Real GDP growth captures the contemporaneous trend of the economic conditions, while actual-to-potential GDP represents the state of the business cycle. For political ideology, we include the median DW-Nominate score from Carroll et al. (2009) for the House, Senate, and the President. The DW-Nominate ideology variables range from liberal (-) to conservative (+), so increases in the ideology score represents a move to a more conservative position. We also include their polarization variable that represents the absolute difference in the median democrat from the median republican in congress. Finally, we include a measure representing the ideological divide between congress and the President by taking the absolute difference from the median DW nominate score in Congress from that of the president. Summary statistics, variable notes, and data sources are described in Table 2. Results

11

Table 3 displays the estimation results of equations (1)-(3), with all specifications including the main variables of interest. With respect to the deficit, the only budget process variable which has a statistically significant effect in the full specification is BEA90, which indicates that on average it reduced the deficit by 1.1% of GDP. The statistical significance of BEA90 is sensitive to model specification, but the point estimates across the restricted and unrestricted models are similar. The CBICA, by contrast, requires control variables to produce a sign that is consistent with deficit control and even then it is statistically insignificant. The SPAYGO/BCA period is correlated with higher deficits by 2.6% of GDP, but again this is sensitive to model specification and is inferred from a very short time horizon. The economic and ideological variables seem to play a more substantive role in determining deficits, at least when judged in terms of statistical significance. Deficits decline as the economic business cycle approaches full employment, consistent with an effort towards countercyclical macroeconomic policy. A standard deviation increase in actual to potential GDP (2.76) reduces the deficit by 1.61% of GDP. 8 As the House of Representatives becomes more conservative, the deficit declines with a standard deviation increase in the median representatives DW-Nominate score correlated with about a 1.12% decline in the deficit. 9 Senate ideology is not statistically significant while presidential ideology is associated with a statistically significant deficit expansion as it becomes more conservative. Congressional polarization between parties results in deficit expansion, as a standard deviation increase in polarization is associated with an additional 1.2% increase in the deficit. 10 A standard deviation increase in the ideological divide between the president and congress reduces deficits by -0.35%. 11 Turning attention from deficit to total spending in Table 3 yields a similar set of results. The only stable budget process variable is BEA90, which in the full specification is associated with a $182 billion reduction in real total outlays. CBICA is now a statistically significant reducer of outlays, but both sign and precision are sensitive to model specification. No other budget process variable is statistically significant, and the other economic and political variables are similar in their findings to those for the deficit. Incremental spending, or the percent change

8

Calculation: 2.76x(-.0584)=-1.615 Calculation: 0.19x(-5.816)=-1.123 10 Calculation: 0.11x(10.712)=-1.230 11 Calculation: 0.18x(-1.990)=-0.352 9

12

in outlays, is the final set of results and it produces the fewest statistically significant findings. BEA90 is again negative in both specifications with significance level that is sensitive to variable choice. CBCIA is correlated with positive growth while the Balanced Budget and Emergency Deficit Control Act is correlated with a 3.2% decline in outlays. Unlike the previous specifications, all of the political ideology variables are statistically insignificant. Summary of Findings The time series analysis of Table 3 demonstrates that budget control acts and process features generally have very tenuous correlations with fiscal sustainability. Furthermore, ideological variables frequently appear to be significant and whose exclusion generally affects the point estimates on the budget process variables. If ideology was independent to the budget process governing features there would only be implications for the standard errors rather than sign flipping. To further explore this possibility, the next section uses the synthetic control method to estimate the effect of BEA90 on fiscal sustainability in a post-2002 counterfactual period. Implementing this method with BEA90 is motivated by the fact that it is the only budget process variable associated with fiscal sustainability in all of the time series specifications. We also argue that the history of this act is particularly useful for the synthetic control analysis as a means of drawing a more causal inference. 4. Synthetic Control Analysis of Congressional Budgets The time series estimates presented in the previous section provides a useful initial attempt to identify the effect of BEA90 on federal spending in the sense that when the BEA90 was “on” the federal government held tighter deficit standards and expenditure standards than when it was “off” as it was post-2002. The public choice critique, though, renders this inference projected onto the post-2002 period somewhat obsolete. For example, it is possible that an underlying preference for lower spending led to the passage of BEA90. If the purpose of the BEA90 was to affect the terms of trade in a budget process where the general preference included a desire for deals that facilitated deficit reduction, then indeed the negative effect of BEA from 1992 to 2002 shown in the times series results would be qualitatively correct even if the point estimate over attributed its actual influence. However, to project the possible deficit reducing effect of BEA90 during the 1992 to 2002 period into the 2002 period and forward is not necessarily an internally 13

valid inference in the public choice critique that process features facilitate exchanges. The expiration of the Act could represent that elements of the act no longer served this process, and that surrendering the familiar budget process was an acceptable sacrifice to permitting expenditure growth. In such a counterfactual, political actors will have to learn the new formal and informal rules of “the budget game” over how to bargain most effectively for their expenditure goals. The history of the BEA90 expiration is suggestive of such an instance, in that the deficit and revenue neutrality of BEA90 made the executive’s desire for a new wave of tax cuts and the passage of Medicare Part D more difficult. According to Paul O’Neill (in Suskind, 2004: 192), then Treasury Secretary, in December 2002 Vice President Cheney declared that “Regan proved deficits don’t matter….We won the midterms. This is our due.” But had the Congress and president been able to continue in subsequent years with their preference for deficit growth, the familiarity of BEA90 era rules would facilitate trades of that form and be associated with greater expenditures rather than less. Addressing this identification problem requires a regime switching model that endogenously alters the process according to ideological preferences to judge against a counterfactual. The absence of similar budget control processes or suitable instruments for such a purpose leads this paper to consider an alternative comparative case study approach. This strategy is also problematic because the USA differs significantly from every other country, which has been a barrier to other research. 12 One approach that has recently been advanced for cases such as these is to identify a set of countries for which a linear combination is comparable to the US. This linear combination represents a “synthetic” United States which behaves similarly in a given policy era that can then serve the purpose of constructing a counterfactual in the new policy era. 4.1 Synthetic control method The synthetic control method developed by Abadie et. al. (2003, 2008), allows us to estimate weights such that the weighted combination of countries creates a synthetic version of the US.

12

For the synthetic control variable, we use panel data on central government finances from the International Monetary Fund’s Government Finance Statistics. These measures of revenues and expenditures differ from the receipts and outlays used in the time series analysis that were drawn from the OMB. However, the pairwise correlation for the U.S. on these fiscal variables is 0.98.

14

This synthetic US is then compared to the actual US in order to identify the effect of BEA90’s expiration in 2002. 𝐾 𝐾 Following Abadie et. al. (2008) we define 𝑋1 = (𝑍1 , 𝑌�1 1 , . . . . . , 𝑌�1 𝑀 )′ as a (k x 1) vector

𝐾 of pre-intervention characteristics for the US. Y is expenditure, 𝑌�1 𝑖 is a linear combination of

pre-intervention expenditures, and Z is a vector of variables that predict expenditure. Similarly, we define 𝑋0as a (K x J) matrix that contains the same variables as X 1 for countries not affected by BEA90. The objective is to select a vector of weights 𝑊 = (𝑤2 , . . . . , 𝑤𝑗+1 )′ such that

||𝑋1 − 𝑋0 𝑊|| is minimized, 𝑤𝑗 > 0 and ∑𝐽+1 𝑗=2 𝑤𝑗 = 1. We specify

�|𝑋1 − 𝑋0 𝑊|� = �(𝑋1 − 𝑋0 𝑊)′ 𝑉(𝑋1 − 𝑋0 𝑊), where V is a (k x k) symmetric positive

semidefinite matrix chosen to minimize the mean squared prediction error of expenditure for the pre-intervention periods. Because our research question asks whether BEA90 affects budget deficits in the post2002 era, we would ideally like to use the deficit normalized by GDP as our outcome variable. However, the trend in this variable is not smooth, which makes it difficult to find a synthetic match. For this reason, we instead use expenditure as our outcome variable. Because expenditure in the US is higher than that of every other country in our sample, it is not possible to find a synthetic match when using nominal expenditures. In other words, no combination of other countries will match the US because the US has the largest value in every year. Therefore, we operationalize our specifications using two standardized measures of expenditure: expenditure per capita and an expenditure index that measures levels normalized against a country’s own expenditure level in 1993. The latter measure sets expenditure in 1993 equal to one and measures expenditure for each of the subsequent years relative to the value in 1993. Creating the synthetic US requires predictors of our outcome variable of interest. The goal is to include variables that are good predictors of expenditure in vector Z. These include per capita GDP and growth in GDP to capture level of development and the business cycle respectively. Countries generally increase expenditures during economic down turns due to automatic stabilizers and other stimulus fiscal policies. Similarly, higher income countries generally have higher levels of expenditure. Share of population age 0 to 14 and 65 and older are included to account for the fact these demographic groups places greater spending pressures on 15

governments. We also include political alignment between executive and lower chamber of congress as this may affect the ease with which budgets are passed. Corruption is another institutional feature of a country that is likely to affect expenditure. Finally, we include total revenues, which is standardized using the same procedure as for expenditures. It is important to note that the sole purpose of the covariates is to predict pre-intervention expenditures; there is no need to establish a causal relationship. Implementation of the synthetic method also requires a set of countries that could potentially serve as controls. Together, this group of countries is called the donor pool. There are several restrictions on the donor pool. First, every country has to have complete data on the outcome variable. Therefore, we begin by restricting the sample to countries that have data on central government expenditure for the period 1993 to 2008. Second, we can only include a country if it has at least one non-missing value in the pre-treatment period for each variable in Z. A country that has all missing values for at least one variable is excluded from the analysis. Finally, and most importantly, the donor pool cannot include any country that implemented an expenditure related policy change at the same time as the US. Additionally, any country that depends directly on US expenditure has to be excluded. These restrictions are necessary in order to prevent any confounding effects from biasing our estimated treatment effect. Of the countries that survive the first two data restrictions, Israel is the only one that arguably depends greatly on US expenditure policy. We exclude Israel from the analysis. These restrictions leave us with a sample of 25 potential donors: 12 OECD and 14 Non-OECD (See Table 4 for the list of potential donors). 4.2 Synthetic Control Results This section describes our main findings. We begin by describing the results for expenditure per capita. This is followed by a series of robustness tests including the use of expenditure growth relative to 1993 as the outcome of interest as well as two placebo tests. Each of the models is implemented using a donor pool consisting of 25 countries for the period 1993 to 2008. Since BEA90 ended in 2002, we set 2002 as our treatment year and define 1993 to 2001 as the pretreatment period and 2003 to 2008 as the post-treatment period.

16

Figure 2 plots the trend in expenditure per capita for the US and the mean of central government expenditures per capita among the other countries in our sample. The figure shows that the two experienced mostly similar trends in expenditure per capita. There is a little more variation in the pre-treatment period for the other countries, and expenditure in the US began rising a few years before that of the other countries. The most striking observation, though, is the fact that the US has much higher expenditure per capita than the average country in the sample. There also does not seem to be any noticeable differences in the post treatment trends. However, this simple observation of trends does not allow us to say anything about the impact of BEA90 on expenditures in the US. We therefore implement the synthetic control method which allows us to synthesize a counterfactual US, which we then compare with the actual US. The result of this exercise is presented in Table 5 and Figure 3. First, Table 4 shows that the synthetic US is comparable to the actual US on the predictors included in the model. The synthetic has higher revenue per capita, political alignment, corruption and unemployment, but lower per capita GDP. There is no observable difference in share of population below 14 years old or above 65 years old, or GDP growth. The synthetic US is made of Croatia, Iceland, and Norway, with weights of 38.6%, 60.2%, and 1.1%, respectively. Figure Y shows that both versions of the US track each other fairly closely in the pre-treatment period. However, we observe a clear difference in the trends post BEA90; the actual US is spending a lot less than it would have spent in the presence of BEA90. In other words, the removal of BEA90 led to a decrease in spending. These results continue to hold when we measure expenditure relative to expenditure in 1993. These results are presented in Figure 4. Again, we observe very similar trends in the pre-treatment period and a very large difference in the post treatment period. 13 The evidence presented so far suggest that removing BEA90 lowered expenditure relative to the control. While these differences appear to be large in magnitude, it is not clear how likely they would have been realized by random chance. To gain insight on this, we follow Abaidie et. al. (2008) and run a series of placebo tests. For each of these tests, we assume the treatment took place in one of the other countries in the donor pool. In other words, we repeat the synthetic control method with the US as part of the donor pool and each of the other countries is sequentially assigned the treatment status. We then take the difference between the expenditure 13

The synthetic control is comprised of Chile (5.8%), Norway (41.7%), Romania (37.9%), and Switzerland (14.5%).

17

for each country and its synthetic, and plot the trend in these differences in Figures 5 and 6 for expenditure per capita and growth in expenditure relative to 1993, respectively. Panel A of Figure 5 shows the trends for all countries; the US is shown in solid black while each of the placebo runs is shown in grey. The other panels successively removes countries whose mean squared prediction errors is much larger than that of the US. 14 The trends in Panel A show that the treatment effect identified for the US could have happened by chance. In other words, several other countries that did not have a major expenditure reform in 2002 also saw large declines in expenditure. However, many of these countries have MSPE that are much larger than the US. The MSPE for the US is 543 and there are five countries with MSPE greater than 1000. The effect for the US is more obvious relative to the other countries when we remove those countries with MSPE greater than 1000 (Panel B). However, only when we remove countries with MSPE of more than 400, does the effect for the US seems like it could have occurred for reasons other than chance. We find much stronger results when we examine the growth in expenditure relative to 1993 (Figure 6). Again, Panel A shows the analysis for all countries while Panels B to D shows the results after removing countries with MSPE larger than the US. The MSPE for the US is 13.3 and the average, median and maximum MSPE in the sample is 32.7, 22.6, and 277.9, respectively. The effect for the US stands out once we remove the eleven countries with MSPE greater than 25 (Panel B), and is even more obvious when we remove those with MSPE of more than 15. 4.3 Summary of Findings The synthetic control findings are consistent with the view that altering the budget process rules, formal or informal, at the expiration of the BEA90 reduced the ability of the budget stakeholders to capitalize on political exchanges. If the BEA90 era had been extended into the new era that deemphasized deficit control and valued expenditure growth, then even more of these trades could have occurred. This does not suggest that BEA90 actually increased deficits or controlled expenditures, but it does deemphasize the significance of the process for fiscal sustainability. For

14

The mean squared prediction error measures the accuracy of the pre-treatment fit of expenditure; i.e., how closely does expenditure in the synthetic US come to that of the actual US.

18

fiscal sustainability, these results suggest that budget processes are neither necessary nor sufficient in American federal government. 5. Conclusions This paper articulates a simple positive political economy theory of the American federal budget, namely that it employs processes that facilitate political exchanges for preference satisfaction. In this view, normative objectives like maintaining fiscal sustainability do not emerge directly from budgeting theory, and as a result theories of budget process cannot be judged as having failed to predict those fiscal outcomes. Budget process theory (e.g. incrementalism) can only be judged by its applicability to eras of political preferences. We demonstrate in a 1962-2012 time-series analysis of deficits, expenditures, and annual expenditure growth in the U.S. that budget process features have little to no relationship with these fiscal outcomes and that they are highly sensitive to the inclusion of political ideology variables. The one exception to this is the Budget Enforcement Act of 1990, which governed fiscal year outcomes from 1992 to 2002 and is strongly associated with fiscal sustainability. While the BEA may very well have facilitated political exchange towards fiscal sustainability in its era, we use a synthetic control analysis to demonstrate that continuing familiar process and means of exchange into an era unconcerned with fiscal control would have actually increased spending. This pattern of findings is consistent with the public choice model of politics-as-exchange applied to the budget process. The implication for research in public budgeting is to reconsider how process theory should be evaluated and expanded upon. If the era of incrementalism ended, as many scholars have suggested, then rather consider it a “failure of budget theory” it should be studied for how incrementalism created value for actors in one era of fiscal exchange but not another. Did incrementalism informally serve to raising the value of new program deals by giving confidence to budget stakeholders that completed negotiations would not be revisited? If so, then why did this cease to be a valuable feature for political exchange the led to the departure of incrementalism? The continued expansion of institutional comparative case study research will be important in the development of these hypotheses within a theory of political exchange for the budget process. REFERENCES 19

Abadie, Alberto and Javier Gardeazabal. 2003. “The Economic Costs of Conflict: A Case Study of the Basque Country.” American Economic Review (March): 113-132. Abadie, Alberto, Alexis Diamond, and Jens Hainmueller. 2008. “Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program.” Journal of the American Statistical Association 105 (490): 493-505. Albouy, David. 2013. “Partisan Representation in Congress and the Geographic Distribution of Federal Funds.” The Review of Economics and Statistics 95(1): 127-41. Anderson, Sarah E. and Jonathan Woon. 2014. “Delaying the Buck: Timing and Strategic Advantages in Executive-Legislative Bargaining over Appropriations.” Congress & The Presidency 41(1): 25-48. Berry, Christopher R., Barry C. Burden, and William G. Howell. 2010. “The President and the Distribution of Federal Spending.” American Political Science Review 104(4): 2010. Berry, William D. 1990. “The Confusing Case of Budgetary Incrementalism: Too Many Meanings for a Single Concept.” The Journal of Politics 52(1): 167-96. Bozeman, Barry and Jeffrey D. Straussman. 1982. “Shrinking Budgets and the Shrinkage of Budget Theory.” Public Administration Review 42(6): 509-15. Buchanan, James M. 1987. “Tax Reform as Political Choice.” Journal of Economic Perspectives 1(1): 29-35. Caiden, Naomi. (1984). “The New Rules of the Federal Budget Game.” Public Administration Review 44(2): 109-118. Carroll, Royce, Jeffrey B. Lewis, James Lo, Keith T. Poole, and Howard Rosenthal. 2009. “Measuring Bias and Uncertainty in DW-Nominate Ideal Point Estimates via the Parametric Bootstrap.” Political Analysis 17: 261-75. Davis, Otto A., M. A. H. Dempster, and Aaron Wildavsky. (1966). “A Theory of the Budgetary Process.” The American Political Science Review LX(3): 529-547. Dezhbakhsh, Hashem, Soumaya M. Tohamy, and Peter H. Aranson. 2003. “A New Approach for Testing Budgetary Incrementalism.” The Journal of Politics 65(2): 532-58. Edin, Per-Anders and Henry Ohlsson. 1991. “Political Determinants of Budget Deficits: Coalition Effects Versus Minority Effects.” European Economic Review 35: 1597-603. Joyce, Philip G. and Robert D. Reischauer. 1992. “Deficit Budgeting: The Federal Budget Process and Budget Reform.” Harvard Journal on Legislation 29: 429-53. Larcinese, Valentino, Leonzio Rizzo, and Cecillia Testa. 2006. “Allocating the U.S. Federal Budget to the States: The Impact of the President.” The Journal of Politics 68(2): 447-56. Mikesell, John L. 2014. Fiscal Administration, Analysis and Applications for the Public Sector. Ninth Edition. Boston: Wadsworth Cengage Learning. 20

Natchez, Peter B. and Irwin C. Bupp. 1973. “Policy and Priority in the Budgeting Process.” American Political Science Review 67(3): 951-63. Office of Management and Budget. 2015. The President’s Budget for Fiscal Year 2016: Historical Tables. Washington: OMB. (https://www.whitehouse.gov/omb/budget/Historicals) Pitsvada, Bernard T. and Frank D. Draper. 1984. “Making Sense of the Federal Budget the Old Fashioned Way-Incrementally.” Public Administration Review 44(5): 401-407. Roubini, Nouriel and Jeffrey D. Sachs. 1989. “Political and Economic Determinants of Budget Deficits in the Industrial Countries.” European Economic Review 33: 903-38. Rubin, Irene. 2007. “The Great Unraveling: Federal Budgeting, 1998-2006.” Public Administration Review 67(4): 608-17. Schultze, Charles. 1968. The Politics and Economics of Public Spending. Washington, D.C.: The Brookings Institution. Shi, Min and Jakob Svensson. 2006. “Political Budget Cycles: Do They Differ Across Countries and Why?” Journal of Public Economics 90: 1367-89. Suskind, Ron. 2004. The Price of Loyalty: George W. Bush, the White House, and the Education of Paul O’Neill. Simon & Schuster Wildavsky, Aaron. 1964. The Politics of the Budgeting Process. Boston: Little, Brown. Wildavsky, Aaron and Richard Fenno. 1966. The Power of the Purse. Boston: Little, Brown & Co. Woo, Jaejoon. 2003. “Economic, Political, and Institutional Determinants of Public Deficits.” Journal of Public Economics 87: 387-426.

21

Figure 1: Federal Budget Process

Table 1: Federal Budget Acts and Process Features, 1962-2014 Balanced Congressional Congressional Budget and Statutory Appropriations Presidential Budget Budget Budget & Pay-As- Budget Emergency FISCAL Passed by Start Resolution Budget Enforcement You-Go Control Impoundment Deficit YEAR of Fiscal Year Late (Days) Control Act Control Act Act Act Passed Act 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 0 1 0 1 0 1 1 1 0 0 0 0

-1 -6 -6 0 0 0 0 0 -2 0 -6 -8 0 0 6 -12 -1 -13 -7 0 -4 0 0 -5 0 0 0 45 0 21 0 -5 66 0 0 0 3 0 0 0 63 0 0 0 0 0 0 0 98 0 7 7 64

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

1

0

0

0

1

1

0

0

0

1

1

0

0

0

1

1

0

0

0

1

1

0

0

0

1

0

0

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

1

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

0

0

1

0

0

1

1

1

0

0

1

1

1

0

0

1

1

Table 2: Summary Statistics of U.S. Time Series Data, 1962-2014 Variable Name

Mean

Std. Dev.

Min

Max

2.64

2.43

-2.33

9.80

Total Outlays ($billions)2

1,808.30

850.00

598.99

3,517.70

Total Receipts ($billions)3 % Change Outlays4

1,542.75 3.38

662.92 4.02

558.91 -4.00

2,795.60 16.58

3.33

6.06

-17.57

16.80

0.75

0.43

0

1

0.09

0.30

0

1

0.21

0.41

0

1

0.06

0.23

0

1

0.58

0.50

0

1

0.06

0.23

0

1

0.11

0.32

0

1

2.14

2.18

-2.13

8.50

3.10

2.14

-3.42

7.82

99.19

2.76

93.43

105.50

0.00

0.19

-0.18

0.45

-0.09

0.11

-0.25

0.12

0.07

0.45

-0.51

0.52

0.64 0.46

0.11 0.18

0.50 0.11

0.94 0.81

Deficit (% of GDP)1

% Change Receipts5 Congressional Budget and Impoundment Control Act6 Balanced Budget and Emergency Deficit Control Act7 Budget Enforcement Act

8 9

Statutory Pay-As-You-Go Act Congress Passed Budget Resolution10 All Appropriations Passed by FY Start

11

12

Presidential Budget >7 Late

13

President's Deficit (% of GDP) 14

Real GDP Growth (%)

Actual to Potential GDP

15

16

House Ideology

17

Senate Ideology

18

President's Ideology 19

Polarization Presidential-Congress Ideological Divide20

Definitions and Sources: 1. Budget Deficit: Actual fiscal year expenditures minus revenues divided by gross domestic product (Source: Office of Management and Budget, 2015). 2. Congress Passed Budget Resolution: Equals ‘1’ if the fiscal year of the deficit was based on a budget resolution passed by congress; else zero (Source: Congressional Research Service, 2014). 3. All Appropriations Passed by FY Start: Equals ‘1’ if ….; else zero (Source: Mikesell, 2014: 122). 4. Presidential Budget >7 Late: Equals ‘1’ if the president’s budget was delivered to congress more than 7 days late; else zero (Source: House Budget Committee, 2014). 5. President’s Deficit (% of GDP): President’s proposed fiscal year expenditures minus revenues divided by gross domestic product (Source: Compiled from the President’s Budget produced annually by the Office of Management and Budget.). 6. Mandatory % of Total Outlays (lagged): Previous fiscal year’s non-discretionary share of total outlays (Office of Management and Budget, 2015). Real GDP Growth (%). Inflation adjusted gross domestic product fiscal year growth (Source: Office of Management and Budget, 2015). Actual to Potential GDP: Potential GDP is an estimate of the level of GDP attainable when the economy is operating near full employment, so that the ratio of actual to potential becomes an indicators of the economy in terms of business cycle (Source: Congressional Budget Office, 2015). House Ideology: Median first dimension DW-Nominate Score from the House of Representatives for the congress responsible for the fiscal year budget (Source: Carroll et al. (2009)). Senate Ideology: Median first dimension DWNominate Score from the House of Representatives for the congress responsible for the fiscal year budget (Source:

Carroll et al. (2009)). Polarization: Absolute value difference between median democrat and median republican of the congress (Source: Carroll et al. (2009)). Presidential-Congress Ideological Divide: The absolute distance between the president and the median representative of congress in their DW-Nominate Score.

Table 3: Time Series Regression Results on U.S. Deficit, 1962-2013 Deficit (% of GDP) Congressional Budget and Impoundment Control Act of 1974

Balanced Budget and Emergency Deficit Control Act of 1985

Budget Enforcement Act of 1990

Statutory Pay-As-You-Go Act 2010/Budget Control Act 2011

Congress Passed Budget Resolution

All Appropriations Passed by FY Start

Presidential Budget >7 Late

Total Outlays ($billion)

% Change Outlays

2.719***

-0.845

197.458

-177.727*

-0.126

3.209

(0.626)

(0.894)

(175.749)

(107.919)

(1.952)

(2.532)

-0.473

0.739

-134.248

90.059

-1.453

-3.208**

(0.519)

(0.674)

(84.621)

(80.573)

(1.057)

(1.401)

-1.753

-1.130***

-416.025***

-182.670***

-2.333***

-1.858

(1.200)

(0.430)

(142.066)

(57.965)

(0.854)

(1.432)

-0.939

2.595*

145.627

128.474

-5.733***

3.808

(0.702)

(1.376)

(270.279)

(187.517)

(1.982)

(5.468)

0.341

0.389

51.465

71.591

-0.142

-1.110

(0.423)

(0.500)

(173.819)

(63.612)

(1.406)

(1.687)

-0.194

-0.288

-11.527

0.210

0.414

0.973

(0.447)

(0.411)

(108.177)

(50.027)

(0.868)

(1.170)

-0.223

-0.068

140.096

-22.944

-1.028

-1.514

(0.382)

(0.306)

(145.180)

(41.677)

(1.535)

(1.313)

Real GDP Growth (%)

0.188**

15.211

-0.503

(0.090)

(15.427)

(0.313)

-0.584***

-60.959***

0.945***

(0.121)

(15.940)

(0.256)

-5.816**

-619.250**

-14.172**

(2.356)

(259.423)

(6.168)

-0.744

-265.750

12.651

(2.595)

(317.401)

(7.860)

President's Ideology

1.173***

105.390**

0.782

(0.352)

(42.250)

(1.291)

Polarization

10.712**

3,412.853***

2.500

(4.506)

(967.363)

(7.557)

-1.990**

-207.322

-5.336

(0.959)

(141.534)

(4.900)

Actual to Potential GDP

House Ideology

Senate Ideology

Presidential-Congress Ideological Divide

AR(1)

0.719***

-0.022

(0.107)

(0.257)

Total Receipts ($billion)

1.151***

0.848***

(0.113)

(0.183)

% Change Receipts

Intercept

T Log-Likelihood

-0.213*

-0.279**

(0.118)

(0.108)

0.921

54.486***

-70.791

4,516.641**

5.326***

-87.087***

(0.657)

(12.983)

(85.989)

(1,872.421)

(1.496)

(25.930)

53

51

53

51

52

51

-91.60

-71.67

-369.21

-316.70

-136.12

-125.72

chi2

189.867

733.307

1,084.952

3,210.938

28.828

337.640

AIC

203.207

177.349

758.417

667.406

292.246

285.448

BIC

222.910

210.190

778.119

700.247

311.758

318.289

Note: Robust standard errors reported in parentheses. Statistical significance indicated at the 1% (***), 5% (**), and 10% (*) level.

Table 4: List of Potential Donor Countries Country Name

Country Name

Country Name

Australia

Iceland

Switzerland

Bulgaria

India

Thailand

Canada

Lithuania

Tunisia

Chile

Norway

United Kingdom

Costa Rica

Peru

United States of America

Croatia

Republic of Korea

Uruguay

Czech Republic

Romania

Denmark

Singapore

Dominican Republic

South Africa

Hungary

Sweden

Figure 2: U.S. and Potential Donor Country Real Per Capita Central Government

10000 8000 6000 2000

4000

Expenditure per capita

12000

Expenditures, 1990-2012

1990

1994

1998

2002

2006

2010

Year

USA

Other Countries

Notes: Vertical line indicates the elimination of BEA90. Broken line shows the trend in the average per capita expenditure for countries in the donor pool.

Figure 3: U.S. and Synthetic U.S. Real Per Capita Central Government Expenditures,

14000 11000 5000

8000

Expenditure per capita

17000

1993-2007

1990

1994

1998

2002

2006

2010

Year

USA

Synthetic USA

Notes: Vertical line indicates the elimination of BEA90. Broken line shows the trend in per capita expenditure if BEA90 had remained in place after 2002.

Table 5: Predictors of per capita expenditure for U.S. (Treated) and Synthetic U.S. Treated Revenue per capita

Synthetic

5,705.94

5,801.39

36,275.00

29,504.29

21.66

21.83

Political Alignment

0.25

0.96

Corruption

0.67

0.77

12.51

12.22

Unemployment

5.61

6.83

GDP growth

0.03

0.03

GDP per capita Share of Population (0 to 14 Yrs old)

Share of Population (older than 65)

Notes: Reported are the means of each predictor for the USA (treated) and its synthetic match.

4 3 2 1

Expenditure growth relative 1993

5

Figure 4: Expenditure growth relative to 1993 for the USA

1990

1994

1998

2002

2006

2010

Year

USA

Synthetic USA

Notes: Vertical line indicates the elimination of BEA90. Broken line shows the trend in expenditure growth relative to 1993 if BEA90 had remained in place after 2002.

Panel B: RMSPE