Download The Impact of Government Consumption Cuts on Sovereign Yield Spreads and more Lecture notes Macroeconomics in PDF only on Docsity!
Does austerity pay off?∗
Benjamin Born, Gernot J. M¨uller, Johannes Pfeifer
August 2014
Abstract Austerity measures are frequently enacted when the sustainability of public finances is in doubt. Such doubts are reflected in high sovereign yield spreads and put further strain on government finances. Is austerity successful in restoring market confidence, bringing about a reduction in yield spreads? We employ a new panel data set which contains sovereign yield spreads for 26 emerging and advanced economies and estimate the effects of cuts of government consumption on yield spreads and economic activity. The conditions under which austerity takes place are crucial. During times of fiscal stress, spreads rise in response to the spending cuts, at least in the short-run. In contrast, austerity pays off, if conditions are more benign.
Keywords: Fiscal Policy, austerity, sovereign risk, yield spreads, confidence, panel VAR, local projections JEL-Codes: E
∗Prepared for the SAFE Research Conference “Austerity and Growth: Concepts for Europe”, June 2014. Born: University of Mannheim and CESifo, born@uni-mannheim.de, M¨uller: University of Bonn and CEPR, gernot.mueller@uni-bonn.de, Pfeifer: University of Mannheim, pfeifer@uni-mannheim.de. We thank our discussants Nicola Fuchs-Sch¨undeln, Alessandro Gioffr´e, Josef Hollmayr, and Klemens Hauzenberger. We also thank Kerstin Bernoth and seminar audiences at Bonn, Heidelberg, Naples, the Bundesbank/DFG/IMF Workshop “Credit frictions and default in macroeconomics” and the ifw/Bundesbank Workshop “Fiscal Policy and Macroeconomic Performance” for helpful discussions. Andreas Born, Diana Sch¨uler, and Alexander Scheer have provided excellent research assistance. Gernot M¨uller also thanks the German Science Foundation (DFG) for financial support under the Priority Program 1578. The usual disclaimer applies.
“Debt is a slow-moving variable that cannot—and in general should not— be brought down too quickly. But interest rates can change much more quickly than fiscal policy and debt.” (C. Reinhart and K. Rogoff 2013)
“[F]inancial investors are schizophrenic about fiscal consolidation and growth. They react positively to news of fiscal consolidation, but then react negatively later, when consolidation leads to lower growth—which it often does.” (O. Blanchard 2011)
1 Introduction
In the years following the global financial crisis, many European governments have been implementing sizeable austerity measures: spending cuts and tax increases in order reduce budget deficits. These measures were taken in order to confront mounting concerns about rising levels of public debt or outright solvency issues. In fact, in a number of euro area countries sovereign yield spreads relative to Germany started to take off by 2010, arguably leaving policy makers with no alternative course of action.^1 Yet a further deterioration of financial market conditions, coupled with dismal growth performance in the following years, lead many observers to question the wisdom of austerity. In this paper, we ask whether austerity pays off, that is, whether it actually helps restoring market confidence in the sustainability of public finances as reflected by sovereign yield spreads. We focus on how financial markets respond to austerity measures or, more specifically, on the response of yield spreads and sidestep the issue of how such measures impact the actual health of government finances. In fact, while the response of fiscal indicators such as the level of sovereign debt is of first-order importance in this regard, it generally does not provide a sufficient statistic for assessing the sustainability of debt. For the willingness and the ability of governments to honor a given level of debt obligations depends on a number of country-specific, partly unobserved factors, such as the ability to raise taxes. The same level of debt may thus have very different implications for debt sustainability in different countries (Bi, 2012). Sovereign yield spreads, instead, provide more comprehensive picture, both because they reflect a broader assessment of market participants and because (^1) Historically, in addition to primary surpluses, output growth as well as negative real interest rates have contributed to the reduction of debt-to-GDP ratios (Hall and Sargent, 2011). Real interest rates in turn may have been depressed due to “financial repression” (Reinhart and Sbrancia, 2014). While it is unclear to what extent these factors will play an important role in stabilizing debt levels in the years to come, they are arguably no viable means in order to meet market pressures instantaneously.
which are available only for parts of our sample. Moreover, our measure of yield spreads measures the real borrowing costs of governments. We establish a number of basic facts regarding yield spreads. First, they vary considerably across time and countries. In some instances they are slightly below zero, in others they are as high as 70 percentage points. A total of about 1500 observations also allows us to compute the empirical density function. It increases sharply at low levels of spreads, as the number of observations for which spreads are high is limited. Second, yield spreads co-move negatively with economic activity. The correlation of yield spreads and current output growth is negative in all countries of our sample, but Sweden. Third, there is no systematic correlation pattern of spreads and government consumption. In a second step, we provide estimates on the effects of austerity. We focus on the effects of government consumption cuts, as identification is somewhat less controversial in this case than in case of tax hikes. Specifically, we assume that government consumption is predetermined within a given quarter. This assumption goes back to Blanchard and Perotti (2002) and is rationalized by the fact that changes in government spending cannot be agreed upon without a considerable decision lag. We collect quarterly data for government expenditure while drawing on earlier work by Ilzetzki, Mendoza, and V´egh (2013), extending their data set to include observations up to the year 2013. For some countries, our observations for both quarterly government consumption as well as sovereign yield spreads date back to the beginning of the 1990s. We pursue alternative econometric strategies to obtain estimates for how government consumption impacts the economy and, eventually, sovereign yield spreads. Following a large empirical literature on the fiscal transmission mechanism, we rely on estimated panel vector autoregressions (VAR) to compute impulse response functions. In addition, we also employ local projections which, following Jord´a (2005), have become very popular in recent years. For both models we compute unconditional estimates for the effects of fiscal shocks as well as estimates conditional on the state of the economy. In this regard, local projections stand out in terms of flexibility (Auerbach and Gorodnichenko, 2013a). Our main results can be summarized as follows. We find that austerity, or more precisely, cuts of government consumption, tend to raise sovereign yield spreads, if we do not condition our estimates on fiscal stress. At the same time output declines considerably. Upon closer inspection, however, these results mask considerable heterogeneity. If we condition the effects of spending cuts on the absence of fiscal stress, we find that spreads decline. At the same time, we find fiscal multipliers much reduced. In fact, we cannot reject the hypothesis
that government consumption leaves economic activity unaffected in the absence of fiscal stress. On the other hand, in the presence of fiscal stress spending cuts raise spreads and depress economic activity considerably. Fiscal stress episodes thus tend to dominate the overall sample. We also find that spreads tend to decline in response to spending cuts in the medium to long run. Our results are based on exogenous variations in government consumption, while austerity is typically considered to be an endogenous response to, say, financial market developments. That said, note that identifying an exogenous variation in government consumption is key to isolate the impact of austerity on the variables of interest per se , rather than the joint effect of financial market developments cum austerity. Still, it is certainly possible that austerity measures impact the economy in different ways than a “regular” fiscal shock—perhaps because they are implemented under special circumstances. Conditioning the effects of spending cuts on the state of the economy is one strategy to address this concern. Alternative, and complementary, approaches to assess the effects of fiscal consolidation episodes include case studies, notably the classic analysis of Giavazzi and Pagano (1990). Yet another approach goes back to Alesina and Perotti (1995), recently applied by Alesina and Ardagna (2013). It identifies (large) fiscal adjustments as episodes during which the cyclically adjusted primary deficit falls relative to GDP by a certain amount. Finally, fiscal consolidations have also been identified on the basis of a narrative approach (Guajardo, Leigh, and Pescatori, 2011; Devries, Guajardo, Leigh, and Pescatori, 2011). Our analysis also provides new results on the effects of government spending cuts on sovereign yield spreads for a large panel of advanced and emerging economies. Related studies include numerous attempts to assess the effects of fiscal policy on interest rates. In particular, Ardagna (2009) finds that interest rates tend to decline in response to large fiscal consolidations. Laubach (2009) investigates how changes in the U.S. fiscal outlook affect interest rates. Finally, Akitoby and Stratmann (2008) use a similar measure for sovereign yield spreads as we use in the present paper. They focus on emerging market economies, however, and assess the contemporaneous impact of fiscal variables on spreads within a given year. The remainder of the paper is organized as follows. Section 2 details the construction of our data set. In this section, we also establish a number of basic facts regarding the time-series properties of sovereign yield spreads and their relationship to government consumption and output growth. In Section 3 we discuss our econometric specification and identification strategy. We present the main results of the paper and an extensive sensitivity analysis in
Table 1: Basic properties of government consumption-to-output ratio
Country first obs last obs min max mean std Belgium 1995.00 2013.25 0.24 0.25 0.25 0. Denmark 1991.00 2013.25 0.25 0.29 0.27 0. Finland 1990.00 2013.25 0.18 0.26 0.21 0. France 1986.00 2013.25 0.23 0.27 0.25 0. Greece 2000.00 2011.00 0.17 0.23 0.18 0. Hungary 1995.00 2013.25 0.21 0.28 0.23 0. Ireland 1997.00 2013.25 0.16 0.18 0.17 0. Italy 1991.00 2013.25 0.19 0.22 0.20 0. Netherlands 1988.00 2013.25 0.23 0.28 0.25 0. Poland 1995.00 2013.25 0.17 0.21 0.18 0. Portugal 1995.00 2013.25 0.19 0.22 0.20 0. Slovenia 1995.00 2013.25 0.17 0.22 0.19 0. Spain 1995.00 2013.25 0.16 0.22 0.18 0. Sweden 1993.00 2013.25 0.07 0.12 0.09 0. United Kingdom 1986.00 2013.25 0.21 0.28 0.23 0. Argentina 1993.00 2013.25 0.12 0.15 0.13 0. Chile 1989.00 2012.75 NaN NaN NaN NaN Colombia 2000.00 2013.25 0.15 0.17 0.16 0. Ecuador 2000.00 2013.25 0.10 0.13 0.12 0. El Salvador 1994.00 2013.25 0.06 0.09 0.07 0. Malaysia 2000.00 2013.00 NaN NaN NaN NaN Peru 1995.00 2013.25 NaN NaN NaN NaN South Africa 1993.00 2013.25 0.13 0.18 0.15 0. Thailand 1993.00 2013.25 0.07 0.11 0.09 0. Turkey 1998.00 2013.25 0.09 0.12 0.11 0. Uruguay 1988.00 2013.25 NaN NaN NaN NaN Notes: Government consumption is consumption of the general government except for Chile, El Salvador, Malaysia, Peru, and Sweden, where it refers to central government consumption. For Chile, Malaysia, Uruguay, and Peru, we do not have information about the level of quarterly non-interpolated government consumption.
eliminating fluctuations in yields due to changes in real interest rates, inflation expectations and the risk premia associated with them. In addition to a default risk premium, yield spreads may still reflect liquidity premium and, if duration differs or drifts, a term premium (see e.g. Broner, Lorenzoni, and Schmunkler, 2013). However, we try to minimize the term premium by constructing the yield spread on the basis of yields for bonds with a comparable maturity and coupon.^4 Moreover, any liquidity premium is likely to be small in (^4) We focus on long-term rates whenever possible. As they are closely linked to the average of expected future short-term rates, they are a more appropriate measure of governments’ refinancing costs than short term-term rates. Assessing the effects of austerity on the term structure is beyond the scope of the present
our sample. As a result, yield spreads should reflect primarily financial markets’ assessment of the probability and extent of debt repudiation by a sovereign. We obtain spreads following three distinct strategies. First, for a subset of (formerly) emerg- ing markets we directly rely on J.P. Morgan’s Emerging Market Bond Index (EMBI) spreads which measure the difference in yields of dollar-denominated government or government- guaranteed bonds of a country relative to those of U.S. government bonds.^5 Second, we add to those observations data for euro area countries based on the “long-term interest rate for convergence purposes”. Those are computed as “yields to maturity” according to the International Securities Market Association (ISMA) formula 6.3 from “long-term government bonds or comparable securities” with a residual maturity of close to 10 years (ideally 9.5 to 10.5 years) with a sufficient liquidity (see, for details, European Central Bank, 2004). In case more than one bond is included in the sample, simple averaging over yields is performed to obtain a representative rate. For this country group, we use the German government bond yield as the risk-free benchmark rate and compute spreads relative to the German rate.^6 Finally, we also make use of the issuance of foreign currency government bonds in many advanced economies during the 1990s and 2000s to extend our sample to non-euro countries and the pre-EMU period. In particular, drawing on earlier work by Bernoth, von Hagen, and Schuknecht (2012), we identify bonds denominated in either US dollar or Deutsche mark of at least 5 years of maturity by developed countries. We compute the spread of yields for those bonds relative to the yields of US or German government bonds of comparable maturity and coupon yield in order to have comparable duration and thus term premia.^7 Whenever possible, we aim to minimize the difference in coupon yield to 25 basis points and the difference in maturity to one year. In order to avoid artifacts introduced by trading drying up in the last days before redemption, we omit the last thirty trading days before study. (^5) Note that inclusion of a bond into the EMBI requires a minimum bond issue size of $500 million, assuring that the liquidity premium compared to US bonds is not too large. Moreover, we rely on stripped spreads (Datastream Mnemonic: SSPRD), which “strip” out collateral and guarantees from the calculation. 6 The bonds used for computing the “long-term interest rate for convergence purposes” are typically bonds issued in euro, but under national law. In this regard they differ from the securities on which the EMBI is based, which are typically issued under international law. This difference becomes important if the monetary union is believed to be reversible. In case of exit from the EMU, the euro bonds will most likely be converted into domestic currency bonds, implying that they should carry a depreciation/exchange rate premium that is absent in the international law bonds. Still, even during the height of the European debt crisis, reversibility risk accounted for a small fraction of sovereign yield spreads in Greece (Kriwoluzky, M¨uller, and Wolf, 2014). 7 Yields on individual bonds are based on the yield to maturity at the midpoint as reported in Bloomberg or the yield to redemption in Datastream.
Italy
1990 1995 2000 2005 2010
0
1
2
3
4
5
6
7
8
9 Spread Yield Benchmark
B1 B1+B2 B2 ECB Long Term Convergence Rates
Figure 2: Construction of the Italian yield spread series.
four countries, it displays the yields of foreign currency bonds jointly with those of the associated benchmark bonds. For three countries (Italy, Denmark, UK), we consider bonds denominated in US dollars, while for Greece we consider a bond issued in Deutsche mark.^10 Note that yield spreads are typically small relative to the level of yields and vary considerably over time. For Italy and Greece, data on foreign currency bonds allow us to extend the sample to include observations prior to the introduction of the euro. In case of Denmark and the UK, they allow us to compute common-currency yield spreads, although those countries are not members of the euro zone. Figure 2 details the construction for the case of Italy. Until 1991, only one foreign bond is (^10) Italy: 10year $US bond issued on 08/02/1991 with coupon 8.75% (XS0030152895); benchmark bond: US 10 year Treasury note issued on 15/11/1990 with coupon 8.25% (US912827ZN50). Denmark: 10year $US zero coupon bond issued on 8/6/1986 (GB0042654961); benchmark bond: US 10 year Treasury note issued on 15/08/1998 with coupon 9.25% (ISIN: US912827WN87). UK: 10year $US bond issued on 12/9/1992 with coupon 7.25% (XS0041132845); benchmark bond: US 10 year Treasury note issued on 06/05/1992 with coupon 7.5% (US912827F496). Greece: 10year DEM bond issued on 11/13/1996 with coupon 6.75% (DE0001349355); benchmark bond: German 15 year bond issued on 04/10/1996 with coupon 6.25% (DE0001135010).
available. Starting in 1992, we obtain a second bond and compute the yield spread as the average over those of both bonds. When the first bond matures in 1997, we are left with one bond until 1999. From that point on, we use the long-term convergence bond yields provided by the ECB.
Table 2: Basic properties of sovereign yield spreads
Country first obs last obs min max mean std ρ (∆ yt, st ) ρ (∆ gt, st )
Belgium 1991.75 2013.25 0.04 2.53 0.45 0.44 -0.38 -0. Denmark 1988.50 2002.50 0.02 1.93 0.57 0.42 -0.17 -0. Finland 1992.25 2013.25 -0.04 0.80 0.27 0.18 -0.46 -0. France 1999.00 2013.25 0.02 1.35 0.27 0.32 -0.33 0. Greece 1992.25 2013.25 0.15 23.98 2.80 5.24 -0.59 -0. Hungary 1999.00 2013.25 0.10 6.05 1.79 1.66 -0.63 -0. Ireland 1991.75 2013.25 -0.04 7.92 1.04 1.79 -0.23 -0. Italy 1989.00 2013.25 -0.07 4.68 0.77 0.98 -0.41 -0. Netherlands 1999.00 2013.25 0.00 0.67 0.19 0.17 -0.63 -0. Poland 1994.75 2013.25 0.42 8.71 1.97 1.43 -0.01 -0. Portugal 1993.25 2013.25 0.00 11.39 1.27 2.62 -0.44 -0. Slovenia 2006.00 2013.25 -0.17 5.13 1.59 1.59 -0.42 -0. Spain 1992.50 2013.25 0.01 5.07 0.71 1.15 -0.61 -0. Sweden 1986.00 2009.50 -0.95 2.95 0.90 0.94 0.33 -0. United Kingdom 1992.75 2007.75 -0.03 0.64 0.29 0.17 -0.18 0. Argentina 1993.75 2013.25 2.04 70.78 15.80 18.74 -0.09 -0. Chile 1999.25 2013.25 0.55 3.43 1.45 0.58 -0.45 0. Colombia 1997.00 2013.25 1.12 10.66 3.65 2.19 -0.38 -0. Ecuador 1995.00 2013.25 5.02 47.64 12.58 9.08 -0.12 0. El Salvador 2002.25 2013.25 1.27 8.54 3.33 1.36 -0.47 0. Malaysia 1996.75 2013.25 0.46 10.55 1.84 1.45 -0.62 0. Peru 1997.00 2013.25 1.14 9.11 3.60 2.01 -0.36 -0. South Africa 1994.75 2013.25 0.70 6.52 2.28 1.24 -0.43 -0. Thailand 1997.25 2006.00 0.48 5.55 1.51 1.11 -0.55 0. Turkey 1996.25 2013.25 1.39 10.66 4.19 2.44 -0.31 -0. Uruguay 2001.25 2013.25 1.27 16.43 4.07 3.18 -0.36 -0. Notes: spreads st are average of daily observations per quarter, measured in percentage points.
Table 2 provides information on the coverage of our spread sample and some basic descriptive statistics.^11 Spreads st are measured in percentage points and vary considerably across (^11) Figure A.1 in the appendix displays the time series on a country-by-country basis.
(^00 10 20 30 40 50 60 )
1 Full Sample Advanced Emerging
Figure 3: Sovereign yield spreads: empirical distribution function (CDF). Notes: hor- izontal axis measures spreads in percentage points. Vertical axis measures fraction of observations for which spread exceeds value on the horizontal axis. Solid line displays CDF for full sample (total number of observations: 1497), dashed-dotted line: advanced economies only (783), dashed line: emerging economies only (714).
on specific identification assumptions which are imposed within a particular econometric framework.
3 Econometric framework
In this section we discuss the econometric framework used to establish the effects of austerity on sovereign yield spreads. We first discuss identification and then turn to how we condition on the economic environment.
3.1 Identification
In terms of fiscal policy measures we focus on the dynamic effects of exhaustive government consumption for reasons of data availability. We obtain identification by assuming that, within a given quarter, government consumption is predetermined relative to the other variables included in our regressions. This assumption goes back to Blanchard and Perotti (2002) and has been widely applied in the empirical literature on the fiscal transmission mechanism. It is plausible, because government consumption is unlikely a) to respond automatically to the cycle and b) to be adjusted instantaneously in a discretionary manner by policy makers. To see this, recall that government consumption, unlike transfers, is not composed of cyclical items and that discretionary government spending is subject to decision lags that prevent policymakers from responding to contemporaneous developments in the economy.^15 Still, influential work by Ramey (2011b) and Leeper, Walker, and Yang (2013), has made clear that identification based on the assumption that government spending is predetermined may fail to uncover the true response of the economy to a government spending shock whenever such shocks have been anticipated by market participants. Of course, the notion that fiscal policy measures are anticipated, because they are the result of a legislative process and/or subject to implementation lags is plausible.^16 To address this issue, we follow Ramey (2011b) and Auerbach and Gorodnichenko (2013a), and consider a specification of our model where we include forecast errors of government consumption, rather than government spending itself. Given data availability, we show—for a subset of our sample—that results do not change qualitatively relative to our baseline case. A number of promising alternatives to identify shocks to government spending have been pursued in the literature. Ramey (2011b) relies both on war dates and forecast errors for government spending. The latter are available for a subset of countries in our sample. Below (^15) Anecdotal evidence suggests that this holds true also in times of fiscal stress. For instance, in November 2009, European Commission (2009) states regarding Greece: “in its recommendations of 27 April 2009 ... the Council [of the European Union] did not consider the measures already announced by that time, to be sufficient to achieve the 2009 deficit target and recommended to the Greek authorities to “strengthen the fiscal adjustment in 2009 through permanent measures, mainly on the expenditure side”. In response to these recommendations the Greek government announced, on 25 June 2009, an additional set of fiscal measures to be implemented in 2009.... However, these measures... have not been implemented by the Greek authorities so far.” In fact, it appears that significant measures were put in place not before 2010Q1, see Greece Ministry of Finance (2010). 16 Still, whether or not this invalidates the identification assumption is a quantitative matter (Sims, 2012). Results by Beetsma and Giuliodori (2011), Corsetti, Meier, and M¨uller (2012b), and Born, Juessen, and M¨uller (2013), for instance, suggest that the issue is of limited quantitative relevance as far as shocks to government spending are concerned.
where μi and αi are vectors containing country-specific constants and time trends. The matrices Ak capture the effect of past realizations on the current vector of endogenous variables. νi,t is a vector of reduced form residuals. We estimate model (3.1) by OLS. Identification is based on mapping the reduced-form innovations νi,t into structural shocks:
εi,t = Bνi,t , with εi,t^ iid ∼ (0 , I ).
In the present context, identifying shocks to government consumption under the assumption that it is predetermined boils down to equating the first element in νi,t with a structural fiscal shock.^18 Recently, local projections have been a popular tool to complement VAR analysis. As argued by Jord´a (2005), local projections are more robust to model misspecification as they do not impose cross-equation restrictions as in sVAR models. Moreover, local projections prove highly flexible in accommodating a panel structure. Finally, and most importantly, they offer a very convenient way to account for state dependence—the focus of our analysis below. Earlier work by Auerbach and Gorodnichenko (2013a) has illustrated this in the context of fiscal policy. More specifically, they suggest a panel smooth transition autoregressive (STAR) model on which we rely below. Defining the vector Xi,t = [ gi,t yi,t si,t ]′, the response of a variable xi,t + h at horizon h to gi,t can be obtained by locally projecting xi,t + h on time t government spending and a set of control variables/regressors. That is, the following relation is estimated:
xi,t + h = αi,h + βi,ht + ηt,h
- F ( zi,t ) ψA,hgi,t + [1 − F ( zi,t )] ψB,hgi,t
- F ( zi,t ) Π A,h ( L ) Xi,t − 1 + [1 − F ( zi,t )] Π B,h ( L ) Xi,t − 1 + ui,t. (3.2)
Here αi,h and βi,ht are a country-specific constant and a country-specific trend, respectively. ηt,h in turn captures time fixed effects, which we do not allow for in the VAR model above. ui,t is an error term with strictly positive variance. L denotes the lag operator. At each horizon, the response of the dependent variable to government spending is allowed to differ across regimes “A” and “B“, with the ψ -coefficients on the gi,t terms indexed accordingly. Similarly, Π∗ ,h ( L ) is a lag polynomial of coefficient matrices capturing the impact of control variables in each regime. We estimate (3.2) using OLS, assuming that government spending (^18) As a practical matter, we impose a lower-triangular structure on B , attaching no structural interpreta- tion to the other elements in εi,t.
is predetermined (see also Auerbach and Gorodnichenko, 2013b). Conceptually it is convenient to distinguish two polar regimes which give rise to possibly different dynamics after a fiscal impulse. These polar cases are characterized by F ( zi,t ) being equal to zero and one, respectively. It is quite unlikely, however, that actual economies operate in either of these regimes. Rather, they tend to be more or less close to one of the two. This is captured in the estimation, as the projection of the dependent variable at each horizon is a weighted average, whereby the fiscal shock as well as the control regressors are allowed to impact the dependent variable differently. The weights, in turn, are a function F (·) of an indicator variable zi,t which provides information of how close the economy is to one of the two regimes. By using this weighted average, all observations between the two polar help in identifying the two regimes. In our estimation below we use lagged yield spreads zi,t = si,t − 1 as an indicator variable in order to measure how closely an economy operates to a regime of “fiscal stress”. Using the lagged value of the spread assures that the indicator is orthogonal to our identified government spending shocks. We weigh regressors on the basis of the country-group specific empirical CDF (see Figure 3 above). Formally, we have
F ( zi,t ) = N^1
∑^ N j =
(^1) zj <zi,t (3.3)
where 1 denotes an indicator function and j indexes all country-time observations in the respective country group. As an alternative to the empirical CDF, one may assume a specific parametric function in order to map the indicator variable into specific weights.^19 Using the empirical CDF, however, has two advantages. First, there are no degrees of freedom in specifying the transition function. Second, the polar cases are now given by states of the world that were actually obtained in-sample.
4 Results
We now turn to our estimates of the effects of government consumption cuts. Our main focus is the dynamic response of sovereign spreads to such cuts. Still, as argued above, because the adjustment of output is likely to be a key determinant of spreads, we also (^19) Auerbach and Gorodnichenko (2012) use a logistic cumulative density function F ( zi,t ) = (^) 1+exp(exp(−− γzγzi,ti,t )) as their transition function so that P rob ( z < z ¯) = F ( z ¯). The free parameter γ was chosen using extraneous evidence that the US is in recession 20% of the time.
Government spending
−1.4 0 1 2 3 4 5 6
−1.
−
−0.
−0.
−0.
−0.
0 GDP
−1 0 1 2 3 4 5 6
−0.
−0.
−0.
−0.
0
0.4 Spread (basis points)
−40 0 1 2 3 4 5 6
−
0
20
40
60
80
(a) Local projection Government spending
−1 0 5 10 15 20
−0.
−0.
−0.
−0.
0
0.4 GDP
−1.5 0 5 10 15 20
−
−0.
0
0.5 Spread (basis points)
−40 0 5 10 15 20
−
0
20
40
60
(b) Vector autoregression
Figure 4: Dynamic response to a government spending shock in linear model. Notes: con- fidence bounds represent 90 percent standard errors. Horizontal axis represent quarters. Vertical axes represent deviation from pre shock level in terms of trend output and basis points (spread). Top panels show results based on local projections, bottom panels show results based on VAR.
Finally, in the third column, we present estimates for the dynamic response of spreads to the cut in government consumption. Here we find that spreads do, in fact, increase in response to the spending cut. The impact response varies between 25 basis points (LP) and 30 basis points (VAR). For both specifications we find a maximum effect of about 40 basis points. The VAR response shows that the spread returns to its pre-shock level, after mildly undershooting it for an extended periods. It thus appears that austerity doesn’t pay off: spending cuts fail to reassure investors about the sustainability of public finances.^21
4.2 Fiscal stress vs benign times
The above result obtains for the entire sample, possibly masking heterogeneity of eco- nomic circumstances which may matter for how spreads respond to austerity, both across (^21) It is also worth pointing out that the continuous movements of spreads over time do not violate market efficiency. Spreads move with fundamentals in basic models of sovereign default (see, e.g., Arellano, 2008).
Government spending
−1.4 0 1 2 3 4 5 6
−1.
−
−0.
−0.
−0.
−0.
0 GDP
−1 0 1 2 3 4 5 6
−0.
−0.
−0.
−0.
0
0.4 Spread (basis points)
−40 0 1 2 3 4 5 6
−
0
20
40
60
80
(a) Linear model Government spending
−1.4 0 1 2 3 4 5 6
−1.
−
−0.
−0.
−0.
−0.
0
No StressStress
GDP
−2.5 0 1 2 3 4 5 6
−
−1.
−
−0.
0
1 Spread (basis points)
−200 0 1 2 3 4 5 6
−
0
100
200
300
(b) Conditional with country group-specific CDFs
Figure 5: Dynamic response to a government spending shock derived from local projections (two lags). All standard errors are clustered robust 90% standard errors.
time and countries. In particular, earlier studies show that fiscal policy may affect the economy differently in “bad times” (Bertola and Drazen, 1993; Perotti, 1999). And in- deed, recent evidence established by Corsetti, Meier, and M¨uller (2012a), Auerbach and Gorodnichenko (2013a) and Ilzetzki, Mendoza, and V´egh (2013) suggests that the gov- ernment spending multiplier on output tends to be relatively low, if debt is high. This is particularly relevant, as austerity is often enacted in response to concerns about the sustainability of debt. However, as discussed above, public debt per se is an insufficient statistic to assess the sustainability of public finances, because fiscal capacity varies strongly with a number of country-specific factors. Instead, sovereign yield spreads provide more comprehensive information regarding the extent of “fiscal stress”, both because of the underlying—arguably broader—assessment of financial market participants and, not least, because of the immediate budgetary consequences. In what follows we therefore estimate the non-linear model (3.2) relying on the spreads as an indicator variable and their empirical CDF (3.3) as weighting function. We thus allow the effects of spending cuts cuts to differ depending on whether they are in a regime of fiscal stress, evidenced by high spread or not (“benign times”). Recall, however, from
