The effect on economic growth when public debt is restrictive

(1)

The effect on economic growth when

public debt is restrictive

Thesis Supervisor: Damiaan Chen

Katherinne Ortiz Espinoza 5908590

Bachelor Thesis July 24, 2014

(2)

Abstract

Does the effect of certain channels on the economic growth differ and does it become negative when the public debt level of a country exceeds the Maastricht treaty debt limit of 60%? The encompassing channels are: private savings, private investments, total factor productivity, capital stock and inflation. This research statistically examines this question for the countries Germany, France, Italy and Belgium, in the period of 1970-2013.

Overall, an absence of a structural break caused by exceeding the Maastricht treaty debt limit of 60% was found. However, it appeared to exist in France and Italy in 2003; in this year the public debt-to-GDP ratio was higher than the debt limit of 60%. This finding implies the existence of a difference in effect of the channels on the economic growth. Since Italy exceeded the debt limit of 60% earlier in 1982, there may be suggested that other factors provoke a structural break in a country’s debt level.

(3)

A turning point of the effect of the public debt on the economic growth at a debt-to-GDP ratio higher than 90% was found by Reinhart and Rogoff in 2010. During the years afterwards, the possible negative relationship between these constructs was more deeply investigated. Such a turning point i.e. debt threshold was also confirmed by the research of Afonso and Jalles (2013). In that research statistical evidence was found for a debt threshold at a debt to GDP ratio of 58%. From these results it may be concluded that public debt has a positive effect on economic growth, up to the point where a certain debt-GDP-ratio is reached. Beyond this point public debt is related to lower economic performance.

In 1992 members of the European Union signed the Maastricht treaty. A debt limit of 60% of GDP and annual deficit limit of 3% of GDP was set as a requirement to take part in the new single currency Euro. It was agreed upon that countries with higher debt levels could also introduce the Euro provided that the development of their public debt was structurally decreasing towards the 60% limit, the so-called convergence. In the past years many Euro countries exceeded the Maastricht treaty debt limit of 60% (Checherita-Westphal & Rother, 2012). It is essential to gain more insight into the consequences this may have. Therefore, understanding the effect of public debt levels above 60% of GDP on economic growth is needed in the interest of policy makers, the private sector and future generations.

Economical literature and empirical research mentioned different channels that are causing the possible negative relationship between the high public debt and the economic growth. First, inflation can be seen as a negative influencer on economic growth (Elmerdorf & Mankiw, 1998). Second, Schclarek (2004) found statistical evidence on the negative effect of private capital accumulation. Third and fourth, Kumar and Woo (2010) suggested two different channels that could have an important influence on this negative relationship: higher tax rate and long-term interest rates. Fifth, Checherita-Westphal and Rother (2013) found a significant relationship between total factor productivity,

(5)

private savings and public savings, negatively affecting the relationship between the high public debt and the economic growth. These are just a few examples of the possible influencing channels. It can be concluded that there is no consensus on the channels that cause the negative effect of high public debt on growth.

The purpose of this research was to statistically test whether the effect of certain channels on economic growth differ and even become negative when public debt levels are higher than the Maastricht treaty debt limit of 60% for specific Euro countries. Germany, France and Italy are accounted as the three largest economies of the euro area in terms of nominal GDP. In the past years Belgium is known to carry a big debt load and was added as control country. Therefrom, the research question becomes:

‘Does the effect of certain channels on economic growth differ and become negative when public debt levels are higher than the Maastricht treaty

debt limit of 60% for Germany, Italy, France and Belgium?’

To answer this question, chapter 2 will review the existing theoretical models, a discussion of the scientific empirical research and methodological issues will be addressed. Chapter 3 will describe the data and methodology used to conduct this research. Chapter 4 presents the descriptive statistics and results of statistical research. Lastly, chapter 5 will conclude the findings of this research.

(6)

2. Literature review

This chapter will summarize the economic and the empirical literature on the relationship between public debt and economic growth. First, the theoretical models will be discussed. In the second section of this chapter the findings of empirical research will be presented. Followed by the methodological issues that arise when doing empirical research. The last section will highlight the gaps in the literature.

2.1 The theoretical models

In economic literature different views on the relationship between public debt and economic growth are proposed. Some economist argue the positive impact of public debt, while others belief it mainly has a negative effect on economic growth.

Robert Barro (1979) suggests a negative relationship between public debt and growth. An increase in public debt eventually will be financed with an increase in taxes. This leads to a decrease of disposable income, in turn lowering potential output (Reinhart & Rogoff, 2010).

The conventional view states that government debt has a positive effect in the short run, but a negative effect in the long run.In the short run the economy is Keynesian. The issuance of government debt increases aggregate demand and economic growth through the public expenditure multiplier. The crowding-in effect arises. Investors perceive the future as positive, because of the induced government investments. This leads to an increase in private investments. Contrary, in the long run the economy is classical. Public debt has a negative effect on economic growth because capital is crowd out. In the long run wages and prices will eventually adjust to the increased liquidity levels, giving rise to inflation. An increase in inflation raises uncertainty about future inflation. This leads to a decrease of investments, causing a decline in output (Elmerdorf & Mankiw, 1998).

The Ricardian Equivalence proposes a neutral effect of government deficit and public debt on economic growth. It assumes that consumers are forward looking when making their consumption decisions. The government deficits can either be financed

(7)

through taxes or issuing of bonds. Bonds eventually must be repaid by raising taxes in the future. Hence, public debt will have no effect on consumption, private investments and growth in the long run (Pan & Wang, 2010).

Reinhart and Rogoff (2010) state it is essential whether debt is built up in wartime or peacetime. Debt build up during wartime causes fewer problems since it comes to a natural close in peacetime. Peacetime debt explosion likely leads to a decline in economic development. It reflects an unstable political economy and tends to last for a longer period of time.

According to Nautet and van Meensel (2011) the criterion of intergenerational neutrality should hold when discussing public debt. This criterion is violated if public debt increases, because more wealth from future generations is transferred to current generations. The criterion states that the net costs each generation makes should be equal. In order to equalize this great difference the government should invest in for example education and infrastructure so future generations will also benefit. The current generation should also partly pay for the future costs of pensions and health care due to population ageing.

Mussolini et al. (2013) propose a theoretical model where the positive effect of government expenditure on economic growth is smaller when the debt-to GDP ratio is higher. On the one hand government expenditure has a positive effect on economic growth because it leads to an increase in production. On the other hand a negative effect arises if these expenditures are financed with an increase in tax or public debt, because this leads to a decline in savings. Now less capital is available for investments. This tends to push up interest rates, which in turn increases the government debt burden even more. However, if the conditions of a healthy fiscal situation and indebtedness are met, an increase in productive expenditures will increase economic growth.

2.2 Empirical studies

The different theories mentioned in the first section provide contrary predictions on the effect of public debt on economic growth. In this section empirical research on this topic will be discussed. The empirical literature for countries in the euro area is limited. Abbas and Christensen (2007) suggested three main reasons for this. First of all, reliable and

(8)

comparable data on public debt are missing for many countries. Secondly, there is no general agreement on whether public debt is an endogenous or exogenous variable. Finally, until recently public debt was not seen as a problem for advanced economies. There is also a lack of systematic evidence. Some find evidence for a negative or non-linear relationship, whereas other say there is no relationship between high public debt and economic growth (Kumar & Woo, 2010; Checherita-Westphal & Rother, 2012).

2.2.1 Negative or non-linear effects of public debt

A turning point of the effect of the public debt on the economic growth at a debt-to-GDP ratio higher than 90% was for the first time found by Reinhart and Rogoff in 2010 (Reinhart & Rogoff, 2010). The data covered 70 countries in Africa, Asia, Europe, Latin America and North America over a 200 years period. However, the method used is simple correlation statistics. Also no other growth variables are taken into account.

Kumar and Woo (2010) also found empirical evidence for an inverse relationship on the impact of high public debt on growth. Above 90 % of GDP evidence is found for non-linearity. The sample consisted of 38 advanced and emerging countries over the period 1970-2007. Two main approaches were used in their research, growth regression and growth accounting.

The results of research done by Cecchetti et al. (2011) showed a clear link between high debt and low economic growth. A threshold at a debt-to-GDP ratio of 86% is found. 18 OECD countries were examined for the period 1980-2010. The methodology employed is the non-linear panel method.

Checherita-Westphal and Rother (2012) were the first who focussed only on Euro area countries. Evidence is found for a concave relationship between public debt and economic growth. At debt levels above 90-100% of GDP growth rates are lower.

Research was done on twelve old Euro countries for the period 1970-2008.

The results of Baum et al. (2013) showed that at debt-to-GDP ratios above 67% the initially positive impact of public debt almost decreases to zero. For ratios above 95% the effect even becomes negative. Their research contributed in the following way. Firstly, the use of the dynamic threshold panel approach ensured consistent results. Secondly, this research focussed more on the short-run instead of long run impact.

(9)

Finally, the special attention to EMU data provided the opportunity to make policy statements.

Aristovnik et al. (2011) found support for a non-linear impact of debt on annual GDP per capita. A sample of 25 EU countries was used, whereby a distinction was made between ‘old’ members and ‘new’ members. The period covered for ‘old’ members is 1980-2010 and for new members 1995-2010.

Afonso and Jalles (2013) found statistical evidence for a debt threshold at a debt-to-GDP ratio of 58% for euro area countries. The sample consisted of 155 developed and developing countries over a period of 38 years starting from 1970 until 2008.

2.2.3 No significant effect of public debt

Schclarek (2004) found no significant relationship between public debt and growth for emerging and advanced countries. This result is in contrast with the empirical researches mentioned above. The sample consisted of 24 countries over the period of 1970-2000. After replicating the research done by Reinhart and Rogoff (2010), Ash et al. (2013) conclude there is no relationship between high government debt and economic growth. They found several methodological issues: coding errors, selective exclusion of data and different weighting of the data.

Panizza and Presbitero (2013) also found a weak relationship. It was stated that the negative correlation as proposed by Reinhart and Rogoff doesn’t imply causation. A decline in economic growth could also lead to an increase in public debt.

Pescatori et al. (2014) found more evidence for the importance of the debt trajectory instead for the actual debt level. Countries with high debt-to-GDP ratios that deleveraged public debt grew as fast as countries with low debt-to-GDP ratios. The causation from economic growth to debt was weakened by analysing data for a longer time span. The dataset covered the period from 1875 until 2010 for 19 advanced economies.

(10)

2.3 Gaps in empirical research

After going through the literature an important gap needs to be highlighted. There is no clear consensus on what channels cause the possible negative relationship between high public debt and economic growth. In the economic literature different channels are mentioned.

Kumar and Woo (2010) suggested two different channels that have an important influence on this negative relationship: higher tax rate and long-term interest rates. A higher tax rate in order to service an increase in public debt leads to a decrease in disposable income and savings. Savings are negatively correlated with short-term interest rates. A decrease in savings causes an increase in long-term interest rates. High long-term interest rates negatively affect economic growth. The neo-classical approach is followed that high long-term interest rates transmit a decline in public investments, private investments, consumption and labour supply causing a decrease in growth

Another channel mentioned is inflation. There is a risk that governments intentionally let inflation rise to lower the debt burden which also has a negative effect on growth. An increase in inflation raises uncertainty about future inflation. This leads to a decrease of investments, causing a decline in output (Elmerdorf & Mankiw, 1998).

Egert (2012), Cechetti (2010), Nautet and van Meensel (2011) agree that more pressure is put on economic activity if higher public debt leads to more sovereign risk. This results in a rise of government bond yields at medium and long run maturities. Besides more sovereign risk, uncertainty about the future arises. This in turn affects economic development. To sum up, the channels mentioned in economic literature are interest rates, taxes, savings, investments, capital accumulation, inflation, uncertainty and long-term sovereign yield.

For euro area countries here is not much empirical evidence on these variables. Schclarek (2004) found evidence on the channel of private capital accumulation. The sample consisted of 59 developing countries and 24 industrial countries. Kumar and Woo (2010) found statistical evidence on investment for advanced countries. Checherita-Westphal and Rother (2013) focused only on transmission channels for countries within the euro area. A significant relationship is found for total factor productivity, private savings and public savings. Other variables examined were private investments, public

(11)

investments, sovereign long-term real rates and sovereign long-term nominal rates. Chen (2014) identified three channels for the United Kingdom: national savings, real investments and capital stock. The impact of long-term interest rates was also examined

but not found to be significant.

Baum et al. (2013) suggested that more research has to be done on this topic. At high public debt levels multiple channels could have a negative impact on economic growth. Also Aristovnik et al. (2014) emphasized that research on the transmission mechanism should be extended.

The purpose of this research is to statistically test whether the effect of certain channels on economic growth differs and even becomes negative when public debt levels are higher than the Maastricht treaty debt limit of 60% for the specific Euro countries. Therefrom, the research question becomes: ‘Does the effect of certain channels on economic growth differ and become negative when public debt levels are higher than the Maastricht treaty debt limit of 60% for Germany, Italy, France and Belgium?’

(12)

3. Data and methodology

The purpose of this thesis is to find statistical evidence for a difference in effect of certain channels on economic growth when the public debt-to-GDP ratios exceed the Maastricht treaty debt limit of 60%. First, the measurement of economic growth and public debt will be clarified. Followed by a description of the data and the models used. The last part will cover the methodology and gives a summary of the statistical tests that will be performed.

3.1 Transmission channels of public debt on economic growth

The following section will define economic growth, public debt and transmission channels. In this research economic growth was used as dependent variable. Economic growth is the difference of the annual gross domestic product (GDP) of a country in year

t compared to the previous year, divided by GDP of the previous year.

Economic growth = !"!!!!"!!!!

!"!!!! × 100% (3.1)

GDPt is denoted as the nominal GDP in year t and GDPt-1 the nominal GDP of the

previous year. Secondly, total public debt consists of domestic and foreign liabilities of a government. It is measured as a ratio to GDP (Kumar & Woo, 2010).

Public debt-to-GDP ratio= !"#$!%& !"#$%& !"#$%"&' !"#$!!"#$!%& !"#$%& !"!"#$%& !"#$_{!"#$%&' !"#} (3.2)

Lastly, the definition of transmission channels is given. These are the processes through which public debt affects economic growth. Literature suggested that multiple channels simultaneously cause a negative effect on growth (Checherita-Westphal & Rother, 2010; Nautet & Van Meensel, 2011).

(13)

3.2 Data description

This research focused on a sample of four Euro area countries: Germany, France, Italy and Belgium, in the period of 1970-2013. Time series data was used and consisted of 44 observations. For some variables there are less observations due to the unavailability of data. The first three countries were selected based on the size of their economies. These countries accounted the three largest economies of the Euro area in terms of nominal GDP. Belgium carried a big debt load over the period 1970-2013 and was added as control country. To be able to compare multiple countries GDP per capita growth (gdp_cap_g) was used. Data on GDP per capita growth was found in the Wold databank. Data on the debt-to-GDP ratio is found in the Reinhart and Rogoff database.

The main channels that were investigated are: private savings, private investments, total factory productivity, inflation and net capital stock. A selection was made out of the channels discussed in section 2.2.4. First, private savings are the parts of the income that is saved by companies and household. Second, when capital assets are purchased by anyone else than the government to generate income, it is called private investments. Third, total factor productivity (TFP) is the part of output that cannot be explained by the amount of capital and labour used in production. It measures how productive and efficient output is generated. Fourth, inflation is an indicator of the increase in general prices. Lastly, net capital stock is the value of fixed assets in use minus depreciation.

Data on yearly inflation is obtained from the database of the OECD. The data for the remaining variables was collected from the European commission AMECO database following the example of Checherita and Rother (2012). Data on private savings, private investment and net capital stock was measured in units of Euros. To avoid non-stationary, the yearly growth rates for the variables debt-to-GDP ratio (debt_gdp_g), private savings (sav_priv_g), private investments (inv_priv_g) and net capital stock (net_cap_stock_g) were computed by calculating:

!!"!!!"!!

!!"!! ×100% (3.3)

However, total factor productivity (tfp_g) was measured as an index (2005=100). It was calculated by dividing the weighted sum of total output by the weighted sum of all costs of capital and labour used. Total factor productivity growth occurs when the difference

(14)

between the weighted sum of total output and the weighted sum of all costs increases. Checherita-Westphal and Rother (2012) calculated the total factor productivity growth rate in the following way:

TFP growth rate = !"# !"#$% !"#$!!!"# !"#$% !"#$!!!

!"# !"#$% !"#$!!! × 100% (3.4)

The Euro crisis (2007-2013) could have significant effects on growth. To separate the influence of the crisis a dummy variable (d_cri) was added. Where the dummy (d_cri) equals 0 for years before 2007 and 1 otherwise. Besides this dummy another dummy variable was added in the German model. The dummy (d_91) will account for the effect of the reunification of East and West Germany on economic growth. The AMECO database split data on Germany into two periods. The first period (1970-1990) covered the timespan before the reunification of East and West Germany. The data only included numbers of West-Germany. The second period (1991-2013) covered data of the total economy of Germany. The dummy (d_91) equals 0 in the period before 1991 and equals 1 otherwise.

3.3 Research methodology

This study first identified the channels that caused a significant effect on economic growth by performing Ordinary Least Squares (OLS) regression. In order to perform OLS five assumptions were made.

#1. Linear in parameters 𝑦! = 𝑥!𝛽 + 𝜀! (3.5)

#2. Random sample of n observation 𝑦_!: 𝑥_!, 𝑥_!, … 𝑥_! (3.6)

#3. Zero conditional mean 𝐸 𝜀! 𝑋) = 0 (3.7)

#4. No perfect collinearity 𝑣𝑎𝑟 𝜀! 𝑈 = 𝜎! (3.8)

(15)

3.3.1 Regression model

The following equations presented the model that was estimated for the sample countries.

France, Italy and Belgium

gdp_cap_g = β0 + β1*sav_priv_g + β2 *inv_priv_g+ β3 * tfp_g + (3.10)

β 4 * net_cap_stock_g + β5* infl + β6 * D_cri +εi

Germany

gdp_cap_g = β0 + β1*sav_priv_g + β2 *inv_priv_g+ β3 * tfp_g + (3.11)

β4 * net_cap_stock_g + β5* infl+ β6 * D_cri + β7 * D_91 + εi

The variable GDP per capita growth (gdp_cap_g) was the dependent variable. The independent variables used were the channels selected from the economic literature and empirical research. As mentioned before, the dummy variable (d_91) was added to the German model to capture the categorical effect of the reunification of West and East Germany. The dummy variable (d_cri) tried to capture the additional effect on economic growth caused by the euro crisis of the period 2007-2013. This simple regression model was proposed to test whether the explanatory variables had a significant effect on economic growth.

3.3.2 Chow test

Revising the empirical literature it became clear that the high public debt-to-GDP ratio might have a negative effect on economic growth. Public debt had a positive effect on economic growth until it reached a certain debt-to-GDP ratio. Thereafter public debt was related to lower economic performance. This implied that certain channels had a negative effect on economic growth when public debt is high.

The Chow test was performed to see whether a significant difference of the effect of the independent variables existed beyond the GDP level of 60%. The debt-to-GDP ratio of 60% is the debt limit as set by the European Union in the Maastricht treaty. Members of the European Union signed the Maastricht treaty in 1992. Besides the debt limit of 60% it states that countries are allowed to have a deficit of 3% of GDP.

(16)

The formula of the Chow test is defined in the following way:

!!!!!!!!! !!!!!_! !!!!!!!!!

!!!!!!!! (3.12)

𝐹_(!,!_!_!!_!_!!!) =

SSEp = Sum of squared error term for the pooled model

SSE1 = Sum of squared error term for respectively period 1

SSE2 = Sum of squared error term for respectively period 2

k = Number of estimated parameters N1 = Number of observation for period 1

N2 = Number of observation for period 2

In figure 1.a a simple regression of the form yi= β0 + β1xi + εi without structural break is illustrated. When statistical evidence is found for a structural break, the analysis is conducted with two separate models as shown in figure 1.b. The slope β0 and the coefficient β1 are different in the period before and after the break. This implies a difference in effect of the independent variable xi on yi.

Figure 1 Illustration of a structural break

In this research first the data of period 1 and period 2 were pooled. The regression model is shown in equation 3.12. By performing OLS regression the sum of squared error term for the pooled model (SSE ) was found.

(17)

gdp_cap_g i= β0 + β1*sav_priv_g + β2 *inv_priv_g+ β3 * tfp_g + (3.13)

β4 * net_cap_stock_g + β5* infl + εi)

Secondly, a model for the first period was conducted. The first period consisted of the data of the variables of the years where debt-to-GDP ratio was lower than 60%. The model for period 1 was defined in the following way:

gdp_cap_g i= d_1* (β0 + β1*sav_priv_g + β2 *inv_priv_g+ β3 * tfp_g + (3.14)

The dummy variable (d_1) equals 0 when debt-to-GDP ratio is higher than 60% and 1 otherwise. By performing OLS regression the sum of squared error term for period 1 (SSE1) was found.

Thirdly, the model for period 2 covered the data where debt-to-GDP ratios were higher than 60%. The model for period 2 was expressed as:

gdp_cap_g = d_2* (β0 + β1*sav_priv_g + β2 *inv_priv_g+ β3 * tfp_g + (3.15)

The dummy variable (d_2) equals 0 when debt-to-GDP ratio is beneath the Maastricht treaty debt limit of 60%. Also here OLS regression was used to find the sum of squared error for period 2 (SSE2).

Lastly, the F-statistic was calculated using the Chow test formula. The critical values at 10%, 5% and 1% significance level for each individual country were extracted from the table of the F-distribution. A structural break was found when the F-statistic exceeded the critical values. It implies a significant difference in effect of the independent variables on economic growth between period 1 and period 2.

(18)

4. Results

Economic literature and empirical research identified multiple channels by which high public debt has a possible negative impact on the economic growth. In this chapter the results on the investigated channels private savings, private investments, total factor productivity, capital stock and inflation, are presented. The first section provides the descriptive statistics of the main variables. Secondly, the testing results of the independent variables will be shown. Lastly, the results of the analysis whether there exists a structural break beyond the debt-to-GDP ratio of 60% will be presented.

4.1 Descriptive statistics

Table 1 shows that Germany has the highest GDP over the period 1970-2013, on average. Belgium, on the other hand, has the smallest economy. When looking at the public debt-to-GDP ratio, France has the lowest average ratio of 43.9% and Italy the highest ratio of 87.3%.

Table 1: Descriptive statistics GDP (MRD/EUR) and debt-to-GDP ratio %

Germany Franee Italy Belgium

Variables obs mean Std.

dev obs mean Std.dev obs mean Std. dev obs mean Std.dev

GDP 44 1748.5 490.4 44 1310 342.2 44 1129.9 266.3 44 188.4 _107.7

Debt/GDP 44 47.1 19.2 44 43.9 22.5 44 87.3 29.3 44 85.5 26.1

The difference between Germany and France, and Italy and Belgium in the evolvement of the debt-to-GDP ratio can be seen in figure 2, in which the left y-axis belongs to the blue line and depicts the debt-to-GDP ratios in the period 1970-2013. On the right y-axis the percentage GDP growth is shown in the green bars for the same period. Germany and France kept their deb-to GDP ratio close to the Maastricht treaty debt limit of 60% until in 2008 it started to increase to ratios far above 60%. On the other hand, the debt-to-GDP ratio of Italy and Belgium increased steadily beyond the 60% limit.

Germany exceeded the Maastricht treaty debt limit of 60% in 1998. The years afterwards, the debt-to-GDP ratio stayed close to 60%. However, debt-to-GDP ratio increased rapidly from 65% to 78% in the period 2008-2013.

(19)

a. Germany b. France

c. Italy d. Belgium

Figure 2 GDP growth and public debt as % of GDP (1970-2013)

Figure 2.b shows that France did not exceed the 60% debt level until 2003. In the period

of 2003-2007 debt-to-GDP growth rates slowed down, pushing the debt-to-GDP ratio close to 60% again. Since 2008 an excessive grow from 63.8% till a 90.6% debt-to-GDP ratio is observed.

A different debt trajectory can be seen for Italy and Belgium. Over the period 1970-2013 Italy and Belgium continuously carried a big debt load, reaching respectively debt-to-GDP values of 120% and 118%.

The current debt limit of 60% was already violated by Italy in 1982. Thereafter, the debt-to-GDP ratio kept steadily increasing until 2000. The following period of deleveraging was between 2000-2008 and in 2004 it even reached a point below 100%. Since 2008 the debt-to-GDP ratio started to increase again.

Not only Italy, but also Belgium exceeded the 60% in the early 80’s and kept on increasing until 1994. From 1994 until 2007 the debt-to-GDP ratio was brought back to

(20)

until 85%. Figure 2d shows that the public debt-to-GDP ratio started to increase again in 2008. However, not as much as in Germany, France and Italy.

As mentioned above, in 2007-2013 the debt-to-GDP ratio rapidly increased in the four sample countries. The financial and economic crises were the main reason to this sharp increase. First, the U.S. market suffered great losses. Then the banks in Europe were exposed to these losses because of the high integration between the markets. Recapitalizations were necessary to prevent the collapse of the banking sector. Banking regulation was mainly a national responsibility and as a consequence the bail-outs were largely publicly funded, thereby raising public debt (Lane, 2012). Furthermore, governments tried to stimulate the economy by increasing spending in order to offset the downward effects on consumption and private investment. At the same time government revenues were decreasing, mainly caused by the downturn in demand pushing up debt-to-GDP ratios even more.

Table 2 describes the sample of independent variables. The growth rates of the

independent variables private savings, private investments, total factor productivity, capital stock and inflation were used. The average growth rate of private savings for Belgium is 5.98%, which - compared to the other countries - is the highest value. Interestingly, the high average rate of inflation in Italy of 7.05%, which is far above the ECB inflation target of 2%.

Table 2 Descriptive statistics independent variables (1970-2013)

Germany France Italy Belg.

Variables obs Mean Std.ev obs Mean Std.dev obs mean Std.dev obs mean Std.dev

Pr. sav 43 4.36 6.47 35 2.86 4.60 33 2.65 7.10 43 5.98 7.39 Pr. Inv. Tfp Capital Inflation 43 43 43 43 5.99 0.87 1.98 2.88 7.73 1.55 0.87 1.89 43 43 43 43 3.78 0.88 2.68 4.74 4.83 1.27 1.03 4.08 43 43 43 43 3.82 0.74 2.35 7.05 6.44 1.73 5.94 1.15 43 43 43 44 6.35 0.99 2.12 3.92 7.39 1.52 0.89 2.99

4.2 Testing results OLS regression

In the following section the regression results of the four countries will be presented. The basic estimation technique was Ordinary Least Square regression. The dependent variable was GDP per capita growth. This simple regression model was proposed to test whether the independent variables had a significant effect on economic growth. The results are

(21)

presented in table 3. Summarizing the most important results; a significant effect of the total factor productivity and the private investments was found for all sample countries. Further, all five explanatory variables were significant for the countries France and Belgium. Lastly, the reunification of West and East Germany and the recent Euro crisis had a significant effect on the German economy.

Taking a closer look at table 3, it can be noted that the total factor productivity coefficients were found significant at a 1% level for the sample countries. If the total factor productivity increases with 1%, the GDP per capita will increase with 1.1090% in France, for example, ceteris paribus.

Table 3 Regression results

Germany France Italy Belgium

Dependent variable GDP per capita growth Independent variables Constant 1.0565 -1.1262 0.1449 0.0284 (0.3754)*** (0.4670)** (0.5994) (0.1960) Private savings -0.0246 -0.0337 -0.1693 -0.0269 (0.0173) (0.0146)** (0.3101) (0.0123)** Private investments 0.036 0.0526 0.0944 0.0517 (0.0194)* (0.0220)** (0.0313)*** (0.0093)***

Total factor productivity 0.9597 1.1090 0.9928 1.0441

(0.0859)*** (0.0831)*** (0.1166)*** (0.0512)*** Capital stock 0.4248 0.8514 0.1695 0.4556 (0.1811)** (0.2209)** (0.3256) (0.0922)*** Inflation -0.2363 -0.0852 0.0198 -0.0983 (0.0695)*** (0.0216)*** (0.0468) (0.0223)*** Dummy crisis 0.9774 -0.1247 -0.5557 -0.2772 (0.3184)*** (0.4670) (0.4680) (0.1883) Dummy 1991 -0.7119 (0.2456)*** R-Squared 0.9295 0.9443 0.8969 0.9667 Adj R-squared 0.9150 0.9320 0.8721 0.9610

Note: * ,** and *** indicate significance at 10%, 5% and 1% respectively

Interesting is the high effect total factor productivity has on economic growth for the four sample countries. Table 4 shows the correlation between total factor productivity and GDP per capita growth. The high correlation could be explained by how total factor productivity is measured.

(22)

Table 4 Correlation between TFP and GDP per capita growth

Correlation between TFP and GDP per capita growth

0.9279 0.9140 0.8827 0.9359

As mentioned in section 3.2 total factor productivity was measured as an index (2005=100). It was calculated by dividing the weighted sum of total output by the weighted sum of all costs of capital and labour used. When total output increases, everything else being equal, the total factor productivity index increases too. It implies an increase in technological progress occurred that drove economic activity up. As predicted by economic literature a positive relationship between the total factor productivity and economic growth can be confirmed.

The effect of private investments on GDP per capita growth was found significant as well for all countries, even though this effect was relatively much smaller than the effect of the total factor productivity. If the private investments increase with 1%, the GDP per capita will increase with 0.0526% in France, for example, ceteris paribus.

Furthermore, the results show a positive relationship between capital stock and GDP per capita growth. The coefficients were statistically significant at 5% level for Germany and France, for Belgium at 1% significance level.

The coefficients of inflation had a negative sign for Germany, France and Belgium and were found significant at 1% level. Interesting is that Italy was an exception, since it had a positive coefficient and is statistically insignificant. The effect of inflation on economic growth can either be negative or a positive. A negative effect occurs when uncertainty over future inflation arises, leading to a decrease in investments. On the other hand, a positive effect arises when the central bank decreases real interest rates to offset the inflation effect. A decrease in real interest rates causes a rise in investments.

The coefficients of private savings were found significant at a 5% level for France and Belgium. Private savings also seemed to have a negative relationship with economic growth. In economic literature this negative relationship is explained as follows: an increase in savings leads to a decrease in disposable income. A lower disposable income leads to lower consumptions, causing a decrease in GDP.

(23)

The dummy variable 1991 was found significant at 1% level for Germany. This implies that the regime switch after the unification of West and East Germany had a significant effect on economic growth. The supplement intercept in the period 1991-2013 is -0.7119. The euro crisis of the period 2007-2013 also had an additional effect on economic growth in Germany since the dummy variable crisis is found significant at 1% level. Here the supplement intercept is 0.9774. The dummy variable crisis was highly insignificant for France, Italy and Belgium. This implies that the crisis in those three countries had no additional effect on economic growth other than the effect of the independent variables.

4.3 Results Chow test

It is possible that a difference exists in the effect of the independent variables on economic growth when public debt is above the Maastricht treaty debt limit of 60%. The chow test was performed to see whether a structural break exists beyond the debt-to-GDP level of 60%. If statistical evidence was found for a structural break the data was split up in two periods. The first period consisted the data of the years where debt-to-GDP ratio was lower than 60%. The second period covered the data where debt to GDP was higher than 60%. Two separate regressions were run. The results of the Chow test are presented in the table below.

Table 5 Chow test results

As shown in table 5, the chow test was significant at a level of 10% only for France. Therefore, evidence was found for the existence of a structural break. The sample of France was split up in two sub periods. In the first period public debt-to-GDP ratio is lower than 60%. The second period consist the part of the sample were debt-to-GDP ratio

N1 N2 N1 N2 N1 N2 N1 N2

Observations 27 15 24 10 11 31 11 32

Chow break point Year 1998 Year 2003 Year 1982 Year 1981

(24)

is more than 60%. The coefficients of the independent variables were estimated using Ordinary Leas Square regression for the two sub periods.

As shown in table 6, in the first period all dependent variables except private savings were statistically significant at 1%. The effect of 1% increase of total factor productivity on GDP per capita is the biggest with 1.0845%. In the second period only the coefficients of private investments and total factor productivity were found to be significant at a level of 1%.

When looking at the coefficients of total factor productivity and private investments, respectively they became smaller and bigger in the second period. It could suggest that at public debt-to-GDP ratios above 60%, the effect of total factor productivity on GDP per capita decreases. Whereas, the effect of private investments on GDP per capita more than doubles. A possible explanation could be the fraction of external public debt a country has.

Table 6 Regression results two sub periods France France

Dependent variable Period 1 Period 2

GDP per capita growth < 60% > 60% Independent variables Constant -0.8552 (0.5108) 0.5472 (0.6060) Private savings -0.0435 -0.0013 (0.0158) (0.0238) Private investments 0.0708 0.1648 (0.0247)*** (0.0351)*** Total factor productivity 1.0845 0.8232

(0.1019)*** (0.0808)*** Capital stock 0.8007 -0.0912 (0.2431)*** (0.3119) Inflation -0.1004 -0.2372 (0.0234)*** (0.1183) R-squared 0.9251 0.9964 Adj R-squared 0.9043 0.9920

When government deficits are mainly financed with external debt, a bigger part of future income will go to entities outside the country. Companies are less motivated to become

(25)

more productive, since foreign investors will have more advantage (Patillo, 2004). However, the effect of private investment on economic growth increases at higher debt levels. This could be explained through the crowd-in effect of private investment. The crowd-in effect arises when an increase in public debt leads to an increase in public expenditure. Investors consider the future of the economy to be more positive and increase private investment.

4.3.1 Robustness check – Breakpoint in 2003

It could be possible that the structural break found for France was not caused because the Maastricht treaty debt limit of 60% was exceeded. Another event could have caused a breakpoint in 2003. To check whether this could be the case the Chow test was

performed again. The same breakpoint of 2003 was set for Germany, Italy and Belgium.

Table 7 Chow test results with a breakpoint in 2003

Table 7 shows the results of the Chow test. The Chow test was significant at a level of

10% for Italy. Statistical evidence was found for a structural break. In 2003 the debt-to-GDP ratio of Italy was 99%. This suggests that the structural break found for France probably is caused by a different event. A possible explanation could be the introduction of the Euro in 2002. Due to this introduction a structural break in France and Italy in 2003 is found, because in these two countries the effect of the independent variables on economic growth was significant different.

The regression results for Italy are presented in table 8. The results are not discussed further, since the structural break found seems not to have been caused by exceeding the Maastricht treaty debt limit of 60%.

Germany Italy Belgium

N1 N2 N1 N2 N1 N2

Observations 32 10 22 10 32 10

Chow break point Year 2003 Year 2003 Year 2003

(26)

Table 8 Regression results two sub periods Italy Italy

GDP per capita growth < 2003 > 2003 Independent variables Constant -0.4209 (0.5847) -1.9033 (1.3360) Private savings -0.0060 -0.0741 (0.0278) (0.1235) Private investments 0.0783 -0.0348 (0.0277)** (0.2396) Total factor productivity 0.9916 1.1009

(0.1074)*** (0.4610)* Capital stock 0.6535 0.3238 (0.3174)* (0.6733) Inflation -0.4209 0.6732 (0.5847) (0.4564) R-squared 0.8864 0.9099 Adj R-squared 0.8509 0.7973

4.3.2 Robustness check – Breakpoint in 1992

Instead of looking at when the Maastricht treaty debt limit was exceeded. Lastly, it was researched whether a structural break was found after the Maastricht treaty was signed. In 1992 members of the European Union signed the Maastricht treaty. The debt limit of 60% of GDP was set as a requirement to take part in the new single currency Euro. It was agreed upon that countries with higher debt levels could also introduce the Euro provided that the development of their public debt was structurally decreasing towards the 60% limit, the so-called convergence. This could have caused a change in regime. The breakpoint used was 1992, the year the members of the European Union signed the Maastricht treaty. The Chow test results are shown in table 9. Statistical evidence at a 10% level was found for Germany. Implying a structural break exists in 1992. The regression results for the period before 1992 and the period after 1992 are presented in

(27)

Table 9 Chow test results with a breakpoint in 1992

As mentioned in section 3.2 the AMECO database split data on Germany into two periods. The first period (1970-1990) covered the timespan before the reunification of East and West Germany. The data only included numbers of West-Germany. The second period (1991-2013) covered data of the total economy of Germany. It seems more plausible that this structural break found for Germany is caused by the reunification of East and West Germany. Since no evidence was found for France, Italy and Belgium, it can therefore be concluded that signing of the Maastricht treaty did not have a significant influence on the effect of the independent variables on economic growth.

Table 10 Regression results two sub periods Germany Germany

GDP per capita growth < 1992 > 1992 Independent variables Constant 0.6016 (0.4031) 0.8464 (1.4489) Private savings 0.0036 -0.0047 (0.0207) (0.0793) Private investments 0.0497 0.1188 (0.2051)** (0.2140) Total factor productivity 1.0173 0.6949

(0.1251)*** (0.4773) Capital stock 0.3139 -1.8678 (0.2258) (2.3036) Inflation -0.1964 1.1618 (0.0773)** (1.0863) R-squared 0.8879 0.9600 Adj R-squared 0.8663 0.9101

N1 N2 N1 N2 N1 N2 N1 N2

Observations 21 21 13 21 11 21 21 21

Chow break point Year 1992 Year 1992 Year 1992 Year 1992

(28)

Summarizing the results of this chapter, firstly a significant effect of the total factor productivity and the private investments was found for all sample countries. Secondly, all five explanatory variables were significant for the countries France and Belgium. Thirdly, the reunification of West and East Germany and the recent Euro crisis had a significant effect on the German economy. Fourthly, statistical evidence was found for the existence of a structural break in 2003 for France. However, statistical evidence for a structural break was also found for Italy in 2003. Suggesting that the structural break found for France probably is caused by an event other than exceeding the Maastricht treaty debt limit of 60%. Lastly, statistical evidence was found for a structural break in 1992 for Germany. However, it is probably caused by the reunification of East and West Germany. In the next chapter the findings will be concluded.

(29)

5. Conclusion

The purpose of this research was to statistically test whether the effect of certain channels on economic growth differs and becomes negative when public debt levels are higher than the Maastricht treaty debt limit of 60% for Germany, France, Italy and Belgium. The channels investigated were private savings, private investments, total factor productivity, capital stock and inflation. The Chow test was performed to test whether a structural break exist. In this research no statistical evidence was found for a structural break caused by exceeding the Maastricht treaty debt limit of 60%.

The results show that a structural break existed in 2003 for France. In that year the public debt-to-GDP ratio exceeded the Maastricht treaty debt limit of 60%. Implying a difference in effect of the channels on economic growth. However, no statistical evidence on a negative relationship between the channels and economic growth was found. Thereafter, a robustness check was performed which provided evidence for a structural break in Italy in 2003. Combining this with the fact that the debt limit of 60% was exceeded in Italy in 1982, there may be suggested that the structural break found for France is probably caused by an event other than the overdraw of the Maastricht treaty debt limit. A possible explanation could be the introduction of the Euro in 2002. Due to this introduction a structural break in France and Italy in 2003 is found, because in these two countries the effect of the independent variables on economic growth was significant different.

Even though, the overall existence of a structural break could not be statistically proven by this research, along with the existing literature, suggestions for its absence would be: firstly, exceeding the debt limit of 60% might offset the effect of the independent variables on economic growth, and thereby decreases chances of a structural break. Secondly, statistical evidence for a structural break could be found at higher debt-to-GDP ratios. Existing literature points out the following: Reinhart and Rogoff (2010) found a turning point at debt-to-GDP ratios higher than 90%. Whereas Checherita-Westphal and Rother (2012) found statistical evidence that public debt levels above 90-100% of GDP have a negative impact on growth rates. Afonso and Jalles (2013) were the only who found a turning point of the effect of public debt on economic growth at a

(30)

debt-to-GDP ratio of 58%. By that means, it is questionable whether a structural break can be defined at a 60% debt level. Thirdly, other events, like the Euro introductions, could consequently lead to a structural break.

This study is limited by the use of yearly data instead of quarterly data, namely with the latter a more in debt investigation could be possible. However, this kind of data was not available for all the variables and transforming the yearly data into quarterly would have given a biased result Thereby a sample of only 44 observations (1970-2013) was used. Furthermore, The breakpoint of each country was assumed to be in the year the debt-to-GDP ratio of 60% was exceeded. Lastly, the use of other independent variables could have been given statistical evidence for the structural break caused by exceeding the Maastricht treaty debt limit of 60%.

The use of other independent variables to examine the effect on economic growth when public debt is restrictive, gives an opening to future research. As said, structural breaks may be caused by other events. Further statistical research could be done to bring unknown break levels to the surface, instead of examining a defined level. Lastly, the use of other sample countries to investigate certain channels on economic growth could be an extension of existing literature.

(31)

6. References

Abbas, S. A., & Christensen, J. (2007). The role of domestic debt markets in economic growth: an empirical investigation for low-income countries and emerging markets. International Monetary Fund working papers. pp. 1-42

Aristovnik, A., Mencinger, J., & Verbic, M. (2014). The Impact of Growing Public Debt on Economic Growth in the European Union, Available at SSRN, pp. 1-14 Afonso, A. & Tover Jallas, J. (2013), Growth and productivity: the role of government

debt, International review of economics and finance 25, pp. 384-407 Ash, M., Herndon, T. & Pollin, R. (2013), Does high public debt consistently stifle

economic growth? A critique of Reinhart and Rogoff, Political Economy

Research institute, pp. 1-23

Baum, A., Checherita-Westphal, C. & Rother, P. (2013), Debt and growth: new evidence for the euro area, Journal of international money and finance 32, pp. 809-821

Cecchetti, S., Mohanty, M. & Zampolli, F. (2011), “The real effects of debt", BIS

Working Papers No. 352. pp. 1 – 33

Checherita-Westphal, C. & Rother, P. (2012), The impact of high government debt on economic growth and its channels: an empirical investigation for the euro area, European economic review 56, pp. 1392-1405

Chen, S. S. (2014), Public Debt and its Impacts on Output: A Long-Horizon Perspective, Available at SSRN, pp. 626-643

Egert, B. (2012), Public debt, economic growth and non-linear effects; Mythe or reality?,

Available at SSRN, pp. 1-29

Elmendorf, Douglas W. and Mankiw, Gregory N. (1999), “Government debt”, in J. B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics, volume 1 of

Handbook of Macroeconomics, chapter 25, pp. 1615–1669

Herndon, T., Ash, M., & Pollin, R. (2013). Does high public debt consistently stifle economic growth? A critique of Reinhart and Rogoff. Cambridge Journal of

Economics, pp 257-279.

Kumar, M. S. & Woo, J. (2010). Public debt and growth, International Monetary Fund

working paper, pp. 1-46.

Lane, P.R. (2012), The European Sovereign debt crisis, Journal of economic perspective

26, pp. 49-68

Meensel, L. van & Nautet, M. (2011). Economic impact of the public debt.

Economic review 2, pp. 7-19

Misztal, P. (2010), Public debt and economic growth in the European Union, Journal of

Applied Economic Sciences (JAES), (13), pp. 292-302

Mussolini, C.C. & Teles, V.K., (2014), public debt and the limits of fiscal policy to increase economic growth, European economic review 66, pp. 1-15

Pan, H., & Wang, C. (2013). Co-movement of Government Debt and Economic Growth in the Euro-area: A Bayesian Dynamic Factor Model Analysis. International

Economic Journal, 27(4), pp. 625-643.

Panizza, U., & Presbitero, A. F. (2013). Public debt and economic growth in advanced economies: A survey. Swiss Journal of Economics and Statistics,149(2), pp.175-204.

(32)

Pattillo, C., Poirson, H., & Ricci, L. (2004). What Are the Channels Through Which External Debt Affects Growth?, International Monetary fund working papers, pp. 1 -34

Pescatori, A., Sandri, D., & Simon, J. (2014). Debt and Growth: Is There a Magic Threshold?, International Monetary Fund (No. 14-34), pp 1-19.

Reinhart, C. M., & Rogoff, K. S. (2010). Growth in a Time of Debt,

American Economic Review, 100(2), pp. 573–578

Schclarek, A. (2004), Debt and economic growth in developing and industrial

(33)

7. Appendix

Variable abbrev Variable name/discription Source

GDP_cap_g GDP growth per capita World databank

Debt_gdp Debt to gdp ratio (%) Reinhart and Rogoff

database saving_priv Gross savings private sector

growth rate %

AMECO

inv_priv Gross Fixed capital

formation private sector growth rate %(UIGP)

AMECO

inv_publ Gross Fixed capital

formation public sector growth rate % (UIGG)

AMECO

tfp_g Growth rate of Total factor

productivity, calculated based on TFP_index (2005=100)

AMECO

Net_cap_stock_g Net capital stock growth %

(OKND) AMECO

infl Inflation, consumer prices

(annual %) OECD

D_91 Dummy variable

0= period before 1991 1= otherwise

D_Cri Dummy variable

0= period before 2007 1= otherwise

The effect on economic growth when public debt is restrictive