By Habits or Choice? Discretionary Spending in the Oecd

Transcription

1 University of Siena From the SelectedWorks of riccardo fiorito June, 2013 By Habits or Choice? Discretionary Spending in the Oecd riccardo fiorito, University of Siena Available at:

2 Riccardo Fiorito By Habits or Choice? Discretionary Spending in the Oecd Treasury Department Rome, June GOALS Length, depth and spread of the last international contraction (2008-?) - and the differences in the recovery pace (Oecd, 2013) - made and make popular the debate on fiscal stimuli and spending multipliers. Yet, this debate does not seem conclusive: probably because before evaluating the effect of government spending on the economy, one should evaluate first if there are (and which are) spending variables under some policy control. Following Coricelli and Fiorito (2013) [henceforth: CF (2013)], I will discuss here the following preliminary question:

3 Which is the government spending share that can be considered discretionary in a number of similar countries? This will be evaluated linking discretion to reversibility and temporariness of government expenditure and without estimating the effect of this or other type of spending on the economy. The analysis focusses on government spending as revenues normally reflect their cyclical tax bases, and discretion is virtually confined to occasional tax rate changes and negligible lump-sum receipts. Empirically, we use detailed, annual, general government NIPA variables ( ) for 15 Oecd countries, chosen only when providing data long enough for a minimal time-series analysis: Austria, Belgium, Denmark, Finland, France, Iceland, Ireland, Italy, Japan, Netherlands, Norway, Spain, Sweden, UK and the US. 2

4 2. FISCAL DISCRETION IN THE EMPIRICAL LITERATURE Isolating discretionary policies is complicated as every spending component combines automatic, inertial and discretionary elements. The empirical research, however, is roughly based on three main approaches: (1) cyclically adjusted government balances (2) residuals from feedback equations (3) event chronology ( narrative approach ). 2.1 Cyclically adjusted balances Output gap elasticity is used to remove automatic stabilization from actual (CAB) or primary (CAPB) balances (Blanchard, 2003; Girouard and André, 2005). Despite their wide use by international organizations, CAB (CAPB): Do not account for recessions nor for differences in government size. Have a limited ability to remove cyclical fluctuations as Table 1 clearly shows comparing adjusted and unadjusted balances. 3

5 Table 1- Primary Balances and Cyclically Adjusted Primary Balances (CAPB) Country Range (1) Primary Balances (2) CAPB Correlation Volatility Persistence Volatility Persistence (1)-(2) Austria (5) (5).84 Belgium (8) (8).97 Denmark (8) (8).97 Finland (8) (8).95 France (8) (8).88 Iceland (5) (5).94 Ireland (5) (5).84 Italy (8) (8).97 Japan (7) (7).97 Netherlands (8) (8).91 Norway (8) (8) 1.00 Spain (8) (8).96 Sweden (8) (8).95 UK (8) (8).96 US (8) Source: Oecd, Economic Outlook. Primary balances and CAPB are percentage ratios to actual and potential GDP, respectively. Persistence is given by the Ljung-Box test. 4

6 2.2 Residuals from estimated spending equations Fatás and Mihov (2003) first estimated gov.t consumption/gdp changes as a function of real GDP changes and of a set of controls in a panel of developed and developing countries. Afonso et al. (2008) follow a similar approach, estimating expenditure and revenues in levels. Recently, Corsetti et al. (2012) estimate an Oecd panel where regression controls include government debt and exchange rate variables. In practice, all these studies estimate a government consumption equation and interpret its residuals as a measure of the discretionary spending component. However: Approximating discretion via estimated residuals confines discretionary spending to an unpredictable shock, in a non-var framework. The fact that estimated residuals are white does not imply per se discretion. On the contrary, it is plausible that discretionary spending aims at improving though temporarily - the economy: i.e. that discretionary interventions should only be less cyclical than automatic stabilizers are. 5

7 2.3 Event studies The narrative approach measures directly policy decisions, taking discretion from Laws, policy interventions or intentions: See: Ramey and Shapiro (1998); Romer and Romer (2010) postwar reconstruction on tax legislation in the US. Recently, Ramey (2011) evaluates spending news. Limits of the event approach: Policy statements (even Laws!) are policy intentions that not always result in approved budget decisions. Often, proposals are not precise enough in terms of the involved funding. Further, approved budget decisions do not necessarily imply that actual spending is the same as the approved spending in the reference year (Elmendorf, 2011). Linguistic problems confine analysis to the US only, despite two recent IMF studies (IMF, 2010; Devries et al., 2011) provide cross-national evidence. Finally, events are sort of dummies that cannot reconstruct how fiscal policy works in the light of expenditure composition and of inside and outside lags (Blinder, 2006). 6

8 3. DISCRETIONARY SPENDING: A NEW MEASURE Discretionary and non-discretionary spending are artificial constructs, having no reference to actual data. Yet, they are important for evaluating stabilization policies. This happens because automatic stabilizers do not require policy makers to know the current state of the economy and, more importantly, to have political consensus. Discretionary spending assumes instead knowledge of the current and perspective economic conditions. Discretion is generally based on the political commitment to improve, more or less immediately, the state of the economy. Table 2 Fiscal Policy Lags SPENDING AUTOMATIC DISCRETIONARY INSIDE LAG: From perception to decision (A B) None Recognition + Decision OUTSIDE LAGS: From actual spending to macroeconomic effects (B C) B C A B B C (B = B + e); e = Lag between decision and actual spending 7

9 3.1 Conceptual requirements Our measure of discretionary spending (CF, 2013) rests on economic rather than legal grounds. Namely, we rest on a few, intuitive, assumptions, that are approximated and tested via simple univariate properties of each government spending component. i) DISCRETION CANNOT BE INERTIAL Since actual policy decisions do not simply update or confirm past decisions, Discretionary spending should be less persistent than automatic stabilizers ii) DISCRETION DOES NOT IMPLY OBLIGATIONS This requirement should help to exclude a priori those variables, mostly characterized by several types of obligations (e.g. debt payments, wages, pensions). iii) DISCRETION MUST BE TEMPORARY AND REVERSIBLE Policy interventions should reflect discrete and reversible choices, i.e. temporary spending decisions that democratic governments can legally start and cancel. Commonly used measures of discretionary spending DO NOT satisfy these criteria. 8

10 3.2 Empirical implementation Besides debt, government spending includes many types of obligations (wages, pensions etc.), regardless of their legal, contractual or plainly moral nature. Empirically, the absence or the presence of less stringent obligations should not only imply a lower persistence but also a higher volatility relative to more automatic types of spending We consider the volatility and persistence of each spending variable by taking cyclical variables (HP-filter) in volume to look at their stationary time-series properties. The persistence is evaluated by the Ljung-Box (LB) statistics, whose value increases with the dependence on past. Volatility is measured via the sdev of all zero-mean variables. Combining LOW persistence and HIGH volatility should provide a reasonable signal for detecting discretion as earlier defined, independently of any particular theoretical assumption. This is not a contradiction if volatility originates from abrupt spending decisions rather than from long transmission of the interventions over time since the persistence of a timeseries affects also its variance, as we can see in the simplest AR(1) case: 9

11 (1) y(t) = ρ*y(t-1) + e(t), e(t) ~ iid (0, σ 2 ), 0 < σ 2 <, (2) Var(y) = σ 2 /(1 - ρ 2 ), 0 < ρ < 1. To account for this possible link between persistence and variance, we introduced a deflated volatility indicator, correcting volatility by persistence (LB) to better approximate the volatility due to abrupt (discretionary) interventions only: (3) DVol = Vol/LB. 10

12 3.3 Candidate variables Candidate variables satisfying in principle our criteria are the following: Subsidies (TSUB), Purchases (CGNW) and Capital spending (KT). We exclude from discretion not only Interest payments, but also Compensation of the employees and Transfers (SSPG), which are in most cases dominated by pensions. This means that we exclude from discretionary spending the non-pension component (e.g., mainly unemployment and welfare benefits), which behave more as a persistent cyclical component than as an occasional labor market intervention. Thus, our virtual government discretionary spending (GD) is obtained adding - both in nominal an real terms - the following components where Capital spending (KT) sums separate Fixed investment (IG) and Capital transfers (TKPG): (4) GD = TSUB + CGNW + [IG + TKPG]. 11

13 4. RESULTS Results refer to single variables first (see for details Table 4 in CF cit.) and then to discretionary and non-discretionary aggregates (Tables 3-4, here). 4.1 Background data To help interpretation, I summarize first a few background data on government spending composition and patterns in the Oecd sample [Tables 2a, 2b and 3 in CF cit.]: Overall, discretionary spending share declined over time, although the Great Recession reversed this tendency, mainly in those countries with smaller deficits and debt. Government consumption (CG = WAGES + CGNW) is about everywhere the largest spending variable ( 40%-50%), followed almost everywhere by Social security ( 30%- 40%) which contains Pensions and Welfare expenditures. Usually, pensions dominate except than in the Northern countries, Ireland and the UK. Capital spending is the third ( 10%), much smaller and volatile, component. Investment is low except for Japan. Finally, Government size tends to fall over time and is rather small in Japan where, however, the (gross) debt/gdp ratio is indeed very high. 12

14 4.2 Discretionary spending by variable Leaving more details to CF Table 4, I summarize here major results: While there are obvious differences by variable and country, most of the reported statistic conform to our assumption that discretionary spending should be more volatile, and less persistent than automatic spending is. In general, persistence results differ more by country and by variable. Deflating volatility from persistence makes stronger this overall evidence. In most cases Capital transfers (KT) are the most volatile spending variable and this is maintained using the deflated volatility index (DVol). Conversely, Fixed investment has some persistence induced by the time-to-build technology. The last discretionary variable in our scheme is given by Subsidies. Given the policy nature of this variable, cross-countries differences are not surprising. Within non-discretionary spending, the Welfare component is generally characterized by a high degree of cyclicality (Adema et al., 2011) which is inconsistent with the standard view that this type of expenditure should reflect temporary interventions only. 13

15 4.3 Aggregate discretionary spending The discretionary spending share (GD) is about 1/3 of total spending (Table 3). This is much larger than the share obtained from estimated residuals, but is much smaller than typically assumed in the earlier econometric models. Although there are significant differences across countries, the non-discretionary share (GN) is generally increasing over time though this is reversed in the Great Recession. Looking at the components of discretionary spending, patterns across countries are similar: purchases are about 2/3 of discretionary spending and are always and everywhere the largest component, generally increasing in the last period. Capital spending ranks second, at about the 20% of the aggregate, with country differences reflecting unusual episodes (Ireland) or structural patterns (US, Japan). In all cases, Subsidies are the third and smallest discretionary component (5%- 10%). The components of non-discretionary spending (GN) differ more markedly among countries, partially because of Social security and Interest spending differences. 14

16 Table 3 - Discretionary (GD) and non-discretionary spending (GN) by period Country GD GN GY GD GN GY GD GN GY GD GN GY Austria Belgium Denmark Finland France Iceland Ireland Italy Japan Netherlands Norway Spain Sweden United Kingdom United States Source: OEC,D EO90 cit. Average data; GD = Discretionary spending; GN = Non-discretionary spending; GY = Gen. Govt. Spending/Nominal GDP 15

17 4.4 Comparing Different Aggregates In Table 4 we compare the persistence and volatility indicators relative to both discretionary and non-discretionary government spending aggregates in each country, considering also total and primary spending. Regarding the comparison between primary and total government spending, the main finding is that discretionary spending (GD) is always more volatile than nondiscretionary spending (GN), and more volatile as well than primary and total government spending. Furthermore, non-discretionary spending is generally less volatile than primary (GP) and total government spending (GT). While total expenditure by definition cannot separate its components, it is interesting to note that primary spending (GP) i.e. the first empirical proxy for discretion does not comply with our criteria, since not only is more volatile than the GN aggregate but often is also more persistent. 16

18 Deflating volatility from persistence, the Dvol index does not affect the volatility ranking: discretionary spending volatility is still the highest in all cases but Norway, confirming our priors even in a stronger way. The persistence indicators, reported in Table 4, broadly support the volatility results, though in a weaker way. In ten out of fifteen cases, the non-discretionary spending aggregate is more persistent then the discretionary aggregate. In the remaining cases, however, inertia in discretionary spending is either the same as non-discretionary spending (in Spain), or higher (in Belgium, Italy, Norway and Sweden). By looking at the individual variables, this result could either reflect a rising share of somewhat inertial purchases or the time needed to activate and complete fixed investment decisions. 17

19 Table 4 Persistence and volatility of Different Types of Government Spending Austria (1) Vol (2) LB (3) = (1)/(2) Dvol Belgium (1) Vol (2) LB (3) =(1)/(2) Dvol Denmark (1) Vol (2) LB GD GD GD GN GN GN GP GP GP GT GTOT GTOT Finland (3) = (1)/(2) Dvol France Iceland GD GD GD GN GN GN GP GP GP GT GT GT Ireland Italy Japan GD GD GD GN GN GN GP GP GP GT GT GT Netherlands Norway Spain GD GD GD GN GN GN GP GP GP GT GT GT Sweden UK US GD GD GD GN GN GN GP GP GP GT GT GT Source: OECD EO90 cit. All indicators refer to deflated cyclical deviations from an annual HP trend; Vol is the standard deviation of the cyclical data. LB(p) is the Ljung-Box statistics where p = T/4 is the number of autocorrelations and T is the number of data points. The LB(p) statistics is used as an indicator of persistence; Dvol = Vol/LB is an indicator of deflated volatility, i.e. of the volatility not induced by the estimated persistence. 18

20 5. CONCLUSIONS In most countries, discretionary spending is about 1/3 of the total. The majority of spending is more automatic and historically rises, with the exception of last crisis. The fact that policy decisions involve only 1/3 of total spending does not say too much about the quality or the efficacy of the decision. Our evidence suggests only that economic policy changes are not too frequent, regardless if they are good or bad. Probably, economic policy should focus more on reducing the reasons of persistence. Inertia contributes to make government spending procyclical since built-in stabilizers do not refer to theoretical cycles but to the way in which actual economies evolve: in the Oecd, about 90% of cases growing and facing contractions otherwise (Fiorito, 2013). This procyclical (or acyclical) pattern stimulates spending when it is not necessary and prevents spending in the few recession cases, especially when fiscal constraints due to past inertia make difficult since not credible - promises of temporary spending. 19

21 REFERENCES Adema W., P. Fron and M. Ladaique (2011), Is the European Welfare State Really More Expensive? Indicators on Social Spending ( ) and a Manual to the Oecd Social Expenditure Database, Oecd WP # 124. Afonso A., L. Agnello and D. Furceri (2010), Fiscal Policy Responsiveness, Persistence and Discretion, Public Choice, 145, Blanchard O.J. (1990), Suggestions for a New Set of Fiscal Indicators, Oecd WP # 79. Blinder A.S. (2006), The Case Against the Case Against Discretionary Fiscal Policy, in Kopcke, R., G. Tootell and R. Triest (eds.), The Macroeconomics of Fiscal Policy, The MIT Press, Coricelli F. and R. Fiorito (2013), Myths and Facts about Fiscal Discretion: A New Measure of Discretionary Expenditure, Documents de Travail du Centre d Economie de la Sorbonne, Corsetti G., A. Meier and G.J. Müller. (2012), What Determines Govenment Spending Multipliers?, IMF Working Paper 12/

22 Devries P., R. Guajardo, D. Leigh and A. Pescatori (2011), A New Action-Based Data Set of Fiscal Consolidation, IMF Working Paper 11/128. Elmendorf, (2011), Discretionary Spending, Congressional Budget Office, United States. Fatás A. and I. Mihov (2003), The Case for Restricting Fiscal Policy Discretion, Quarterly Journal of Economics, 118(4), Fiorito R. (2013), Business Cycles and Recessions in the Oecd Area, Modern Economy, 4, Girouard N. and C. André (2005), Measuring Cyclically Adjusted Budget Balances for Oecd Countries, Oecd WP # 434. IMF (2010), Will It Hurt? Macroeconomic Effects of Fiscal Consolidation, in World Economic Outlook, Ch. 3, October, Oecd (2013), Economic Outlook # 93, May [Preliminary version]. Ramey, V.A. (2011), Identifying Government Shocks: It s All in the Timing, Quarterly Journal of Economics,

23 Ramey, V.A. and M. D. Shapiro (1998), Costly Capital Reallocation and the Effects of Government Spending, Carnegie-Rochester Conference on Public Policy, 48(1), Romer C. and D. Romer (2010), The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks, American Economic Review, 100: