Fiscal Consolidation and Economic Growth in the Eurozone
Marijke Samantha Dekker
Student number: 10315233 E-mail: m.s.dekker@icloud.com
Date: 11-08-2016
University of Amsterdam Faculty of Economics and Business
Master: Economics (MSc. ECON)
Specialization: International Economics and Globalization Course: Master Thesis International Economics and Globalization
Supervisor: Drs. N.J. Leefmans Second Reader: Dr. D.J.M. Veestraeten
Statement of Originality
This document is written by student Marijke Samantha Dekker who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents
Abbreviations 1
Abstract 2
1. Introduction 3
2. The Relationship between Fiscal Consolidation and Economic Growth 6
2.1. Channels 6 2.2. Conditional Impact 7 2.2.1. Consolidation Composition 7 2.2.2. Debt Levels 8 2.2.3. Business Cycles 9 2.2.4. Euro Membership 10
3. Other Factors Influencing Economic Growth 11
3.1. Impact Fiscal Consolidation on Economic Growth 12
3.2. Determinants of Economic Growth 13
4. Data Selection and Measurement 18
4.1. Economic Growth/Income 18
4.2. Fiscal Consolidation 19
4.3. EMU Membership 20
4.4. Debt Levels 21
4.5. Business Cycle Levels 21
4.6. Human Capital 21
4.7. Institutional Environment 22
4.8. Monetary Policy 24
4.9. Accumulation of Technological Know-How 24
5. Methodology 25
5.1. Endogeneity Problem 26
5.2. Regression Model 27
5.3. Robustness Tests 28
6. Analysis and Robustness 29
6.1. Correlations 30
6.2. Regression 32
6.2.1. Total Fiscal Consolidation 32
6.2.2. Tax-based Fiscal Consolidation 35
6.2.3. Spending-based Fiscal Consolidation 37
6.3. Overall Results 37
6.4. Robustness 38
7. Conclusion and Recommendations 40
7.1. Recommendations 42
References 45
Abbreviations
CAPB Cyclically Adjusted Primary Budget Balance
CB Central Bank
ECB European Central Bank
EMU Economic and Monetary Union
EU European Union
FC Fiscal Consolidation
FE Fixed Effects
GDP Gross Domestic Product
OECD Organization for Economic Cooperation and Development
OLS Ordinary Least Squares
R&D Research and development
SB Spending-based
SGP Stability and Growth Pact
TB Taxed-based
UnDebt Unsustainable Debt Dummy
Abstract
This research investigates the effect of fiscal consolidation on economic growth in the Eurozone. The analysis is based on 17 OECD countries in the period 1978-2009 and the methodology is based on pooled OLS regression analyses. The literature indicates that the relationship between fiscal consolidation and economic growth is dependent on four conditions. First, the composition of fiscal consolidation in tax increases and spending cuts. Spending cuts are expected to have a positive effect, while tax increase will negatively affect economic growth. Second, fiscal consolidation in a country that has an unsustainable debt level will have a more positive effect on economic growth. Third, fiscal consolidation is expected to have a negative effect on economic growth in countries in which the potential GDP is higher than actual GDP (i.e. a negative output gap). Fourth, fiscal consolidation is more likely to have a positive effect on economic growth in the Eurozone because investors have a lower debt tolerance for EMU member countries. This thesis is the first to test these conditions by taking into account interaction effects. This research only finds statistically significant evidence for the fourth claim. The result of the analysis is that fiscal consolidation in EMU member countries has a significant positive effect on economic growth when controlling for monetary policy and human capital, institutional environment, and R&D spending.
1. Introduction
The global financial crisis and the European debt crisis have increased debt levels in developed countries substantially. According to Abbas et al. (2014) sovereign debts are approaching historical records. The average general government consolidated gross debt of EMU member countries jumped from 64.9 percent of GDP in 2007 to 78.3 percent of GDP in 2009 (Eurostat, 2015). This is illustrated by figure 1: a high increase of sovereign debt is visible from 2007 onwards. The high debts result in multiple problems. First, high debts, especially debt above the threshold of 90 percent debt-to-GDP (Reinhart & Rogoff, 2010), are thought to have a negative impact on economic growth. Second, higher debt levels increase interest rates. Higher interest rates result in bigger debt services and thereby reduce the government’s spending capacity. Moreover, public interest rates are correlated with private interest rates and thereby reduce private investment (Vranceanu & Basancenot, 2013). Thus, higher public interest rates may also reduce private investment. Third, countries with high debts are more sensitive to future shocks, since they are limited by their budget constraints to cope with these shocks. Lastly, “when debt is high there is a risk of falling into a bad equilibrium caused by self-fulfilling expectations” (Abbas et al., 2014, p. 70).
Source: Graph constructed by author based on data from Eurostat (2015). Indicator General Government Gross Consolidated Debt as Percentage of GDP.
,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 200,00 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 Pe rc en ta ge of G D P
Figure 1: Government Debt
Euro area (19 countries) Belgium Germany Ireland Greece Spain France Italy Netherlands United Kingdom
The Economic and Monetary Union (EMU) member-states have an additional problem due to the Stability and Growth Pact (SGP), which indicates that the debt-to-GDP levels should be below 60%. The SGP is introduced to ensure the stability of the EMU and its main objective is fiscal discipline. Non-compliance can lead to sanctions. As soon as the Euro debt crisis started, austerity became Europe’s policy consensus: “Given that excessive debt seemed to be the problem, fiscal consolidation became the dominant theme” (Collignon, 2013, p. 7). However, austerity introduced in the EMU countries has led to slow recovery.
Thus, unsustainable debt is bad and according to the SGP, debt-to-GDP levels should be reduced. However, austerity can have a negative effect on direct output. In the current literature there is no consensus on the size and direction of the fiscal multiplier. “As many recent papers have highlighted, the response of the economy to these fiscal adjustments has varied substantially” (Ardagna, 2004, p. 1047). Moreover, most of the research is focused on developing countries (Checherita-Westphal & Rother, 2012). This thesis contributes to the literature by focusing on fiscal consolidation in OECD countries. Furthermore, different strands of literature focus on certain conditions which influence the effect of fiscal consolidation on economic growth. This thesis is the first to test these conditions by taking into account interaction effects. The effect of fiscal consolidation on economic growth is argued to depend on (1) the output gap, as fiscal consolidation during a recession or a time with a large output-gap will more likely have a negative effect on economic growth (Blanchard & Leigh, 2013; Collignon, 2013; Jayadev & Konczal, 2010); (2) the debt level, as fiscal consolidation when debt levels are unsustainably high can have a positive effect on economic growth since it reduces the crowding out of private investment more than that it increases the negative direct effect of reduced spending (Dreger & Reimers, 2013; Reinhart & Rogoff, 2010; Vranceanu & Basancenot, 2013); (3) the composition of the fiscal consolidation, as reduced government spending can increase private investment, since it reduces the crowding out effect. In contrast, increases in taxation can reduce private investment because this can lead to a higher demanded pre-tax wage (Alesina & Ardagna, 2009; Ardagna, 2004); and (4) EMU membership.
With respect to the fourth condition, Vranceanu and Basancenot (2013) and De Grauwe (2011) argue that being part of the EMU makes countries more vulnerable, since they cannot monetize debt. EMU countries do not have individual control over their interest rate, exchange rate or the European Central Bank (ECB). If investors fear the default of an EMU country, they will sell their bonds for euros. These investors will then invest those euros
somewhere else in the Euro-zone. This means that liquidity will drain from the country in trouble, a sudden stop, which could result in a liquidity crisis. The Central Bank (CB) of an individual EMU country is not allowed to print money and therefore cannot help in this crisis. Moreover, because all EMU countries are fixed to the same exchange rate, an individual EMU member also cannot depreciate its currency to boost its economy and change its relative price. Therefore, this research will specifically focus on the effect of fiscal consolidation on economic growth in EMU-member states. The research question is: How does fiscal
consolidation affect GDP growth in the Eurozone and how does this depend on the output gap, the debt-level, and the composition of fiscal consolidation?
To analyze the research question, regression analysis will be conducted. The first regression will include total fiscal consolidation as main explanatory variable. In the second and third regression total fiscal consolidation will be split into tax-based fiscal consolidation and spending-based fiscal consolidation respectively. Each of the regressions will include interaction effects of fiscal consolidation with (1) an EMU membership dummy, (2) an unsustainable debt dummy and (3) the output-gap. Moreover, each regression includes controls for human capital, institutional environment, investment in R&D, monetary policy (the interest and exchange rate), GDP per capita, and lag of GDP growth. In this way, the three regression analyses will incorporate different insights of the literature and make it possible to test whether the composition of fiscal consolidation is important for its impact on economic growth and whether the impact of fiscal consolidation on economic growth is dependent on certain conditions. To see whether the regression results are robust, several indicators of variables will be changed slightly in additional regressions.
This research uses the indicator for fiscal consolidation recently constructed by Devries et al. (2011) as indicator for fiscal consolidation in the regression analyses. Devries et al. (2011) selected episodes of fiscal consolidation, which were only motivated by the desire of policymakers to reduce the budget deficit or improve government financial sustainability rather than to restrain domestic demand for cyclical reasons. This is important with respect to endogeneity. This database consists of 173 observations of fiscal consolidations in 17 OECD countries (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, Portugal, Spain, Sweden, the United Kingdom and the United States) during the time period 1978-2009. The observations are divided into three measures: total fiscal consolidation, tax-based fiscal consolidations and spending-based consolidation.
(2011) as starting point. All additional variables will be collected for the same 17 OECD countries during the same time period of 1978-2009.
The following chapter offers a literature overview on the effect of fiscal policy on economic growth. The third chapter will research other factors influencing economic growth and fiscal consolidation. Chapter four will elaborate on the way in which variables will be measured and indicate the sources from which the data is collected. The fifth chapter contains a description of the methodology. In the sixth chapter, the analysis of the regression model is put forward and several robustness tests are included. The last chapter will conclude the research and give additional recommendations for future research.
2. The Relationship between Fiscal Consolidation and Economic Growth
In order to answer the research question ‘How does fiscal consolidation affect GDP growth in
the Eurozone and how does this depend on the output gap, the debt-level, and the composition of fiscal consolidation?’, this chapter will review on available research on this topic. In
section 2.1. attention will be paid to possible channels between fiscal consolidation and economic growth. In section 2.2. four matters are explained that influence this relationship between fiscal consolidation and economic growth.
2.1. Channels
Fiscal consolidation can have a direct negative impact on output since direct government spending declines, or if taxes increase, private spending declines. Ardagna (2004) and Alesina and Ardagna (2009) argue that fiscal consolidation can also have an indirect effect on economic growth through two channels. The first channel is based on the expectation-view: if agents believe that future more disruptive fiscal adjustments are prevented, total fiscal consolidation, whether through increasing taxes or reducing spending, has a positive wealth effect. The higher wealth and reduced need for pre-cautionary hoarding will increase demand and induce economic growth. Moreover, this timely consolidation can reduce risk premiums on interest rates. This is called the credibility effect (Abbas et al., 2014). Lower interest rates induce economic activity in two ways. First, lower interest rates directly decrease the interest-bill and therefore a larger share of the budget is available for productive public spending. Second, lower interest rates indirectly reduce private sector borrowing costs. This will induce private investment, consumption, and economic growth. Thus, according to the first channel the indirect effect of fiscal consolidation is that it is
expansionary and thus has a positive effect on economic growth.
The second channel is based on the labor market view. Whether the indirect effect through this second channel is positive or negative depends on the type of fiscal consolidation, i.e. either lowering spending or increasing taxes. Fiscal consolidation will, by lowering spending, reduce public sector employment and possibly unemployment benefits. The decline in unemployment benefits will make being unemployed less attractive to agents. This will result in more agents searching for jobs. Moreover, layoffs in the public sector will increase the group of unemployed even further. The increasing size of the group of unemployed has a downward pressure on the wage and increases marginal productivity of labor. Lower wages increase profits and competitiveness and thereby induce private investment and economic growth. Fiscal consolidation by decreasing government spending will thus have a positive expansionary effect. Fiscal consolidation, by increasing (labor) taxes, will have the opposite effect. Increased taxes will reduce the after-tax wage. Because of labor unions, a higher pre-tax wage will be demanded. This will squeeze profits and reduce investments, competitiveness and economic activity. Whether fiscal consolidation will have a negative or positive effect according to the labor market view thus depends on the composition of the consolidation (Ardagna, 2004; Alesina & Ardagna, 2009).
2.2. Conditional Impact
Further research on the effect of fiscal consolidation or in general of fiscal changes on economic growth elaborates on the idea that the impact of fiscal consolidation or fiscal changes is conditional upon several factors. This literature can be divided into four broad strands: (1) consolidation composition matters, (2) debt levels matter, (3) business cycles matter, and (4) Euro membership matters. Each of these groups will now be explained.
2.2.1. Consolidation Composition
The first strand argues that the composition of fiscal consolidation affects the relationship between economic growth and fiscal consolidation (Alesina et al, 2012, Alesina & Ardagna, 2009; Ardagna, 2004). Ardagna (2004) argues that, due to the labor market view, the composition of fiscal policy is important. Reductions in spending have an indirect positive effect on economic growth through the labor market channel, while tax increases are expected to have a negative effect according to this channel. Moreover, the direct effect of both reduced spending and increased taxes is negative, while the indirect effect of both tax
increases and spending reductions according to the expectations view is positive. The total fiscal consolidation effect would depend on the size of all separate effects. Alesina et al. (2012) agree with Ardagna (2004) that the composition of fiscal policy is extremely important. They find a small spending-multiplier: reductions in spending have small effects on output. However, increases in tax-rates result in persistent recessions. Thus, the tax-based multiplier is large. The channel through which the differing results come into play is the response of private investment. Thus, a reduction in government spending increases private investment, while increases in taxation reduces private investment. This result is consistent with the theory that government spending ‘crowds out’ private investment.
In conclusion, tax-based consolidations are expected to have a negative effect on economic growth according to the labor market view and the direct negative effect, although according to the expectations view higher taxes could have a positive effect on economic growth. The effect of spending-based consolidations depends on the size of the negative direct effect and the positive indirect effect through the labor market and expectations view.
2.2.2. Debt Levels
The second strand of literature argues that the impact of fiscal consolidation is dependent on the debt level. There is a non-linear relationship between government debt, especially debt-to-GDP levels, and economic growth (Checherita-Westphal & Rother, 2012; Dreger & Reimers, 2013; Reinhart & Rogoff, 2010; Vranceanu & Basancenot, 2013). Research on the nonlinear effect of government debt on economic growth started with Reinhart and Rogoff (2010). They argue that when the debt-to-GDP level is higher than 90 percent, economic growth will be adversely affected. Vranceanu and Basancenot (2013) agree and find a negative fiscal spending multiplier in the case of high debt-to GDP levels: the negative effect of crowding out private investment is larger than the positive direct government spending effect. Dreger and Reimers (2013) support this argument. They furthermore indicate when public debt is unsustainable: the nominal output growth is smaller than the nominal interest rate and there is no primary budget surplus to make up for this difference. IMF (2010) also underwrites this effect. They indicate that a deficit reduction is less contractionary in countries with high default risks due to high debt-to-GDP levels.
In conclusion, this section argues that when debt levels are unsustainably high, fiscal consolidation will have a more positive effect on economic growth since it reduces the
crowding out effect more than that it increases the negative direct effect of reduced spending.
2.2.3. Business Cycles
The third strand of literature argues that the state of the economy or the business cycle influences the effect of fiscal consolidation on economic growth (Blanchard & Leigh, 2013; Collignon, 2013; Jayadev & Konczal, 2010; Leão, 2013; Riera-Crichton et al., 2014). Collignon (2013) indicates that the recent austerity in the Eurozone had a negative impact on economic growth because the austerity measures were introduced in a time in which many EMU member-states had negative output-gaps. During a positive output-gap, demand exceeds supply. This results in inflationary pressures, which can be stabilized by austerity. However, a negative output-gap indicates a lack of demand. This has a negative effect on prices, which reduces private investment. Subsequently, potential output capacity and employment fall. “In that case, stimulating demand by increasing private and public spending is required to stabilize the economy” (Collignon, 2013, p. 9). Austerity would, however, inhibit the closing of the output gap, diminish productive capacities and reduce potential GDP by disincentivizing investment.
Gravelle and Hungerford (2013) agree and argue that fiscal consolidation in advanced countries has not been expansionary. They find, by reviewing and further analyzing the findings of popular research in favor of the expansionary effect of spending cuts on economic growth (such as Alesina & Ardagna, 2009) that fiscal consolidations have only been successful when there was almost no output-gap. Leão (2013) researches the question what the effect of an increase in government spending is on the debt-to-GDP level when there is less than full employment by extensively calculating Keynesian multipliers and comparing the minimum values of the multiplier that makes an increase in government spending reduce the debt ratio with the actual multipliers in the United States (US) and the European Union (EU) as a whole. The actual multipliers are from Romer and Romer (2010) and Nakamura and Steinsson (2011). He finds that below full employment a rise in government spending can even result in a lower debt-to-GDP level, since a rise in government spending will increase GDP (the denominator). Higher GDP means higher tax revenues and lower government transfers. Thus, the increase in debt, because of government spending, will only partially translate in an increase in debt (the nominator). This will result in an uncertain effect on the debt-to-GDP ratio. However, Leão (2013) finds evidence of large fiscal multipliers that agree with his argument: the actual multipliers in the case of below full employment are bigger than
the minimum values of the Keynesian multiplier that makes government spending reduce the debt ratio.
Blanchard and Leigh (2013) argue that the expected economic growth since the 2007 financial crisis, due to fiscal consolidation in 26 European countries, has been overestimated. Actual economic growth in these countries is much lower. This is especially true for the early years (2010-2011) of the Euro debt crisis. The findings of Riera-Crichton et al. (2014) explain this. They argue that estimations of fiscal multipliers are biased because they ignore the direction of the change in government spending in an expansion or recession. It is usually assumed that government spending goes up in a recession. However, like current austerity in the Eurozone, government spending can be pro-cyclical as well. When controlling for these influences, the true long-term spending-multiplier for recessions is 2.3 (instead of 1.3 without controlling). Thus, the multiplier for increases in spending is higher than the multiplier of reductions in spending. Moreover, the multiplier for an increase in spending during extreme recessions is much larger. “While cutting spending during typical recessions reduces output by less than one, doing so in extreme recessions reduces output by more than one … this would imply that debt to GDP ratios would increase in response to cuts in fiscal spending” (Riera-Crichton et al., 2014, p. 20).
Jayadev and Konczal (2010) moreover argue, in their descriptive study of the research by Alesina and Ardagna (2009) on expansionary and successful fiscal consolidations that the only two cases in which fiscal consolidations during a recession have not resulted in a deeper slump have been due to currency depreciation and interest rate declines. Most countries that reduce their deficit during recession are left with either lower average economic growth or higher debt-to-GDP levels. Thus, this section indicates that fiscal consolidation during a recession or a time with a large output-gap will have a negative effect on economic growth.
2.2.4. Euro Membership
The Eurozone is an area of countries that introduced the Euro currency and is member of the EMU. All EMU members share a “single monetary policy, a single monetary authority, a single currency and coordinated macroeconomic policies” (Verdun, 2010, p. 325). Euro member countries are expected to be more vulnerable to high debt levels (De Grauwe, 2011; Vranceanu & Basancenot, 2013). This is because they handed over their monetary sovereignty to the EMU or European Central Bank (ECB). As part of the EMU they no longer can print money to finance their debt. Moreover, they cannot depreciate their currency to
become more attractive or use the exchange rate to keep the money supply stable. This makes them vulnerable to a ‘sudden stop’ of liquidity. Additionally, they do not control the interest rate. These conditions can make investors’ confidence loss in the ability of the government to service their debt a self-fulfilling prophecy: when investors lose confidence, they will sell their bonds and invest their Euro’s elsewhere in the monetary union. This creates a sudden stop of liquidity for the affected country and increases the interest rate. This loss of liquidity will make it hard for the country to roll over its debt. Thus, the liquidity crisis is evolved to a solvency crisis (De Grauwe, 2011).
De Grauwe (2011) indicates one additional problem of EMU member countries. Because these countries are part of a monetary union, they can no longer use counter-cyclical policies. An example of a counter-cyclical policy is increasing spending in a time with low economic growth, such as a recession. During a recession, budget deficits will increase, because lower economic activity will diminish the value of the collected taxes. If countries would like to introduce a counter-cyclical policy, the budget deficit will increase further. Investors are more likely to lose their confidence with a high budget deficit. This will increase the interest rate. Since the public and private interest rate is correlated, this will reduce investment and intensify the recession. Thus, an EMU member is forced into a bad equilibrium. This seems to hold for all countries, however the threshold for debt tolerance is much lower in EMU countries. Moreover, according to the Maastricht treaty budget deficits are not allowed to be above three percent of GDP and public debt should be lower than 60 percent of GDP.
It is thus clear that EMU countries are more vulnerable to high debt levels. The debt tolerance for EMU member countries is lower than for non-EMU members. Fiscal consolidation, which will reduce the budget deficit and government debt, would therefore likely have a more positive effect on economic growth in an EMU-member country.
3. Other Factors influencing Economic Growth
Chapter 2 gave an overview of the relevant research on the effect of fiscal consolidation on economic growth. This chapter focuses on the control variables that have been used in two types of literature. The first is the literature on the impact of fiscal consolidation on economic growth. The second is the literature on the determinants of economic growth. These variables should be controlled for when researching the question ‘How does fiscal consolidation affect
GDP growth in the Eurozone and how does this depend on the output gap, the debt-level, and the composition of fiscal consolidation?’.
3.1. Impact Fiscal Consolidation on Economic Growth
Romer and Romer (2010) argue in their research on the (macroeconomic) effects of tax changes on economic growth that it is crucial to include a lagged value of output growth itself in the regression. A lagged value of output growth helps to control for business cycle changes. Moreover “because many factors affecting output growth are likely to be serially correlated, including lagged growth is an easy way to control for a multitude of other influences” (Romer & Romer, 2010, p. 781). Romer and Romer (2010) argue that including lagged growth is a way to test for the hidden motivation of policymakers. “One worry is that even though policymakers may say they are changing taxes for reasons unrelated to current and prospective macroeconomic conditions, perhaps the democratic process causes such changes to be correlated with economic performance” (Romer & Romer, 2010, p. 781). For example, a policymaker calling for tax cuts is more likely to be popular, and thus elected, in times with low demand and low economic growth.
Ardagna (2004) studied the determinants and channels through which fiscal contractions influence the dynamics of the debt-to-GDP ratio and GDP growth and also includes multiple variables as control variables in her real per capita growth regression. These variables are the ratio of the government deficit, the ratio of public debt to GDP, a measure of the weighted average real per capita GDP growth rate (t-1) of G7 countries and the real per capita growth rate (t-1) of the country in question. Moreover, she indirectly controls for political factors, such as whether the government in power is left or center in the political spectrum and whether the party in parliament has a single majority. Additionally, Ardagna (2004) argues that fiscal consolidation often happens in conjunction with other changes in policy. The most important example hereof is monetary policy. Not controlling for monetary policy would mean that the effects of fiscal policy could be biased. Ardagna (2004) therefore includes three additional variables in her regression. Namely, “lagged values of the rate of growth of M2, of the changes in the short-term nominal interest rate, and of the rate of growth of the nominal exchange rate” (Ardagna, 2004, p 1067). Abbas et al. (2014) also mention that monetary policy and inflation can help when dealing with debt and fiscal consolidation. Monetary policy can help inducing demand by lowering the interest rate and increasing liquidity in the economy. Inflation can erode the real value of debt. Alesina et al. (2012)
agrees and mentions that monetary policy (including exchange rate movements) could influence the effect of fiscal consolidation on economic growth. Beetsma and Giuliodori (2011) further support this claim by including the long-run nominal interest rate and the real effective exchange rate in their VAR analysis on the effect of government spending shocks. Additionally, Checherita-Westphal and Rother (2012) underwrite the importance of the long-term real interest rate and the real exchange rate by including it as control variables in their Fixed Effects regression analysis on the non-linear impact of government debt on per capita GDP growth in twelve EU countries.
These findings on the control variables used in research on economic growth, debt and fiscal consolidations are summarized in table 1. These findings clearly indicate that this research must consider controlling for changes in monetary policy by including information on the exchange rate and the interest rate. Moreover, Romer and Romer (2010) make a great point for including a lag of GDP growth additional to initial GDP per capita, as it helps to control for business cycle changes.
Table 1: Control variables used in fiscal consolidation research
Article Controlling for Consists of
Abbas et al. (2014) Monetary policy Inflation
Romer and Romer (2010)
Lag of GDP growth Ardagna (2004) Government finances
Growth controls Political factors Monetary policy
Government deficit, Debt-to-GDP ratio Lag of GDP growth, lag of GDP growth in G7 countries
Left, center, majority
Lag of M2 growth, short-term nominal interest rate, nominal exchange rate
Alesina et al. (2012) Monetary policy Exchange rate Interest rate Beetsma and
Giuliodori (2011)
Monetary policy Long-run nominal interest rate Real effective exchange rate Checherita-Westphal
and Rother (2012)
Monetary policy Long-term real interest rate Real exchange rate
3.2. Determinants of Economic Growth
This section reviews the literature on the determinants of economic growth. Economic growth is usually measured by the change in GDP. GDP is “the total income earned domestically, including income earned by foreign-owned factors of production” (Mankiw, 2010, p. 579). Barro (1991) investigated which variables could explain economic growth in a cross sectional
regression on 98 countries in the 1960-1985 period. Barro (1991) argues that according to neoclassical growth models initial GDP per capita is negatively correlated with economic growth because of diminishing returns of capital. “Poor countries, with low ratios of capital to labor, have high marginal product of capital and thereby tend to grow at high rates” (Barro, 1991, p. 407). However, the assumption of diminishing returns of capital could be untrue, especially if you include human capital. In endogenous growth models human capital triggers technological progress. Countries with high human capital can introduce new goods more rapidly and thereby grow faster. Moreover, higher human capital makes it easier for new (foreign) ideas and technologies to spread and to be incorporated in the economy. Furthermore, higher human capital tends to reduce fertility rates because it increases the opportunity cost of having children. For example, the time spend with the children could be better spent earning a higher wage due to higher human capital. “In effect, people shift from saving in the form of children to saving in the form of physical and human capital” (Barro, 1991, p. 422).
Barro (1991) also mentions (besides initial GDP per capita, human capital and fertility rate) two other variables that influence economic growth. First: political instability. Political instability in his paper is measured as the number of revolutions and assassinations in a country. Political instability negatively impacts property rights and thereby reduces investment and economic growth. Second, according to Barro (1991) market distortions, especially distortions of market prices of capital goods, negatively affect economic growth by reducing investment. If the prices of capital goods are higher due to the distortions, the return to this capital will be relatively lower and will thereby not induce investment.
Sala-I-Martin (1997) sets out to find which variables are correlated with economic growth. He found 62 variables that were significantly correlated with economic growth in at least one of the research studies he read. To test these variables, he used the following regression: ߛ = ߙ+ߚ௬ݕ + ߚ௭ݖ+ߚ௫ݔ+ ߝ. In this regression ࢟ is a vector of variables that are included in each regression and are thought to be robust. These variables are “level of income”, “life expectancy” and “primary-school enrollment rate”. The last two variables are measures for human capital. In the regressionࢠ is the variable of interest to be tested for significance. Moreover,࢞is a combination of three variables picked from the pool of 58 remaining variables.
In total nearly 2 million regressions were conducted in this way. Multiple variables were found to be significant. These variables can be organized into nine groups: (1) regional variables; (2) political variables (rule of law, political rights, civil liberties); (3) religious variables; (4) market distortions and market performance (real exchange rate distortions, standard deviation from the black market premium); (5) type of investment (equipment investment, non-equipment investment); (6) primary sector production (fraction of primary products in total exports, fraction of GDP in Mining); (7) openness; (8) type of economic organization (degree of capitalism); and (9) former Spanish Colonies. Multiple important variables were found to be insignificant as well, such as government spending, the inflation rate and variance, scale effects, tariff restrictions and the black market premium. According to Sala-I-Martin (1997) government spending would probably affect economic growth, but in a non-linear fashion.
The cases studied in this thesis are OECD countries and are therefore more uniform than the cases in the research of Sala-I-Martin (1997). Therefore, there is no need to control for certain differences in religious variables, regional variables, such as Sub-Saharan Africa or Latin America, or former Spanish Colonies. Moreover, certain variables are quite similar in the developed countries that are part of the OECD.
Bassanini, Scarpetta and Hemmings (2001) researched the link between economic growth and policy settings and institutions in OECD countries with a panel regression. They found that human capital in OECD countries is extremely important for economic growth. Moreover, they argue that inflation has a negative effect on economic growth. According to them inflation will reduce investment in physical capital because the high inflation needs to be earned back with higher profitability of the investment. Also, high variability of inflation will create uncertainty. This uncertainty will change the composition of investment towards projects with lower returns and thereby reduce economic growth.
Mankiw, Romer and Weil (1992) also argue that differences in the level of human capital can explain differences in economic growth among countries and therefore introduced the augmented Solow model. In the augmented Solow model not only physical but also human capital is introduced. Nonneman and Vanhoudt (1996) found through their ordinary least squares estimation analysis that human capital is important, but the augmented Solow model still does not have a good explanatory power in OECD countries. They argue that the Solow model should be further augmented by introducing the accumulation of technological
know-how in the model. Bassanini et al. (2001) conducted a panel regression on OECD countries to study the role of policies and institutions on economic growth. They also emphasize the importance of the accumulation of technological know-how or R&D.
Henisz (2000) points out the link between economic growth and the institutional environment in a country. The institutional environment consists of political institutions, the extent of government corruption, checks and balances, protection of property rights and other political and civil rights or liberties. According to Henisz (2000) there are two direct links between institutional environment and economic growth. First, in countries without a strong institutional environment investor uncertainty is quite high. A low institutional environment can result in random modifications in taxation, regulation or other relevant policies. Faced with this uncertainty “investors may choose to either invest in safeguards against policy changes, demand higher and more immediate returns, or alter the nature of their investments – to include not investing at all” (Henisz, 2000, p. 2). The latter is typically the case with long-term investments, which could be a strong driving force for economic growth and development. The second link indicates that in countries with a low institutional environment economic returns can easily be gained through the political process due to corruption. If this is the case, agents will allocate their resources to politics instead of business and economic investment. This relocation will negatively affect economic growth.
The two links between economic growth and institutional environment show according to Henisz (2000) that it is important to include variables that capture institutional environment as control variables in research on economic growth. Examples of such variables are “the extent of the rule of law, political stability, democracy or the degree of political constraints” (Henisz, 2000, p. 20).
Knack and Keefer (1995) also argue that the security of contractual and property rights are extremely important for private investment and thereby economic growth. Moreover, government efficiency with public goods provision and policy creation is also important for economic growth. The absence of these rights or efficiency will discourage investment.
Table 2: Determinants of Economic Growth
The above-described literature on possible determinants of economic growth is summarized in table 2. Several indicators are mentioned multiple times. First, all articles, except for Knack and Keefer (1995), mention human capital as an important explanatory variable. However, there does not seem to be a clear consensus on how human capital should be measured. This will further be discussed in the next chapter. Second, Barro (1991),
Sala-I-Article Explanatory variables Consists of Control variables
Barro (1991) Initial GDP per capita Human capital
Political instability Market distortions
Primary, secondary school enrollment, fertility rate, mortality rate
Assassinations, revolutions Distortions of market prices of capital goods Africa Latin America Sala-I-Martin (1997) Level of income Human capital Regional variables Political variables Religious variables Market distortions Primary sector production Openness Type of economic organization
Former Spanish colonies
Life expectancy, primary school enrollment
Rule of law, political and civil rights
Exchange rate, black market premium Not applicable Bassanini et al. (2001) Human capital Inflation Accumulation of know-how
Domestic and business expenditure on R&D Population growth Government size Financial development Openness Mankiw et al. (1992)
Human capital Not applicable
Nonneman and Vanhoudt (1996) Human capital Accumulation of technological know-how
Domestic expenditure on R&D and real domestic investment
Population growth
Henisz
(2000) Human capitalForeign exchange rate
Institutional environment Corruption, checks and balances, protection of property rights, black market premium, political and civil liberties
Not applicable Knack and Keefer (1995) Property rights Government efficiency Not applicable
Martin (1997), Henisz (2000) and Knack and Keefer (1995) all mention variables that are related to institutional environment, political stability and the protection of property rights. Other important variables that are included are level of income/initial GDP per capita (Barro, 1991; Sala-I-Martin, 1997), the exchange rate (Henisz, 2000; Sala-I-Martin, 1997) and the accumulation of technological know-how as measured by expenditure on R&D (Bassanini et al, 2001; Nonneman & Vanhoudt, 1996). It is clear that all these variables must be considered as control variables
In conclusion, all the important variables that apply to the research question ‘How
does fiscal consolidation affect GDP growth in the Eurozone and how does this depend on the output gap, the debt-level, and the composition of fiscal consolidation?’ are identified. The
dependent variable is economic growth. The important explanatory variables are (1) fiscal consolidation (including spending-based and tax-based fiscal consolidation), (2) an interaction variable of fiscal consolidation and EMU membership, (3) an interaction variable of fiscal consolidation and unsustainable debt, (4) an interaction variable of fiscal consolidation and the business cycle. Moreover, several variables need to be considered as control variables: human capital, institutional environment, monetary policy, GDP per capita and lagged growth, and the accumulation of technological know-how.
4. Data Selection and Measurement
All crucial variables to answer the research question ‘How does fiscal consolidation affect
GDP growth in the Eurozone and how does this depend on the output gap, the debt-level, and the composition of fiscal consolidation?’ are identified. This chapter will indicate how these
variables will be measured. Moreover, where the data for each indicator is originated from is also specified. This thesis is dependent on the narrative approach database for fiscal consolidation of Devries et al. (2011). Data will be selected for all 17 OECD countries in this database (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, Portugal, Spain, Sweden, the United Kingdom and the United States) during the time period of 1978-2009.
4.1. Economic Growth/Income
The dependent variable economic growth will be measured by GDP growth (annual %) in the year in question. Moreover, it is necessary to include lagged growth and initial income as
control variables. First, lagged growth will be measured by GDP growth in year t-1. Second, GDP per capita (current US$) will measure the initial income of a country. All data on GDP growth will come from World Bank (2015a). Data on GDP per capita comes from World Bank (2015b).
4.2. Fiscal Consolidation
The fiscal consolidation data in this thesis are similar to the ones used in Alesina et al. (2012). They did research on whether fiscal consolidation would cause output losses. They indicate that the key difficulty in research in this topic is finding the changes in government spending or taxes that are not a response to the state of the economy. This means that the changes in government spending or taxes are exogenous. Narrative approaches to data gathering can identify these exogenous changes. Alesina et al. (2012) argue that the narrative approach also helps with a possible omitted variable bias: “shocks identified via a narrative method are model independent and therefore are not affected by the possibility that some variables might be omitted in the estimation” (Alesina et al., 2012, p. 8).
Alesina et al. (2012) use the database constructed by Devries et al. (2011). Devries et al. (2011) developed a dataset of fiscal consolidation in 17 OECD countries in the period of 1978-2009. This database was developed with the specific purpose of furthering the research on the impact of fiscal consolidation on macroeconomic variables. It is introduced as an alternative to the often used cyclically adjusted primary budget balance (CAPB) as measure of fiscal consolidation. The primary balance is the “government net borrowing or net lending excluding interest payments on consolidated government liabilities” (OECD, 2005, p. 1). This primary balance is cyclically adjusted “to filter the impact of cyclical movements [in the state of the economy] on fiscal variables and assess the ‘underlying’ fiscal stance” (Fedelino, Ivanova, & Horton, 2009, p. 2).
According to Devries et al. (2011) cyclically adjusted primary balances are a problematic proxy for fiscal consolidation because of two reasons. First, although the primary balance is cyclically adjusted, the measure is still likely to suffer from measurement error. “Cyclical adjustment typically fails to remove the impact of sharp swings in economic activity and asset prices from fiscal data, resulting in changes in the CAPB that are correlated with economic activity but are not necessarily linked to policy action” (Devries et al., 2011, p. 3). The second problem according to Devries et al. (2011) is reverse causality. It is not clear
whether changes in the CAPB are the result or the cause of changes in fiscal policy. These two problems give rise to an endogeneity problem.
To overcome these problems, Devries et al (2011) took a different approach. They used a historical approach that looked at the intention and actions of governments. This is also called the ‘narrative approach’. They indicated the intentions with qualitative research on historical sources and policy documents. The sources they researched include IMF reports, OECD economic surveys, central bank reports and other country-specific sources. In these sources they looked for two main motivations. Namely, the motivation to reduce the budget deficit to maintain or improve financial sustainability of the government and the motivation to mitigate cyclical changes. The fiscal consolidation episodes they identified are a result of (past) decisions and therefore unlikely to be linked with other variables or developments affecting growth. The financial impact on the budget of the government is measured in percent of GDP. Possible issues with endogeneity are considered in the methodology chapter.
The final result of this process is a database of fiscal consolidation in 17 OECD countries in the period of 1978-2009. The data is of annual frequency. The 17 OECD countries include “Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, Portugal, Spain, Sweden, the United Kingdom and the United States” (Devries et al., 2011, p. 4). In total there are 173 observations of fiscal consolidation. These observations are indicated in three ways: total fiscal consolidation, tax-based fiscal consolidation and spending-tax-based fiscal consolidation. This makes it possible to test the claim of Alesina et al. (2012), Alesina and Ardagna (2009) and Ardagna (2004). They argue that tax-based consolidations will have a negative effect on economic growth, while spending-based consolidations have a small positive effect on economic growth
4.3. EMU Membership
As mentioned, an interaction variable of fiscal consolidation and EMU membership is included in the regression. EMU membership will be measured by a dummy variable, which is coded one if the Euro is adopted. All EMU-countries in the sample adopted the Euro (Austria, Belgium, Finland, France, Germany, Ireland, Italy, the Netherlands, Portugal, Spain) in 1999 (European Commission, 2015), so the dummy is coded one for these countries since 1999.
4.4. Debt Levels
Multiple authors (Checherita-Westphal & Rother, 2012; Dreger & Reimers, 2013; Reinhart & Rogoff, 2010; Vranceanu & Basancenot, 2013) claim that there is a non-linear relationship between debt-to-GDP levels and economic growth. This literature constructs the following hypothesis: when debt levels are unsustainably high, fiscal consolidation will have a more positive effect on economic growth since it reduces the crowding out effect. To accommodate this hypothesis an interaction variable of fiscal consolidation with an unsustainable debt-dummy is included in the regression analysis.
As indicated, Reinhart and Rogoff (2010) argue that debt-levels above 90 percent of GDP will adversely affect economic growth. Checherita-Westphal and Rother (2012) find a similar value of 90 to 100 percent of the debt-to-GDP ratio in their research on debt and economic growth in the Eurozone. This thesis will follow Reinhart and Rogoff (2010) by picking 90 percent of to-GDP as threshold value for unsustainability. Thus, the unsustainable debt-dummy is coded 1 if the debt-to-GDP ratio is at least 90 percent.
This thesis follows Reinhart and Rogoff (2010) by defining public debt as gross government debt. The indicator used is general government gross debt as a percentage of GDP. Data will be collected from IMF (2015a).
4.5. Business Cycle Levels
As indicated, many articles (Blanchard & Leigh, 2013; Collignon, 2013; Jayadev & Konczal, 2010; Leão, 2013; Riera-Crichton et al., 2014) argue that the state of the economy influences the effect of fiscal consolidation on economic growth. They claim that fiscal consolidation during a recession or a time with a large output-gap will have a negative effect on economic growth. To test this hypothesis, an interaction variable of fiscal consolidation with the output-gap is included in the regression analysis. The output output-gap is the difference between actual and potential GDP. The output gap is measured in percentage of the potential GDP. Data for this indicator is collected from IMF (2015b).
4.6. Human Capital
A standard control variable in research on economic growth concerns human capital. Barro (1991) uses primary and secondary school enrollment rates as proxies for human capital. These proxies can be criticized because they could be evidence of the investment in human
capital, rather than the initial stock of human capital. This gives the problem of reverse causality. It is argued that higher human capital will increase economic growth, however, it is also plausible to argue that economic growth will increase the investment in human capital. Barro (1991) tested whether this was true by changing the date in which the level of human capital proxies was measured to ten years prior and after his initial measurement date and found that these measures were highly correlated. Barro (1991) thus argues that the school enrollment rate proxies the initial level of human capital rather than the investment in human capital.
Another value that could be a proxy for human capital is the literacy rate. The literacy rate indicates the level of human capital instead of the investment in human capital without a doubt. However, the literacy rate is quite hard to measure uniformly across countries. Therefore, Barro (1991) argues that school enrollment rates are more accurate and consistent.
This thesis will follow Barro (1991) in selecting the primary and secondary school enrollment rates as indicators of human capital. This research will further extend this by also including tertiary school enrollment rates. In order to minimize the individual additional regressors, the school enrollment rates will be included as one average variable. Each of the enrollment rates will have equal weight. Thus, human capital will be measured by adding primary, secondary, and tertiary school enrollment rates and dividing it by three1. Primary
school enrollment will be measured by World Bank (2015c). Secondary school enrollment will be measured by World Bank (2015d). Tertiary school enrollment will be measured by World Bank (2015e)
4.7. Institutional Environment
The quality of institutions has often been used as a determinant of economic growth in the literature in recent decades. Knack and Keefer (1995) give an overview of measures that researchers have used to measure institutional environment and the extent to which property rights are protected. These measures are (1) political instability, measured by the number of revolutions or assassinations in a country, and (2) measures of political and civil liberties. Political instability can be used as a proxy for institutional environment and protection of property rights because during political instability instruments for safeguarding property
1Calculating human capital in this way will mainly show the differences in the tertiary school enrollment rates. This is appropriate as in high-income countries “a workforce that is highly educated … is much better prepared to adapt to new technologies, innovate and compete on a global level” (World Economic Forum, 2015, p. 5).
rights are expected to be more fragile. Moreover, political leaders in an instable country are often quickly replaced and need to maximize their political returns in a short period of time. These political leaders will therefore be more likely to seize private goods, mostly because they do not need to bear the future costs.
Knack and Keefer (1995) criticize the measure for political instability in four ways. First, the measure only focuses on unconstitutional changes of those in power. However, even chosen leaders lose their office all the time and therefore also need to maximize their political returns in a short time period. This critique is largely unfounded because constitutionally elected leaders are (more) accountable to their citizens and govern by their vote. Leaders can therefore prolong their time in office by implementing efficient policy and protect the rights and property of voters. The second critique on the measure of political instability is that countries without instability, such as a strong dictatorship, still could have low protection of property rights and inefficient policies. The third critique is that political instability such as assassinations might be easy to measure, however, it only marginally can explain the effect of institutions on property rights. It is therefore a crude proxy. The fourth critique of Knack and Keefer (1995) is that of reverse causality. Political instability can threaten the protection of property rights but the opposite can also be true. Unsecure property rights can lead to chaos and thereby political instability. Thus in conclusion, political instability, measured by the number of revolutions or assassinations, in itself is not an appropriate measure for policy efficiency and the protection of property rights.
A better measure of institutional environment, which includes political stability and measures of political and civil liberties, is the Polity IV indicator (Marshall, Gurr, & Jaggers, 2014). The Polity IV project tries to code the institutional features of countries. The Polity IV indicator consists of two variables that are subtracted from each other. The first variable is institutionalized democracy. An institutionalized democracy has three crucial components:
One is the presence of institutions and procedures through which citizens can express effective preference about alternative policies and leaders. Second is the existence of institutionalized constraints on the exercise of power by the executive. Third is the guarantee of civil liberties to all citizens in their daily lives and in acts of political participation (Marshall, Gurr, & Jaggers, 2014, p. 14).
Concepts that are important for economic growth such as the protection of property rights, checks and balances, and the rule of law are incorporated in these components. The
institutionalized democracy variable can have a value between 0 and 10 in which 0 means no institutionalized democracy and 10 indicates a full institutionalized democracy (Marshall, Gurr, & Jaggers, 2014).
The second variable is institutionalized autocracy. An institutionalized autocracy consists of a “sharply restricted […] competitive political participation. Their chief executives are chosen in a regularized process of selection within a political elite, and once in office they exercise power with few institutional constraints. Most modern autocracies also exercise a high degree of directiveness over social and economic activity” (Marshall, Gurr, & Jaggers, 2014, p. 15). The autocracy variable also has a score between 0 and 10 in which 0 indicates the country is not autocratic and 10 indicates that the country is completely autocratic (Marshall, Gurr, & Jaggers, 2014).
The Polity IV indicator is constructed by subtracting the autocracy score from the democracy score. The resulting value is between -10 and 10. -10 indicates a completely autocratic polity without any protection of civilians or constraints on the executive. A score of 10 indicates a complete democratic polity with civil rights and institutionalized constraints. This indicator is converted slightly to better accommodate the needs of time-series analyses in case of missing values. This is the Polity2 IV indicator (Marshall, Gurr, & Jaggers, 2014). This thesis will use the Polity2 IV as indicator for the institutional environment. Data will be collected from the Integrated Network for Societal Conflict Research (2015).
4.8. Monetary Policy
Following Beetsma and Giuliodori (2011) and Checherita-Westphal and Rother (2012) this thesis will identify the impact of monetary policy by controlling for changes in the interest rate and the exchange rate. As in Checherita-Westphal and Rother (2012) the interest rate will be measured by the real interest rate. This is the interest rate that is inflation adjusted. Data will come from the World Bank (2015f). The exchange rate will be measured by the real effective exchange rate index. Data is collected from the World Bank (2015g).
4.9. Accumulation of Technological Know-How
Following Bassanini et al. (2001) this thesis will measure accumulation of technological know-how by the investment in research and development. Data will be collected for the indicator ‘research and development expenditure, percentage of GDP’ from World Bank (2015h).
In conclusion, table 3 shows the variables, indicators and data sources that will be used in the main regression analysis. To check whether the outcomes of the analyses are robust, the regression analyses are also redone with certain changes to the indicators. These changes are explained in chapter 5.3.
Table 3: Indicators used for regressions
Variable Indicator Comments Source
Economic
growth/income GDP GrowthGDP per capita Lag of GDP Growth
Dependent variable Control variable Control variable
World Bank (2015a) World Bank (2015b) World Bank (2015a) Fiscal
consolidation
Total fiscal consolidation Tax-based fiscal consolidation Spending-based fiscal
consolidation
Constructed with the narrative method
Devries et al. (2011)
EMU membership EMU membership dummy Coded 1 when Euro
is adopted
European
Commission (2015) Unsustainable debt General government gross
debt, % of GDP dummy
Coded 1 if threshold of 90% is reached
IMF (2015a)
Business cycle Output-gap as % of GDP IMF (2015b)
Human Capital Primary enrollment rates
Secondary enrollment rates Tertiary enrollment rates
Constructed as one variable with equal weights
World Bank (2015c) World Bank (2015d) World Bank (2015e) Institutional
environment
Polity2 IV Constructed from a
democracy and autocracy score
Integrated Network for Societal Conflict Research (2015) Monetary policy Real interest rate
Real effective exchange rate
World Bank (2015f) World Bank (2015g) Accumulation of
technological know-how
Research and development expenditure, % of GDP
World Bank (2015h)
5. Methodology
The resulting database consists of 173 observations of 17 OECD countries (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, Portugal, Spain, Sweden, United Kingdom, United States) in 1978-2009. This is an unbalanced panel dataset, which indicates that there are missing values. This is especially the case for the debt-to-GDP ratio and research and development expenditure. Before turning to the methodology, it is important to mention the endogeneity problem.
5.1. Endogeneity Problem
Endogeneity indicates a correlation between a variable and the error term. Endogeneity can be the result of omitted variables bias, reverse causation and measurement error. Alesina et al. (2012) argue that there is no endogeneity problem when including fiscal consolidation as an explanatory variable for economic growth when the narrative approach is chosen. The measure of fiscal consolidation does not depend on the business cycle because this “is ruled out by the way narrative shocks are identified” (Alesina et al., 2012, p. 25). The narrative approach of Devries et al. (2011) identifies two motivations for fiscal consolidation: (1) to reduce the budget deficit to maintain or improve financial sustainability and (2) to restrain public demand for cyclical reasons. Only fiscal consolidation episodes that can be categorized under the first motivation are included in the dataset. The fiscal consolidation episodes in the database are, therefore, not used to influence the business cycle. However, it is important to keep in mind that financial sustainability is not a standalone thing. Financial sustainability, and the budget deficit, are in part determined by economic growth. If economic growth is low, tax income is smaller which means that a financial sustainability is in jeopardy. Low economic growth can thus be an indirect (and implicit) cause of the motivation to reduce the budget deficit to maintain or improve financial sustainability. This could imply a possible reverse causation in the dataset if Devries et al. (2011) were unable to identify this indirect link in the historical documents. As indicated by Romer and Romer (2010) including lagged growth is a good way to test for the hidden motivation of policymakers.
Alesina et al. (2012) indicate that there could be an endogeneity problem between the separate indicators tax-based financial consolidation and spending-based financial consolidation and the business cycle. The type of fiscal consolidation chosen to reduce the deficit could be dependent on the state of the economy. In this research this could be a problem as well, especially when including an output-gap dummy. To test this possible endogeneity problem, Alesina et al. (2012) constructed a measure of the business cycle. The measure of the cycle is “the deviation of output from its Hedrick-Prescott trend” (Alesina et al., 2012, p. 25). They then run a binary choice (panel) probit regression, including dummies for tax-based and spending-based financial consolidation. They find a McFadden R square of 0.001 from the regression between tax-based fiscal consolidation and the business cycle. The McFadden R square is a little bit larger for the regression between spending-based fiscal consolidation and the business cycle, namely 0.01. The coefficient variable is -0.16 and the standard error is 0.07. Thus, it is a little bit more likely that spending-based fiscal
consolidation is correlated with the business cycle. However, the slight significance completely disappears when including time dummies. These time dummies incorporate common random shocks. Alesina et al. (2012) therefore argue that there is no endogeneity problem with using these variables.
5.2. Regression Model
This thesis will conduct three main analyses and compare the results. In the first analysis total fiscal consolidation is used as the main explanatory variable. In the second analysis tax-based fiscal consolidation is the main explanatory variable. The third analysis will use spending-based fiscal consolidation as the main explanatory variable. As indicated three interaction effects are considered. Namely, (1) the interaction between fiscal consolidation and EMU membership, (2) interaction between fiscal consolidation and the output gap, and (3) interaction between fiscal consolidation and unsustainable debt. In the second and third analysis these interaction variables are recomputed with tax-based fiscal consolidation and spending-based fiscal consolidation respectively. Moreover, in each analysis the control variables will be included separately to understand the effect of the control variables on the relationship between the explanatory variables and economic growth. Hereto the control variables are separated into three groups:
1. Monetary policy: interest rate, exchange rate 2. GDP per capita, lag of GDP growth
3. Human capital, institutional environment, R&D spending.
With panel data, two types of regression models are most common. Namely, pooled ordinary least squares (OLS) regression and fixed effects regression (McManus, 2011; Park, 2009; Stock & Watson, 2012). Pooled ordinary least squares regression assumes that each observation is independent of each other. Thus, Australia in 1987 is seen as an uncorrelated observation with Australia in 2000. This means that it is assumed that there is neither a time trend nor a country trend. Each year entry of a country is seen as independent. Thus, pooled OLS assumes the following model:ݕ=ߚ+ ߚଵݔଵ+ ߚଶݔଶ+ ݑ. In which ݑ is the error term. The error term is assumed to be homogeneous. Since the model and variables that need to be estimated are unlikely to be endogenous, pooled OLS regression could be an appropriate option. Whether this option is appropriate is also dependent on autocorrelation, which will be checked with the Durbin Watson statistic.
The second type of regression is a fixed effects (FE) regression. Fixed effects regression assumes that there are differences between units (i.e. countries) that are important and need to be controlled for. These differences do not change over time. FE regression assumes the following model: ݕ௧= ߚଵݔଵ,௧+ ߚଶݔଶ,௧+ ߙ+ ݑ௧. Here,ߙ represents the country specific effects. In effect, the fixed effects regression is a pooled regression model including dummy variables to control for country differences.
Thus, shortly said, pooled OLS assumes that there are no differences between entities. FE regression assumes that there is variation, which needs to be controlled for. This thesis found that in an initial analysis, in which the most important variables were included, that the results of the pooled OLS and Fixed Effects regression are extremely similar. These regression results are included in table 13 and 14 in appendix 1. Thus, it can be assumed that there is no obvious country fixed effect in the dataset. This thesis will therefore only conduct pooled OLS regressions2.
5.3. Robustness Tests
To make sure that the regression results are robust, the regression analyses will be redone with different calculations for some of the variables, namely the dependent variable economic growth, the EMU membership dummy and the debt sustainability dummy. Table 4 shows the indicators that are changed in the additional robustness analyses.
First, the dependent variable will be measured by a three-year average of GDP growth. This average will be calculated by adding GDP growth in the year t, t+1 and t+2 and dividing it by three. This is similar to Alesina and Ardagna (2009) and is used to accommodate for medium term effects of fiscal consolidation.
Second, the EMU membership dummy will also be recoded. Checherita-Westphal and Rother (2012) argue that before the adoption of the Euro, there was a pre-EMU convergence phase that started when the Maastricht Treaty came into force in 1993. Therefore, the EMU membership dummy will be recoded as 1 since the year the Maastricht treaty came into force. For all countries in our sample that introduced the Euro, the Maastricht treaty came into force
2The dataset of this thesis consists of 173 observations and 18 variables of interest. Including as many as 17 additional regressors (for country fixed effects) reduces the degrees of freedom immensely, risking undefined model statistics. Therefore, the simpler model of pooled OLS is chosen. For the same reason, year dummies (meaning 29 additional regressors) are not included in the regression. Also, for some years (such as 1978, 2001, 2009) only one observation is included in the dataset.
in 1993. This dummy will thus be used instead of the EMU dummy used in the main regression coded 1 since the adoption of the Euro in 1999.
Third, this thesis will recode the unsustainable debt dummy. The value 1 is given for a debt-to-GDP ratio above the threshold of 60 percent, rather than the 90 percent used in the main regression. The 60 percent threshold is copied from the Maastricht convergence criteria and might give some insight in why European policymakers decided on this random threshold.
6. Analyses and Robustness
This chapter includes both a descriptive data analysis and the main regression results. Examination of the data showed two interesting points. First of all, fiscal consolidation consisted for 57.76% out of spending-based adjustments. Only 42.24% was based on tax-based fiscal consolidation.
Moreover, when splitting the dataset by whether a country is an EMU-member or not, some interesting descriptive statistics came to light. First of all, average GDP growth is lower in EMU member countries. Additionally, the average fiscal consolidation is also somewhat smaller in EMU member countries. Moreover, the average debt-to-GDP ratio is more than 10 percent smaller in EMU member countries. Lastly, the average output-gap is positive in EMU member countries. This means that actual GDP is actually bigger than potential GDP. This not the case in non-EMU member countries. They have a negative average output-gap. The findings are summarized in table 5 and 6.
Concept Indicator Comments Source
Economic growth Three-year average of GDP
growth
Average calculated from t, t+1 and t+2
World Bank (2015a)
EMU membership EMU membership dummy Coded 1 if the
Maastricht Treaty (1993) is in force
European
Commission (2015) Unsustainable debt General Government Gross
Debt, % of GDP dummy Coded 1 if thresholdof 60% is reached IMF (2015a)