The effect of public debt on economic growth

The effect of public debt on economic growth

OECD countries over the period 1980-2009. By using a fixed effects panel growth model augmented with a debt variable, a short-term nonlinear effect of public debt on economic growth is found. However, after adding time fixed effects, the debt variables become insignificant and do not indicate a short-term effect of public debt on growth anymore. When using a five-year average growth rate there is also found a nonlinear effect of debt on growth. Although after adding time fixed effects the quadratic debt term becomes insignificant, indicating a positive effect of public debt on economic growth. The fact that time fixed effects makes the coefficients of public debt insignificant might be explained by multicollinearity. As the results found in this paper are different from the results found in the paper of Checherita and Rother (2012), it might be interesting for further research to extend the model used in this paper with lagged dependent variables and an IV-regression.

Faculty of Economics and Business Sanneke van Walsem

10447474

Economics and Finance Field: Macroeconomics 12 ECTS

Supervisor 1: C. G. F. van der Kwaak Supervisor 2: R. M. Teulings

Faculty of Economics and Business Amsterdam, 15/07/2016

2 Statement of Originality

This document is written by Student Sanneke van Walsem who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Contents

1. Introduction ... 4

2. Literature Review ... 5

2.1 The Importance of Debt ... 5

2.2 Previous Literature ... 6

2.2.1 Theoretical Literature ... 6

2.2.2 Empirical Literature ... 8

3. The Trend of Debt ... 10

4. Methodology ... 12

4.1 Description and Selection of the Data ... 12

4.2 Model Specifications ... 13 5. Results ... 14 6. Discussion ... 16 7. Conclusion ... 17 8. References ... 19 9. Appendix ... 21

4

1. Introduction

Over the past decades public debt to GDP ratios were growing vast and rose to historically high levels that have not been measured since the second world war (Reinhart & Rogoff, 2011). In these past decades many countries were facing high public deficits, causing a sharp increase in public debt and a worsening of countries’ fiscal situations (Mencinger, Aristovnik & Verbic, 2014). According to Tanzi and Schuknecht (1997) the accumulation of debt was accompanied by a large increase in public expenditures. Additionally, the recent financial crisis has put considerable pressure on public finances in many countries, in particular on public debt (Checherita & Rother, 2012). During crises governments issue debt to support insolvent banks and other financial institutions (Rosas, 2006). Furthermore they issue public debt to support aggregate demand (Gagnon & Hinterschweiger, 2011). As a result, many countries are still facing high amounts of debt with little prospect for improvement due to ageing populations and high levels of governments expenditure. However, public debt can stimulate the economy if it is used carefully. It allows for redistributing taxes and consumption intergenerational and it can be used as a stimulus in times of economic downturns. Although, if the level of public debt is high it could be deleterious for economic growth. It might limit the ability of a government to borrow and to follow a countercyclical fiscal policy (Cecchetti, Mohanty, & Zampolli, 2011). According to Cecchetti et all. (2011) debt is a relevant determinant of growth. History has shown that high amounts of public debt lower economic growth and lead governments to reduce budget deficits that generally increase unemployment. Therefore, the relation between public debt and economic growth is of great interest to policymakers. Hence, this paper investigates the effect of public debt on economic growth. Although total debt consists of private and public debt, this paper only assesses the impact of general public debt on economic growth.

In order to show the effect of public debt on economic growth as good as possible this paper has selected 18 OECD countries over the period 1980-2009, as this period was the start of a significant increase in debt. Based on a conditional convergence equation used by Checherita and Rother (2012), this model “relates the GDP per capita growth rate to the initial level of income per capita, the investment/saving-to-GDP rate and the population growth rate” (p. 1395). Moreover, a debt term and a quadratic debt term will be added to the model, as this paper is particularly interested in the non-linear effect of public debt on economic growth. This paper extends the model of Checherita and Rother (2012) by controlling for banking crises, to take in consideration the effect of a banking crisis on economic growth.

5 Furthermore, it is important to mention that there could be endogeneity issues in the regression models. Particularly, the problem of causality is important as low economic growth can cause higher levels of debt.

The structure of this thesis is as follows. Section 2 begins with the literature review, to provide an overview of earlier studies that have been done. Section 3 consists of a data description and an explanation of the specific model that will be used in order to empirically asses the effect of public debt on growth. Section 4 describes the trend of debt and section 5 describes the results of the model. In section 6 the results are discussed. Finally, section 7 concludes and answers the research question.

2. Literature Review

2.1 The Importance of Debt

If public debt is used prudently it can be stimulating for the economy for multiple reasons. Public debt gives governments the opportunity to finance government spending and to spread taxes over time. Moreover, public debt can be used to transfer a tax burden from present to future generations (Cecchetti et al., 2011).For example, if the present generation fights a war this will also be in the interest of future generations, so they should carry a part of the tax burden (Mankiw, 2012). Because a part of the tax burden is shifted to future generations, generations that are currently living are better off and can consume more (Elmendorf & Mankiw, 1999). So the intertemporal redistribution of taxes will provide a possibility to increase intertemporal welfare of the community (Cecchetti et al., 2011). Lastly, government debt eases the credit conditions that firms and households face by providing liquidity services. As a result this crowds in private investment (Woodford, 1990). So for the reasons provided above, public debt helps stabilizing the economy in times of uncertainty and economic downturns and growing public debt is closely associated with improvements in welfare (Cecchetti et al., 2011).

Nevertheless, growing public debt is not without risk. In times of economic downturns governments often prefer to follow a countercyclical fiscal policy or to act as a lender of last resort. However, the ability of the government to borrow might be constrained. When the economy falls into a recession, the government’s ability to stabilize the economy depends on two things. First, it depends on their level of debt. Second, it depends on the way their creditors perceive the ability of the government to increase taxes so they can service their debt and eventually repay it. So higher levels of public debt can limit the government in following

6 a countercyclical fiscal policy or the ability of the government to act as a lender of last resort (Cecchetti et al., 2011).

Moreover, if governments are facing high levels of debt along with high costs of debt servicing they might want to reduce their debt level. They can decrease their debt level through growth, but in worse cases they have to decrease their debt by inflating it away, by running budget surpluses and by repudiation. Although this eliminates the debt problem, these options, excluding the one where debt will be worked down by growth, will launch new problems (Dornbusch & Draghi, 1990; Ostry, Ghosh, & Espinoza, 2015).

Furthermore, a higher level of debt makes the ability of a borrower to repay more sensitive to shocks, such as volatility in interest rates and revenues. So a higher debt decreases the probability of repayment (Cecchetti et al., 2011). If creditors believe that the government is unable to repay their full investment, they will stop providing new lending to the government. This might lead to the default of a country. The debt crisis of Mexico in 1994-1995 is an example of this (Cole & Kehoe, 1996).

Therefore, public debt at low levels improves welfare and provides economic stability, but when the level of public debt is high it increases real volatility and is deleterious for economic growth (Cecchetti et al., 2011).

2.2 Previous Literature

There are multiple papers who have focused on the impact of public debt on growth. However, this remains an important field of research, because many governments are still facing increasing public debt to GDP ratios caused by the economic crisis. Moreover, there is still little prospect for improvement of debt to GDP ratios. The existing literature can be divided in two parts, namely theoretical literature and empirical literature.

2.2.1 Theoretical Literature

The theoretical literature focuses on the relationship between public debt and economic growth and is indicating a negative relation. Growth models that are augmented with public agents who issue debt to finance producer goods or consumption show this negative relationship between public debt and growth, especially in a neoclassical context (Checherita & Rother, 2012).

Aschauer’s growth model (2000) proposed a non-linear impact of public debt via public infrastructure capital on economic growth. As governments use part of their debt to finance public infrastructure capital, a higher debt first has a positive effect on economic

7 growth up to a certain threshold. After this threshold the effect turns out to be negative. Elmendorf and Mankiw (1999) also state that public debt has a positive effect on economic growth in the short-run, as an increase in aggregate demand raises national income. However, in the long-run public debt has a negative effect on growth, as it reduces capital stock. Elmendorf and Mankiw (1999) show how a rise in debt will lead to a decrease in private capital stock by supposing the government keeps their spending constant and decreases tax revenues, whereby it creates a budget deficit and decreases public saving. They argue that private saving will not rise as much as public saving falls, so that national saving decreases. As the sum of private and public saving equals domestic investment and net foreign investment, a decrease in savings will decrease investments in the current country as well as in foreign countries. Decreasing investment in the domestic country will reduce the domestic capital stock, which in turn would lower incomes and economic growth.

In the theoretical literature are also pointed out multiple channels through which public debt has an effect on economic growth. According to Elmendorf and Mankiw (1999), the long-term interest rate is an influential channel through which an increase in public debt can affect economic growth. They state, just as Modigliani (1961), that accumulating public debt increases long-term interest rates, because a decrease in private capital rises its marginal productivity. This increase in interest rates might crowd out private investment, whereby it can moderate economic growth. In addition, an increase in public debt yields may lead to a higher net flow of funds from the private to the public sector. This flow out of the private sector could potentially cause a rise in private interest rates and a decline in private expenditures. This is the case for households as well as for firms (Elmendorf & Mankiw, 1999).

Krugman (1988) focused on the problem emerged by a debt overhang. In case of a debt overhang the country’s external debt (debt that is borrowed from foreign lenders) is substantial, which means there is a probability that foreign creditors will not receive their full investment back. In other words, debt of a country is expected to exceed the present value of potential resources. Krugman (1988) argues that an external public debt overhang has an effect on economic growth via private investment. A higher public debt means an increase in the service of debt in proportion to GDP. So, some of the returns from investing in the domestic economy will be used to repay foreign creditors. (Clements, Bhattacharya & Nguyen, 2003). Another thing is that a higher public debt could form expectations that the government would service its debt with distortionary taxation (Agénor and Montiel, 1996).

8 These two facts discourage investment of new investors, and in turn discourage economic growth (Clements et al., 2003; Krugman, 1988).

Another channel through which public debts affects economic growth is uncertainty. Countries are currently in a situation where they could only service their level of debt with new borrowing. However, since the uncertainty about large debtors’ ability to repay its debt, it would be difficult to attract new borrowing (Krugman, 1988).

Inflation could also be a channel through which public debt affects economic growth. If the government finances its budget deficit by printing new money it can cause inflation (Mankiw, 2012). In addition, Mankiw argues that some governments that face a high debt might create inflation to decrease their amount of debt in real terms. However there is little evidence for this in developed countries. According to Barro (1995) high inflation decreases the GDP growth rate.

Finally, according to Reinhart and Rogoff (2010) it is important to monitor how debt accumulates. A debt buildup in times of peace is more dangerous for long term growth and inflation than accumulating debt resulting from war. There are two explanations for this. First, high government spending during a war is only temporary. If the war is over, the accumulation of debt comes to an end. Second, after the war economic growth tends to be high, because a war allocates workers and capital in the direction of a civilian economy (Reinhart and Rogoff, 2010).

2.2.2 Empirical Literature

The empirical literature on the effect of public debt and economic growth has grown extremely large and was a famous field of research the last few years. Reinhart and Rogoff (2010) have recently done an influential research to this relationship. They found that the relation between public debt and growth is not strong at low levels of debt, but a public debt to GDP level above 90 percent lowers the GDP growth rate by one percent. Furthermore, they did not find a difference in the relationship of public debt and economic growth among emerging and advanced economies.

Herndon, Ash and Pollin (2014) criticized the findings of Reinhart and Rogoff and argued that the relationship between public debt and economic growth found by them is inaccurately measured. The reason for this are coding errors, omission of data and different weighting and transcription of data. If they properly recalculate the effect of public debt on economic growth they found a different effect. According to them, if a country has a public debt level above 90 percent of GDP, the growth rate is not seriously different than when

9 public debt to GDP ratios are lower. Furthermore, Herndon et al. (2014) challenge the threshold of the public debt to GDP ratio, found by Reinhart and Rogoff, after which there will be a non-linear effect on growth. They also find a non-linear effect in the relationship between public debt to GDP and economic growth, just as Reinhart and Rogoff, but only in public debt to GDP levels of 0-30 percent and 30-60 percent. Those public debt to GDP levels are not relevant for the current debate about the debt policy, and hence they argue that according to the errors in findings we should reconsider the austerity measures in the world.

Moreover, some other studies have focused on the relationship between public debt and economic growth. Those studies (Smyth and Hsing, 1995; Mencinger, et al. 2014) point out a negative correlation between public debt and economic growth, and find that this correlation becomes particularly strong after a specific threshold of the debt to GDP ratio. Although most studies agree on the nonlinear effect of public debt on growth, they differ in the findings of the debt threshold. Kumar and Woo (2010) provided evidence for a concave relation between public debt and growth over the period 1970-2007 and state that a higher level of initial public debt has an almost in proportion negative effect on subsequent growth. Cecchetti et al. (2011) examined this relationship over the period 1980-2006 and found that a public debt to GDP ratio above 85 percent is harmful for economic growth. Baum, Checherita and Rother (2013) investigated the effect of public debt on growth in the Euro area and found a nonlinear relationship between public debt and growth where a public debt to GDP ratio above 95 percent lowers long-term growth rates.

When Panizza and Presbitero (2014) studied the relation between public debt and economic growth in advanced countries they did not find evidence for a negative effect of public debt on economic growth. Accordingly, they argue that the relation between debt and growth should not be used as an argument in favor of minimizing budget deficits. However, as they do not find a direct effect of public debt on growth, they argue that not all levels of public debt are sustainable. This is the case when interest payments on the debt exceed GDP.

Additionally, it is important to focus on the reverse causality problem (Pescatori, Sandri & Simon, 2014). If debt and growth are correlated this does not indicate causation. High levels of debt could cause low economic growth, but the other way around, low economic growth can cause high levels of debt. Furthermore accumulating debt could be caused by an omitted variable which has a joint effect on debt and growth. For example, a war could be an omitted variable. It increases public debt and lowers economic growth. To tackle this problem a number of studies, such as Reinhart and Rogoff (2012), analyze the effect of

10 debt on growth by using extended periods of large debt levels. Those studies state that growth is substantially lower during periods of large public debt.

A number of other studies focus on the causality problem by making use of an instrumental variable estimation. In those studies they create an instrumental variable that has a direct effect on debt but has no direct effect on economic growth. One of the studies is done by Panizza and Presbitero (2014). If they build a variable that encloses valuation effects accomplished by the synergy between debt indicated in foreign currency and exchange rate fluctuation, they do not find a relation between public debt and economic growth. Hence, a specific threshold could only exist if there is a causal effect of debt on economic growth. It would not be likely to observe a explicit threshold if low economic growth causes high debt. In line with these findings, Pescatori et al. (2014) argue that if a specific debt threshold is observed, it is expected that public debt has an effect on growth.

Empirical literature also points out multiple channels through which public debt affects economic growth. Checherita and Rother (2012) studied the impact of debt on private saving and investment, the public investment rate, total factor productivity and sovereign long-term nominal and real interest rates. They found that public debt has a non-linear impact on private saving, the public investment rate and total factor productivity. Moreover, they also found a robust impact of public debt on sovereign long-term nominal and real interest rates. An increase in the debt to GDP ratio raises sovereign interest rates, which in turn has an effect on economic growth.

3. The Trend of Debt

Figure 1 (Reinhart & Rogoff, 2011, p. 8) illustrates the government debt to GDP in emerging and advanced market economies over the period 1860-2010.

Figure 1: Gross Central Government Debt as a Percent of GDP: Advanced and Emerging Market Economies, 1860-2010

11 Before the 20th century, the growth of debt was quite low and sharp run-ups in debt could largely be attributed to war financing, such as the Napoleonic Wars in the early 1800s or to depressions (Elmendorf & Mankiw, 1999; Reinhart & Rogoff, 2011). Also in the 20th century, two peaks in government debt could be observed that coincide with the two world wars. Increasing debt during peacetime could mostly be attributed to severe or systemic financial crisis. For around two decades, a growing tendency occurs towards increasing government commitment in saving operations. This increases the relation between public debt and financial crashes (Reinhart & Rogoff, 2011). The only periods where the debt to GDP ratio increased sharply during peacetime, were the Great Depression and the 1980s (Elmendorf & Mankiw, 1999)

In the early beginning of the 19th century countries already started to use government debt for other purposes, such as investing in public work or colonial expansion. For example, the United States and France issued a high amount of debt to improve public work. Another reason for the increase in the public debt to GDP ratio was the increase in the size of governments (Checherita & Rother, 2012). In earlier research, Tanzi and Schuknegt (1997) show that bigger governments with higher government expenditures, are facing higher public debt to GDP ratios. Especially after the year 1960 there was a vast growth in government expenditures and activities, as shown by Tanzi and Schuknegt (1997). In that time average government expenditures rose from 27,9 percent of GDP to 44,9 percent of GDP in 1990. At the same time high public debts were created.

Besides the reasons provided above, there were some important developments in the market throughout history. First, since the late ‘1970s and beyond it became easier to borrow, because restrictions on financial markets and lending were eliminated. Together with developments in information science and increased knowledge in financial theory, this liberalization amplified financial innovation (Cecchetti et al., 2011).

Second, since mid-1990 it became easier to maintain a higher public debt because of a significant drop in real interest rates. This drop in real interest rates lowered the costs of servicing debt, which made it easier for governments to borrow. The cause for the lower long-term interest rate is debatable (Cecchetti et al., 2011). According to Bernanke (2005), one of the explanations for lower long-term interest rates is the substantial increase in global saving. He argued that this increase in global saving could, among other things, be attributed to higher saving in industrial countries to pay for the increase in the ratio of retirees to workers.

12 Third, according to Reinhart and Rogoff (2009) the financial crisis that started in 2008 had contributed to the large buildup in debt. During crises governments have to support insolvent banks and other financial institutions (Rosas, 2006). Furthermore governments issue public debt to support aggregate demand (Gagnon & Hinterschweiger, 2011). Also, during a crisis tax revenues fall and government spending on unemployment benefits increase which increases budget deficits (Mankiw, 2012). In the recent crisis this increased average real debt levels by about 75 percent for the countries Iceland, Ireland, Spain, the United Kingdom, and the United States (Reinhart & Rogoff, 2009).

Finally, fiscal and monetary policies of governments also influences the level of debt in a country. Some countries could be facing high levels of debt due to specific policies executed by the government, such as high public spending or tax cuts (Mankiw, 2012).

In order to assess the effect of public debt on economic growth, it is necessary to do an empirical research. The existing literature gives the impression that public debt has a negative effect on growth and that there exists a specific threshold, beyond which public debt has a negative effect on growth. This paper will only focus on the non-linear effect of public debt on growth by using the empirical growth model of Checherita and Rother (2012). However, as the effect of public debt on growth might be affected by a banking crisis, this paper will extend the model of Checherita and Rother (2012) by controlling for banking crises. It will do this by adding a dummy variable of banking crises.

4. Methodology

4.1 Description and Selection of the Data

This paper has selected 18 OECD countries (Austria, Australia, Belgium, Canada, Denmark, Finland, France, Portugal, Germany, Greece, Japan, Italy, The Netherlands, Norway, Spain, Sweden, United Kingdom and the United States) as data from these countries is broadly available and it is properly documented. This paper uses data of databases such as the OECD, IMF and Penn World Tables, covering the period 1980-2009. According to Checherita and Rother (2012) “using this relatively restricted cross-sectional sample also helps mitigating the issue of heterogeneity, which often turns problematic in standard growth regressions” (p.1395).

13 4.2 Model Specifications

An empirical growth model will be employed to account for the impact of the level of the public debt to GDP ratio on the real per capita growth to GDP rate. Based on a conditional convergence equation used by Checherita and Rother (2012), this model “relates the GDP per capita growth rate to the initial level of income per capita, the investment/saving-to-GDP rate and the population growth rate” (p. 1395). Moreover, a debt term and a quadratic debt term will be added to the model, as this paper is particularly interested in the non-linear effect of public debt on economic growth. In addition to the model of Checherita and Rother (2012) this paper will control for banking crises. As a banking crisis might have a negative effect on economic growth, not including a banking crisis in the empirical model might lower the intercept. For this reason, this paper adds a dummy variable for a banking crisis to the model of Checherita and Rother (2012).

Furthermore, control variables will be used to allow for expanding the model beyond a closed-economy form as also used by Checherita and Rother (2012). These control variables are indicators for the openness of the economy as well as a proxy for the external competitiveness. In this paper trade openness and the real effective exchange rate will be taken as a proxy. Moreover, to capture the impact of inflation, a control variable of inflation will be added. In appendix 1 there is shown a list of variables that are used in the different regression models along with the data sources.

In addition, this paper has conducted a Hausman test to test whether there should be used fixed effects or random effects in this estimation. The results of this test are shown in appendix 2 and indicate that fixed effects should be used. These country fixed effects control for variables that are constant over time but differ across countries, whether they are observable or unobservable. For example, country fixed effects can control for economic or social aspects that differ between countries but do not change over time. Moreover this paper has conducted a test to check if time fixed effects are needed. The results of this test are shown in appendix 3 and indicate that time fixed effects should be added in addition to the country fixed effects already used. Time fixed effects control for variables that are constant over time but differ across countries, such as common shocks across countries.

Furthermore, this paper has conducted a modified Wald test to test for heteroscedasticity and a Wooldridge test to test for autocorrelation in panel data. The results of this tests are shown in appendix 4 and 5 and indicate that this paper is subject to heteroscedasticity as well as serial correlation. Therefore, in this estimation heteroscedasticity- and autocorrelation-consistent (HAC) standard errors will be used. HAC

14 standard errors correct for autocorrelation and heteroscedasticity. Finally, in some estimations the five-year average growth rate will be used, to minimize the potential for the endogeneity bias and to analyze the long-term effect of debt on growth and the potential/trend GDP growth rate as a dependent variable (Checherita and Rother, 2012)

The model that will be used is as follows: 𝑔𝑖𝑡+𝑘= 𝛼 + 𝛽 ln (

𝐺𝐷𝑃

𝐶𝑎𝑝)_𝑖𝑡+ 𝜑1𝑑𝑒𝑏𝑡𝑖𝑡+ 𝜑2𝑑𝑒𝑏𝑡𝑖𝑡2 + 𝛿𝑠𝑎𝑣𝑖𝑛𝑔/𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 𝑟𝑎𝑡𝑒𝑖𝑡 + 𝜃𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑔𝑟𝑜𝑤𝑡ℎ_𝑖𝑡 + 𝑐𝑟𝑖𝑠𝑖𝑠_𝑖𝑡+ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠_𝑡+ 𝛼_𝑖 + 𝑣_𝑡+ 𝜀_𝑖𝑡

𝑔𝑖𝑡 = real GDP growth rate per capita

ln (𝐺𝐷𝑃 𝐶𝑎𝑝)_𝑖𝑡

= first difference of the natural logarithm of real per capita GDP 𝑑𝑒𝑏𝑡_𝑖𝑡 = general government debt as a share of GDP

𝑠_𝑖𝑡 = gross national savings as a share of GDP

𝑐𝑟𝑖𝑠𝑖𝑠𝑖𝑡 = dummy variable for banking crisis, 1 for crisis, 0 otherwise

𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑡 = control variables: indicators for the openness of the economy

and external competitiveness, inflation 𝛼𝑖

𝑣_𝑡

= country-fixed effects = time-fixed effects

𝜀_𝑖𝑡 = error term. This error term measures any omitted factors as well as all other factors that have an impact on the growth rate.

So, first the growth rate will be calculated on the basis of the model above with an annual GDP growth rate to capture the short-term effect of public debt on economic growth. Second, the five-year average growth rate will be used, to further minimize the potential for the endogeneity bias and to analyze the long-term effect of debt on growth and the potential/trend GDP growth rate as a dependent variable. Finally, these two models will also be estimated with time fixed effects included.

5. Results

As mentioned in the previous section, panel growth regression models augmented with a debt variable are used to estimate the relation between public debt and economic growth. Precisely,

15 in order to control the influence on economic growth, this paper studies different potential explanatory variables.

First, if the short-term effect of debt on growth is estimated both debt variables (in linear and quadratic form) are significant. The coefficient of the linear public debt variable is positive and the coefficient of the quadratic public debt variable is negative in this sample.

Second, estimating the regression model with a five-year average growth rate also results in significant public debt variables. Just as in the short-term growth model the coefficient of public debt is positive and the coefficient of the quadratic public debt variable is negative. However, the quadratic public debt variable is less significant than in the estimation of the short-term effect of debt on economic growth.

When time fixed effects are added to the short-term effect regression model this results in insignificant debt variables. Moreover, adding time fixed effects to the regression model where the five-year average growth rate is used, also results in an insignificant quadratic term of debt. The linear public debt variable remains significant in this regression model.

However it should be mentioned that adding time fixed effects contributes to the predictability of both regression models, as could be seen from the R-squared.

Table 1: Direct relationship between debt and per-capita GDP growth

Regressor (1) (𝒈_{𝒊𝒕+𝟏}) (2) (𝒈_{𝒊𝒕 / 𝒕+𝟓}) (3) (𝒈_{𝒊𝒕+𝟏}) (4) (𝒈_{𝒊𝒕 / 𝒕+𝟓}) First difference of log_rgdp 0.4558539 *** (0.0726779) -0.0274807 (0.0313256) 0.320812*** (0.0606986) -0.0324227 (0.0366541) Ngs -0.0228107 (0.0534196) -0.0326516 (0.0417826) 0.0219888 (0.0254546) 0.0236666 (0.0390944) debt_gov 0.0508142** (0.0180697) 0.0450368** (0.018723) 0.0225529 (0.0157275) 0.0379376* (0.0201117) debt_gov2 -0.0260671*** (0.0084951) -0.0184503 * (0.0104774) -0.0095049 (0.0057239) -0.0145653 (0.0084046) pop_gr -0.4829728 (0.3703578) -0.733202** (0.3303668) -0.2123293 (0.3241444) -0.5583402 (0.3240359) Crisis -0.0133774*** (0.002866) -0.0008247 (0.0021826) -0.0043254 (0.0028867) -0.0000959 (0.0027868) Cpi -0.0722824*** (0.0214615) 0.008863 (0.0348017) -0.1308045*** (0.0344817) -0.072475 (0.0542378)

16 REER -0.0418161* (0.0213023) -0.0536492 *** (0.0124479) -0.0065769 (0.0114909) -0.0293272** (0.0100932) Openc -0.0716924*** (0.0100799) -0.0257348 * (0.0137509) 0.0298108*** (0.0095905) 0.0182332 (0.0129692) intercept 0.0887963*** (0.0260157) 0.0770103*** (0.0132085) -0.0025972 (0.0169008) 0.0289741* (0.0164645)

Country Fixed Effects Yes yes yes Yes

Time Fixed Effects No no yes Yes

Observations 504 432 504 432

R-Squared Within 0.3758 0.2086 0.6432 0.4733

HAC standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1.

6. Discussion

First, when the model is estimated without time fixed effects (column 2 and 3), there is found a statistically significant non-linear relationship between the public debt ratio and the growth rate per capita for the 18 euro OECD countries included in this paper. As the coefficients of the quadratic public debt variables are negative, this indicates a concave relationship between public debt and economic growth. Furthermore, as the coefficients of the linear debt variables are positive, this confirms the theoretical assumption that debt at low levels has a positive effect on growth, however that beyond a certain threshold this effect becomes negative.

However, after including time fixed effects to the regression model estimated in column 2, both debt coefficients turn insignificant. So according to the results found here there is no relation between public debt and economic growth. Furthermore, adding time fixed effects to the model estimated in column 3 results in a significant coefficient of the linear debt variable and an insignificant coefficient of the quadratic debt variable, which indicates that public debt has a positive effect on economic growth.

The fact that time fixed effects makes the coefficients of public debt insignificant might be explained by multicollinearity. The debt variables might be highly linearly correlated with the time dummies affecting both the coefficient and the standard errors of the debt variables. In appendix 6 this papers uses variance inflation factors (VIF) to detect multicollinearity. Both coefficients of debt have a really high VIF, which suggests that a high degree of multicollinearity is present. Another explanation could be that omitting time fixed

17 effects creates an omitted variable bias whereby the coefficients of debt and debt squared seem higher than they actually are.

The findings that are found in this paper are different from the findings in the paper of Checherita and Rother (2012). They find a significant non-linear effect of public debt on economic growth, also if they include time fixed effects in their regression. The reason for this might be that they have lagged all their dependent variables. It might be the case that adding lagged dependent variables makes the coefficients of debt significant. So, in further research it could be interesting to extend the model that is used in this paper with lagged dependent variables to monitor if the effect of public debt on economic growth becomes significant.

This paper also has some other limitations and areas for further research. First, this paper has not estimated a certain debt threshold, beyond which debt starts to have a negative effect on growth. Second, to address the causality problem and especially the problem of reverse causation, an IV-regression should be used. This was outside the scope of this research. Nevertheless, this paper uses a five-year average growth rate to partially mitigate the problem of endogeneity and especially reverse causation. Lastly, this paper has only focused on the effect of public debt on economic growth. To assess the impact of total debt on growth, it is important to take public as well as private debt in consideration.

7. Conclusion

This paper tries to provide an answer on an important economic question that is of great interest to policymakers: the effect of public debt on economic growth. In order to answer this question there is conducted an empirical study on the effect of public debt on economic growth for 18 OECD countries over the period 1980-2009.

Based on a conditional convergence equation used by Checherita and Rother (2012), this model “relates the GDP per capita growth rate to the initial level of income per capita, the investment/saving-to-GDP rate and the population growth rate” (p. 1395). Moreover, a debt term and a quadratic debt term will be added to the model, as this paper is particularly interested in the non-linear effect of public debt on economic growth. This paper extends the model of Checherita and Rother (2012) as it has controlled for a banking crises. Furthermore, the issue of heterogeneity is mitigated by using a fixed effects panel regression model. In addition, to mitigate the endogeneity issue and in particular the reverse causality problem, five-year averages growth rates are used.

18 Estimating the short-term model without time fixed effects results in significant debt terms, indicating a concave relationship between public debt and economic growth. However, after adding time fixed effects, the coefficients of debt and debt squared become insignificant, indicating that there is no effect of public debt on economic growth.

Estimating the model when using five-years average growth rates also results in significant coefficients of public debt indicating a non-linear relation between public debt and growth. However, after adding time fixed effects the coefficient of the quadratic debt term becomes insignificant, indicating a positive effect of debt on economic growth.

The results that are found in this paper are different from the results found in the paper of Checherita and Rother (2012), as in this paper there is not found an effect of public debt on economic growth after adding time fixed effects. This could be partly explained by multicollinearity. For further research it would be interesting to add lagged dependent variables and an IV-regression to this model to monitor if this results in a significant relation between public debt and growth, as found in other papers.

19

8. References

Agénor, P. R., & Montiel, P. (1996). Development Economics. Editorial: Princeton.

Aschauer, D. A. (2000). Do states optimize? Public capital and economic growth. The annals of regional science, 34(3), 343-363.

Barro, R.J. (1995). Inflation and Economic Growth. NBER working paper No. 5326.

Baum, A., Checherita-Westphal, C., & Rother, P. (2013). Debt and growth: New evidence for the euro area. Journal of International Money and Finance, 32, 809-821.

Bernanke, B. S. (2005). The global saving glut and the US current account deficit (No. 77). Cecchetti, S. G., Flores-Lagunes, A., & Krause, S. (2006). Assessing the sources of changes

in the volatility of real growth. NBER working paper no. 11946

Cecchetti, S. G., Mohanty, M. S., & Zampolli, F. (2010). The future of public debt: prospects and implications. Paper was presented at the Reserve Bank of India’s International Research Conference: Challenges to Central Banking in the context of Financial Crisis. Mumbai: 12-13 February 2010

Cecchetti, S. G., Mohanty, M. S., & Zampolli, F. (2011). The real effects of debt. BIS working paper no. 352.

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(7), 1392-1405.

Clements, B. J., Bhattacharya, R., & Nguyen, T. Q. (2003). External debt, public investment, and growth in low-income countries. IMF working paper no. 03/249.

Cole, H. L., & Kehoe, T. J. (1996). A self-fulfilling model of Mexico's 1994–1995 debt crisis. Journal of international Economics, 41(3), 309-330.

Dornbusch, R., & Draghi, M. (1990). Public debt management: theory and history. Cambridge University Press.

Elmendorf, D. & Mankiw, N. G. (1999). Government Debt, in Taylor, J. and Woodford, M. (eds.), Handbook of Macroeconomics 1, part 3: 1615-1669.

Gagnon, J. E., & Hinterschweiger, M. (2011). The Global Outlook for Government Debt over the next 25 years. Peterson Institute for International Economics.

Herndon, T., Ash, M., & Pollin, R. (2014). Does high public debt consistently stifle economic growth? A critique of Reinhart and Rogoff. Cambridge journal of economics, 38(2), 257-279.

20 Krugman, P. (1988). Financing vs. forgiving a debt overhang. Journal of development

Economics, 29(3), 253-268.

Kumar, M., & Woo, J. (2010). Public debt and growth. IMF working paper no. 10/174 Mankiw, N. G. (2012). Macroeconomics (8th ed.). New York: Worth Publishers.

Mencinger, J., Aristovnik, A., & Verbic, M. (2014). The impact of growing public debt on economic growth in the European Union. Amfiteatru economic, 16(35), 403.

Modigliani, F. (1961). Long-Run Implications of Alternative Fiscal Policies and the Burden of the National Debt. Economic Journal, 71 (284), 730-755.

OECD (2016), Long-term interest rates (indicator). doi: 10.1787/662d712c-en (Accessed on 23 June 2016)

Ostry, M. J. D., Ghosh, M. A. R., & Espinoza, R. A. (2015). When should public debt be reduced?. International Monetary Fund.

Panizza, U., & Presbitero, A. F. (2014). Public debt and economic growth: Is there a causal effect?. Journal of Macroeconomics, 41, 21-41.

Pescatori, A., Sandri, D., & Simon, J. (2014). Debt and growth: Is there a magic threshold? (No. 14-34). International Monetary Fund.

Reinhart, C.M., & Rogoff, K. S. (2009) The aftermath of financial crises. NBER Working Paper No. 14656

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

Review, 100(2), 573-578.

Reinhart, C. M., & Rogoff, K. S. (2011). A decade of debt. NBER working paper no. 16827

Reinhart, C. M., Reinhart, V. R., & Rogoff, K. S. (2012). Public debt overhangs: Advanced-economy episodes since 1800. The Journal of Economic Perspectives, 26(3), 69-86. Rosas, G. (2006). Bagehot or bailout? An analysis of government responses to banking

crises. American Journal of Political Science, 50(1), 175-191.

Smyth, D. J., & Hsing, Y. (1995). In search of an optimal debt ratio for economic growth. Contemporary Economic Policy, 13(4), 51-59.

Tanzi, V., & Schuknecht, L. (1997). Reconsidering the fiscal role of government: the international perspective. The American Economic Review,87(2), 164-168.

Woodford, M. (1990). Public debt as private liquidity. The American Economic Review, 80(2), 382-388.

21

9. Appendix

Appendix 1

Tabel 2: Data description and sources

Variable Variable name/description Source

Growth Real growth per capita in % of GDP Penn World Tables 7.0

Rgdp Real GDP per capita Penn World Tables 7.0

Ngs Gross national savings in % of GDP IMF World Economic Outlook Crisis Banking crisis, 1 for crisis, 0 otherwise Reinhart, C.M.

http://www.bis.org/publ/work352.htm debt_gov General government, all liabilities (in %

of GDP)

IMF, OECD, national sources Debt_gov2 The square of general government debt

(in % of GDP)

IMF, OECD, national sources Pop_gr Population growth in % (1980 is basis

year)

Penn World Tables 7.0 Cpi Annual changes in consumer prices in

%

World Bank World Development Indicators

REER Real Effective Exchange Rate, based on ULC, relative to rest 67 countries

Bruegel Datasets

http://bruegel.org/publications/datasets /real-effective-exchange-rates-for-178-countries-a-new-database/

Openc Trade openness in % of GDP (in current prices)

Penn World Tables 7.0

22 Appendix 2 Hausman test ---- Coefficients ---- (b) fixed (B) random (b-B) difference sqrt(diag(V_b-V_B)) S.E. First difference of log_rgdp 0.4558539 0.4739453 -0.0180914 0.0107897 ngs -0.0228107 0.0043 -0.0271107 0.0259841 debt_gov 0.0508142 0.0327279 0.0180863 0.0086632 debt_gov2 -0.0260671 -0.0173788 -0.0086883 0.0031768 pop_gr -0.4829728 0.023327 -0.5062998 0.2628333 crisis -0.0133774 -0.0113487 -0.0020287 cpi -0.0722824 -0.0479571 -0.0243252 0.0194387 REER -0.0418161 -0.024146 -0.0176702 0.0062235 openc -0.0716924 -0.0132814 -0.058411 0.0090458 b = consistent under Ho and Ha; obtained from xtreg

B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic

chi2(9) = (b - B) ' [ (V_b-V_B) ^ (-1) ] (b-B) = 70.01

Prob>chi2 = 0.0000

23 Appendix 3

Test to see whether time fixed effects are needed

( 1) 1982.Year = 0 ( 2) 1983.Year = 0 ( 3) 1984.Year = 0 ( 4) 1985.Year = 0 ( 5) 1986.Year = 0 ( 6) 1987.Year = 0 ( 7) 1988.Year = 0 ( 8) 1989.Year = 0 ( 9) 1990.Year = 0 (10) 1991.Year = 0 (11) 1992.Year = 0 (12) 1993.Year = 0 (13) 1994.Year = 0 (14) 1995.Year = 0 (15) 1996.Year = 0 (16) 1997.Year = 0 (17) 1998.Year = 0 (18) 1999.Year = 0 (19) 2000.Year = 0 (20) 2001.Year = 0 (21) 2002.Year = 0 (22) 2003.Year = 0 (23) 2004.Year = 0 (24) 2005.Year = 0 (25) 2006.Year = 0 (26) 2007.Year = 0 (27) 2008.Year = 0 Constraint 7 dropped Constraint 10 dropped Constraint 12 dropped Constraint 13 dropped Constraint 17 dropped Constraint 18 dropped Constraint 19 dropped Constraint 20 dropped Constraint 22 dropped Constraint 25 dropped F( 17, 17) = 65.39 Prob > F = 0.0000

24 Appendix 4

Modified Wald test for groupwise heteroscedasticity in fixed effect regression model H0: sigma(i)^2 = sigma^2 for all i

chi2 (18) = 355.80 Prob>chi2 = 0.0000

Appendix 5

Wooldridge test for autocorrelation in panel data H0: no first-order autocorrelation

F( 1, 17) = 77.519 Prob > F = 0.0000

25 Appendix 6

Test for multicollinearity

Variable VIF 1/VIF

debt_gov 33.45 0.029895 debt_gov2 26.14 0.038256 Year 1981 1.94 0.516534 1982 1.96 0.510871 1983 1.99 0.503350 1984 2.01 0.496521 1985 2.05 0.488263 1986 2.07 0.482838 1987 2.08 0.481302 1988 2.06 0.484920 1989 2.04 0.489122 1990 2.05 0.488859 1991 2.09 0.478888 1992 2.16 0.461895 1993 2.27 0.440457 1994 2.27 0.441017 1995 2.30 0.434129 1996 2.31 0.432127 1997 2.29 0.436287 1998 2.29 0.437386 1999 2.24 0.446929 2000 2.20 0.455007 2001 2.17 0.460110 2002 2.19 0.457502 2003 2.19 0.456772 2004 2.19 0.456499 2005 2.18 0.458950 2006 2.15 0.465453 2007 2.12 0.472748 2008 2.17 0.461036 FEtid_2 2.32 0.431193 FEtid_3 1.98 0.504621 FEtid_4 1.94 0.516057 FEtid_5 1.91 0.524823 FEtid_6 2.17 0.460743 FEtid_7 2.28 0.438376 FEtid_8 2.16 0.462192 FEtid_9 1.89 0.528546 FEtid_10 2.43 0.412198 FEtid_11 1.91 0.523775 FEtid_12 2.06 0.485904 FEtid_13 2.02 0.494529 FEtid_14 1.96 0.510606 FEtid_15 1.98 0.505473 FEtid_16 1.89 0.529347 FEtid_17 1.94 0.516392 FEtid_18 1.91 0.523675 Mean VIF 3.28