The Relationship between Public Debt and Economic Growth in the Eurozone

(1)

The Relationship between Public Debt and Economic Growth in the

Eurozone

Peter J. van der Noord1 University of Groningen

June 15th 2016 Supervisor: Richard Klijnstra

Abstract

This study applies various estimation techniques to analyze the relationship between public debt and subsequent economic growth in the Eurozone. Increasing public debt is associated with lower subsequent economic growth. The effects are stronger on the medium- to long-term compared to the short-term. This negative effect of public debt appears to be much larger for GIIPS countries. In addition, the financial crises of the past years have led to highly increased public debt levels and decreasing or stagnating economic growth. We find no significant evidence to explain the relationship in crisis years.

Keywords: Public Debt; Government Debt; Economic Growth; Eurozone; Financial Crises; GIIPS.

JEL classification: F33, F34, H63, O11, O47

(2)

2 1. Introduction

(3)

3 government revenues that have hit advanced and emerging market economies

alike” (Reinhart & Rogoff, 2010, pp. 574). There remain to be many different views on the impact of debt on economic growth and the best way to deal with debt especially after the financial crises.

Figure 1.1: Cumulative Increase in Real Public Debt, 2007-2009.2

Much has been written on economic growth, especially in the recent years. However, many of the empirical studies have failed to include the recent crises in their research, which is a shame considering the gigantic impact it has had on the entire world. Since Europe has had to deal with quite a lot more than merely the global financial crisis (i.e. the European debt crisis), we are very interested to examine the Eurozone for this study. Reinhart and Reinhart (2015) have shown that crises are typically associated with lower subsequent medium-term growth. We will examine if the large degree of debt accumulation is responsible for the lower medium-term growth. If we are successful, our study together with

(4)

4 additional empirical literature in the field could potentially aid policy makers in

the Eurozone as well as the European Central Bank (ECB) in devising a long-term vision on public debt and economic growth. Policy makers and governments in the Eurozone, together with the ECB should work more closely together to devise a strategy to increase economic growth in the aftermath of the recent crises.

This paper has multiple purposes regarding the topic of public debt and economic growth. Firstly, it sketches the changing situation of public debt and economic growth in Eurozone countries since 1995 up till now. Secondly, the goal is to empirically examine the effects of the initial public debt of Eurozone countries on their subsequent growth of real GDP per capita. Thirdly, we will hunt for various patterns in our data. For instance, what has the global financial crisis of 2007/2008 done to impact the relationship between public debt and economic growth, and also what was the impact of the European debt crisis? Furthermore, we are interested to see if there are differences between countries who have been hit hard by the European debt crisis and other countries. We will look at Greece, Italy Ireland, Portugal, and Spain (GIIPS-countries) separately to view if the relationship of debt on economic growth there shows similar patterns as in our entire sample or if there are significant differences. The idea to look at the GIIPS-countries separately came from the paper by Popov and van Horen (2015) who looked at syndicated bank lending during the European debt crisis. In addition, Aspergis and Cooray (2016) look at the influence of debt uncertainty in economic growth for the GIIPS countries. They have found a significant negative effect of debt uncertainty on economic growth in these countries.

2. Literature Review

(5)

5 debt in the short run can stimulate output and aggregate demand, but in the long

run crowds out capital and reduces output (Elmendorf and Mankiw, 1999, provide an extensive literature overview). Aizenman, Kletzer and Pinto (2007) argue that a country’s growth rate in the steady-state will be higher when there is a binding exogenous debt limit or when there are constrained borrowing policies. In other words, growth in the country will be higher when public debt is limited. In addition to the above, Reinhart and Rogoff (2009) provide a remarkably extensive overview of economic crises and public debt, highlighting patterns prevalent over several centuries.

(6)

6 on the subject have yielded quite similar results, despite the different data samples

of countries and periods. Caner, Grennes, and Koehler-Geib (2010) confirm the negative relationship for a sample of 101 developing and developed countries. Panizza and Presbitero (2012) examine the relationship for 17 OECD countries and Cecchetti, Mohanty, and Zampolli (2012) for 18 OECD countries in the same time period of 1980-2005. Checherita and Rother (2012) survey a sample of 12 Eurozone economies for the period between 1970 and 2008, where Baum, Checherita-Westphal, and Rother (2013) also examine 12 Eurozone economies, but in their research for the period of 1990 to 2010.

A substantial amount of the conducted research has focused on finding some sort of threshold level for which the negative relationship between debt and GDP becomes significant. Reinhart and Rogoff (2010) in their research containing forty-four countries and a time period of over two hundred years have concluded that a threshold level of debt-to-GDP of roughly 90% exists. Their main finding is that the relationship between a country’s debt and GDP is weak when its debt-to-GDP ratio is below 90%. On the other hand countries with a debt-to-debt-to-GDP ratio higher than 90% have a significantly lower median and average growth rate3. Baum et al. (2013) focus on debt sustainability in the euro area and find that the effect of debt on GDP growth in the short run is positive for public debt-to-GDP levels below 67%. When public debt-to-GDP levels are 67% and above, the short-run impact decreases to about zero and is no longer statistically significant. Furthermore, they find that for debt-to-GDP levels above 95%, the impact of additional debt on economic activity, and thereby also economic growth, is negative. Caner et al. (2010) find even lower thresholds for a negative impact of debt on GDP. In their research on developing and developed economies for the period of 1980-2008 they establish a threshold level of 77% for long-run public debt-to-GDP above which the impact of additional debt is negative for economic growth. In addition, they establish an even lower threshold when only looking at

(7)

7 the developing economies, 64% of public debt-to-GDP. We should note that it can

be quite costly for countries to have public debt-to-GDP levels above the threshold. However, we are talking about long-run public debt thresholds. This should be emphasized, since the public debt-to-GDP levels we observe nowadays are potentially far higher than the long-term average as a result of the recent financial crises. Reinhart & Rogoff (2009) have shown that, during the modern era, real government debt in the three years following a banking crisis rises by 86% on average. That is to say that if the real stock of debt is 100 before the crisis, it increases to 186 during the three years following the crisis. In addition, they show that historically a government’s revenue growth is robust in the years leading up to the crisis and declines for an average of three years following the crisis. This explains why the real stock of debt increases in the three years following a crisis as well. Furthermore, the measures taken by the ECB by buying large amounts of government debt could have a substantial impact on a country’s public debt levels. Bond yields in Europe have decreased substantially due to the ECB measures and have served European countries by lowering the interest they have to pay on newly issued loans, due to increased security and stability in the market. This makes it far easier and cheaper for countries to borrow money in the market, which could potentially increase their public debt levels and also decrease the potential negative effect of high debt levels. The thresholds we have presented earlier are all measured for periods before the financial crises (and before the ECB measures) and are therefore not directly applicable to the debt-to-GDP levels of European countries nowadays.

(8)

8 To our knowledge, we are the first to examine European countries separately for a

period including the recent financial crises.

3. Methodology 3.1 Model specification

The main analysis in our paper focuses on the medium and long run relationship between initial government debt and subsequent economic growth in the Eurozone. Our econometric analysis is largely follows the work by Woo and Kumar (2015), who have used a panel of 38 emerging and developed economies for their main sample and full sample of 79 countries for the years 1970-2008. The sample used in this research will focus on the 12 initial Euro-countries for a 20 year period between 1995 and 2014. A more extensive look at the data will be provided later on, where we will also explain and support our data choices.

The baseline panel regression is as follows:

yi,t – yi,t-τ = αyi,t-τ + Xi,t-τβ + γZi,t-τ + ηt + vi + εi,t, (1)

where a period is a three-year or a five-year time interval (i.e. τ = 2, or τ = 4). The end of the period is denoted by t, and the start of the period by t – τ. The country is denoted by i and y denotes the natural logarithm of the real GDP per capita. ηt is

the time fixed effect, vi is the country-specific fixed effect, and εi,t is the

unobservable error term. The initial government debt is denoted by Zi,t-τ, for this

we use the initial government debt as a percentage of the country’s GDP. Lastly,

Xi,t-τ is a vector of financial and economic control variables as used in Woo and

Kumar (2015). It consists of the following 9 different control variables:

- The initial level of real GDP per capita. This is used to capture the catching-up process, and to limit the reverse causality problem.

(9)

9 - The initial level of trade openness in the country. The sum of imports and exports

as a percentage of GDP is used.

- The initial level of inflation as measured by the Consumer Price Index (CPI). We use the natural logarithm of 1 + the inflation rate.

- The initial level of financial market depth. As a proxy we use the liquid liabilities of the country as a percentage of its GDP.

- The initial government size. We measure this by using the government consumption as a percentage of GDP.

- The fiscal deficit of the government. To reflect the notion that a government’s fiscal deficits are negatively associated with long run growth (for further explanation see Fischer 1993; Baldacci et al. 2004).

- The terms of trade growth rates as a percentage of GDP.

- A measure of banking crisis incidence. To reflect the results by Reinhart and Rogoff (2009), who found that banking crises are often accompanied by large surges in government debt for the years following the crisis. In addition, economic growth is slowed down after a banking crisis.

(10)

10 We have made several adjustments to the variables considered by Woo and

Kumar (2015), due to a lack of data after 2008 and 2010 respectively. For human capital we use the years of secondary schooling instead of the Barro and Lee (2011) statistic used by Woo and Kumar (2015) due to a lack of data after 2010. However, we will test the effectiveness of the variable compared to the effectiveness of a benchmark statistic on human capital based on Barro and Lee (2011) and Psacharopoulos (1994) used in the Penn World Table 8.1 (Feenstra, Inklaar, and Timmer, 2015) to justify this decision. In addition for the measure of banking crisis incidence we cannot rely solely on the work by Reinhart and Rogoff (2009), as they do not observe banking crises after the year 2008. Therefore we use the five-year CDS spread of the country as a proxy for banking crisis incidence for the years after 2008. We construct a dummy variable which attains a value of 1 when the country has experienced a banking crisis that year and a value of 0 when it has not. Whenever the five-year CDS spread in a country in one year is higher than 400, that is 400 basis points, we conclude that is has experienced a banking crisis.

3.2 Hypotheses

We will formulate three independent hypotheses for our research. These are mostly based on the findings of several previous studies on the subject. We expect to find similar outcomes as previous studies did, even though our dataset is vastly different. Therefore, our first hypothesis will become:

H1: A country’s public debt has a negative influence on its economic growth in the

following three to five years.

(11)

11 Therefore, we expect the presumed negative effect of public debt on economic

growth to be significantly larger. Hence, our second hypothesis is formulated as follows:

H2: The negative relationship between public debt and economic growth is

stronger in GIIPS countries.

The final hypothesis is focused around the crisis years. Many researchers have investigated the influence of financial crises on economic growth and the role that debt has played in this relationship. Reinhart and Rogoff (2009) have shown that public debt rises significantly in the years following a crisis and that economic growth is usually lagging behind. We formulate our final hypothesis as follows:

H3: Public debt has a more substantial negative impact on economic growth in

years of crisis.

We expect that the presumed negative effect of public debt on economic growth to be significantly larger in the years following a crisis, not only for the GIIPS countries, but for our entire sample.

3.3 Regression analysis

(12)

12 different estimation techniques we aim to improve the robustness of our results.

We use an OLS regression, a Fixed Effects panel regression (FE), and a robust regression (RR). Our regressions will be carried out over two different time frames, three and five years respectively. The periods will be overlapping following the work by, among others, Cecchetti et al (2012). We use the RR to protect ourselves against potential biasing effects through outliers. Moreover, the RR is employed, as it is not sensitive to heteroscedasticity. Furthermore, we will carry out our OLS and FE regressions with and without time fixed effects, meaning that we will have a total of five different analyses. We will include time fixed effects in two of our analyses to account for the possibility that global factors will affect public debt and economic growth at the same time, which could potentially bias our results towards finding a stronger relationship. Although, as Woo and Kumar (2015) argue, the global factors could also be correlated with the fiscal or economic variables in our regression, which could lead to an understatement of the effects of these variables. We are interested in all five different analyses to arrive at the most robust results possible.

4. Data

(13)

13 the period of 1995-2014. Table A1 in the appendix provides an overview of the

countries included in our sample.

The reason we use a sample starting in 1995 is simply because of the availability of data. We were aiming to construct a dataset which would allow us to look at the relationship between public debt and economic growth with the model provided by Woo and Kumar (2015). Since many of the statistics were only available from 1995 onwards, we chose to start our sample there. In addition, we would like to analyze the financial crises of the past several years. Therefore, our latest data observations needed to be as recent as possible, which is why our sample runs until 2014. For the interested reader, we provide an overview of the descriptive statistics in the appendix in table A2.

4.1 Stylized facts

To start analyzing the data we will first solely look at the data on GDP per capita (resembling economic growth) and government (public) debt. By simply plotting the initial level of public debt against the growth in GDP per capita over the next three and five years respectively, we can see that the relationship is negative. This means that the higher the initial level of public debt is the lower subsequent economic growth will become. This holds for both the three- and the five-year time period. Figures 4.1 and 4.2 show the negative relationship between initial public debt and subsequent growth in GDP per capita. The five-year time period clearly shows a more divergent pattern, although the negative relationship remains.

(14)

14 well. It would be very interesting to see whether or not the accumulation of public

debt will decrease over the next several years, or if the public debt will continue to rise in the future.

Left: Figure 4.1: Initial government debt and growth of GDP per capita over the next three years (τ

= 2). Right: Figure 4.2: Initial government debt and growth of GDP per capita over the next five years (τ = 4).

Figure 4.3: Average Public Debt from 1995-2014

50 60 70 80 90 100 96 98 00 02 04 06 08 10 12 14 P u b li d D e b t (% o f G D P ) Years -10,000 -7,500 -5,000 -2,500 0 2,500 5,000 7,500 10,000 0 40 80 120 160 200

Initial level of government debt (% of GDP)

G ro w th o f G D P p e r c a p it a o v e r n e x t 3 y e a rs -8,000 -4,000 0 4,000 8,000 12,000 16,000 0 20 40 60 80 100 120 140 160

Initial level of government debt (% of GDP)

(15)

15 The next step we take is calculating all correlations of the variables used in

our regression analysis to check for potentially troublesome values. The values can be viewed in table 4.1. We find high correlations between Liquid Liabilities (Financial Market Depth) and Trade Openness. This is possibly due to increasing imports. An increase in imports one for one increases Trade Openness and increases the need for liquid liabilities. In addition, countries who engage in much trading have either a surplus leading to increased savings and investments or a deficit leading to an increased demand for (short-term) debt. Both of these lead to an increase in the liquid liabilities of the country. Since the high correlations may bias the explanatory power of our control variables we decide to take out Liquid Liabilities. We have run several regressions with both Liquid Liabilities and Trade Openness and several with only one of the two and have observed the explanatory power of Trade Openness to be higher. Therefore we will not use Liquid Liabilities in the estimation of our model.

(16)

16 Table 4.1: Correlation Matrix of explanatory variables and GDP per capita, 1995-2014.

GDP per Capita Public Debt Human Capital Trade Openness Log (1+Inflation) Liquid Liabilities Government Size Fiscal Deficit Terms of Trade Growth Banking Crisis Incidence GDP per Capita 1.0000 -0.5167 0.3723 0.9049 -0.1851 0.8116 -0.2356 0.4354 -0.0554 -0.1256 Public Debt -0.5167 1.0000 -0.0160 -0.4305 -0.1018 -0.4158 0.4700 -0.4287 0.0218 0.4444 Human Capital 0.3723 -0.0160 1.0000 0.1132 -0.2191 0.0128 0.3197 0.1978 0.0036 -0.0826 Trade Openness 0.9049 -0.4305 0.1132 1.0000 -0.0533 0.8244 -0.3572 0.3543 -0.0317 0.0033 Log(1+Inflation) -0.1851 -0.1018 -0.2191 -0.0533 1.0000 -0.0506 -0.2937 -0.0346 -0.1429 -0.1407 Liquid Liabilities 0.8116 -0.4158 0.0128 0.8244 -0.0506 1.0000 -0.3167 0.2314 0.0057 0.0241 Government Size -0.2356 0.4700 0.3197 -0.3572 -0.2937 -0.3167 1.0000 -0.2715 0.0581 0.1343 Fiscal Deficit 0.4354 -0.4287 0.1978 0.3543 -0.0346 0.2314 -0.2715 1.0000 0.0184 -0.2269

(17)

17 5. Results

In this chapter we will report our results on the analyses of the full sample, which can be viewed in 5.1, on the analyses of the GIIPS-countries separately, which can be viewed in 5.2, and on the analyses of the crisis-years in 5.3.

5.1 Full sample analysis

Five-year full sample analysis

The analysis for the five-year periods contains 16 different periods for 12 countries between 1999 and 2014. We start in 1999, since we have data available from 1995 onwards. Hence, the 1999 five-year growth in GDP is calculated from 1995-1999.

We present the main results of our regression using the five-year time period (τ = 4) in table 5.1. The OLS without period fixed effects is shown in column 1 and shows the results we expected to find. Initial debt has a significant negative effect on economic growth. A 10 per cent increase in initial public debt levels leads to a decrease in economic growth of 1.209 per cent over the five year period. Other researchers typically find a 0.1-0.2 per cent decrease in economic growth per year. Our results are larger, though this is most probably due to the inclusion of crisis-years in our sample. Almost all other researchers have not analyzed samples including years after 2008. Column 4 shows the OLS with period fixed effects to account for global factors that could potentially impact both economic growth and public debt levels. Quite logically the relationship is similar, though smaller that without period fixed effects. A 10 per cent increase in initial public debt levels in this case leads to a decrease in economic growth of 0.688 per cent per five-year period.

(18)

18 per cent increase in initial public debt levels leads to a slowdown in economic

growth of 0.983 per cent.

The Fixed Effects (FE) models show slightly different results compared to the other analyses. We report the FE results in columns 2 and 5 of table 5.1. First of all, the effect is insignificant at the 1%, 5%, and 10% level, and the coefficients are smaller than in the other cases. Although it is clear that higher initial public debt has a negative impact on economic growth in the following five years, the significance of the results varies along the estimation technique we employ. Fixed effects and period fixed effects have a large positive impact on the R2. Hence, the OLS model with period fixed effects performs best in explaining the relationship between initial public debt and growth.

Looking at the signs of the control variables, we encounter little surprises. Higher human capital leads to higher economic growth, except for in the FE analyses. Although human capital is not very significant in the FE analyses. More trade openness leads to higher economic growth as well, as does increased government size (government expenditure) when significant. Terms of trade growth is of the positive sign as well, though mostly insignificant. Initial inflation as well as the initial fiscal deficit have shown to be insignificant in all the analyses in table 5.1, and their sign changes along the different estimation techniques. As expected, the incidence of banking crises has a negative effect on economic growth, though it is mostly insignificant in these analyses.

Three-year full sample analysis

(19)

19 Table 5.1: Full sample panel regression five-year period (τ = 4) - Initial public debt and economic growth, 1999-2014

Dependent variable: Growth of real GDP per capita over five years

(1) (2) (3) (4) (5)

Explanatory variables OLS FE RR OLS FE

Initial real GDP per capita -0.0010*** -0.0027*** -0.0008*** -0.0006*** -0.0024***

(-6.3984) (-17.0645) (-5.9289) (-7.8363) (-13.2384)

Initial human capital 6.1957*** -11.0300 6.0619*** 5.5834*** -10.8363*

(5.9665) (-1.4837) (3.2733) (9.3241) (-1.7892)

Initial trade openness 0.1652*** 0.2816*** 0.1254*** 0.1133*** 0.3328***

(5.5648) (4.5417) (5.1270) (6.9228) (7.8746)

Initial inflation rate -0.1142 -0.0506 0.4557 -0.6285* -0.3086

(-0.1341) (-0.0972) (1.0041) (-1.7010) (-0.6725)

Initial government size 0.3652** -0.8316 0.3482** 0.0108 -0.2875

(2.2434) (-1.5850) (2.2997) (0.1049) (-1.0097)

Initial fiscal deficit 0.4468 -0.7236 0.1858 0.3542 -0.0297

(1.5556) (-1.3338) (1.0041) (1.2302) (-0.0763)

Terms of trade growth 1.0607 0.8499* 1.4605** 0.2634 0.1303

(1.1984) (1.7263) (2.3608) (0.5141) (0.2750)

Banking crisis incidence -7.0338 -3.8635 -9.6033* -1.5434 -2.9617

(-1.3487) (-0.7770) (-1.9306) (-0.4218) (-0.6930)

Initial public debt -0.1209*** -0.0741 -0.0983*** -0.0688*** -0.0448

(-4.8279) (-1.6215) (-3.5062) (-4.1147) (-0.9595)

Number of observations 188 188 188 188 188

R2 0.2462 0.7380 0.1506 0.7177 0.8526

Time fixed effects No No N/A Yes Yes

Notes

OLS is the Ordinary Least Squares Regression, FE is the Fixed Effects regression, and RR is the Robust Regression t-statistics are in parentheses. For the RR the z-statistics are in parentheses

(20)

20 A 10 per cent increase in initial public debt levels leads to a decrease in economic

growth of 0.514 (column 1), 0.428 (column 3), or 0.317 (column 4) per cent per three-year period. Coefficient signs for the analyses are as expected when significant, which is exactly the same as what we found for the five-year period. The FE regressions yield more insignificant results, though the sign for initial public debt is still negative.

Implications

We should be careful when interpreting our results, since we can only say that for three of our five estimations techniques our initial hypothesis is true. For the other two techniques, FE with and without period fixed effects, the hypothesis cannot be accepted nor rejected. Remember that the initial hypothesis we formulated was:

H1: A country’s public debt has a negative influence on its economic growth in the

following three to five years.

This hypothesis holds for three out of our five analyses. For the other two it is likely to hold, but the results of our regression analysis are insignificant for those estimation techniques. Our small sample size hinders us in increasing the robustness of our results. If we would have had access to a larger dataset, we could have tested the power of the FE technique on various samples to ensure its effectiveness. Furthermore, as we will see from further analysis the FE is negatively affected by small sample periods, which could be a possible explanation for its insignificant results in these analyses. Hence, it is relatively safe to say that the initial debt of a country does indeed have a negative impact on its subsequent growth in the following three to five years.

5.2 GIIPS-countries analysis

(21)

21 Table 5.2: Full sample panel regression three-year period (τ = 2) - Initial public debt and economic growth, 1997-2014

Dependent variable: Growth of real GDP per capita over three years

(1) (2) (3) (4) (5)

Initial real GDP per capita -0.0005*** -0.0014*** -0.0004*** -0.0003*** -0.0011***

(-6.0385) (-8.0855) (-5.8644) (-6.5693) (-8.7371)

Initial human capital 3.1736*** -1.4829 3.4820*** 2.8833*** -1.1489

(5.6825) (-0.4164) (3.3125) (7.3191) (-0.3828)

Initial trade openness 0.0766*** 0.1428*** 0.0628*** 0.0573*** 0.1601***

(5.5274) (5.5066) (4.7359) (6.7108) (7.1186)

Initial inflation rate -0.6881 -0.5985* -0.6974*** -0.4506 -0.1708

(-1.4123) (-1.6893) (-2.7645) (-1.4156) (-0.4956)

Initial government size 0.0301 -0.6197** -0.0378 -0.0481 -0.3265*

(0.2962) (-2.4337) (-0.4752) (-0.9099) (-1.8327)

Initial fiscal deficit 0.2165 -0.4089 0.1304 0.1818 -0.0907

(1.4100) (-1.4216) (1.3238) (1.2377) (-0.3892)

Terms of trade growth 0.6580 0.4481 0.9640*** 0.0954 -0.0690

(1.3148) (1.4352) (2.7735) (0.2554) (-0.2750)

Banking crisis incidence -3.9968* -4.1766* -4.0989*** -1.8997 -3.4925

(-1.8189) (-1.9013) (-2.6757) (-0.9568) (-1.5083)

Initial public debt -0.0514*** -0.0238 -0.0428*** -0.0317*** -0.0005

(-4.0913) (-0.8838) (-2.7417) (-3.4079) (-0.0226)

R2 0.2801 0.5976 0.1808 0.7221 0.8118

Notes

(22)

22

Five-year GIIPS analysis

As in the full sample analysis, the analysis for the five-year periods for the GIIPS countries contains 16 different periods between 1999 and 2014. The results for these analyses are reported in table 5.3.

Compared to the full sample analysis we find much larger results, which are of very high significance, except for the simple OLS model in column 1. Disregarding the OLS analysis (column 1), we find that a 10 per cent increase in initial public debt leads to decrease in subsequent five-year economic growth of 3.503 per cent for the FE model (column 2), 3.295 per cent for the OLS model with period fixed effects (column 3), and 2.611 per cent for the FE model with period fixed effects (column 4) respectively. These analyses indicate that for the five-year analysis growth in the GIIPS countries is much more affected by high initial public debt than growth in the other Eurozone countries. As for the coefficients of the control variables, they are mostly of the expected sign when significant. Only in the OLS analysis with period fixed effects, fiscal deficits have a positive influence on economic growth. We should not read too much into it, since it only appears in one out of four analyses, though it is possibly due to an increase in government spending which boosts growth, but is financed through a fiscal deficit. Looking at the significance of the results we obtain in conjunction with high R2 in column 2, 3, and 4, we can conclude that fixed effects have a significant positive impact on the performance of our model.

Three-year GIIPS analysis

As in the full sample analysis, the analysis for the five-year periods for the GIIPS countries contains 18 different periods between 1997 and 2014. The results for these analyses are reported in table 5.4.

(23)

23 Table 5.3: GIIPS-countries panel regression five-year period (τ = 4) - Initial public debt and economic growth, 1999-2014

(1) (2) (4) (5)

Explanatory variables OLS FE OLS FE

Initial real GDP per capita -0.0017*** -0.0037*** -0.0017*** -0.0027***

(-8.7273) (-17.9008) (-10.0827) (-10.3644)

Initial human capital 10.1810* -19.0514*** 24.8726*** 20.4467***

(1.6906) (-2.2193) (7.5372) (2.8410)

Initial trade openness 0.2687*** 0.1486 0.1511*** 0.1727**

(4.8367) (1.6362) (6.6943) (2.0967)

Initial inflation rate -0.0238 -1.3989*** -3.9431*** -2.2886***

(-0.0241) (-2.7726) (-12.9357) (-3.8209)

Initial government size 0.3341 -0.9311* 0.9814** 0.7472*

(0.3824) (-1.6843) (2.1170) (1.9770)

Initial fiscal deficit 0.8136 -0.8045 1.0366** 0.4377

(1.2419) (-1.5226) (2.3881) (1.1019)

Terms of trade growth 1.2630 -0.2440 -0.0117 -0.8685

(0.9175) (-0.5594) (-0.0067) (-0.3829)

Banking crisis incidence -7.4775 10.1904*** 4.4079 2.4452

(-0.7886) (3.2417) (1.5458) (0.7432)

Initial public debt 0.0234 -0.3503*** -0.3295*** -0.2611***

(0.1306) (-4.5815) (-4.5031) (-3.4885)

Number of observations 79 79 79 79

R2 0.4114 0.8566 0.9038 0.9411

Time fixed effects No No Yes Yes

Notes

(24)

24 Table 5.4: GIIPS-countries panel regression three-year period (τ = 2) - Initial public debt and economic growth, 1997-2014

(1) (2) (4) (5)

Explanatory variables OLS FE OLS FE

Initial real GDP per capita -0.0008*** -0.0017*** -0.0006*** -0.0012***

(-6.3482) (-6.6629) (-3.7371) (-7.4487)

Initial human capital 4.9217** -6.9800** 7.6866*** 10.3422***

(2.3394) (-2.2193) (2.7472) (2.9763)

Initial trade openness 0.1162*** 0.1135 0.0751*** 0.2027***

(3.8293) (1.3388) (6.1019) (3.7862)

Initial inflation rate -0.4085 -0.7901* -1.7687*** -1.1353*

(-0.6637) (-1.7105) (-3.5496) (-1.9989)

Initial government size -0.1827 -0.9886** -0.0390 -0.1199

(-0.3988) (-2.2115) (-0.1252) (-0.4048)

Initial fiscal deficit 0.2192 -0.6062 0.0725 -0.0621

(0.6335) (-1.6202) (0.2999) (-0.2379)

Terms of trade growth 0.6302 0.0182 0.8121 -0.2322

(0.7031) (0.0326) (0.8998) (-0.2105)

Banking crisis incidence -6.1358** 0.4363 -3.3451** -2.7791

(-2.1264) (0.1951) (-1.9587) (-1.3798)

Initial public debt 0.0060 -0.0661 -0.0642 -0.0631

(0.1008) (-1.0834) (-1.2579) (-1.4480)

Number of observations 89 89 89 89

R2 0.4319 0.7109 0.8567 0.8870

Time fixed effects No No Yes Yes

Notes

(25)

25

Implications

It is safe to say that the impact of initial public debt on subsequent economic growth in the GIIPS countries is negative for the five-year period and is still ambiguous for the three-year period. Remember that our initial hypothesis was:

H2: The negative relationship between public debt and economic growth is stronger in

GIIPS countries.

We have seen that this hypothesis is very much true for the five-year period. The effects are up to three times stronger in GIIPS countries. However, the effects of initial public debt do not become clear until after at least three years. We therefore cannot confirm the hypothesis for the short-term three-year period.

5.3 Crisis analysis

Five-year crisis analysis

The analysis for the five-year periods contains 3 different periods for 12 countries between 2012 and 2014. We start in 2012, since this calculates the economic growth over the period 2008-2012. The global financial crisis for almost all European countries started in 2008 and flowed into the European debt crisis since the end of 2009. Results for the five-year period crisis analysis are presented in table 5.5. Even more than in the full sample analysis, we find relatively ambiguous results.

(26)

26 Table 5.5: Crisis-years panel regression five-year period (τ = 4) - Initial public debt and economic growth, 2012-2014

(1) (2) (3) (4) (5)

Initial real GDP per capita -0.0004 -0.0016*** -0.0003 -0.0004* -0.0016***

(-1.6439) (-56.8280) (-1.3895) (-1.8057) (-33.7322)

Initial human capital 8.8066** -14.4987** 8.4310*** 8.7530** -14.0448**

(2.4789) (-2.3731) (2.9799) (2.3321) (-2.9745)

Initial trade openness 0.0621** 0.0078 0.0539 0.0565*** 0.0475

(2.3751) (0.3668) (1.5820) (3.4767) (1.1459)

Initial inflation rate -4.2766*** 0.7634*** -2.3779*** -4.7287*** 0.8011**

(-8.0162) (5.8074) (-3.6311) (-6.0782) (2.6587)

Initial government size 0.2062*** 0.4105 0.2577 0.2467*** 0.5864

(18.6554) (1.2029) (1.3776) (3.9862) (1.4681)

Initial fiscal deficit 0.8610** 0.0614 0.2064 0.8457** 0.2738

(2.3242) (0.1866) (1.1080) (2.3765) (0.6454)

Terms of trade growth -1.9504*** 0.7868*** -1.0484 -1.1051 0.5024***

(-6.0197) (5.4797) (-1.4290) (-0.8656) (11.0707)

Banking crisis incidence 8.8056* -0.4957 -1.0946 7.7572* 0.1767

(1.8313) (-0.5321) (-0.3517) (1.8881) (0.3037)

Initial public debt -0.1092*** 0.0892*** -0.0690* -0.1043** 0.0662***

(-2.9331) (5.0446) (-1.7736) (-2.3634) (6.2889)

R2 0.7072 0.9943 0.5022 0.7208 0.9952

Notes

(27)

27

Three-year crisis analysis

For the three-year period we conducted the analysis on the same time period as the five-year period. This means we were interested in the initial data from 2008-2010, meaning that our sample runs from 2010-2012. In addition, we ran the regressions for a time period of 2010-2014, though that analysis yielded merely insignificant results.

Results for the time period of 2010-2012 can be found in table 5.6, whereas the results for the time period of 2010-2014 can be found in the appendix in table A4. Similar to the five-year analysis we find quite ambiguous results. In this case only the OLS (column 1) and the OLS with period fixed effects (column 4) yield significant negative results. All other analyses yield insignificant results. Again, this ambiguity likely stems from the small sample size, which annuls the robustness of our results.

Implications

Our analysis of the crisis with the current model has not been successful in finding robust results. We have seen a steep incline in the accumulation of debt and decreasing or stagnating economic growth for the years in and after the financial crisis and European debt crisis. However, our regression models have failed to show a robust relationship between the two. Therefore we cannot accept our final hypothesis:

H3: Public debt has a more substantial negative impact on economic growth in years of

crisis.

(28)

28 Table 5.6: Crisis-years panel regression three-year period (τ = 2) - Initial public debt and economic growth, 2010-2012

(1) (2) (3) (4) (5)

Initial real GDP per capita -0.0002*** -0.0013** -0.0001 -0.0003*** -0.0012**

(-3.0912) (-2.5255) (-0.7754) (-3.6942) (-2.4793)

Initial human capital 4.2075 12.1514 2.3954 4.4732 11.2429

(1.2968) (0.8143) (1.0936) (1.4924) (0.8210)

Initial trade openness 0.0410*** 0.0319 0.0235 0.0425*** 0.1190*

(6.9955) (1.3306) (0.8900) (3.3395) (2.1422)

Initial inflation rate -2.6921*** 1.1070 -2.2745*** -2.7734*** 1.2938*

(-4.7721) (1.6304) (-4.4861) (-3.7173) (1.9648)

Initial government size 0.2936*** 2.0266 0.3343** 0.3159*** 2.2447*

(5.7909) (1.5048) (2.3080) (7.5706) (1.7796)

Initial fiscal deficit 0.5699** 1.1808 0.3642** 0.5516** 1.4359

(2.1870) (1.1726) (2.5257) (2.2070) (1.4338)

Terms of trade growth -0.3210 1.5627*** -0.3942 0.2334 1.3204***

(-0.7252) (3.6581) (-0.6940) (1.0453) (4.0859)

Banking crisis incidence 0.1758* -10.3246 -3.1972 -0.7699 -7.8004

(0.0352) (-1.6283) (-1.3268) (-0.1573) (-1.2084)

Initial public debt -0.0637* 0.0299 -0.0361 -0.0672* -0.1143

(-1.7108) (0.2890) (-1.1988) (-1.7818) (-0.7620)

R2 0.7088 0.9434 0.5622 0.7169 0.9531

Notes

(29)

29 5.4 Limitations and further research

One relatively obvious limitation of our study is the small sample size. As we focus on the Eurozone, we only include 12 countries of which most data is only available from 1990 onwards and for some variables even from 1995 onwards. It would be interesting to see what results will be obtained from a larger dataset, which would also include the recent crises. It is quite likely that our results are slightly bloated due to the dataset containing a relatively large number of years in crisis. In addition, the Fixed Effects model we have employed did not always yield the expected results, which might very well be due to the limited years in our sample as well as the aforementioned large number of years in crisis. We looked for signs of autocorrelation and heteroscedasticity for individual countries instead of for the sample as a whole, due to estimation limitations. This could potentially slightly bias the standard errors in our sample. In the future this problem might be circumvented through increased capabilities of statistical software. Our research has used five different estimation techniques, but future researchers might want to look at various other estimation techniques to increase the robustness of estimation results even further. Furthermore, we were not able to observe the long-term effects of the recent crises. If we would jump several years into the future, we would have much more data to analyze and we could examine the medium- to long-term effects of the crises on economic growth. In addition, it would interest us to see if the large accumulation of debt will decrease in the future, and what would trigger it.

(30)

30 6. Conclusion

(31)

31 References

Aizenman, J., Kletzer, K., Pinto, B. 2007. Economic growth with constraints on tax revenues and public debt: implications for ﬁscal policy and cross-country diﬀerences. NBER Working Paper no. 12750.

Aspergis, N., Cooray, A. 2016. Debt uncertainty and economic growth: evidence from five highly indebted Eurozone countries. Applied Economics Letters, 23(3), 171-174.

Baldacci, E., Hillman, A., Kojo, N. 2004. Growth, governance and ﬁscal policy transmission channels in low-income countries. European Journal of Political Economy, 20, 517–49.

Barro, R., Lee, J. 2011. A new data set of educational attainment in the world, 1950–2010. NBER Working Paper no. 15902.

Baum, A., Checherita-Westphal, C., Rother, P. 2013. Debt and growth: new evidence for the euro area. Journal of International Money and Finance, 32, 809–21.

Caner, M., Grennes, T., Koehler-Geib, F. 2010. Finding the tipping point—when sovereign debt turns bad. World Bank Working Paper no. 5391.

Cecchetti, S., Mohanty, M., Zampolli, F. 2012. Achieving growth amid ﬁscal imbalances: the real eﬀects of debt. In Achieving Maximum Long-Run Growth, A Symposium Sponsored by The Federal Reserve Bank of Kansas City.

Checherita, C., and Rother, P. 2012. The impact of high and growing government debt on economic growth: an empirical investigation for the Euro area. European Economic Review, 56(7), 1392–405.

Elmendorf, D., and Mankiw, N. 1999. Government debt. In J. B. Taylor and M. Woodford (eds), Handbook of Macroeconomics. North-Holland, Amsterdam, pp. 1615-1669.

Feenstra, R., Inklaar, R., Timmer, M. 2015. The next generation of the Penn World Table. American Economic Review, 105(10), 3150-3182.

Fischer, S. 1993. The role of macroeconomic factors in growth. Journal of Monetary Economics, 32(3), 485– 512.

Popov, A., van Horen, N. 2015. Exporting sovereign stress: evidence from syndicated bank lending during the euro area sovereign debt crisis. Review of Finance, 19, 1825-1866.

(32)

32 Reinhart, C., Reinhart, V. 2015. Financial crises, development, and growth: a long-term

perspective. The World Bank Economic Review, 1-24.

Reinhart, C., Reinhart, V., Rogoff, K. 2012. Public debt overhangs: advanced economy episodes since 1800. Journal of Economic Perspectives, 26(3), 69-86.

Reinhart, C., Reinhart, V., Rogoff, K. 2015. Dealing with debt. Journal of International Economics, 96, 43-55.

Reinhart, C., Rogoff, K. 2009. This Time Is Different: Eight Centuries of Financial Folly. Princeton University Press. New Jersey, Princeton.

Reinhart, C., Rogoff, K. 2010. Growth in a time of debt. American Economic Review, 100(2), 573-578.

(33)

33 Appendix

Country List

All countries that started using the Euro from the start are included in our sample. Table A1: Country list

Descriptive Statistics

Table A2: Descriptive statistics of key variables

Mean Median Maximum Minimum Std. Dev. Observations

GDP per Capita 31932.80 30570.44 83710.84 13614.79 13967.43 233 Public Debt 70.54 65.04 178.40 6.08 33.00 233 Human Capital 41.08 43.20 68.10 10.70 13.03 233 Trade Openness 100.77 70.42 374.15 37.11 69.88 233 Log(1+Inflation) 2.16 2.11 8.45 -1.71 1.31 233 Liquid Liabilities 108.40 90.45 399.11 35.18 69.49 233 Government Size 47.25 47.30 65.57 30.90 5.78 233 Fiscal Deficit -2.94 -3.01 6.86 -32.30 4.17 233

Terms of Trade Growth -0.05 0.02 2.90 -4.24 0.82 233

Banking Crisis Incidence 0.06 0.00 1.00 0.00 0.25 233

(34)

34 Table A3: Heteroscedasticity and autocorrelation in individual countries, 1995-2014

Null hypothesis of homoscedasticity. Null hypothesis of no autocorrelation

(1) (2) (3) (4) (5)

Countries F-statistic Obs.*R2 Scaled explained SS F-statistic Obs.*R2

Austria 0.5557 6.2145 1.4562 4.9609 13.8988*** (0.7860) (0.6232) (0.9934) (0.1096) (0.0076) Belgium 4.2193** 13.2519 1.9139 1.4818 10.6232** (0.0368) (0.1035) (0.9835) (0.3888) (0.0311) Finland 1.4073 9.8658 0.8638 1.6518 11.0037** (0.3328) (0.2746) (0.9990) (0.3545) (0.0265) France 0.8260 7.6209 1.3994 1.8908 10.8363** (0.6009) (0.4714) (0.9943) (0.2503) (0.0285) Germany 0.5644 6.4409 1.0974 2.6322 12.6055** (0.7777) (0.5980) (0.9976) (0.2938) (0.0134) Greece 3.4048* 13.3802 1.5100 3.3155 13.9033*** (0.0748) (0.1461) (0.9971) (0.2449) (0.0076) Ireland 1.9450 11.6675 0.7369 1.5831 12.9542** (0.2400) (0.2327) (0.9998) (0.5288) (0.0115) Italy 1.6286 10.4081 1.5796 0.7013 7.7316 (0.2671) (0.2375) (0.9913) (0.6408) (0.1019) Luxembourg 2.9614 11.9688 0.8384 1.7465 6.9926** (0.1015) (0.1526) (0.9991) (0.2850) (0.0303) Netherlands 1.4420 9.8676 0.5804 3.6757 13.2039** (0.3372) (0.2744) (0.9998) (0.2251) (0.0103) Portugal 0.1916 3.5719 0.4827 0.5378 8.2912* (0.9860) (0.9373) (1.0000) (0.7315) (0.0815) Spain 0.5718 6.3233 0.6600 0.9682 9.0159* (0.7749) (0.6111) (0.9996) (0.5322) (0.0607) Notes

Column (1) - (3) test for heteroscedasticity through the Breusch-Pagan-Godfrey test, column (4) and (5) test for autocorrelation through the Breusch-Godfrey test. For autocorrelation testing four lags are used

(35)

35 Table A4: Crisis-years panel regression three-year period (τ = 2) - Initial public debt and economic growth, 2010-2014

(1) (2) (3) (4) (5)

Initial real GDP per capita -0.0002** -0.0018*** -0.0001 -0.0002 -0.0017***

(-2.0140) (-5.1388) (-0.8862) (-1.5345) (-4.6581)

Initial human capital 2.5773 0.6093 1.6730 2.4655 -1.3811

(0.9421) (0.1087) (0.8034) (0.8414) (-0.2477)

Initial trade openness 0.0441*** 0.0986* 0.0296 0.0375*** 0.1384***

(3.9912) (1.7526) (1.3593) (2.9179) (6.9616)

Initial inflation rate -2.8636*** -0.9922 -2.4822*** -2.8919*** -0.7656

(-12.8847) (-1.4693) (-5.6761) (-6.3927) (-0.9723)

Initial government size 0.1493** -0.2332 0.1555 0.1135 0.1650

(2.2294) (-0.9683) (1.4682) (1.2682) (0.3535)

Initial fiscal deficit 0.6017*** -0.0171 0.4330*** 0.5181** 0.3294

(3.5859) (-0.1341) (3.3460) (2.6187) (0.8304)

Terms of trade growth -0.7334** 0.3613 -0.6902 -0.6664 -0.3250

(-2.4873) (0.9570) (-1.2816) (-1.1989) (-0.6054)

Banking crisis incidence 2.2681 -1.8644 0.2016 1.1692 -1.9421

(0.8852) (-1.3375) (0.1357) (0.4186) (-1.2947)

Initial public debt -0.0408 0.0404 -0.0192 -0.0447 0.0753

(-1.0987) (1.5220) (-0.7058) (-1.5345) (1.5983)

R2 0.5889 0.8614 0.3869 0.6349 0.8798

Notes