A note on expansionary austerity in Brazil*

Abstract This article tests the expansionist austerity hypothesis for Brazil between 2000/01 and 2020/12 through Nonlinear Autoregressive Distributed Lags (NARDL) models. Among the results, a long-term relationship was verified between GDP, gross and net public debt, total, direct and indirect taxes, in addition to total, mandatory and discretionary expenses. With regard to the Error Correction Models, the short-term shocks that affect the Brazilian economy dissipate, however, the adjustment dynamics is too slow for all the fiscal variables. Finally, the analysis of dynamic multipliers reports that austerity policies in Brazil, undertaken on the expenditure, tax or debt side, are expansionist. JEL Codes E32, H62, H68


Introduction
In the two decades between 2000 and 2020, Brazilian fiscal policy went through distinct moments.Throughout the 2000s there was economic growth reconciled with primary surpluses.However, from the mid-2010s on, the Brazilian economy saw its growth significantly reduce.At the same time, the country started to present countless fiscal difficulties, the primary result presented a long sequence of deficits, and the public debt expanded strongly.In this context, the authorities are faced with a difficult dilemma: to use fiscal policy to stimulate activity or preserve the country's fiscal position?
This paper has two objectives, the first is to analyze the short and long run effects of austerity policies on economic growth in Brazil between January 2000 and December 2020; the second is to investigate whether different austerity instruments have different effects on GDP.To this end, the effects of gross and net public debt, total, direct and indirect taxes, and total, mandatory and discretionary spending on GDP will be estimated using Nonlinear Autoregressive Models Distributed Lags (NARDL).The article is based on two hypotheses: i) that austerity policies can be expansionary, and ii) that different fiscal policy instruments, whose objective is to maintain the long-term equilibrium of public accounts, exert asymmetric effects on GDP.
The article is divided into five sections starting with this introduction.In section 2 the literature will be presented.In section 3 the Brazilian case will be analyzed.Section 4 will present the empirical strategies.In section 5 the results will be presented and discussed.Finally, in the final section some concluding remarks will be made.

Literature
The hypothesis of expansionary austerity has recently been raised by the contribution of economists Alberto Alesina; Carlo Favero and Francesco Giavazzi through numerous contributions that culminated in the book Austerity: When It Works and When It Doesn't (Alesina, Favero, & Giavazzi, 2019).Austerity can be defined as a set of policies whose goal is to make fiscal policy sustainable, austerity policies can take three forms: i) tax hikes; ii) cutting public spending, and iii) some combination of both strategies.Such policies would be dispensable if governments used fiscal policy more assertively.The authors studied 184 austerity plans comprised in a sample of 16 Organization for Economic Cooperation and Development (OECD) economies between 1981 and 2014.Of these, 57 plans were undertaken predominantly on the revenue side, while 113 were predominantly on the expenditure side. 1  In a traditional macroeconomic context, fiscal policy is subject to Musgrave and Miller's (1948) logic of "built-in flexibility" and exhibits compensatory movements in the face of swings in inflation and employment.Such an approach suggests that tax cuts and spending increases be adopted during periods of deflation and unemployment.For the remaining periods, taxes should go up and expenses should be cut.According to Persson and Svensson (1989) and Tabellini and Alesina (1990), politicians in office have the incentive to accumulate public debt, forcing their successors to implement adjustment policies.This creates the so-called deficit bias of fiscal policy.
In a Keynesian2 approach, fiscal expansions are important elements in expanding output.This model, however, privileges the multiplier effects of spending vis-à-vis taxes, since taxes affect disposable income and hence aggregate consumption, while spending is exogenous and affects all demand components and hence income as a whole.Going back to Alesina et al. (2019), austerity policies can comprise a single period, or span several years over the horizon, so that long periods characterized by several institutional reforms can be grouped together as a single austerity measure.On this, these authors raise six points that need to be considered when dealing with expansionary or contractionary effects of fiscal policy: i) Phase of the economic cycle -The authors argue that confusions can arise from the difficulties of estimating the effects of a fiscal shock separately from the economic cycle effects.They argue that fiscal contractions can occur at the apex of an economic boom, in which case it is difficult to know if what caused the recession was the fiscal contraction, or if this would be a natural movement of the cycle.
ii) Expectations -The simple Keynesian model is static, so that expectations do not interfere with the outcome of fiscal policy.When considering expectations, spending cuts can signal future tax reductions and increases in permanent disposable income.
iii) Other macroeconomic policies -Austerity policies are rarely undertaken alone; they are usually accompanied by, or exert an effect on, monetary and exchange rate policy.In this sense, the authors argue that fiscal contractions can result in looser monetary policies and a more devalued exchange rate Alesina, Ardagna, Perotti, and Schiantarelli (2002).
iv) Supply side -Although in the traditional modeling, fiscal policy is treated as a demand element, its effects also have repercussions on the supply side because it changes incentives.Daveri, Tabellini, Bentolila, and Huizinga (2000) argue, for example, that taxes on capital discourage investment.On the other hand, poorly calibrated social security expenses can discourage savings and have repercussions on capital accumulation.
v) International context -Adjustment policies can be adopted in periods of worldwide recession, or the opposite, expansions can be undertaken in periods of international growth.
vi) Institutional Reforms -The authors argue that austerity policies are usually undertaken together with institutional reforms that increase long run productivity and can be expansionary.
When it comes to the empirical literature on the effects of fiscal policy on growth, the results are inconclusive.The variation in the multiplier effects of fiscal shocks on output led Leeper (2010) to call estimates of multiplier effects "alchemy."Other authors have also focused on understanding whether austerity policies are always recessive.For example, Alesina and Ardagna (2010) argue that austerity policies can be expansionary as long as they are committed to the expenditure side.Perotti (2013), on the other hand, argues that adjustments conducted mostly on the expenditure side have less recessionary effects than those committed on the revenue side.Alesina, Barbero, Favero, Giavazzi, and Paradisi (2015) have no conclusive results about the effects of austerity policies, for these authors, the effects of fiscal consolidations on output can be recessionary or expansionary and are conditional on the instruments used.

Austerity in Brazil
Austerity policies, in Brazil, have a much shorter history in relation to the countries analyzed by Alesina et al. (2019).The work of these authors covers episodes of austerity starting in 1980, In Brazil, the concern with the state's financial health only began to gain institutional contours after the 1988 Constitution.Historically in the country there has been a predominant principle that development would depend on a more active action by the State.
Since the 1988 Constitution, and under the need to stabilize the hyperinflation that plagued the country, macroeconomic and fiscal policy gained new institutional arrangements, in the articles between 163 and 169 that deal exclusively with public finances.Article 167 item III of the constitutional letter, for example, determines that financing through National Treasury credit operations must not be used to pay for operating expenses and, therefore, must not exceed capital expenditures.This rule, popularly known as the golden rule, is intended to prevent the Brazilian government from going into debt to pay current expenses.
In 1996, aiming to correct the fiscal deficits and simultaneously finance social services foreseen in the Constitution, the executive proposed Constitutional Amendment 12/96, which created a Provisional Contribution on Financial Transactions (CPMF), initially at a 0.2% rate.However, under the macroeconomic problems that followed, in 1999 the tax was extended and its rate increased to 0.38%.Other measures such as the increase of the Cofins rate from 2% to 3% also contributed to the fiscal consolidation of that year.
From then on, a new phase of rule-based fiscal policy began in Brazil.As recalled by Afonso (2016), in 1998, the federal government presented the draft of the Fiscal Responsibility Law (LRF).Such rule aimed to discipline public finances in Brazil acting on several fronts and can be considered a paradigm shift in Brazilian history, fiscal sustainability began to be seen, for the first time, as a conditioning factor for development.Therefore, the development model that prevailed in Brazil in the 50 years prior to the Constitution, rules were created to facilitate financing the development conducted by the State, from the 1990s on, rules started to be built to block non-fiscal sources of financing.
Still under the context of the macroeconomic turbulence suffered in the late 1990s and under the guidance of the International Monetary Fund (IMF), Brazil adopted a more flexible macroeconomic regime, characterized by a formalized floating exchange rate regime, as recalled by Salomão, Santos, and Reis (2021) by decree 6545/1999.The change in the exchange rate regime is of utmost importance due to the indexations in that period of the public debt to the Dollar.The inflation targeting regime was also instituted in that period through decree 3088/99.The consecutive currency devaluations coupled with an excessively tight monetary policy produced considerable expansions of public debt causing suspicions about the long-term solvency of the Brazilian government.It was also in this context that the regime of primary surplus targets established by the 1999-2001 Action Plan emerged in Brazil.Primary surplus targets, however, are commonly criticized because they are too rigid rules, preventing the expansion of public spending to smooth out the economic cycle in a single fiscal year.In Brazil, changing the primary surplus target depends on the opening of supplementary credits that can only be authorized by the legislature.
Still in 2000, another important matter was proposed and approved in the fiscal field, the Untying of Union Revenues (DRU), regulated by Constitutional Amendment 27/2000.The Brazilian Constitution universalized a relevant set of social rights and created tax revenues that should be applied to this end.However, such restrictions caused excessive fiscal rigidity, making it very difficult to allocate resources.The approval of the DRU gave the Executive Branch freedom to handle, according to its priorities, up to 20% of its previously bound tax revenues.
Measures on the expenditure side were also taken in order to keep the budget balanced.The main target of corrective measures aimed at smoothing the trajectory of spending was social security.In 1998 the executive presented to the Congress the PEC 20/1998 that aimed at correcting the trajectory of social security expenses.The proposal forbade the accumulation of retirement pensions for the workers of the statutory system, including military personnel.The initial project of welfare reform foresaw the adoption of a minimum age for CLT workers, but Congress vetoed such change on that occasion and, in a context of increasing life expectancy in the country, the executive was compelled to adopt measures that would postpone the request for welfare benefits by workers.Because of this, Law 9876/99 was sanctioned, instituting the social security factor.By this new rule, workers who applied for retirement were submitted to a factor that gave them higher pensions, the longer the period of contribution and the age of the applicant.
In 2003, social security returned to the macroeconomic policy scene and a new reform was approved through PEC 40/03.The target of this new reform was the public servants who lost, in practice, the right to parity in their pensions.Up to this date, public servants retired with the salary of the position they held while still working; the right to parity allowed that adjustments to workers still working were automatically granted to retired servants.This evidently imposed difficulties in financing pension benefits in the face of demographic changes.As of 2003, with the new rules, the workers that request retirement kept the right to full salary, but lost the right to parity.
During the transition from the 1990s to the 2000s, other measures such as state asset adjustment, financial restructuring of states and municipalities, and the use of non-recurring revenues were adopted to consolidate the Brazilian macroeconomic arrangement.Starting in 2004, still under the maintenance and deepening of the previous framework, Brazil's fiscal problems were mitigated by a long period of economic growth experienced by the country.Figure 1 shows the evolution of the primary result and the country's economic growth.
Even so, the country was not exempt from following through on its fiscal institutionality.In 2006 there was a change in the method for determining primary surplus targets, which were previously disclosed as a percentage of GDP and as of 2007 would be disclosed in nominal values.At the end of 2007, Congress decided not to renew the CPMF tax; the absence of this tax would cause a loss, according to the National Treasury Secretariat, of R$78 billion in public accounts.This measure was offset by the 0.38 percentage point increase in the rates of the Tax on Financial Transactions (IOF) whose law 5143/66 empowers the National Monetary Council (CMN) to alter, also by raising the rate of the Social Contribution on Net Profits (CSLL) for the financial sector from 9% to 15%, which occurred through Provisional Measure 413/08 that altered article 3 of Law 7689/88.
As seen in Figure 1, throughout the 2010s there was a reversal of primary results from surpluses to deficits.In the middle of that decade, revenue growth  began to show signs of exhaustion, but expenses maintained their trajectory of expansion, pressured by mandatory spending.On the revenue side, the situation was further aggravated by the progress of sectorial tax breaks given in the scope of Reintegra, a program instituted by Law 12546/11 within the Plano Brasil Maior (PBM).Initially aimed at exporting manufacturing industries and, as recalled by Garcia, Sachsida, and Ywata (2018), aiming to increase the competitiveness of the industrial sector, this program granted tax benefits to exporting sectors, such as credits to be discounted from their tax liabilities related to taxes such as IPI, PIS and COFINS, and also exempted the social security contributions of these companies.
On the expenditure side, in Brazil, a set of fiscal rules have been implemented over the past decades, aiming to contain public spending.However, numerous other non-fiscal rules put pressure on the involuntary expansion of spending.An example of this is Law 12382/11 which imposed that the minimum wage should be adjusted by inflation measured by the INPC of the previous year, plus the GDP growth rate of two lagged years.A rule of this nature has a high fiscal impact, as it impacts budget items such as personnel and payroll taxes, social security, unemployment insurance and salary bonuses, continued benefit (BPC), among others.
This new context characterized by growth slowdown, tax cuts and compulsory expansion of public spending, led fiscal policy to strong difficulties that were further deepened by accounting maneuvers called by Gobetti and Orair (2017) as creative accounting.The trick was used by the government to meet the primary target approved in the Annual Budget Law (LOA), as the Treasury is a partner of public companies and banks, at the end of the fiscal year, the government needed to meet the primary target provided for in the LOA, in this case, in order not to contingent expenses, the Treasury did not transfer to official banks funds from programs such as Bolsa Família and Minha Casa, Minha Vida, forcing them to pay these expenses to the beneficiaries.This behavior made the Treasury meet the fiscal target of the LOA, but put the financial health of these banks at risk.This combination of factors led fiscal policy to a deficit condition, as can be seen in Figure 1(B).This deficit was caused by both the contraction of revenues and the trajectory of public spending.
Faced with such difficulties, new fiscal measures have since been proposed.Once again the social security system was the target of alterations, and a new reform was proposed based on the benefits of public servants.This reform, regulated by Law 12618/12 instituted the complementary social security system for public servants, ending the full retirement for those entering the public service, who are now subject to the same retirement cap as the INSS and giving them the option to complement their retirement via pension funds related to their careers: i) funpresp-exe for executive branch workers; ii) funpresp-leg for the legislative branch; and iii) funpresp-jud for judges and other judiciary careers.
In 2015, again under major economic turbulence, the debate about austerity policies intensified, especially when unexpectedly the 2014 primary result was disclosed by the STN with a primary deficit of R$32 billion.For the first time since the creation of the primary target regime the primary result of the federal government was in deficit.Cavalcanti, Vereda, Zanderer, and Rabelo (2019) attribute the intensity and duration of the 2014-2016 three-year recession to the expectational shock caused by the 2014 and 2015 primary deficit.They argue that agents did not expect deficits of that magnitude and this new information took them by surprise, deteriorating the confidence picture in the economy.
Given these difficulties, new fiscal measures have begun to be considered.On the revenue side Law 13169/15 increased the CSLL rate from 15% to 20% for companies in the financial sector, and also instituted a 17% rate for stock exchange and capital markets companies.The Provisional Measure 669/15 also recovered part of the social security rates unburdened in 2011 by the aforementioned Reintegra program.Taxes on cold beverages were also raised.
In 2016, in the face of political turbulence and acute macroeconomic imbalances, the government proposed a new fiscal rule, the public spending cap, regulated by Constitutional Amendment 95/2016.The rule, which came into effect in 2017, imposed as a ceiling for the growth of the Union's primary expenditure, the inflation verified until July of the previous year.The spending cap would prevent real growth in public spending and would be in force in Brazil for 20 years, and could be revised after the first 10 years of effectiveness with modifications via supplementary law.This measure, however, was subject to two criticisms: the first is that in a context of involuntary growth of mandatory spending, discretionary spending would be the adjustment variable Salomão and Silva (2023).The second criticism is that this is a rule designed to work well in a context of low and stable inflation; in periods of high inflation, the ceiling is shifted upward, and may stimulate nominal growth in public spending.
In 2017 the Long-Term Rate (TLP) was instituted through Law 13486/17 to replace the former Long-Term Interest Rate (TJLP).These interest rates are instruments for financing long-term investment via the National Bank for Economic and Social Development (BNDES).However, as the bank's main funding source is the National Treasury, which is financed via debt securities among which are the National Treasury Notes type B (NTN-Bs) that are indexed to inflation, with the TLP also indexed to inflation, the fiscal impacts of the bank's business were reduced, bringing funding costs closer to the rate charged on loans.Still under intense fiscal difficulties, in 2019 the social security came back on the radar of macroeconomic policy through CA 103/19 which, among other things, instituted the minimum age of 65 for men and 62 for women.
The analysis of austerity policies summarized until ago reveals that the legislative treatment of measures on the revenue and expenditure side is asymmetric.As it is possible to see, as a rule, measures that aim to discipline the behavior of public spending are addressed via constitutional amendment, while measures related to tax expansions are regulated via complementary law, or even via provisional measure.Given that constitutional amendments require a qualified majority for legislative approval and complementary laws require only a simple majority, in Brazil the political effort to adjust the budget via tax increases is lower than that for spending cuts.For no other reason, the national tax burden went from 27% of GDP in the second half of the 1990s to 34% today.

Method and data
The empirical analysis begins with an evaluation of the data used in the estimations.The estimates include a period between January 2000 and December 2020 for monthly data.All variables were accumulated over 12 months, deflated at constant December 2020 prices and logarithmically transformed.A detailed breakdown of the variables is available at Appendix A.
Once the data and its treatment are known, it is necessary to check its behavior regarding the presence of unit root, besides the order of integration.This will be done through three stationarity tests, the Augmented Dickey-Fuller (ADF) and Phillips Perron (PP) and the KPSS.The accepted estimates for 1% significance and all series are integrated in order I(1).Once the data and their respective behaviors are known, the estimations of the tests will be carried out by the method of nonlinear autoregressive models distributed lags (NARDL).Which are an extension of the ARDL models with cointegration analysis.The method contributes to the hypothesis of the article since it captures an asymmetric relationship of GDP from positive and negative shocks on revenues, expenditures and public debt.
In all, 8 models will be estimated, whose explained variable is GDP.Each of these models is adapted to the Brazilian fiscal reality.As stated, the nonlinearities will be given from the positive and negative shocks in the fiscal variables debt (gross and net); taxes (total; direct and indirect) and expenses (total; mandatory and discretionary), as represented in equations ( 1) and ( 2): For the purposes of the analysis proposed here, there is interest in estimating the effects on GDP from shocks in the most varied fiscal instruments.Therefore, expenditures can be replaced by any fiscal variable.To estimate these effects, 8 models based on equation (3) will be tested: Equation ( 3) is the generic representation of Error Correction Models (NARDL-ECM), where the superscripts + and − indicate positive and negative shocks respectively.In them,  is the intercept and  as a time trend.The family of coefficients  consist of the model's slope parameters.The variables accompanied by the summation indicator, meanwhile, are the long run coefficients.Given this representation, the empirical models to be estimated have the "GDP" as Dependent Variable and the specifications about the fiscal instruments are respectively: Gross Debt, Net Debt -Models 1 and 2, respectively; Total Taxes, Direct Taxes, and Indirect Taxes -Models 3, 5, and 6, respectively; Total Expense, Mandatory Expense, and Discretionary Expense -Models 4, 7, and 8, respectively.

Results
The analysis of the results begins with the diagnostic tests.Starting by checking for asymmetries between tax variables and GDP that justify the use of NARDL models.This was done via the Wald Test that tests for possible short-run and medium-run nonlinearities, the results are shown in Table 1.It is possible to verify long run nonlinearities in models 3, 4 and 8.That is, in the long run, GDP reacts asymmetrically to shocks in total taxes, total expenses and discretionary expenses.In the short run, gross debt showed asymmetric behavior.With these results it is possible to notice that GDP reacts asymmetrically both to tax shocks and too expenditure and public debt shocks.
Having analyzed the asymmetries, we must now move on to the diagnostic tests.Table 2 presents a detailed list of the estimations performed.Starting with the lags selected by Akaike's criterion (AIC).The LM test coefficients significant at 5%, indicate absence of autocorrelation in all estimated models, and eventual heteroscedasticity was corrected by White's method.Table 2 also presents an analysis of parameter stability using the Cumulative Recursive Sum of Residuals (CUSUM) and Cumulative Sum of Squared Residuals (CUSUMQ). 3he tested models are stable, with the exception of the test for the CUSUM of models 4 and 7, where a small, localized instability was verified.
The NARDL models also rely on cointegration analysis (bounds testing approach) useful to find out whether the variables tested are related in the long run.Such tests are also shown in Table 2, indicating that the variables that make up the model cointegrate in the long run.This dynamic attests that, in the long run, the GDP of the Brazilian economy behaves in the same way as fiscal aggregates such as debt (gross and net), taxes (disaggregated between direct and indirect), and expenditures (disaggregated between mandatory and discretionary).This indicates that despite the numerous crises that have befallen the country over the period, many of them with a strong fiscal component, the  numerous austerity plans reported above, together with institutional rules that have emerged, have maintained the relationship between fiscal aggregates and GDP.
Table 3 presents the estimated long run coefficients for positive and negative shocks of the fiscal variables on GDP.Starting with model 1, there was statistical significance at 5% in the negative shock of the gross debt on the GDP, showing that when there are drops in the gross debt, activity also retracts.There was also statistical significance at 5% and positive sign in the control variables concerning net investment and terms of trade, which positively influence the GDP.Model 2 (net debt) did not present statistical significance.
Regarding the tax shocks presented in Table 3 in models 3, 5 and 6, which test respectively the effects of total, direct and indirect taxes, there was again statistical significance.Starting with total taxes, the estimated model shows a positive sign and significance at 1% for positive and 10% for negative shocks.Regarding direct taxes, there was statistical significance at 1% in the positive shock, all variables had a positive sign.Finally, with regard to indirect taxes, there was statistical significance at 10% in the positive shock and positive sign.This shows that in the long run, fiscal consolidations on the revenue side are expansionary.However, the long-term expansionist effects of an austerity program based on indirect taxes is greater vis a vis what is estimated for direct taxes.
Unlike Alesina et al. (2019), who found a negative relationship of austerity plans undertaken on the tax side and activity dynamics, the results here point  to a positive dynamic.At first glance this may seem counterintuitive, however, austerity plans may be accompanied by interest rate declines and exchange rate devaluations, they may also be associated with institutional reforms that stimulate growth.In the case of direct taxes, it is also possible to establish hikes with a view to increasing progressivity, which might not have detrimental effects on GDP dynamics.This was done in models 4, 7, and 8, which tested the effects of total, mandatory, and discretionary spending, respectively.When it comes to total and mandatory expenses, there was no statistical significance in any variable tested.This means that in the long run, these expenditure items do not influence activity.With regard to discretionary spending, however, there was a positive sign and statistical significance at 10% for the positive shock on output.Discretionary spending includes items considered by the literature to have a high multiplier effect, among which are public investments and transfers to households with high propensity to consume.Therefore, statistical significance and a positive sign were expected.In this specification, there was also statistical significance at 10% for net investment.
Once the long-term relationships are verified, it is necessary to pay attention, now, to the short-term dynamics that will be given by the analysis of the Error Correction Models (ECM), on the Table 4.Given the aforementioned cointegration relationship verified above, short-term positive and negative shocks may cause temporary deviations from this trend, the ECM analysis shows how the short-term adjustment dynamics of the variables tested will occur.In  all models tested, without exception, it is possible to verify an adjustment dynamic that is too slow.Starting with models 1 and 2, one sees that only 2% and 1% respectively of debt shocks on growth return to their long run trend in the first month.On the tax side, a slow dynamic of adjustment was again verified.Starting with model 3 that tests the shocks of total taxes on activity, only 2% of the tax shocks return to their trend in the first month.In direct taxes, reported in model 5, only 3% return to their long-run trend in the first month.Finally, indirect taxes, checked in model 6, have the slowest adjustment dynamics, only 1% of indirect tax shocks return to trend in the first month.
On the expenditure side, the dynamics are even slower than on the tax side.Only 0.7% of total expenditures verified in model 4, return to the trend in the first month.As for mandatory expenditures, described in model 7, the adjustment is even slower, so that only 0.08% of the shocks on mandatory expenditures return to the long-term trend in the first month.Finally, as for discretionary spending there is a less slow adjustment in relation to the other items on the expenditure side, about 1% of the shocks of discretionary spending on output return to their trend in the first month.
It is absolutely understandable that the dynamics of adjustment of expenditures in relation to GDP are slower than the dynamics of taxes.As seen in section 3, expenditures in Brazil are regulated in a constitutional instance, while taxes are regulated infra-constitutionally, by complementary laws or provisional measures.This difference means that in critical moments, adjustments on the tax side can be made more quickly, since, as a rule, infra-constitutional matters only depend on a simple majority for approval in the Brazilian Congress, while expenses depend on amendments to the Constitution, which, as far as it concerns, are only approved by a qualified majority of two-thirds of the members of both the House and the Senate.
The NARDL models also include an analysis of the dynamic multipliers, presented here in figures 2, 3 and 4. Beginning with Figure 2, which depicts models 1 and 2 that test the effects of gross and net public debt on GDP, the analysis of figures 2(a) and 2(b) brings some evidence: first, positive asymmetry was verified in the case of gross debt and negative in the case of net debt.Second, positive shocks in debt, both gross and net, are recessive after a few months, showing that higher debt leads to reduce GDP growth.Furthermore, negative debt shocks are recessive at first, but expansionary after a few months.When comparing the effects of negative debt shocks on GDP, it can be seen that in the case of gross debt, expansionary effects occur more quickly and to a greater extent than net debt effects.Reductions in gross debt lead to GDP expansion after the 5 th month, while reductions in net debt lead to GDP expansion in twice as long.
When analyzing the effects of taxes on economic activity, the dynamic multipliers reported in Figure 3 show a quite heterogeneous behavior among the estimated models.Starting with total taxes (Figure 3(a)), it was verified negative asymmetry and that negative shocks have high recessive effects on GDP, while positive shocks are recessive at the first moment, but expansionary after the 9 th month.As for direct taxes, the estimated asymmetry is again negative, becoming positive after one year.Finally, Figure 3(c) shows the effects on activity from shocks in indirect taxes, this time symmetrical behavior was verified, and negative shocks in these taxes are recessive, while positive shocks are expansionary.With these results, in Brazil, one can say that there is evidence of expansionary austerity on the tax side.This result is distinct presented by Alesina et al. (2019) for whom austerity undertaken on the revenue side are always contractionary.However, this is related to some specificities of the Brazilian economy.Starting with direct taxes, in Brazil they are levied on individuals with high monthly incomes and on legal entities.In this case, cuts in personal income tax may not stimulate consumption, but rather savings allocated to assets with no bearing on the activity, such as real estate or government bonds.In the case of legal entities, many are framed in special tax regimes such as Simples Nacional, or Presumed Profit, direct tax breaks for these companies do not necessarily mean new or increased investments, but may mean, again, accumulation of resources in cash or increases in mark ups.
Figure 4 shows the estimated dynamic multipliers for the nexus between total, mandatory, and discretionary spending relative to GDP, respectively The results of the dynamic multipliers estimated on the expenditure side, agree with the results of Alesina et al. (2019).That is, fiscal adjustments undertaken on the expenditure side can be expansionary after some time after their implementation.The results reported here go further, first because in Brazil, due to specificities already mentioned, austerity plans undertaken on the tax side can also be expansionary.Moreover, austerity plans focused on mandatory expenditures are expansionary in the first moment.

Conclusions
This paper tested the effects of austerity policies on the GDP of the Brazilian economy between January 2000 and December 2020.Several conclusions can be drawn from the results.First, there is a long-term relationship between GDP and fiscal variables such as gross debt, net debt, total direct and indirect taxes, and total mandatory and discretionary spending.This highlights that the numerous crises experienced in the two decades, austerity measures were efficient in preserving the long-term relationship of fiscal aggregates.
Secondly, it was seen that in the long run public expenses have little effect on GDP, except for discretionary expenses that positively influence GDP.With regard to taxes, it was verified that total taxes, as well as direct and indirect taxes, exert long-term effects on the GDP.As for the short-term dynamics, it was found that shocks of the fiscal variables on GDP, return to their long-term trend.However, the short-term adjustment is quite slow.It was also seen that shocks on the expenditure side are more resilient in relation to tax and debt shocks, and this is related to the normative environment of the Brazilian economy, whose public spending is constitutionalized and depends on great legislative effort to have its trend modified.
Finally, the analysis of dynamic multipliers pointed to evidence favorable to the expansionary austerity hypothesis in Brazil.It was found that negative shocks in the gross, net public debt are expansionary after a few months.It was also seen that tax shocks can be expansionary after some time, and finally, austerity measures on the expenditure side are expansionary, especially plans focused on mandatory expenditures.
The results of this paper are important in the current context of the country characterized by excessively low growth rates and high public debt.In this context it is common to see calls for the removal of fiscal rules that discipline fiscal policy in order to stimulate growth.With these results, such proposals may be counterproductive.Either because positive shocks to public spending inhibit output, or because this may lead to expansions in the tax burden and public debt, contributing to a reduction in GDP.Or because this may alter the long-term dynamics of the fiscal aggregates, putting Brazil's macroeconomic stability at risk.

- A -Union Primary Result (% of GDP) and GDP Growth Rate (In % per year) B -Federal Government Revenues and Expenditures (In R$)
Source: National Treasury Secretariat (STN) and National Accounts System (SCN).

Table 1 .
Wald Test -Short and Long-Term Asymmetries - Note: P-value in square brackets. 0 : asymmetric relation.

Table 3 .
NARDL Models -Long Term Results