Fiscal Policy and Recessions

Transcription

1 University of Siena From the SelectedWorks of riccardo fiorito May, 2012 Fiscal Policy and Recessions riccardo fiorito, University of Siena Available at:

2 1. Goals and motivations Recessions are different from negative cycles Negative (and positive) cycles involve 50% of cases. Recessions are much less frequent This has important implications for fiscal policy Riccardo Fiorito University of Siena Fiscal Policy and Recessions May 2012 Here, Oecd annual data are used to show: the frequency and the depth of postwar recessions Number of countries involved in major recession episodes Government Spending: Really countercyclical? Always useful? Discretionary spending: measure, timing and possible effects.

3 Graph 1: Textbook business cycle phases F trend Actual output B E output gap (positive) C output gap (negative) A D Output gaps ( y t - y* t ) refer to the areas ABC and CDE DGAP = d t - d* is the growth gap, i.e. the difference between actual (d t ) and potential growth (d* t ) Recessions d t < d* t occur in phases BC and CD. Phase1 = AB: d t > d* t and y t > y* t ρ(dgap, YGAP) >0 Phase2 = BC: d t < d* t and y t > y* t ρ(dgap, YGAP) <0 Phase3 = CD: d t < d* t and y t < y* t ρ(dgap, YGAP) >0 Phase4 = DE: d t > d* t and y t < y* t ρ(dgap, YGAP) <0 2

4 2. What the data say instead? Table 1: Business cycle phases and recessions in a number of Oecd countries Country Number of observations Phase1 AB (44%) Phase2 BC (3%) Phase3 CD (9%) Phase4 DE (44%) Recessions (12%) Austria (.42) (.08) (.50) Belgium (.47) (.03) (.07) (.43) Denmark (.46) (.05) (.13) (.33) Finland (.45) (.10) (.45) France (.47) (.03) (.06) (.44) Iceland (.47) (.07) (.13) (.33) Ireland (.35) (.05) (.10) (.50) Italy (.45) (.02) (.08) (.45) Japan (.48) (.02) (.09) (.41) Netherlands (.51) (.02) (.07) (.39) Norway (.44) (.04) (.52) Spain (.40) (.09) (.51) Sweden (.40) (.04) (.11) (.45) UK (.32) (.07) (.10) (.50) US (.46) 2 (.04) 5 (.10) 20 (.40).14 Legend : Oecd, Economic Outlook database. Numbers in parenthesis denote the percentage frequency of each phase. 3

5 A few stylized facts summarize what Table 1 and other analyses (Fiorito, 2012) say instead: Expansion phases (AB+DE) dominate (88%). AB and DE have the same frequency (.44) Recession phases (BC+CD) are much less frequent (12%) Within recessions, the BC portion (3%) occurs much less than the CD phase does (9%) Overall, in the full Oecd sample (Table 2), annual recessions involve about the 10% of cases Strong recessions - i.e. GDP contractions of at least 2% (Table 2) - are about the 4% of cases. Excluding last crisis, recessions and strong recessions are 8.2% and 2.5% of cases, respectively Recovery cases (DE) (d t > d* t ) still belong to negative output gap cases when fiscal expansion is recommended.. Major recession episodes Number of countries involved (Table 3) first oil shocks 12 countries over Early 80s: a mixture between the 2 nd oil shock and a (regional) Northern countries crisis: 8 (1981), 6 (1982) cases 3. Early 90s crisis: EMS fall Lira, Peseta and UK Pound devaluation 12 countries 4. Last crisis peak in 2009: 27 Oecd countries over 30 have been involved! Length of recessions (quarterly data) Most recessions last about 1-2 years. Using quarterly data (Table 4), recessions are more frequent but are still less frequent than negative output cycles (50%). A distinction between contractions and recessions (at least two consecutive quarters, is also made (Shiskin, 1974). 4

6 Table 2 Recessions in the Oecd ( ) Country Recessions Strong recessions Country Recessions Strong recessions Australia 61, 83, 91 Korea 80, Austria 77, 81, Luxembourg 75, 81, 09 75, 09 Belgium 75, 93, Mexico 82, 83, 86, 95, 09 83, 86, 95, 09 Canada 82, 91, 09 82, 91, 09 Netherlands 75, 81, 82, 03, Czech Republic New Zealand 67, 68, 77, 78, 91, 08, 67, 68, 78, Denmark 75, 80, 81, 93, 08, 09 81, 09 Norway 88, Finland 76, 90, 91, 92, 93, 09 91, 92, 09 Poland France 75, 93, Portugal 75, 83, 84,93, 03, 09, 75, 93, Germany 67, 75, 82, 93, 03, Slovenia Greece 74, 81, 82, 83, 87, 93, 74, 87, 09, 10, 11 Spain 81, 93, 09, , 10, 11 Hungary 92, 93, 09 92, 09 Sweden 77, 81, 92, , 09 93, Iceland 61, 83, 92, 02, 09, 10 92, 02, 09, 10 Switzerland 75, 76, 82, 91, 93, 03, Ireland 86, 08, 09, 10 08, 09 Turkey 79, 80, 94, 99, 01, 09 80, 99, 01, 09 Italy 75, 93, 08, 09 75, 09 United Kingdom 74,75, 80, 81, 91, 09 80, 09 Japan 74, 98, 99, 02, 08, 09, United States 74, 75, 80, 82, 91, * Legend: Oecd, Economic Outlook Database. Data for 2001 stem from # 90 Economic Outlook (December 2011). Strong recessions indicate at least 2% real GDP contractions with respect to previous year. 5

7 Table 3 Number of countries involved in recession episodes Year Recession Strong recessions Year Recession Strong recessions /26 1/ /28 3/ /26 4/ /30 2/ /26 1/ / / /30 1/ /26 1/ /30 23/ /28 3/ /34 2/ /34 1/ * 4/34 2/34 Source: See Table crisis in Asia for non-oecd countries is not calculated; * The 2012 forecast refers to the # 90 Oecd Economic Outlook (December 2011). Table 4: G-7 recessions frequency and length for quarterly data * Country Sample Number of observations Canada United States Japan United Kingdom Germany France Italy Number of contractions Number of recessions Longest recession phase (.18) 17 (.14) (.12) 10 (.08) (.28) 15 (.12) (.16) 15 (.12) (.26) 13 (.16) (.13) 10 (.08) (.26) 22 (.18) *Source: Oecd, Economic Outlook database (s.a. real GDP data, March 2012). 6

8 3. A Possible Reconciliation between Growth and Cyclical Views Let both actual (y) and potential output (y*) be a unit root process: (1) y t = d t + y t-1, (2) d t = d + a(l)e t, e t ~ iid(0, σ 2 ), a 0 = 1 (3) y* t = d* t + y* t-1 (4) d* t = d + a*(l)e* t, e t ~ iid(0, σ 2* ), a * 0 = 1, σ 2* <σ 2 <. Variables are in logs and then cycles (Coricelli-Fiorito, 2009) can be expressed as the sum of the growth gap (d t d* t ) and of the previous cycle, it being c t-1 = y t-1 y* t-1 : (5) c t = (d t d* t ) + ( y t-1 y* t-1 ), d* t > 0, where d* can be obtained by the HP filter. Implications: Recessions (d t <0) are more volatile than normal changes since the growth gap (d t d* t ) variance must be necessarily bigger once potential growth (d* t ) is positive.. c t = (d t d t * ) business cycle changes = growth gap. 7

9 Fig. 1 Canada: Business cycles (CCAN), actual (DCAN) and potential (DCANHP) rates of growth CCAN DCAN DCANHP Fig. 2 United States: Business Cycles (CUS), actual (DUS) and potential (DUSHP) rates of growth CUS DUS DUSHP 8

10 Fig. 3 Japan: Business Cycles (CJAP), actual (DJAP) and potential (DJAPHP) rates of growth CJAP DJAP DJAPHP Fig. 4 United Kingdom: Business Cycles (CUK), actual (DUK) and potential (DUKHP) rates of growth CUK DUK DUKHP 9

11 Fig. 5 Germany: Business Cycles (CGER), actual (DGER) and potential (DGERHP) rates of growth CGER DGER DGERHP Fig. 6 France: Business cycles (CFR), actual (DFR) and potential (DFRHP) rates of growth CFR DFR DFRHP 10

12 Fig. 7 Italy: Business cycles (CIT), actual (DIT) and potential (DITHP) rates of growth CIT DIT DITHP 4. Fiscal Implications Automatic Stabilizers Advantages: Data release takes time and the present must be predicted too! Automatic stabilizers avoid the need of forecasting and even more important - of Obtaining political consensus. Limits: Apply in nominal terms and refer to actual data rather than to estimated output gaps. If they apply too much to government spending, spending becomes heavy, procyclical and difficult to cut. Why is it so? Again, because expansions are much more frequent than contractions! 11

13 Government spending: really countercyclical? Traditional views: government spending is countercyclical in developed countries, being procyclical in developing ones because of inefficiency and corruption (Talvi - Vegh, 2000; Alesina, Campante, Tabellini, 2008). Careful analysis of disaggregated spending (Fiorito, 1997; Lane, 2003) shows instead that government spending is not countercyclical in the Oecd area too, though there are differences by variable and country. The simple fact that government debt is high shows that government spending is not countercyclical but often acyclical or inertial. Why? Positive and negative cycles should offset each other as in Graph 1! Wrong remedies: cyclically adjusted (primary) balances (CAPB) Basically, CAPB corrects actual balances for the output gap: (5) f*(t) = f(t) - α*[output gap]. This is done to make fiscal policy more countercyclical (stabilizing) and to measure discretion (Blanchard, 1990), implicitely meant as a business cycle responsive tool. The widely accepted value α ½ for the output gap correction (Girouard-André, 2005) implicitly assumes a zero government spending elasticity to the business cycle and a unit tax elasticity. 12

14 5. Discretionary spending is measured in 3 ways: Cyclically adjusted deficits Regression residuals from fiscal variable estimates (Fatás and Mihov, 2003) Event Studies (Romer and Romer, 2010; Ramey, 2011). 5.1 Cyclically adjusted deficits Despite actual data in Tables 1-2, government balance corrections adjust for the cyclical component only However, recently the IMF (WEO, 2010) criticizes the CAPB as stabilization tool. 5.2 Regression residuals: Fatás and Mihov (2003) use in a panel of countries a 2-stage procedure 1 st stage: DCG DY + controls + u(t), where the u(t) residuals should measure discretionary spending. 2 nd stage: The volatility of each country residual (σ 2 i) growth in the 2 nd stage cross section. Government spending is confined to consumption and discretion is measured by DCG residuals 3. Event studies: informations taken by laws, presidential speeches and then made quantitavive: Romer and Romer (2010) postwar reconstruction on tax legislation in the US. In a similar study V. Ramey (2011), obtains smaller VAR multipliers than Blanchard-Perotti (2002) and shows that government purchases help recovery. IMF study (2010) shows that fiscal contraction does not help growth as claimed by Alesina and Ardagna (2010). 13

15 6. A simple alternative Confining discretion to government spending only, definition naturally applies to spending decided to face crises or transitory emergency situations. Thus, discretion should apply more to recessions than to negative cycles to be accomodated by automatic stabilizers. This definition has been used for the Oecd countries by Coricelli and Fiorito (2009) on the basis of few criteria: i) Methods should be easily applied to several countries ii) Different spending items should be evaluated and then aggregated iii) Criteria should not be too subjective or ad hoc. THREE DISCRETION REQUIREMENTS: 6.1 DISCRETION SHOULD AVOID INERTIA: Discretionary policy does not have to be so inertial as many government spending components are (Fiorito, 1997). An implicit requirement is also that spending components must be evaluated separately and cannot be confined to government consumption only. 2. DISCRETION REQUIRES NO OBLIGATION: Discretionary spending should not reflect any type of obligation, regardless if legal, contractual or even moral: typical obligations are not only the payment of debt interests but also payment of employees (usually, about 2/3 of government consumption) and pensions (often the highest transfer). 3. DISCRETIONARY SPENDING SHOULD NOT BE PERMANENT The third requirement is that discretionary spending must be temporary and revocable: as when a specific spending aims at obtaining an immediate goal and does not need to keep on after the goal is achieved. 14

16 These requirements: do not imply that discretionary spending must be a white noise process: they only imply that discretionary spending should be volatile but also less persistent than real GDP. This can also imply a conflict since volatility is increased by persistence as several data show. In particular, discretionary spending should be less persistent than automatic stabilizers are, though some persistence can be induced by the fact investment spending projects take neceassirily time. To measure discretionary spending, Coricelli and Fiorito (2009) sum (and also deflate) the following variables: 1. Government purchases (CGNW), i.e. about 1/3 of government consumption in the Oecd countries. Purchases buy on the market the intermediate inputs, necessary for providing government services. 2. Capital expenditure (IGAA+TKPG): potentially, is a growth enhancing factor (e.g. infrastructures). Thus, single projects must be activated only when necessary and then stopped once the project is accomplished. 3. Welfare and unemployment insurances (SSPG PENSIONS): empirically, this (heterogeneous) variable can be calculated as the difference between household transfers and pensions. These transfers (in Italy, e.g. CIG) must be temporary and conditional to make more flexible labor market adjustment. 4. Finally, firm subsidies (SUBSIDIES) are also added, though usually their share is small or negligible. MAIN RESULT: This measure of discretionary spending (GD) roughly corresponds to 1/3 of total spending so that most of government spending seems to be automatic (GN), i.e. resulting more from habits than from decisions. 15

17 Table 5: Discretionary (GD) and non discretionary (GN) % spending given gov.t spending/gdp (GY) share* 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: Average data based on nominal ratios and on national samples (first subperiod); GY = total government spending/gdp ratio. 16

18 What Table 5 basically shows? In most cases, discretionary spending (GD) is about 1/3 of total spending In some cases (AT, BE, FR, SW) the discretionary share is about constant over time In other cases GD (DE FI, I, UK, US) share tends to fall over time, strongly in Japan. Overall, the tendency to increase the government spending/gdp ratio (GY) is due to an increase of the automatic (GN) component, probably because of social security. In the last decade, Japan has the smallest G/Y ratio but also the highest GD share (.50%) while Northenrn countries such as Denmark and Sweden have the highest G/Y share but a small (D) or a normal (SW) GD share. Other set of results can be drawn from the following Table 6 and 7 where more details are available: As expected, discretionary spending is always more volatile than automatic spending but is less persistent just in about half of cases. There is in other words, an apparent difficulty in revoking decided expenditures. In some cases, the goal of reducing deficit/debt ratios contributed to increase the non-discretionary share. Looking at the components for a smaller number of countries (Tables ) there are interesting differences, though capital spending generally conforms to our criteria. It must also noticed that differences in the composition of social spending (SSPG) can affect the aggregate results since pension share is always little volatile. 17

19 Table 6: Discretionary (GD) and Non-Discretionary Spending (GN) Cyclical patterns Country Discretionary spending (GD) Non-Discretionary Spending (GND) GD/GN GDP Volatility Persistence (LB) Correlation with GDP Volatility Persistence Correlation with GDP contemporaneous correlation persistence (LB) Austria ρ(g(0), y(0)) Belgium Denmark Finland ρ[g(+1), y(0)] France Iceland ρ[g(-1), y(0)] Ireland ρ[g(+1), y(0)] Italy ρ[g(+1), y(0)] Japan Netherlands ρ[g(-1), y(0)] ρ[g(-1), y(0)] Norway ρ[g(-1),y(0)] Spain ρ[g(+1),(0)]] Sweden ρ(g(0), y(0)) ρ(g(+1), y(0)) UK ρ[g(-1), y(0)] United States ρ(g(0), y(0)) Legend: Data are from the Oecd (November, 2011) database. All statistics refer to deflated cyclical deviations from the HP trend (6.25 is the smoothing parameter); Volatility is relative to GDP; Cross-correlations with GDP refer to the highest value in the interval [(g(-1),y(0); g(0), y(0); g(+1), y(0)] where g and y denote the spending and the GDP variable, respectively. LB is the - asymptotically ~ χ 2 (p) - Ljung-Box portmanteau statistics where T is the number of observations and p = T/4 is the number of sample autocorrelations. 18

20 Government Spending Components in A Few Countries 1 Table 7.1: France ( ) Variables Volatility Ρ(G(t),Y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.65) WELFARE (.35) Table 7.2: ITALY ( ) Variables Volatility Ρ(G(t),Y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.76) WELFARE (.24) Table 7.3: SPAIN ( ) Variables Volatility ρ(g(t),y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.64) WELFARE (.36) 1 Symbols are those of the Oecd database and are explained at page 15. CGW is the compensation of employees providing government consumption (CG). 19

21 Table 7.3: SWEDEN ( ) Variables Volatility ρ(g(t),y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.44) WELFARE (.56) Table 7.4: UK ( Variables Volatility ρ(g(t),y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.40) WELFARE (.60) Table 7.5: US ( Variables Volatility ρ(g(t),y(t)) Persistence (LB) CGNW IGAA TKPGQ TSUB WELFARE PENSIONS CGW SSPG (Social security) % Shares PENSIONS (.55) WELFARE (.45) 20

22 7. Conclusions The main conclusion is that discretionary government spending should be used in recessions only to be stopped afterwards. This implies that this type of spending is revocable and has also some efficacy. This study does not produce evidence on the efficacy: this is still under scrutiny in a related research. My previous evidence (Fiorito, 1997) on government spending in the G-7 excludes that most government spending items anticipate procyclically real GDP changess: the only variable obtaining this result for all countries is the government purchases and this is also what has been recently obtained by Ramey (2011) for the US. The Italian paradox is that it is impossible help the economy when it could be necessary because most of the excess spending was done when it was not necessary, at least for economic reasons. This is basically the main reason why Italy s government debt is so big and so hindering fiscal discretion. If the policy goal is reducing excess government spending, the focus should be more on cutting the automatic component, especially in recession times. In other words, reforms deal more with affecting the transmission mechanism of automatic stabilizrs than with cutting temporary spending, provided discretionary spending is really temporary and also really useful. 21

23 References Alesina, A., Campante F.R. and Tabellini, G. (2008), Why Is Fiscal Policy Often Procyclical?, Journal of the European Economic Association, September, Alesina, A. and Ardagna S. (2010), Large Changes in Fiscal Policy: Taxes versus spending, NBER WP # Blanchard, O.J. (1990), Suggestions for a New Set of Fiscal Indicators, Oecd WP # 79. Blanchard, O.J. and Perotti R. (2002), An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output, Quarterly Journal of Economics, 113(3): Coricelli, F. and Fiorito, R. (2009), Output Gap, Recessions and Fiscal Discretion, Case Conference, Warsaw, November. Fatás, A. and Mihov, I. (2003), The Case for Restricting Fiscal Policy Discretion, Quarterly Journal of Economics, 118(4), Fiorito, R. (1997), Stylized Facts of Government Finance in the G-7, IMF Working Paper # 97/142. Fiorito, R. (2012), Recessioni, Cicli e Politica Fiscale, Rivista di Politica Economica, 1-3, Gavin, M. and Perotti, R. (1997), Fiscal Policy in Latin America in: B. Bernanke and J. Rotemberg (eds.), NBER Macroeconomics Annual 1997, MIT Press. Girouard, N. and André, C. (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,

24 Hodrick, R. and Prescott, E. (1997), Post-war US Business Cycles: An Empirical Investigation, Journal of Money, Banking, and Credit, 29:1-16. Lane, P.R, (2003), The Cyclical Behaviour of Fiscal Policy: Evidence from the Oecd, Journal of Public Economics, 87: Ramey, V.A. (2011), Identifying Government Shocks: It s All in the Timing, Quarterly Journal of Economics, Romer, C., and Romer, D. (2010), The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks, American Economic Review, 100: Shiskin, J. (1974), The Changing Business Cycle, New York Times, December 12. Talvi, E. and Vegh. C.A. (2005), Tax Base Variability and Procylical Fiscal Policy in Developing Countries, Journal of Development Economics, 78,