The influence of the stability and growth pact’s 3% rule on GDP growth for 11 EU countries

The influence of the Stability and Growth Pact’s

3% rule on GDP growth for 11 EU countries.

Bachelor Thesis

By: Aron Daniël van Scherpenzeel Student number: 10173382

Supervisor: Damiaan Chen

Study: Economics and Business Field: Economics and Finance

Date: 07-07-2014

2

4. Model and Methodology p. 10-13

5. Results & Discussion p. 14-19

6. Conclusion p. 19-20

3

In this investigation the impacts of the implementation of the Stability and Growth Pact in 1992 with respect to GDP growth has been looked in to. With the main focus on the budget deficits and the 3% rule, this paper tried to find the link between deficits above 3% and how economic growth in terms of GDP is affected. Using panel data of 11 EU countries a fixed effects regression model was build for annual as well as for 5-year averages data in the period 1970-2013 in order to answer the question. For every model build the before, after and total results were shown that allowed to compare the stages in the used timeframe. To each of the annual and 5-year averages model a dummy variable was created to test deficits higher than 3%. Although the results found evidence for a break in the model in 1992 using the Chow-test, the implementation of the 3% rule could not be justified since the

coefficient of the dummy variable used to test this turned out to be insignificant. This meant that this paper did support a break in the model, but not that budget deficits higher than 3% affected growth in a negative way

4

During the financial crisis that started in 2008 economies worldwide declined and for several years countries faced no or less economic growth until now. As a result financial markets and governments had to reform in order to alter this economic downturn. One of the ways by which the government could do this was by using fiscal policy to provide a stimulus to the economy. However, despite the fact that many countries tried to find a way to increase their GDP again, the countries within the European Economic and Monetary Union were bound to certain regulation policy.

The reason for this regulation is the Stability and Growth Pact (SGP), which is part of the Treaty on the Functioning of the European Union (Relevant Legal). Part of this pact is the corrective arm, which tries to bring excessive deficits down. When countries either have a government deficit exceeding 3% or a national debt that exceeds 60% of GDP (or is not diminishing sufficiently towards this percentage), sanctions will follow in the form of financial penalties or suspension of Cohesion Fund Financing (Stability).

The implementation of the SGP was not surprising, because as Ball and Mankiw (1995) stated in their article: “No issue in economic policy has generated more debate over the past decade than the effects of government budget deficits” (p.95). Budget deficits and resulting government debts have been rising over the past decades and this trend was mostly accompanied by a rising GDP as well (Checherita & Rother, 2010, p. 7). As stated by Tanzi and Schuknecht (1997), the debt as a percentage of GDP rose from 12% to 43% for a group of thirteen industrialized countries from 1913 to 1990. Average debt even hit 79% at the end of the period for the large governments.

Despite the fact that the debate is still ongoing and many possible

consequences have been explained, the precise effects of budget deficits are not unanimously agreed upon among economist. Nevertheless, a large body of direct and indirect evidence indicates that sustained budget deficits will result in a decline in GDP relative to their level in absence of this deficit (Gale & Orszag, 2003). A simple model used in Gale & Orszag (2003) stated that the decline in the budget outlook for the US between 2001 and 2003 would lead to reduction of $2300 per household, holding other things equal. The channel through which this happens are tax

increases, which result in a decrease in household income. This figure shows that the topic has all the reason to be investigated further and how it affects households on a financial basis.

Although theoretical literature exists on the topic of government debt and budget deficits, concerning their effects on GDP growth, the empirical literature is

5

research in this area, as the financial crisis resulted in the public debt-to-GDP ratios soaring (p. 9). In their paper they analyzed the influence of debt on economic growth, but only shortly mentioned the budget deficit. Especially regarding the influence of the 3% rule of the SGP that the EU has implemented in 1992, it would be of interest to know whether the implementation has resulted in a change in the way budget deficits affect GDP growth and if it was for the better. Does this new monitoring benchmark change the influence of budget deficit on economic growth? In light of this statement this thesis will therefore investigate this issue according to the following research question:

“To what extent did the implementation of the 3% rule for budget deficits of EU countries change the effect they have on GDP growth?”.

This thesis will investigate this question by making a regression model using panel data, whereby the influence of the introduction of the 3% rule will be measured. The first two sections of this thesis will summarize the theoretical literature and then the empirical literature. The theoretical literature is discussed to shed light on the step-by-step consequences of the budget deficits on economic growth, whereas the empirical literature is used to show whether the found evidence is consistent with this theory. In the third section the source and data source and the extent to which they are manipulated will be discussed. In section four the model and methodology will be discussed, where there will be outlined how the research question will be answered and through which model building this is done. After this part the results will be shown and discussed. Here the validity and the extent to which the results answered the research question will be laid out and this will lay the foundation for the last part of the thesis, the conclusion. Concluding the thesis this section will review back on the introduction and the research question. It will elaborate to the extent in which this thesis has answered the research question and brought new evidence into this controversial topic within economics.

2. Literature review

Although deficits and their impact is a much-discussed topic, the literature especially in the empirical field is scarce (Checherita and Rother, 2010, p. 9). It should be stated that this thesis distinguishes theoretical and empirical literature form each other and will be discussed separately. This is for the reasons that theory is always a

6

the resemblance between theory and empirical research is.

2.1 Theoretical literature

Checherita and Rother (2010) mention that the existing theoretical literature tends to point towards a negative relationship between debt (and thus deficits) and economic growth (p.9). This statement is based upon the argument that national debt is a burden for the next generations, though which growth is affected (p. 9). Ball and Mankiw (1995) concur with this statement in their paper and use the national savings identity as a starting point. As stated by Ball and Mankiw (1995) deficits have many effects, but they all start from this identity (p. 96).

 National Saving = Private saving + Public Saving

If the government runs a deficit, meaning they spend more than they gain from their revenues, the public saving is negative and thus national saving falls below private saving (Ball and Mankiw, 1995, p. 97). According to the neoclassical view of Ball and Mankiw (1995) however, they believe that this fall in national income is less than the fall public saving and further adjustments are needed to balance the equation again (Gale & Orszag, 2003). The reason for this is that Bale and Mankiw (1995) believe that the effect of a fall in public saving is partly offset by a corresponding increase in private saving (p. 97)1_{. To see how the fall in savings affects the economy consider} two undeniable accounting identities.

 S = Y – C – G (savings identity)

 Y = C + I + G + NX (National income identity, GDP)2 When substituting Y into S, we obtain the following equation:

 S = I + NX

According to Ball and Mankiw (1995) this equation shows the effects of budget deficits. Reconsidering that national saving as a whole went down; this downfall has to be matched exactly by I, NX, or both. When NX falls, more is imported then

1

Example: if there is a $1 tax cut, Ball & Mankiw (1995) believe that consumers will not entirely spend this $1 benefit, but will save part of this. Hence raising private saving. 2 _{Y= National income, C = Consumption, G = Government purchases, I = Investment, NX =} Net Exports

7

currency, bonds, equity, real estate). In any case, when the country becomes a net importer, it also becomes a net exporter of assets (Ball & Mankiw, 1995, p. 98).

Furthermore, the decline in national saving also reduces the supply of loans available to private borrowers (resulting from the overall savings going down). This results in the interest rates being pushed up, and causing households to reduce their investment (Ball & Mankiw, 1995, p. 99). A reaction from the higher interest rates is also that the currency appreciates, since the domestic assets have become more attractive for foreign investors. Nevertheless does the appreciation affect the goods and service market, since foreign goods become cheaper for domestic residents and domestic goods more expensive for foreigners, causing NX as a whole to go down (Ball & Mankiw, 1995, p. 99).

Suppose that the budget deficit continues for a sustained period, than debt stock will build up as a result of the described effects (Ball & Mankie, 1995, p. 99). This is when economic growth (GDP) and wealth start being affected. In the long run production capacity determines the output and production capacity itself is partly determined by capital (p. 99). Since debt reduces investment, the capital grows at a slower rate than it would have otherwise (Ball & Mankiw, 1995, pp. 99-100). If this happens for long enough, Ball and Mankiw (1995) state that this would significantly reduce the economy’s capacity to produce goods and services. As stated earlier, the deficit reduces savings, which must be met by a decline in either investment of net exports. Although controversy exists over which of the two effects is larger, this is not crucial for our discussion here concerning national income (Ball & Mankiw, 1995, p. 101). This is because if deficits crowd out capital then national income falls because less is being produced, whereas if deficits lead to a trade deficit then just as much is produced, but less income from the production goes to domestic residents (Ball & Mankiw, 1995, p. 101). In the end, whether it goes at the expense of I or NX, GDP will end up lower as a result of the budget deficit.

Besides Ball and Mankiw (1995) there is more theoretical literature that points in the direction of a negative relationship between budget deficits and economic growth. One of the points mentioned by both Modigliani (1961) and Ball and Mankiw (1995) is that the deficits are a burden for the next generations (p. 117). This comes from the fact that accumulating debt from deficits is creating a so-called mortgage against future national income. Meaning that in the end this debt has to be paid back, and this goes at the cost of future taxpayers. Ones this happens, more capital will go towards paying back the debt, and this goes at the expense of investments.

8

speculative view on the further consequences of sustained deficits. Gale & Orszag (2010) wrote that the uncertainty regarding the corrective action that is needed to bring the deficits down could ‘’spook” the economy. As emphasized by Truman (2001) in a case study focusing on the US, he stated that a growing debt might lead to a loss in confidence of the economic policies of a country. This could lead

investors demanding a risk premium as a result and making it harder for the US to attract capital, which is detrimental for the economy (Truman, 2001).

There is also an interesting article from Pasinetti (1998), who discusses the treaty as a whole and the parameters within it. He describes in his article the

controversy that prevails around the 3% rule of the stability and growth pact and the 60% debt-GDP ratio (p. 103-104). He discusses the Treaty of Maastricht (which later became the Treaty on the functioning of the European Union) and whether the parameters were relevantly chosen. He argues that the boundaries set (meaning the 3% rule, the 60% debt-GDP and rate of economic growth) were in fact arbitrarily chosen. These would only specify one point in time, where in fact there exist an infinite number of points sharing the same characteristics (Pasinetti, 1998, p. 113). 2.2 Empirical literature

The article by Checherita and Rother outlines the effect of national debt on economic growth. Even though the ECB paper focuses primarily on debt and this paper on deficit, it is still significant importance to this paper as debt and deficits are highly correlated and the one follows from the other. Checherita and Rother (2010) found evidence that debt and economic growth are highly statistically significantly related in a non-linear way (p. 4). Using a panel data from twelve EU countries over a period of 40 years (1970-2010) they found evidence that public debt and economic growth have a concave relationship (p. 22). The average threshold point they found was around 90-100%, meaning that above this point higher debt levels are associated with lower long-term growth (p. 22). What they also found, which is of importance to this paper, is evidence that budget deficits and economic growth are negatively linearly related (p. 23). Overall this paper suggested additional arguments in favor of debt reduction to avoid undermining growth prospects, which is a conclusion that matches the theory of Ball and Mankiw (1995).

Another paper that investigated the relationship between economic growth and deficits is from Adam and Bevan (2005), focusing on developing countries. This paper used panel data from 45 non-OECD countries, from 1970 to 1999. According to their model there is a threshold effect at 1.5% deficit of GDP, meaning that there

9

They state however that this effect vanishes ones more fiscal contraction is used. Another finding from this article is that the threshold involves not only change in slope, but also change in sign (p. 594). Meaning that there is a range of values for which deficit financing may be growth enhancing. This comes from the fact that this paper investigated developing countries that are not on its steady state growth path (p. 594). Nevertheless does the overall conclusion reasonably coincide with the other research on this topic and does it use methods that will be of value to this paper later on.

In a research by Bleaney et al (1999) on the influence of fiscal policy on growth, conducted on a sample of 22 OECD countries, they found a positive regression coefficient for the variable budget surplus when regressed on per capita growth (p. 171). This evidence suggests that not running a deficit will be growth enhancing based upon this research. Martin and Fardmanesh (1990) obtained the same result in their paper, which investigated the impact of fiscal variables on growth. In a cross-sectional sample of 76 developing countries they found evidence that showed deficits work drag on growth. At the same time were expenditure cuts and deficit-reducing tax increases found to be growth enhancing (p. 239).

Based upon a research from Martin and Fardmanesh (1990), Cebula (1995) wrote a paper on the impact of budget deficits on growth for the US. Using an

Instrumental Variable (IV) method, where the variables are split to make a distinction between endogenous and exogenous variables, he regressed data from 1955-1992 and found evidence in his model that is in line with conclusions of Checherita and Rother (2010), but also Ball and Mankiw (1995). The conclusion was that the federal budget deficits over time, reduce the rate of economic growth for in this case the US with statistical significance (p. 250). It stated that major benefits for the US in terms of real per capita economic growth could be yielded if budget deficits were reduced (p. 250). It also found that this would preferably be done without a further elevating of the income tax, as it was observed that the income tax rate and economic growth were negatively related (p. 250). The paper concluded saying that the concern over the US budget deficit was indeed warranted, and that it requires more than ‘’concern’’ (p. 250).

3. Data

In order to investigate the research question a sample of the 11 most developed countries in terms of GDP within central Europe is used, that all have the Euro as their currency. The countries are the following: Germany, France, Italy, Spain,

10

1970-2013 (43 years) is used to make sure that significant results can be yielded and is not restricted by a too short time frame. This setup is almost similar to what

Checherita and Rother (2010) did in their research on the influence of debt, only is their time frame slightly shorter. Almost all the data in this paper is obtained from the AMECO, which is the annual macro-economic database of the European

Commission. Since the European Commission is an independent body that should act objective when publishing these figures, one may assume the data is reliable and correct. The only variable that isn’t obtained through AMECO is inflation, which comes from the data collection site “quandl”, which is a database collection of macroeconomic data. Quandl stated that the inflation data came from the IMF.

A drawback from this data is that it is not complete for every country or for every variable. Especially for Spain, Luxembourg and Ireland some of the data on government revenue, but also budget deficits only came available from 15 years after the start of the time frame. Spain’s budget deficit data was even published later than this and was only available from 1995 onwards. This is a factor that should be taken into account when analyzing the results as this might affect the outcome in a certain way. Also should it be said that before 1991 Germany was called West-Germany. To avoid missing this whole time series for Germany they took over all the values of West-Germany up until this time. Since the regime in West-Germany is the same as when it became Germany, this is a solid step and the data flows well over from 1991 to 1992.

At last there should be a note on the GDP growth and population growth. Since AMECO only published the variables “GDP” and “population”, this data was edited in excel to obtain the growth rate. Since this technique is (new-old)/old, the data loses one value in the beginning, which is the year 1970. For this reason the data time frame used in the annual analysis is 1971-2013. For the 5-year averages the timeframe will be from 1976 until 2013, for the reason that the 5 previous years determine the value of the 6th _{year. This means that the average of 1971-1975} determine the value of year 1976, so the first 5 years are lost in this way. 4. Model and Methodology

The model that this paper will build is a standard fixed effects panel data model that was among other models also used by Checherita and Rother (2010) and Adam and Bevan (2005). It is preferred above standard regression, as this would require 11 regression models for each country and this would be only country specific. The fixed effects model on the contrary will yield only one model that is applicable to the whole

11

this, is it also appropriate in this situation as the fixed effects option controls for unobservable time-invariant effects that might correlate with the explanatory variables in the model. Each country is likely to have an uncontrollable factor that might disturb with the data, which can be anything ranging from cultural factors, race or even the difference in political systems. Using a fixed effects model this country specific factor is accounted for (Torres). A possible drawback in the case of macro economic data is that there could be some type of cross-country dependency between the countries, causing bias in the results. It should be noted that a side effect of the fixed-effect model is that the dependent variable cannot be investigated on time-invariant causes (Torres).

Although it is reasonable to accept the choice of fixed effects, it would also be possible to use the random effects model instead of the fixed effects model. This model would be appropriate if there is reason to believe that some omitted variables may be constant over time, but vary between countries, where others might be fixed between countries, but vary over time (Introduction to panel data). Although when working with country data the intuition that each country has a different

uncontrollable factor is likely, there is a statistical way of showing this.

The Hausman test is used in stata to support the intuitive choice and thus show that the fixed effects model is the best applicable model for this investigation. With this test first the fixed effects and then the random effects model is run for the whole sample period. Then the Hausman tests if the coefficients in the random effects estimator are the same as the ones in the fixed effects estimator. When the p-value is significant there is enough evidence that this difference exists and the fixed effects model can be used, otherwise the random effects model can also be used. The outcomes of the Hausman test are shown in the table below.

Table 1. Results of the Hausman test. ***significant at 1%

Chi2 _27.91

Prob>Chi2 0.0001***

As seen in the table the Hausman test is highly statistically significant and supports the choice for the fixed effects model over the random affects. At a significance level of 1% this shows to be the best applicable model.

12

Growth Pact has affected the way budget deficits influence GDP growth, the main parameter that will be focused on is budget deficits. However, in order to obtain reliable results there has to be a group of control variables that are also included in the model that should make the model as a whole as reliable as possible. For this model the control variables will be chosen in line with the research of Checherita and Rother (2010) and Adam and Bevan (2005), since they also conducted a similar regression model. This means that the model will use country characteristics and fiscal factors as control variables. The model below shows all the factors that are used:

€

GDP _ g = α

+

β

+

β

deficit + β

pop _ g + β

gfcf + β

rev + β

nomin ali + β

inf lation

αI = fixed effect, changing only per country (i)

β0 = constant

deficit = budget deficit as % of GDP pop_g = annual population growth

gfcf = gross fixed capital formation (also called investment), as % of GDP rev = government revenue as % of GDP

nominali = nominal long-term interest rate

inflation = inflation based on Consumer Price Index (CPI)

As shown in the equation it is a two-dimensional model in both the country and time dimension, hence each variable is being determined by both the “i” and “t”. The coefficient β1 will be of most importance to this investigation, as this will indicate the effect of budget deficits on GDP growth.

This model will be build for both annual changes, and 5-year averages of the dependent variables. This is something that was also used in Checherita and Rother (2010) and Adam and Bevan (2005). The reason why using 5-year averages is relevant is because this sheds more light on the long-term implications, rather than the short-term ones. Also does averaging the data help against the endogeneity problem and the possible bias results that come from this. The endogeneity problem in our model could be explained as followed. Even though we want to establish the effect of deficits on GDP growth, it could be the case that GDP growth does in fact

13

independent variables, but not for the dependent variable GDP growth. Once the model is build and shed some light on the influence of the

indicators, especially budget deficit, the Chow test will be used. The Chow-test tests a break in the model, which is in this case the implementation of the 3%-rule and till be tested on significance. Specifically this means that the model will be broken into two parts, prior to the implementation and after the implementation. What the Chow test then does is testing of the two regression models are statistically significantly different from each other and if this break did in fact change the way the models outcome looks. This is relevant for this investigation to see if the implementation of the 3% rule did in fact change the way in which budget deficits (and the rest of the parameters) influence growth.

Since it will also be relevant to see if the threshold rule of 3% is in fact a significant measure, there will be a dummy variable added. This dummy variable will be called Dummy3% (Deficit>3%) and will interact with the variable deficit. This dummy will therefore be able to show the influence a deficit higher than 3% will have on growth. Depending on the sign of the dummy coefficient, a statistically significant dummy will show what the influence of deficits higher than 3% will do to economic growth, but also reflect back on the rule of the Stability and Growth Pact and argue if the rule is a valid rule or not.

When going over the existing theoretical and empirical literature described above it would be expected that within the model, budget deficits would negatively affect GDP growth. Both Checherita and Rother (2010) and Adam and Bevan (2005) found a statistically significant coefficient on this variable. When doing the Chow test it will thus be expected that the model before and after the break will be statistically significantly different from each other. The Stability and Growth Pact resulted in more regulation with especially the 3% rule, which gives reason to expect that a structural break is the case. Also can it be expected that the interaction term with the dummy variable and the deficit variable will be of negative influence on growth. Since the theoretical and empirical literature gives reason to expect that higher deficits are resulting in lower GDP growth.

14

In table 2 the results of all three annual fixed effects regression models are shown. Model 1 will show the results until the implementation of the SGP (1971-1991), model 2 shows the results from after the implementation of the SGP (1992-2013) and at last will model 3 be shown in the last column for the whole sample from 1971 until 2013.

Table 2. Annual results for model 1,2,3. Dependent variable: GDP growth. *** significant at 1%, ** significant at 5%, * significant at 10%

Variable Model 1 Model 2 Model 3

Deficit (%GDP) -0.8306 (0.2170)*** -0.4658 (0.0953)*** -0.3666 (0.0767)*** Pop_g -5.6461 (0.1659)*** -2.0552 (0.9702)** -3.0489 (0.7360)*** GFCF (%GDP) -0.1951 (0.2903) 0.3707 (0.1620)** 0.4929 (0.1277)*** Revenue (%GDP) 0.6882 (0.2314)*** -0.2190 (0.1585) -0.2680 (0.1012)*** Nominal i -0.3883 (0.2905) 0.2691 (0.1582)* 0.1837 (0.1163) Inflation 0.5258 (0.2166)** 0.3981 (0.2636) 0.3636 (0.1271)*** _cons 49.5126 (13.1652)*** 5.8851 (9.0948) 6.7074 (5.6616) R2 _{0.3435 0.2054 0.3103} F-test on model 11.96*** 12.60*** 35.29***

What is seen in the results for model 1 is that the budget deficit has a negative coefficient that is statistically significant at an alpha of 1%. The negative coefficient implies that along with a 1% increase in the budget deficit of one of the sample countries, a GDP growth of -0.83% is the result. This is a finding that is in line with the results found in other research and thus not a controversial finding. What must be kept in mind when analyzing these results is that these are taking over the beginning of the sample where a lot of data was missing. This means that even though some of the coefficients, but also the model as a whole are significant there is still some room for bias. For model 1 there should be 231 observations in total, but due to the fact

15

used the 99 observations where all the data was available. Even though statistical programs are capable of dealing with missing observations, it should be made clear when presenting the results. It is difficult and not of much value to describe the coefficients of GFCF and Nominal i since these are not significant in model 1. These could just as well have a value of 0, meaning there is not much value that can be attached to this part of the results in model 1.

The coefficient of deficits is for model 2 is also negative, but more importantly statistically significant with an alpha of 1%. In this regression if budget deficit

increases with 1% the GDP growth will go down with 0.47%. For this part of the regression there was almost all the data available. From the possible 242

observations there were 239 observations available. This increases the chances of a fair model highly since there were only 3 missing values. The other coefficients are all taking on signs that were also found in the existing literature. Here is can be seen that GFCF has a positive sign, meaning that a rise of 1% in investment increases growth with 0.37%. Looking from an economical point of view this is something that one would expect, since investments stimulate the economy and thus support GDP growth.

At last for the annual panel data sample we have model 3, which includes the complete data set. In this model almost all coefficients are statistically significant and deficit is one of them. When the government budget deficit would increase with 1% the GDP growth would decline with 0.37% in this model. The other variables such as population growth, GFCF and inflation are also significant at an alpha of 1%. Again GFCF has a positive sign with 0.49, meaning that if there is 1% more investment, there is a corresponding 0.49% higher GDP growth rate. The power of this model is the highest of the three and has the most significant coefficients. The reason why nominal interest rate is not significant is probably because for this variable not all the data was available. Another reason could be that nominal interest rates do not affect GDP growth in such a way that it fits in this model. Nevertheless is it important to include this variable in the model as interest rates are of importance to the economy and are thus highly relevant to include.

When conducting the chow-test for a structural break in the model the results are highly significant. According to the conducted Chow-test the date 1992, when the Stability and Growth pact was implemented, is a break in the growth model for the EU countries. When looking at the coefficients of deficits before and after the break in model 1 and 2:

16

regulation the deficits have had a lower negative impact on GDP growth. However the Chow-test is needed to statistically provide proof of a possible break in the mode. The results of the Chow-test are shown below.

Table 3. Results of the Chow-test. ***significant at 1%

F(7,314) 3.71 Prob>F 0.0007***

The results of the Chow-test above confirm that there is in fact a structural break in the model. In a statistical manner the results show proof that before and after the implementation of the SGP and thus the 3% rule a break occurred and the model changed. It can therefore be said that for the coefficient budget deficits, the break in the model had a positive outcome, as GDP growth is less affected now than before the implementation in 1992.

In the table below the results for Model 4,5, and 6 are shown, where a dummy variable is added. Model 4 is the range of 1971-1991, model 5 is for 1992-2013 and model 6 is for the whole range from 1971-2013. This dummy variable tests if having a budget deficit of more than 3% does in fact influence GDP growth in a positive of negative way.

17

Dependent variable: GDP growth

*** significant at 1%, ** significant at 5%, * significant at 10%.

Variable Model 4 Model 5 Model 6

Deficit (%GDP) -1.8265 (0.3259)*** -0.3350 (0.1911)* -0.5190 (0.1644)*** Pop_g -4.8095 (1.5498)*** -2.1155 (0.9778)** -2.9830 (0.7394)*** GFCF (%GDP) -0.2190 (0.2686) 0.3846 (0.1631)** 0.4712 (0.1294)*** Revenue (%GDP) -0.6952 (0.2140)*** -0.2485 (0.1630) -0.2509 (0.1025)** Nominal i -0.4763 (0.2696)* 0.2834 (0.1594)* 0.1782 (0.1164) Inflation 0.5730 (0.2007)*** 0.3735 (0.2657) 0.3815 (0.1382)*** D3*Deficit 1.0630 (0.2741)*** -0.1525 (0.1932) 0.1658 (0.1581) _Cons 50.7097 (12.1794)*** 6.9988 (8.2236) 6.3075 (5.6736) R2 0.4411 0.2004 0.3189 F-test on model 14.13*** 10.87*** 30.41***

The table above shows the results with an added dummy variable testing the

influence of budget deficits of more than 3%. When looking at model 1 it is seen that the coefficient is 1.06, but more importantly it is significant at 1%. In this case it means that when there is a deficit of more than 3%, GDP growth would increase with 1.06% more. Although theory of the implementation of the rule itself would suggest that a higher deficit than 3% would negatively affect growth, this significant coefficient shows the opposite. However, when looking at model 2 and 3, capturing the last period from 1992-2013 and the complete period, there is a different result. Here both the coefficients are insignificant and this suggests that there would be no difference between a deficit higher or lower than 3% and their effect on GDP growth.

Besides the large set of missing data in model 4, is there another possible explanation for the positive and significant number of the dummy coefficient. In the period 1971-1991 the economy as a whole was growing rapidly and when looking at

18

annually. This went accompanied with larger budget deficits as the government had a large share in the growing economy by their investments. Besides, the debate and concern around budget deficits was not as ongoing as it is nowadays. That is the reason why the higher deficits in model 4 were not only slowing growth down according to the deficit coefficient, but also growth enhancing according to the dummy as the economy was stimulated.

Due to the above reasons this investigation attaches more value to model 5 and 6 and bases its conclusions concerning the 3% rule more on these models. This means that even though theory and the SGP suggest that higher deficits than 3% would negatively affect growth, this experiment found no foundation for the 3% rule of 1992 due to the fact that model 5&6 found a insignificant dummy variable. Table 5. 5-year averages for independent variables for model 7,8,9.

Dependent variable: GDP growth

*** significant at 1%, ** significant at 5%, * significant at 10%. Variable (5-year

averages)

Model 7 Model 8 Model 9

Deficit (%GDP) 1.8326 (0.7510)** -0.0550 (0.2819) -0.0324 (1.0345) Pop_g -8.0953 (3.3779)** -0.1754 (1.3292) -2.4886 (1.0345)** GFCF (%GDP) -0.6926 (0.4406) -0.4835 (0.2543)* -0.2157 (0.2005) Revenue (%GDP) -0.6446 (0.5506) 0.0336 (0.1843) -0.1628 (0.1229) nominali -0.9903 (0.4050)** 0.9137 (0.2188)*** 0.5013 (0.1525)*** inflation 0.6435 (0.2894)** -0.6359 (0.3669)* 0.0496 (0.1496) D3*deficit -1.0205 (0.5690)* -0.4343 (0.2398)* -0.3207 (0.2201) _Cons 57.9411 (25.7433)** 11.3287 (9.6179) 16.5253 (7.1174)** R2 _{0.1431 0.0309 0.1733} F-test on model 3.97*** 6.40*** 12.01***

19

that were looked at to get around the possible problem of endogeneity. For all

independent variables the 5-year averages are taken, but as stated before not for the dependent variable GDP growth. Also is the dummy variable deficit>3 added to the equation. Model 7 is for the first period from 1976-1991, model 8 is from 1992-2013 and at last model 9 is for the whole sample from 1976 until 2013.

When looking at the three models it is seen that only for model 7 the

coefficient of budget deficits is significant, however one would expect the sign to be negative. A possible explanation for this is that in the early years of this sample high growth went accompanied with government stimulating the economy as well. This would mean that deficits were used to stimulate the economy to accumulate growth. It is however interesting that the dummy variable in this model 7 is negative with a coefficient number of -1.03. This would mean that if budget deficits were to be higher than 3% within this period, the growth would be 1.03% lower. Although this

coefficient is only significant at 10%, this is an interesting finding that would support the implementation of the 3% rule. The same significance level is found in model 8 for the dummy variable, only this time the coefficient is -0.43. Here a deficit of more than 3% would mean a lower GDP growth of 0.43%. If there is however looked at the models 9, which captures the whole data period, the coefficients is insignificant. Since in model 9 the dummy is not significant and in model 8 and 9 the models are only significant at 10%, it can not be said that based upon the 5-year averages there is found hard evidence for the implementation of the 3% rule. Besides, are only some coefficients significant and is the overall R2 _{lower than in the previous results. This} means that the date of the 5-year averages does not fit the model as good as the annual data did. It must be said that in model 7 and 9 the sample period went down with 55 observations, as the period of 1971-1975 was lost due to the taking of averages.

6. Conclusion.

In this paper the implications of the implementation of the 3% rule with respect to budget deficits and GDP growth have been investigated, as well as the overall implementation of the Stability and Growth Pact. Although the existing literature both theoretical and empirical points towards a negative relation between deficits and debt, this is only partially confirmed through the found results in this paper. Although the conducted regression model for the annual data found evidence of the negative

20

regression could not support this claim for the long-run. The lack of data available in for example model 1,4,7 and for the 5-year averages the reduced sample size did affect the statistical power of the results. Nevertheless was the Chow-test conducted on the annual data highly important to confirm that the implementation of the Stability and Growth pact in 1992 did in fact result in a break in the model. This means that the corrective arm of the EU does in fact regulate the countries in a better way and this can only result in more financial stability. The variables in the model did thus significantly change before and after the implementation, which was overall in a positive direction for the GDP growth. One way or the other the rules established in this treaty thus reduced the way in which budget deficits affect GDP growth

significantly and that they went down overall. Stricter regulation by the SGP limited countries freedom in their spending and this resulted in the above conclusion of a significantly lower deficit coefficient. Even though the insignificant dummy variable (testing the influence of budget deficits higher than 3%) prohibits us from saying the 3% rule has found foundation for its existence in this investigation, the overall deficits affect growth less after the implementation date of 1992 compared to before 1992.

For a follow up experiment it would be of interest to find a manner in which the missing data could be filled in. Especially for the 1st _{half of the data before the} implementation many sets were missing which is not in favor of the regression. Also could it be of value in the future to see if adding more explanatory variables is useful for the growth models.

21

Adam, C.S., Bevan, D.L. (2005). Fiscal deficits and growth in developing countries. Journal of Public Economics.Vol. 89. 571-597

Ball, L., Mankiw, N.G., (1995). What Do Budget Deficits Do?" Budget Deficits and Debt: Issues and Options. Kansas City: Federal Reserve Bank of Kansas City. 95-119

Bleaney, M.F., Gemmel, N., Kneller, R. (1999). Fiscal policy and growth: evidence from OECD countries. Journal of Public Economics, Vol. 74. 171-190. Cebula, R.J. (1995). The impact of federal government budget deficits on economic

growth in the united states: an empirical investigation, 1955-1992. International Review of Economics & Finance. Vol 4, issue 3. 245-252 Checherita, C., Rother, P. (2010). The impact of high and growing government debt

on economic growth: An empirical investigation for the Euro Area. European Central Bank: Working paper. No 1237. 4-28

Gale, W.G., Orszag, P.R., (2003). Economic Effects of Sustained Budged Deficits. National Tax Journal, Vol 56, issue 3.

Introduction to panel data. Princeton University: Data and statistical services.

http://dss.princeton.edu/online_help/stats_packages/stata/panel.htm

Martin, R., Fardmansh, M., (1990). Fiscal variables and growth: A cross-sectional Analysis. Public Choice, Vol. 64. 239-251.

Menaldo, V., (2011). What is endogeneity bias and how can we address it? University of Washington.

Modigliani, F., (1961). Long-run implications of alternative fiscal policies and the burden of the national debt. Economic Journal, 71 (284). Pp. 730-755 Pasinetti, L.L. (1998). The myth (or folly) of the 3% deficit/GDP Maastricht

‘parameter’. Cambridge Journal of Economics. Vol 22. 103-116 Relevant legal texts and outlines. European Commision. Retrieved from:

http://europa.eu/legislation_summaries/institutional_affairs/treaties/treaties_m

aastricht_en.htm

Stability and Growth Pact. European Commission. Retrieved from:

http://ec.europa.eu/economy_finance/economic_governance/sgp/index_en.ht m

Tanzi, V., Schuknecht, L., (1997). Reconsidering the Fiscal Role of Government: The International Perspective. The American Economic Review, Vol. 87, No. 2, Papers and Proceedings of the Hundred and Fourth Annual Meeting of the American Economic Association. 164-168.

22

Truman, E.M. (2001). The International Implications Of Paying Down The Debt. Institute for International Economics. Policy Brief Number 01-7.