Fiscal policy in Brazil: from countercyclical

Transcription

1 Working Paper 407 Novembro de 2015 Fiscal policy in Brazil: from countercyclical response to crisis Márcio Holland

2 Os artigos dos Textos para Discussão da Escola de Economia de São Paulo da Fundação Getulio Vargas são de inteira responsabilidade dos autores e não refletem necessariamente a opinião da FGV-EESP. É permitida a reprodução total ou parcial dos artigos, desde que creditada a fonte. Escola de Economia de São Paulo da Fundação Getulio Vargas FGV-EESP TEXTO PARA DISCUSSÃO 407 NOVEMBRO DE

3 Fiscal policy in Brazil: from counter-cyclical response to crisis Márcio Holland São Paulo School of Economics at Getulio Vargas Foundation, Brazil 1 This version: November 10 th, 2015 Abstract The main goal of this article is to identify the dynamic effects of fiscal policy on output in Brazil from 1997 to 2014, and, more specifically, to estimate those effects when the output falls below its potential level. To do so, we estimate VAR (vector autoregressive) models to generate impulseresponse functions and causality/endogeneity tests. Our most remarkable results indicate the following channel of economic policy in Brazil: to foster output, government spending increases causing increases in both tax rates and revenue and the short-term interest rate. A fiscal stimulus via spending seems efficient for economic performance as well as monetary policy; however, the latter operates pro-cyclically in the way we defined here, while the former is predominantly countercyclical. As the monetary shock had a negative effect on GDP growth and GDP growth responded positively to the fiscal shock, it seems that the economic policy has given poise to growth with one hand and taken it with the other one. The monetary policy is only reacting to the fiscal stimuli. We were not able to find any statistically significant response of the output to tax changes, but vice versa seems work in the Brazilian case. 1 Corresponding author: marcio.holland@fgv.br. This working paper was partially written during my mandatory cooling-off period when I spent part of this time as a visiting scholar at Columbia University in the City of New York and for this reason I am very grateful to Professors Albert Fishow, Thomas Trebat, Gray Newman, Sidney Nakahodo, Gustavo Azenha, and Daniella Diniz, for having me and giving me all support I needed to develop my research. However, all opinions expressed herein represent those of the author. 1

4 1. Introduction The 2008 international financial crisis put fiscal policy at the forefront of debate, particularly its use in mitigate the painful effects of the crisis on outputs and employment. Probably because of the lack of widely recognized rules, fiscal policy is generally a controversial issue. The economic meltdown with its deep and protracted impact on both goods and labor markets presented the perfect opportunity to approach divergent views about its use. For a few years after the 2008 crash, there was no room for austerity until government debts skyrocketed. Alas, governments had to shift towards fiscal retrenchment, even under economic weakness. The results of recent fiscal policy have been mixed, and its effectiveness remains a disputed issue. On the other hand, a stream of literature has conducted empirical studies with novel methodologies in an effort to identify the dynamic and contemporaneous effects of fiscal policy on outputs. Auerbach and Gorodnichenko (2012) estimate a government purchase multiplier for a large number of OECD countries using a specific form of the STVAR (smooth transition vector autoregressive) model and have identified a fiscal multiplier in both recession and expansion circumstances. The model might be considered a refinement of Blanchard and Perotti s (2002) specification, who used a simplified structural VAR model. In 2014, Mineshima et al. introduced a TVAR (threshold vector autoregressive) model to use when regimes are determined by a transition variable, which is either exogenous or endogenous. More recently, Herwartz and Lütkepohl (2014) proposed a structural vector autoregression model with Markov switching that combines conventional and statistical identification of shocks, which can be useful for future studies on fiscal policy. The main goal of this article is to identify the effect of fiscal shocks on the output in Brazil from 1997 to 2014, and, more specifically, to estimate those effects when the output falls below its potential level, as observed during most of the period following the 2008 international meltdown. We pose an additional question: How low can the output drop when it is already below its potential level (which we call negative initial conditions for the fiscal consolidation program), despite the composition of the fiscal retrenchment? In other words, after the output has bottomed out, how can policymakers quickly instill confidence by cutting expenditures and increasing taxes when these measures were supposed to be implemented an expansionary shock? To do so, we estimate VAR (vector autoregressive) models to generate impulse-response functions and causality tests. Our most remarkable results indicate the following channel of economic policy in Brazil: to foster output, government spending increases, causing increases in both tax rates and revenue and the short-term interest rate. A fiscal stimulus via spending seems quite efficient for economic performance as well as monetary policy; however, the latter operates pro-cyclically, while 2

5 the former is predominantly countercyclical. We were not able to find any statistically significant response of the output to tax changes, but we did find a statistically significant response of tax changes to the output in the Brazilian case. The contributions of this work are threefold: first, it identifies the response of the output to fiscal and monetary policy; second, it estimates the impact of recent fiscal measures on the output; and, third, in line with broader economic literature, it suggests a long-term fiscal plan for Brazil to spur its effectiveness in reducing output losses. This work is divided in the following sections. The next section shows Brazil s recent experience with fiscal policy and its main results. The third section presents the literature and discussions of the effectiveness of fiscal policy. The fourth section extrapolates the impulse-response functions obtained in our research to measure the potential impact of the 2015 fiscal plan on output and confidence. Ultimately, it is the appropriate moment to present the advantages and caveats of our empirical procedures. We are most aware of the many drawbacks presented in this sort of time series analysis. At the end of this section, we suggest a long-term fiscal plan for Brazil using lessons learned from both Brazil s recent experience with fiscal policy and our empirical study. 2. The Brazilian Context Governments across the globe responded to the 2008 crisis with unprecedented expansionary actions in recent economic history. Monetary and fiscal countercyclical actions were implemented to both stymie the contamination of the international crisis in financial systems and to resume growth as soon as possible. From 2008 to 2010, fiscal and monetary stimuli were overwhelmingly recommended. However, since 2010, the focus has shifted to fiscal consolidation in advanced economies. Since then, fiscal results have improved over the world, even though the debt-to-gdp ratios remain high compared with those before the crisis. More recently, the United States has outperformed the euro area, where calling for austerity appears to have fallen out of fashion again, as illustrated in the 2015 Greek case. The fiscal front has deteriorated dramatically in many advanced economies with mixed and far from outstanding achievements. However, comparing the 1929 Great Depression with the 2008 Great Recession, Eichengreen (2015:2) reflects on these crises as follows: as a result of this different 3

6 response, unemployment in the United States peaked at 10 percent in Though this was still disturbingly high, it was far below the catastrophic 25 percent scaled in the Great Depression. Due to this thought, the Brazilian government took a series of countercyclical policies to protect the local economy from crumbling. These policies seemed to work well, at least until The worst of the crisis was absorbed without any major disruption in the Brazilian economic system. Most importantly, the economy resumed growth in the 2 nd quarter of 2009; the unemployment rate did not spike; real wages continued to grow; and consumer and business confidence recovered very quickly. Nevertheless, after a period of recovery until 2013, the overall growth remained disappointing, particularly given the very rapid deterioration in 2014, the expected strong contraction in 2015, and uncertainty about 2016 performance. From late 2014 to early 2015, Brazil launched a tight fiscal program. The pro-cyclical biased fiscal consolidation plan is presumably considered the only plausible policy stance when solvency became the issue rather than economic activity. In Brazil, the diagnosis and prescription have been far from divergent. However, as highlighted by Frankel (2012), a pro-cyclical fiscal policy magnifies the severity of the business cycle. This controversy motivated this research to assess whether such fiscal consolidation policies are expected to hurt GDP growth to a greater extend because the economy is already under contraction. Alternatively, would they spur confidence so that the drag on economic activity could be avoided? There is no doubt that fiscal stimuli were needed at that challenging time, when the Brazilian economy was severely hit by the financial crisis. As for fiscal policy, there were considerable tax exemptions in As a result of the actions taken by the government at that time, the country was able to recover very quickly from the crisis; among other things, Brazil experienced 7.6 percent and 3.9 percent growth in 2010 and 2011, respectively (see table 1). According to the national account s new dataset, released on March 27, 2015 by the National Bureau of Statistics (IBGE), the 2008 crises hit the economy harder and deeper, and the recovery was faster and better because investment resumed quickly and included a greater share of the GDP compared with the previous period of the crisis, regardless of its volatility (see figure 1 and 2). In early 2011, the Federal Administration was able to start applying a fiscal consolidation plan to cool down the economy. Needless to say, solvency had not been a problem in Brazil for a long time, as international reserves have been considerable and more than enough to pay for external liabilities; in addition, the public debt-to-gdp ratio had been decreasing over the years. Thanks to Brazil s reaction to the crash, the general gross debt-to-gdp ratio increased 3.3% from 2008 to 2009, 4

7 which could be considered incredibly low in comparison with the debt dynamics in advanced economies after the crisis. However, the general gross debt-to-gdp ratio is still high in Brazil compared with that of its peers, although it had been relatively stable, even during most of the period of countercyclical fiscal policy (see table 1 and figures 2 and 3 for the fiscal results). Table 1. Brazil: Key Macroeconomic Indicators after the 2008 Crisis ( ) * 2016* Real GDP Change (%) 5,0-0,2 7,6 3,9 1,8 2,7 0,1-3,5-3,0 Unemployment Rate year average (%) 7,9 8,1 6,7 6,0 5,5 5,4 4,8 6,8 7,5 Investment Change (%), eop 12,7-1,9 17,8 6,6-0,6 6,1-4,4-12,0-5,0 CPI Inflation - IPCA (%), eop 5,9 4,3 5,9 6,5 5,8 5,9 6,4-10,5-8,5 Benchmark Interest Rate (%), eop 13,75 8,75 10,75 11,0 7,25 10,0 11,75 14,25 14,25 Current Account (% of GDP) -1,7-1,5-2,1-2,0-2,3-3,4-4,4-4,5-3,5 FDI (US$ billion) 45,1 25,9 48,5 66,7 65,3 64,0 96,9 60,0 50,0 Foreign Reserves (US$ billion) Exchange Rate (Real per USD) eop 2,34 1,74 1,67 1,88 2,04 2,35 2,65 4,0 4,5 Primary Result (% of GDP) 3,3 1,9 2,6 2,9 2,2 1,8-0,6-2,1-0,5 Nominal Result (% of GDP) -2,0-3,2-2,4-2,6-2,3-3,1-6,2-11,1-9,5 Gross G. Govt Debt (% of GDP) 56,0 59,3 51,8 51,3 54,8 53,3 58,9 69,0 72,0 Net Public Debt (% of GDP) 37,6 40,9 38,0 34,5 32,9 31,5 34,1 33,0 34,0 Notes: Unemployment rate is yearly average of the Monthly Employment Survey (PME); CPI is the broad CPI (IPCA); Benchmark interest rate is the target Central Bank interest rate in the end of period; and exchange rate as in the end of period. eop = end of period. * 2015 and 2016 are author s forecasts. Source: Ministry of Finance of Brazil, Central Bank of Brazil, and IBGE. Even with such policies, the primary surplus targets were fully accomplished, at least until (see figure 2). However, by mid-2012 to 2013, the recovery appeared to be weaker than expected, and the Brazilian economic authorities returned to incentives, trying to reignite the economy 3. In 2013, the economy grew again by 2.7%, and the investment grew 6.1%. A wide tax relief program and increasing government expenditures, including a broad financial subsidy for credit for capital goods via public banks, were introduced. The economy reacted reasonably, so the investment to GDP ratio remained relatively stable at approximately 20.5% until at least 2013, despite its high variability (see figure 1). 2 Although the one-off revenues had increased in importance after the 2008 crash, responding, for instance, to 0.74% and 0.85% of GDP in 2009 and 2010, respectively, when the full primary surplus delivered was 1.9% and 2.6% of GDP, respectively, or, in 2013, when 0.68% of 1.8% of GDP was one-off revenue. In 2014, one-off revenue was 0.5% of GDP while the primary deficit was 0.6% of GDP primary surplus is going to be plenty of on-off revenue, as well. 3 At that time, estimates of GDP growth were 2.7% for 2011, instead of 3.9% as reported in the new 2015 dataset, moving downward towards 0.9% in 2012, instead of 1.8% as reported in the new 2015 dataset. Moreover, the share of investment over GDP was sharply declining, but the new 2015 dataset unveiled very stable figures for this indicator. 5

8 Figure 1. Brazil: Gross Formation of Fixed Capital as % of GDP) ,0 20,5 20,6 20,7 20,9 20,0 18,0 16,0 19,1 19,4 18,5 18,3 18,4 17,9 18,0 17,0 17,3 17,1 17,2 16,6 19,1 19,7 19,7 17,8 14,0 12,0 10,0 Source: IBGE, updated in November 17, Note: Investment as % of GDP measured using current values for GDP and gross formation of fixed capital is author s forecast. Figure 2: Nominal Fiscal and Primary Results as Share of GDP (%) ,00 4,00 2,00 - (2,00) (4,00) (6,00) (8,00) (10,00) (12,00) 2,80 (5,20) 3,20 3,30 3,20 3,20 (3,30) (3,30) (4,40) (5,20) 3,70 3,70 3,20 3,20 3, (0,60) (0,50) 1,90 2,60 2,90 2,20 1,80 (2,90) (2,70) (2,00) (2,40)(2,50) (2,30) (3,50) (3,20) (3,10) (3,60) (6,20) (2,10) (11,10) (9,50) 0,50 (7,40) 1,00 (6,00) Nominal Balance Primary Balance Source: Central Bank of Brazil; are author s forecast. However, from 2013 to 2014, the output was not responding at all to tax stimuli or even spending increases. After Brazil graduated to respond to the 2008 crisis using countercyclical fiscal policies (Vègh and Vulletin, 2013), it experienced a rapid deterioration on the fiscal front in Both net and gross public debt increased quickly towards a risky case scenario, so the investment 6

9 grade rating was no longer assured in the medium term 4. Potential output decreased at the same time as the output was well below its declining potential output level. In late 2014 and early 2015, some social benefits were reviewed, and some tax relief was reverted (tables 5.1 and 5.2 detail these measures). The economy had actually started to adjust itself during 2014, as the Central Bank had begun a new tightening cycle from April 2013 to July The target interest rate increased 375 basis points in an interval of one year. Another 325 basis-point increase would take place between October 2014 and July A 700 basis-point monetary shock in an interval of two years is far from negligible. Its primary side effect was a considerable contraction in the domestic credit on durable and capital goods, contributing to an additional drop in consumer and business confidence. Figure 3: Gross General Debt and Net Public Debt as Share of GDP (%) ,0 70,0 66,7 Gross Debt (% GDP) Net Debt (% GDP) 69,0 72,0 73,4 74,0 60,0 50,0 58,9 46,8 61,8 51,5 59,8 60,9 54,2 56,2 56,1 55,5 50,2 47,9 46,5 56,8 56,0 44,6 59,3 51,8 51,3 54,8 53,3 58,9 40,0 30,0 37,6 40,9 38,0 34,5 32,9 31,5 34,1 33,0 35,0 36,0 38,0 20, Source: Central Bank of Brazil; are author s forecasts. Since mid-2014 the Brazilian economy fell into full disarray: a combination of fiscal crisis with strong and prolonged GDP contraction in the midst of a political chaos. There are many plausible explanations for the deceleration of the economic, including the monetary policy shock, associated curbing in household credit for durable goods, and the gradual increase in some tax rates. Meanwhile, the economy suffered from a myriad of other events, such as long-standing and severe drought, corruption scandals involving the largest Brazilian state-run oil company and major entrepreneurs in 4 September 2015, a couple of weeks after the Government had decided to send to Congress the 2016 Budget Law with deficit, the Standard & Poor s downgraded the Brazilian sovereign rating to speculative grade. 7

10 the civil construction sector, low government popularity amid street protests and general disapproval regarding the corruption scandals, several bribery schemes, and the 2014 FIFA World Cup. There are also drivers associated with a profounder phenomenon, as many analysts believe that the previous consumption-based growth model had already been exhausted. Unfavorable terms of trade are also remembered as another key driver of such exhaustion. However, I would highlight the exhaustion of domestic economic cycle as the main guilty for such difficulties, which could dramatize the GDP performance in coming years. Under this circumstance, the economy became very sensitive to either domestic or external shocks. Therefore, fiscal results had been worsening faster than predicted, as the tax revenue had been frustrated in line with the growth downturn. Gross debt, as a percentage of the GDP, increased from 53.3% in December 2013 to 58.9% in December 2014 and, then, soar to 66%, September The gross debt prospects indicate further increases towards over 70% of GDP by the end of 2016 (see figure 3). In addition, surprisingly, the nominal deficit increased from 3.1 to 6.2% of GDP and skyrocketed to 9.34% of GDP in September 2015, a movement that was driven by interest payments, which also increased to 8.9% of GDP. Debt maturity and denomination have deteriorated with the same intensity. As risks are tilted towards deep GDP contraction in 2015 and also in 2016, fiscal sustainability appears to have been an important issue yet again. For instance, in September 2015, the net revenues of the central government decelerated -4.6% year-to-dates comparing to In that month, the government spending also declined -4.0%, in the same terms, although less than the tax frustration. Consequently, the rolling 12-month primary surplus of the central government decreased to -0.5% of GDP in September However, the primary surplus is expected to be delivered will be even worse and can reach 2.1% of GDP. Tax revenue frustration is only partial explanation. In 2015, the government could not take into account the relevant amount of one-off revenues; and also it had to pay the delayed expenses, domestically labeled as fiscal pedaling. Brazil fell into a severe fiscal crisis: the country failures to reach the required primary surpluses, that is, the level necessary to stabilize debt to GDP, this year and even in coming years. Therefore, the gross debt to GDP ratio is expected to soar 72% of GDP in 2016, from 53% of GDP, in This crisis is predicted to last many years. As far as I can tell, there is no solution, not even a light at the end of the tunnel. The causes of such a fiscal situation are closely related to the causes of the GDP contraction. First, I would consider the stronger-than-expected GDP contraction that leads to tax revenue frustration even under tax rates hikes. It could also be explained by the growing government spending 8

11 related to income transfer programs such as pension benefits 5, LOAS 6, Minha Casa Minha Vida (housing program), Bolsa-Família (a conditional cash transfer program), etc. I would also include the reluctance in implementing the appropriate fiscal consolidation plan of early This hesitancy amplified the uncertainty on the economic recovery. Finally, because of the second round of the counter-cyclical fiscal policy implemented in 2012, delayed expenditures had to be settled during Meanwhile, intense realignment in the regulated price provoked a high short-term inflation. It seems there are many plausible explanations for such a drama experienced by the Brazilian economy like a perfect storm. From a long-term perspective, as seen in figure 4, there has been a considerable change in government spending since 2003, as it has been focused on reducing income inequality via increasing conditional cash transfer programs. The federal budget for education and housing has been increasing over the years. Total federal expenditures increased from 14.6% of GDP in 1997 to 18.6% of GDP at the end of Meanwhile, the net total federal tax revenue increased from 15.4% to 19.9% of GDP in 2010 and then decreased to 18.4% at the end of The recent fiscal stimuli combined to increase government expenditures by almost 1% of GDP, while the tax revenue declined from 19.8% to 18.4% of GDP. However, such recent efforts did not ignite growth in Many analysts question whether the overall fiscal multipliers decreased over this period of time. If so, why? Figure 4. Brazil: Net Tax Revenue and Government Spending (% of GDP) ,0 20,0 19,0 18,0 17,0 16,0 15,0 14,0 13,0 Net Tax Revenue (% GDP) Primary Expenditure (% GDP) 18,6 18,7 18,9 18,8 18,3 18,0 17,7 17,0 17,2 17,2 16,6 16,7 16,2 16,0 15,4 15,5 15,3 15,4 16,1 16,0 14,6 14,9 14,0 14,1 14,4 13,7 20,0 18,0 16,5 18,7 18,7 17,1 19,2 17,7 19,0 19,4 18,4 18,7 18,9 18,7 12, Source: IBGE, and Ministry of Finance, Brazil forecast according to the Minister of Planning. 5 The deficit in the general pension system is expected to reach 2.0% of GDP in 2016 from 1.0% of GDP in LOAS is a welfare public policy for elderly with the benefits indexed to minimum wage. Its expenditure reached 0.7% of GDP, in

12 So, what happened with the government expenditure over time? What are the main components of such increases? First of all, according to the table 2, the primary spending increased 2.8% of GDP since 2002, which is practically the same growth level of income transfers to households. Pension benefits show the most relevant increase followed by a welfare benefit for the elderly called LOAS. Part of the growth is related to both generous eligibility criteria and also public policy stance, due to the fact that the benefits are indexed to the minimum wage corrections. As a matter of curiosity, because of this indexation rule, the benefits increased 78%, in real terms, in the last 10 years. At the same time, the number of beneficiaries increased by 9 million people, from 23 million to 32 million. Meanwhile, the government decided to enlarge social programs such as Bolsa Família and Minha Casa Minha Vida. Both are responsible for a 0.63% of GDP variation in the government consumption during this period of time. Table 2. The Main Government Expenditure ( ) % of GDP and Percentage Point (pp) 2002 (% GDP) 2010 (% GDP) 2014 (% GDP) Variation pp Variation pp PRIMARY SPENDINGS 15,58 16,91 18,35 2,77 1,43 Payroll 4,77 4,28 3,98-0,79-0,30 Income Transfer to Households 6,51 8,25 9,31 2,79 1,05 Pension Benefits 5,90 6,56 7,14 1,24 0,58 Unemployment Insurance and Wage Bonus 0,48 0,77 0,98 0,49 0,21 Welfare Benefits (LOAS/RMV) 0,00 0,58 0,70 0,70 0,12 Bolsa-Família (income transfer to poor) 0,13 0,35 0,49 0,36 0,14 Investments 0,82 1,17 1,30 0,48 0,13 Fixed Gross Capital Formation 0,82 1,15 1,04 0,21-0,11 Minha Casa Minha Vida (housing program) 0,00 0,02 0,27 0,27 0,25 Expenditures 3,48 3,21 3,76 0,28 0,55 Health 1,38 1,34 1,42 0,04 0,07 Education 0,43 0,55 0,76 0,33 0,21 Subsidies* 0,16 0,21 0,16 0,01-0,04 Others 1,51 1,11 1,41-0,09 0,31 Net Revenue minus Income Tranfer 11,18 9,87 8,73-2,46-1,14 Source: National Treasure and IBGE. Author s calculation. Note: * Subsides herein is taking into consideration only those due to the correspondent years. Implicit and explicit subsides, that is, the financial and credit subsides, started to be estimated only in 2012; since then they are 0.9% do GDP in annual average, according to the methodology developed by the Secretariat of Economic Policy at the Minister of Finance. Moreover, there appears to be an increase of the financial and credit subsides related to the funding provided by the National Treasure to the state-owned banks. According to the table 2, the subsidies have been stable over time. However, this measurement doesn t take into account the implicit and explicit subsidies, in particular those with delayed payments. Roughly speaking, as part of the counter-cyclical measures, the development bank called BNDES used to lend to the private 10

13 enterprises offering low interest rates and the National Treasure was committed to equalize the interest rates. For instance, the BNDES provided subsidized long-term credit for investments in machineries and in the infrastructure sector. The difference between the benchmark rate Selic and the long-term interest rate called TJLP defines the size of the subsidy offered by the BNDES. The larger the difference between the two rates, the larger the subsidy. According to the calculations conducted by the Secretariat of Economic Policy, those subsidies reached an annual average of 0.9% of GDP in , much higher than the data provided in the table 2. Therefore, more beneficiaries, sizable correction in the benefits and enlarged income transfer programs could be considered the most relevant factors to explain the augmented expenditures. At the same time, tax revenues remained relatively stable until Since 2014 a drop was enough to make vulnerable the annual fiscal results; yet fiscal expansions were no longer financed by tax revenue. The situation became even worse because the deficit in the pension system soared surprisingly fast. After a certain period of stability, the deficit is projected to double from 2014 to 2016, from 1% to 2% of GDP. Pension system is responsible for 40% of the total expenditure followed by payroll with 20%. Its expenditures are foreseen to increase more than $25 billion in a very short interval of time, from 2014 to Figure 5 shows this risky scenario. As can be fairly seen, the spending has increased faster than the revenue. The slow increase in the revenue is because of the weakness of the labor market, which means the higher the unemployment rate, the lower the labor formalization and then the lower the pension revenues. The spending increase, as mentioned before, is due to the generous benefit criteria of eligibility, and an upsurge in the minimum wage, which indexes the benefits. The demography dynamics in Brazil is also a challenging task for the pension system, since the population is ageing quickly. This issue is likely to be the major explanation for the growth in the pension system deficit in the coming years. A comprehensive reform in the system is required. It would embrace much more strict criteria for newcomers such as minimum eligible age, similar treatment for gender, de-indexation of benefits from minimum wage, revision of some special regimes, and also the difference between the pension benefits and the welfare assistance, etc. 11

14 Figure 5. Pension System in Brazil: revenue, spending and deficit (% of GDP) ,50 Pension Revenue Pension Spending Deficit (% GDP) ,73 1, , , , , , ,97 1,03 0, , ,00 2,00 1,50 1,00 0, Source: National Congress The Literature and Our Case Study The discussion on the effectiveness of fiscal policy is mostly associated with the fiscal multiplier, either for government purchases or tax revenue. However, the literature indicates that there is no unique fiscal multiplier. Furthermore, there is scarce literature about emerging market economies and no theoretical support for whether multipliers should be expected to be higher or lower than in the advanced economies (Estevaõ and Samake, 2013; Ilzetzki et al., 2013; Ilzetzki, 2011; IMF, 2008; and Kraay, 2012). Some studies even conclude that multipliers are negative, particularly in the longer term (IMF, 2008) and when public debt is high (Ghosh and Rahman, 2008). One can easily find very divergent fiscal multipliers (see table 2). The multiplier depends on the critical factors, such as trade openness, the exchange rate regime, the fiscal instrument (whether spending or tax-based), the debt level, the monetary policy stance (whether normal or zero-lowerbound), and the state of the economy (whether contracting or expanding). Despite such innumerable factors, the fiscal multiplier is also sensitive to the method of estimation. For instance, the DSGE approach has shown larger multiplier than the VAR approach. However, as highlighted by Mineshima et al. (2014:319), the DSGE model presents difficulties in modeling nonlinearity and does so differently compared with the Taylor rule for monetary policy, as there is no widely accepted fiscal policy (rule) to be included in a DSGE model. 12

15 Table 2. Fiscal Multipliers Survey: the GDP growth response to fiscal shocks Authors Blanchard and Perotti (2002) Christiano et al. (2009) Auerbach and Gorodnichenko (2012) Alesina et al. (2014) Meneshima et al. (2014) Country/ Region United States SVAR United States DSGE OECD countries 17 OECD countries OECD and G7 Countries Methodology Sample GDP growth response to one SD +/- 2S.E innovation on government spending to Peak values from 0.97 to quarter response: 0.45 to 0.55 From 0.8 (no zero bound) to 3.4 (under STVAR Quasi-panel based on a truncated MA representation TVAR to zero bound) From 0.6 (under expansion) to 2.5 (under recession) From close to 0.0 (if spending-based plan) to -3.0 (if tax-based plan) both after 4-quarter response From 0.72 (positive output gap) to 1.22 (negative output gap) after 4-quarter response On the other hand, VAR models are subject to several critics. Commodity-exporting countries, such as Brazil, may experience revenue changes because of booms and busts in the international commodities market, not because of discretionary fiscal policy. As VAR models suffer from the omitted variable problem and required quarterly data might not be available for a long enough time span, they can limit identifying information. In the specific case of the Brazil, the longest possible time span results in 72 observations over the course of 18 years, including the last years of a pegged exchange-rate regime ( ). According to Ilzetzki (2011), the more fixed the exchange rate regime, the larger the fiscal multiplier. Therefore, our results may be biased when we use the full sample ( ). Brazil is considered a closed economy, and this attribute is expected to increase the fiscal multiplier. As Brazilian trade openness does not present relevant changes over time, we do not expect that any sort of influence of such a key variable on the identification of the multiplier using countryspecific VAR models. Although trade openness is highly recommended for many other reasons, if the policymakers are really interested in using a discretionary fiscal policy to obtain any real output effect, the current closed economy makes their fiscal efforts more effective. However, if the effectiveness of the fiscal policy is falls short of policymakers expectations, trade policy should be implemented to achieve a more economic opening. Many efforts have been made to show the importance of the fiscal instruments. As widely known, spending-based fiscal consolidation policy can be more effective than the tax-based policy, 13

16 and the fiscal multiplier of the former is likely higher than that of the latter. Moreover, the procedure of Alesina et al. (2014) involves a simulation of a multi-year fiscal plan rather than of individual fiscal shocks. According to the authors findings, Fiscal adjustments based upon spending cuts are much less costly, in terms of output losses, than tax-based ones and have especially low output costs when they consist of permanent rather than stop and go changes in taxes and spending. As the authors explain, The difference between tax-based and spending-based adjustments appears not to be explained by accompanying policies, including monetary policy. It is mainly due to the different response of business confidence and private investment. Our case study has no multi-year fiscal plan and no fiscal policy adjustment based only on spending cuts. The only goal is to reach the annual announced primary surplus, regardless of the instrument and composition. Late 2014, 1.2% of GDP expressed in terms of the amount of money (e.g., $22 billion) is due as public sector consolidated primary surplus by end of 2015, and there was also a target of 2.0% of GDP for the coming years. It was hard to identify the proportion of spending cuts and tax hikes needed to obtain such surpluses. However, the fiscal measures announced (see table 5.1 and 5.2) defined approximately 2.2% of GDP 7 in overall savings through spending cuts (approximately 1.75% of GDP) and increased taxes (approximately 0.54% of GDP). Alas, these measures did not directly assure a movement from a deficit of 0.6% of GDP, in December 2014, to a surplus of 1.2% of GDP in December Actually, another primary deficit is expected in 2015 and, moreover, surplus for next year is not guaranteed. The government most likely will deliver a deficit of at least 1.1% of GDP, instead of a surplus of 1.2% of GDP announced last December It is a dramatically poorer scenario. But, why such an unpleasant surprise? First, most of the savings come from the 2015 federal budget cut, which is generally overestimated; second, the increase in tax rates does not imply the same increase in tax revenue, which is more related to economic performance 8 ; and, finally, the revision in some social benefits takes time to contribute to fiscal efforts. Moreover, alongside the extreme deterioration of the domestic economic activity pushing down the tax revenue, payments of delayed expenditures have worsened the fiscal balance. Austerity brings more contraction, and such a circumstance is tough to obtain tax revenue. The back and forth economic policy stance during the year also played its role for such drama. It had to include low extraordinary revenue that put a negative bias to the fiscal balance. 7 This value is overestimated as it includes cut in an inflated budget. Taking into account only the structural spending cuts, the plan would retrench only 0.26% of GDP. 8 Using a sample from 1980 to 2014, we estimate the elasticity of tax to income close to 1.5. Moreover, surprisingly, some of the tax hikes can create tax credits because of the cumulative system combined with special tax regimes in the complex Brazilian tax system. 14

17 Our case study is more associated with shifts in fiscal policy over time than with long-term fiscal consolidation programs. The VAR model can capture this short-term dynamic adjustment, although Brazil instead have a fiscal plan that is well designed and communicated for the coming years. In line with Alesina et al. (2014), this sort of plan could result in a shorter recession than would be expected in our case study, as experienced recently. The debt level is very important, especially with the debt threshold is below the international threshold, as in the Brazilian case. According to the economic literature, the lower the debt threshold, the smaller the fiscal multiplier. In Ilzetzki (2011), the fiscal multiplier can eventually become negative when the debt exceeds its threshold. Brazil obtained sound results in terms of net debt levels until at least 2013; the gross debt-to-gdp ratio is higher than those of its peers, and debt maturity has remained a concern. The implicit interest rate of the debt is much higher than the monetary policy rate, which is considered one of the most persistently highest in the world. Because of this debt constraint, Brazil is expected to show a small fiscal multiplier. In other words, fiscal stimuli are welcome during contractions, without losing sight of debt sustainability in the medium term. The state of the economy is one critical factor of the fiscal multiplier. Using regime-switching models, Auerbach and Gorodnichenko (2013) estimated the effects of fiscal policies that might vary over the business cycle. They found considerable differences in the size of spending multipliers during recessions and expansions, with fiscal policy being considerably more effective in recessions than in expansions. As can be seen in the figure 6, Brazil s output is running well below its potential level, as roughly measured by the HP filter, throughout 2014, and it will most likely be the same throughout 2015 and In this scenario, the fiscal multiplier is expected to be larger than the previous years. Puzzlingly, fiscal stimuli recently seem to not being working well at all. As is well known, the effectiveness of fiscal policy is heterogeneous under normal circumstances (Favero, Giavazzi, and Perego, 2011). In the case of conventional monetary policy, fiscal laxity may have a restricted impact on output. Otherwise, countercyclical fiscal policy is likely to smooth the business cycle. The debt level is a constraint in both cases but is most likely a major issue for developing economies. In line with the Easterly s (2013) idea, part of the public debt increase is considered normal in advanced economies. However, in the aftermath of the 2008 turmoil, conventional monetary policy has been used mostly in developing economies, where debt intolerance (Reinhart et al., 2003) is still considered a relevant phenomenon. A simplified debt sustainability assessment unveils such a relevant constraint for the coming years and the peril of a downgrade in the sovereign rating. Under a baseline scenario, gross general government debt is highly likely to reach over 68% of GDP by late Required primary surpluses 15

18 to stabilize debt are higher than that formerly targets announced by Brazilian policymakers for 2015 and 2016; however, it is quite difficult to reach even that 1.2% of GDP target for 2015 and the 2% of GDP target for A bleak slowdown does not allow tax revenue to increase, even with the tax benefit withdrawals announced 9 early In sum, an upward bias for the fiscal multiplier is expected, which is associated with a couple of years under the pegged exchange-rate regime ( ) and with the contractions during the 2008 financial turmoil and since at least the middle of There has been a downward bias for the fiscal multiplier caused by the debt level and the accompanying conventional monetary policy stance. There is also the recessionary bias of the recently announced fiscal measures (in late 2014 and early 2015). Figure 6. Brazil Output Gap - % Quarterly Data ( ) Note: Output gap is measured as the difference between the actual output in log and the estimate output in log according to the HP filter (Lampda = 1,600). 3. The Empirical Model and Findings We roughly estimate the basic VAR model. The specification considers aggregate government purchases in the linear model with no regime shifts or control for expectations, including the following ordering [G T Y i] for Cholesky decomposition: (1) Y t = A(L, q)y t 1 + U t where Y t [T t, G t, X t ) is a three-dimensional vector in the logarithms of quarterly taxes, spending, benchmark interest rate, and GDP, all in real terms. U t [t t, g t, x t ] is the corresponding vector of reduced-form residuals, which generally have nonzero cross correlations. 9 For the recently announced fiscal measures, see tables 5.1 and

19 We then expand our estimations to take into account confidence indicators and disaggregate variables. In Y t [T t, G t, X t ], X t is an n-dimensional vector, including business and consumer confidence, private investment and household consumption 10, besides interest rate and real GDP. We are aware that VAR models have been subject to several criticisms (IMF, 2010; Romer, 2011; and Caldara and Kamps, 2012). DSGE models are alternative approaches, but they also have drawbacks. We are also aware that other key macroeconomic variables could be taken into consideration, such as trade openness, debt level, and financial market deepening. However, the changes over time used in our estimations are negligible. We would highly recommend including them in case of a panel-based empirical analysis, as those variables most likely change across countries. We adopted the VAR Granger causality/block exogeneity Wald tests to examine the causal relationships among the variables. Under this system, an endogenous variable can be treated as exogenous. We used the chi-square (Wald) statistics to test the joint significance of each of the other lagged endogenous variables in each equation of the model and also for the joint significance of all other lagged endogenous variables in each equation of the model. The chi-square test statistics for some variables (X, for example) represent the hypothesis that the lagged coefficients of that variable in the regression equation of another variable (Y, for example) are equal to or different from zero. If equal to zero, that variable (X) is Granger causal for Y at some level of significance, which suggests that Y is not influenced by X. The null hypothesis of block exogeneity is then rejected for all equations in the model. The first group of empirical results is related to a very simplified VAR model and its pairwise Granger causality and block exogeneity Wald tests. Information criteria from Akaike, Hannan-Quinn and Schwarz were used to select the most parsimonious, correct model. Most tables and figures presenting the results appear in the Annex. We have conveniently separated some figures to show alongside the text. We estimated two different VAR models as follows: (1) VAR 2: [Y, S, T, i] 10 These results are not reported herein, as they did not add any relevant analysis, but the results are available upon request. 17

20 where Y is the GDP; S is the government expenditure as a percentage of GDP deflated by IPCA; T is the net tax revenue as a percentage of GDP deflated by IPCA; and i is the Selic interest rate in annual percentage terms. (2) VAR 1: [h, FI, i] where h is the output gap measure, i.e., the difference between the actual GDP and the potential GDP according to the HP filter; FI is the fiscal impulse measure, i.e., the variation of the primary surpluses as a share of GDP; and i is the variation of Selic interest rate in annual percentage terms. The VAR model in specification (2) allows us to simultaneously identify the direction of fiscal and monetary policy. As we split up the sample to cope with the 2008 crash, it is also possible to observe the eventual shift in the economic policy stance before and after the crisis. Then, this second specification is more related to the policy stance, whether countercyclical or pro-cyclical. For instance, fiscal policy could be labeled as countercyclical or pro-cyclical when the sign of the fiscal impulse is equal to or different from (expansionary or contractionary, respectively) the signal of the deviation of the real output from its tendency level (the output gap). Figure 7 illustrates this idea. According to this figure, one can have, for example, countercyclical policy under fiscal adjustments and pro-cyclical stances during output expansion. It is always a matter of the direction of the fiscal policy alongside the business cycle. Figure 7. Fiscal Expansion and Contraction GDP Output Output Gap (-) Output Gap (+) Fiscal Impulse Expansion (-) Countercyclical Pro-cyclical Contraction (+) Pro-cyclical Countercyclical There have been many discussions about countercyclical versus pro-cyclical fiscal policies and when and how much uses each type. However, the first challenge involves assessing how much of effort the policymakers intend to make. The most accurate way to infer this effort is by assessing structural fiscal results instead of conventional ones 11. Before moving on, it is important to consider that the structural fiscal results are still behind the times when the topic is fiscal rules, mainly because they are well known only quarters or even years after the fiscal practice is implemented, as they 11 See Bornhorst et al. (2011) for numerous technical issues associated with this concept. 18

21 depend on many complex calculations. However, if the purpose is to define the intensity of the fiscal policy, they are more appropriate, even though the other methods are not necessarily incorrect. The structural fiscal results can be briefly defined as results that are consistent with the potential output, under the condition of equilibrium in the asset and commodity prices and free of one-off revenues. It may be considered an accurate form to express the fiscal discretionary effect on the aggregate demand. As fiscal results can be influenced by many factors beyond the control of economic authorities, structurally based results consider only the effective fiscal efforts of policymakers. Conventional results, such as primary surpluses, are not capable of measuring such efforts because they depend on others factors, such as the business cycle, and, in turn, changes in asset and commodity prices, changes in output composition, and one-off revenues. Figure 8 illustrates the use of one-off revenues in Brazil, which increased dramatically after the 2008 crisis. One-off revenues includes: concessions revenue in most years, judicial deposit (in 2009), Eletrobras dividend (in 2009 and 2010), Petrobras capitalization (most of 2010 one-off revenue), Sovereign Fund of Brazil (created in 2008 and used in 2012), tax refining program (in 2009, 2011, 2013, and 2014), other one-off revenues, dividends in advance, and FND subsidize. The annual average of one-off revenue from 2009 to 2014 (after the crisis) was 0.6% of GDP when the average of the annual primary surpluses was 1.8% of GDP, representing, then, 1/3 of the annual fiscal results. How much policymakers intend to pledge or cut in terms of stimulus or retrenchment, respectively, during a certain period of time, depends on the fiscal impulse, that is, the difference between the current structural fiscal result and the previous result. Fiscal policy could be labeled as countercyclical or pro-cyclical when the signal of the fiscal impulse is equal to or different from (expansionary or contractionary, respectively) the signal of the deviation of the real output from its tendency level (the output gap). Figure 8: One-Off Revenue, (US$ Billion and % of GDP) 14000, , ,0 8000,0 6000,0 4000,0 1,00 0,80 0,60 0,40 0,20 0,00-0,20 0,74 % GDP 0,02 0,06 0,04 0,04 0,08 0,74 0,85 0,68 0,53 0,50 0, ,0 0,0-2000, ,0 Source: Secretariat of Economic Policy, Minister of Finance of Brazil. Notes: We use 3.0 Brazilian Real per one US Dollar. 19

22 Here, the idea is different for the case of monetary policy. It could be pro-cyclical or countercyclical when the signal of the monetary policy shock (changes in the interest rate) is equal to or different from (contractionary or expansionary, respectively) the output gap. Then, for instance, we label arbitrarily a pro-cyclical monetary policy when the interest rate increases during an expansion. Otherwise, it would be a countercyclical orientation. We also conducted a multivariate pairwise Granger causality test/block exogeneity Wald test derived from the VAR model to examine additional causal relationships between the key variables in the model. This test estimates the χ square value of the coefficients of lagged endogenous variables. The null hypothesis in this test is that the lagged endogenous variables do not Granger cause the dependent variable. Data Descriptions The models are estimated over three samples: first, the full sample spanning from 1997 to 2014 and then two other samples to determine the eventual effect of the 2008 crash: one from 1997 to 2007 and one from 2008 to We are aware that the small size of the sample of times series diminishes the degrees of freedom and the robustness of the estimates. Table 3 shows data descriptions and the sources used here. All series, if necessary, are deflated with the Broad Consumer Price Index (IPCA) and divided by GDP. According to our sample, the GDP has grown an average of 0.73% in quarter-over-quarter terms, which is approximately 3.0% in annualized terms; investment has grown a little faster at 3.2%; and consumption is coupled with the GDP growth (table 4). There are different speeds over time because investment has grown faster and consumption has resumed faster than the GDP after the 2008 turmoil. The average investment as share of GDP is 18.7%, with some increases after the crisis coming to approximately 20%, which could introduce an intriguing question. Net tax revenue as share of GDP has been higher than the government spending-to-gdp ratio. The volatility of the investment rates, as measured by standard deviation (approximately 14% in annualized terms), is quite remarkable. 20