MASTER’S THESIS
The Effects of Political Regimes on the
Expenditure Side of the Greek Economy
1960 2000
Georgios Kouinis, s1580035
Thesis Supervisor: Dr. Gábor Péli
MASTER’S THESIS: THE EFFECTS OF POLITICAL REGIMES ON THE EXPENDITURE SIDE OF THE GREEK ECONOMY (1960-2000)
Dedicated to my father Dimitrios (1944-2000), peacefully sleeping in the arms of his homeland
Acknowledgements: I would like to thank my thesis supervisor Dr. Gábor Péli for his creativeness, his useful comments and his motivating guidance, my research
MASTER’S THESIS: THE EFFECTS OF POLITICAL REGIMES ON THE EXPENDITURE SIDE OF THE GREEK ECONOMY (1960-2000)
ABSTRACT:
Any attempt to describe and analyze the economic system of Greece should be conducted after taking into consideration the unique political regime characteristics and changes in each period of time, which in turn may have resulted in fundamental economic changes. The subject of the present thesis is to create a demand determined macroeconomic model for Greece and to incorporate the political regimes into the fundamental functions in order to isolate their effects on the expenditure side of the Greek economy for the period 1960-2000. The formal presentation of estimations and their results is accompanied by a detailed description of the economic and political conditions predominant in each one of the specific sub-periods. This research empirically supports the argument that political conditions and their changes have a significant impact on the expenditure side of the economy, as well as it reaches important conclusions in the direction of explaining the demand side of the Greek economy.
1. INTRODUCTION
The foundations and specifications of the Greek economic system have always been characterized by great variability and fundamental changes throughout a long period of the Modern Greek history. Furthermore, the Greek society, the political system and institutional structure have always been anything but stable during the last 50 years and one can theoretically assume that the political and institutional changes have had a significant amount of impact to the economic conditions and system of the country, since in Greece, a country closer to the theoretical CME (Coordinated Market
Economy) model of organization (Soskice 1999, Hall & Soskice 2001), the economy has always been based on a system of institutional interconnections although it has never been centrally planned, unlike the case of the Central and Eastern European Economies. From a political and economical point of view, Greece has always been considered part of the Western World, being a strategic ally of Western Europe and the United States during the era of the cold war. The political and economical instability of post-war Greece can be better understood, if the main points of regime changes throughout history are highlighted.
The first years after the devastating Second World War, find Greece amidst a civil war between the leftists and the communists on one side and the right-wing
conservative parties and supporters of the monarchy on the other side. The civil war ends in mid- 1949, with the right-wing party of Greece claiming the victory and settling a monarchic regime to the country. In 1952, Greece becomes a member of NATO. The years that follow are characterized by great instability, political revolts, constitutional monarchy and 23 different governments (mainly conservative) within a period of 15 years before the suppression of democracy in April 1967 and the
settlement of a military dictatorship. Unlike the countries under Soviet control, Greece has survived seven years of military junta approved by the western political world. The dictatorship ends by 1974, as well as the monarchic regime, and
Maastricht treaty. The Social-democratic party is, with a small break of three years (1990-1993), on the government for more than 20 years, and in March 2004, the power turns again to the conservative party, which is on the government until today. Meanwhile, Greece has joined EMU (Economic and Monetary Union) in 2001, and has been one of the first countries to introduce the new Euro currency in 2002.
It is therefore necessary, any attempt to describe and analyze the economic system of Greece, to be conducted after taking into consideration these fundamental political changes, which in turn have resulted to fundamental economic and institutional changes. Furthermore, the challenge for a young scientist to describe the economic conditions of his home country and to try to fit real-life data into economic models he has been taught throughout his career as a student of economics is great. The fact that has mainly motivated my study for the current project is that, as we will more in detail discuss in the existing literature review part of the thesis, there has been little effort made in the direction of describing the Greek economic system and specifically its expenditure side by incorporating these important regime changes, which are unique in the case of our country. Although there have been several papers and articles which attempt to develop modern macroeconomic models in order to partially (specific sectors) or fully (total economic equilibrium) describe the Greek economy, most of them are focused on incorporating and analyzing externalities of a different type, e.g. oil crises or currency depreciation. A real country case-study has not been considered for Greece in the context that the current project attempts to investigate. This is a fact that has drawn my attention and curiosity since the beginning of my research.
Main Subject and Research Question
The main subject therefore of this master’s thesis is an attempt to describe the expenditure side of the Greek economy in the context of a macroeconomic model and afterwards an attempt to jointly monitor the effects of the political, regime changes in Greece, by using the fundamental macroeconomic functions that will have been derived. This can be realized by estimating the economic functions of the expenditure side of the Greek economy, based on models that have been developed by researchers on the field of macroeconomics, by using specifically specialized econometric
The whole concept raises many research questions, and if someone would like to isolate a question as the main one, this would be:
“Do political regimes and their changes have any effect on the fundamental functions of the expenditure side of the Greek economy?”
Other questions that are obvious, mainly regard the effects on the economy of specific political regimes and the economic interpretation of the equations and variables used. A few obvious examples are the following: “Did the regime of military dictatorship have any effect on the general consumption levels of Greek individuals?”, “Did the installment of a social-democratic government affect the levels of private investment?”, “In what extent did the entrance of Greece in the EU affect the Greek import and export levels?”
Thus, it is of great interest trying to incorporate various political regimes into the macroeconomic functions of the expenditure side of the Greek economy and trying to measure their effects.
The model that will be developed in order to estimate the functions of the
2. THEORETICAL BACKGROUND
The first definition we need to theoretically clarify is the one of the “expenditure side of the economy”, a term that will be broadly used in the context of this thesis. Existing literature on general and advanced macroeconomics, define the expenditure side as the side of aggregate demand in an economy, which measures the general income by calculating the overall expenses that private individuals and the government generate in a certain period of time (Mankiw 2002, Romer 2001, Demopoulos 1996). It is therefore implied, that the supply sector is of secondary importance in a model that is expenditure determined.
The main behavioural equations that comprise the expenditure side of an economic system, are the consumption function, the private investments function, the import demand, the export demand and the government expenditure function (Sakellariou & Howland 1993, Demopoulos 1996). The government expenditure function is
generally not estimated, since government expenditure (most commonly appears as G in the economic literature) is treated as a single exogenous macroeconomic variable. On the other hand, an additional equation is often introduced, and this is the one of income from taxation (look for example Sakellariou & Howland 1993, Minford et al. 1984).
What is then the fact that makes the expenditure side of the economy so important, and has motivated the research of the current project to be orientated in this direction? This is mainly the fact that an expenditure determined model, determines the level of GDP (Gross Domestic Product) and the other macroeconomic variables directly. So, changes in an expenditure determined model have a direct effect on the
macroeconomic equilibrium (Demopoulos 1996). On the other hand, the supply side of the economy is mainly based on money market equilibrium, which is generally exogenously determined (Money supply – Central Bank). Furthermore and
specifically for the needs of the current project, it is by far more interesting to monitor the effects of the different political regimes on the “demand side” of the Greek
have been regarded as rather immature and regulated (OECD, World Bank, Ericsson & Sharma 1998). Only during the last ten years there have been serious attempts to promote the role of the stock exchange market as a main source of corporate finance and Greece has also recently experienced a serious crisis of monetary and stock exchange markets (2000). The banking system has always been carrying a substantial part of the burden of financing industry’s needs. In other words, the demand side of the Greek economy is much more fertile for research and fits better the purposes of the current project.
Literature Review
Generating a macroeconomic model is a very careful procedure and requires taking into consideration the specific unique aspects of the economy the model is referring to, and the reasons the specific model is developed for. General macroeconomic models, that jointly monitor the equilibrium of the aggregate demand and supply side of the economy, require several equations and a great amount of variables in order to capture the effects on all sectors and aspects of the economy. Specialized
macroeconomic models on the other hand, are developed in order to monitor specific effects on certain sides of the economy (e.g. models for measuring the Phillips Curve, or the technological regime used in the production).
The most broadly discussed general macroeconomic model, is the IS-LM model, initially developed by Sir John Hicks and Alvin Hansen (1937). It is regarded as one of the most influential and widely used (mostly for educational reasons) models that have been ever developed throughout the history of economic science. It is however generally accepted that the IS-LM model, does not take into account specific
important aspects of modern economies (e.g. inflation rate), cannot capture the subtleties of how the economy works and thus do not give reliable estimates. Any attempt of numerical prediction of policy effects with an IS-LM model, is generally unreliable (Schenk, 2002).
Y=C+I+G – The equation of the IS (Investment-Savings) Curve, Y: The national income
Where:
C=f(Yd,r): C=c0+bYd-cr,
C: Private Consumption, Yd: Disposable Income, r: The interest rate I=f(Y,r): I=i0+aY+dr, I: Private Investment
G=Go, Go: Exogenously determined level of government expenditure. Thus in general: Y=f(r) (Source: Demopoulos, 1996). The LM (Liquidity – Money Supply) part completes the model and represents the supply side of the economy.
The fact that the basic linear approach of the IS-LM model is generally regarded as outdated, makes clear that the economists and econometricians have made several attempts in either expanding or altering the basic model, or developing totally new models based on the aggregate supply and demand theory. A more sophisticated approach of the IS-LM model is the “Liverpool model” developed by researchers at Liverpool University (Minford et al, 1983). Again, it is a full system model,
comprising both demand and supply side of the economy. This specific model is regarded as very important in the literature, because it has introduced the use of logarithms in the basic IS-LM model, as well as various new variables (e.g. real exchange rate, volume of world trade). The expenditure side of the economy is here represented by a new IS curve:
y = φd – αr + kθ + k’Δθ – ηe + xWT + εd
Where, y: the log of real GDP, d: the fiscal deficit as fraction of GDP, θ: log of real finance assets, e: real exchange rate, WT: log of world trade, εd: error term
Other interesting examples of “new-generation” IS-LM models can be found in Fane (1985), McCallum (1989), Koenig (1989, 1993), Auerbach and Kotlikoff (1995) and McCallum and Nelson (1997 and 1998).
The general equilibrium approach provided by classical and “new-generation” IS-LM style models is however regarded by the literature as inapplicable to projects, which are focused only on effects on specific sides or aspects of the economy. The main reason for that is the fact that the basic variables of the demand side in these models, are jointly determined by the equations of the supply side as well. This means that the supply side is also significant in the estimation of the relevant coefficients and thus a simultaneous estimation is required. These coefficients obtained, reflect the equilibrium in the economy as a whole and do not isolate the effects of each one of the sides involved (e.g. look Minford et al 1986, Budd et al 1984).
Reviewing the existing literature, we find out that researchers who wish to capture the effects on one side of the economy only (e.g. the expenditure side, similarly to the current project), choose other paths of macroeconomic modelling. They either
separate the economic system in sectors and they estimate each one of the sectors not jointly (simultaneous estimation) but separately, or they isolate the sector they’re interested in and they estimate separately each one of its fundamental equations. Working this way, allows them to capture several effects in specific equations or specific sectors only, mainly by introducing and testing the significance of dummy variables or by estimating the effects on the same dependent variable but by altering each time the independent variables of the equations. This separation allows for more flexibility in determining the effect of random variables or dummies and it is mainly the methodology that, based on the literature, will be followed throughout the project of the thesis, as we will discuss in the Methodology part of this thesis in detail.
characteristics of each one of the economies into the fundamental macroeconomic functions of each sector, mostly by using dummy variables.
The most relevant of these works, is the one of Den Butter (1991), where the main macroeconomic models for the economy of the Netherlands are presented and analyzed. Mainly this work will serve as the basis for the modelling of the
fundamental functions needed for the purposes of the current project for the Greek economy, and mainly because of the explicit way that the expenditure side of the economy is analyzed separately from each other side (in the specific work: Money and Labour Market).
Den Butter provides us with the estimated equations for the expenditure side of the Dutch economy: 2 ln 12 , 0 )} 100 ln( / ) 100 {ln( 3 , 0 ln 64 , 0 ln 2 , 0 lnc const c 1 y r p 1 m e b − + + + + + = − & − 2 ln 12 , 0 )} 100 ln( / ) 100 {ln( 2 , 1 ln 8 , 0 ln 2 , 0 lni=const+ i1 + y − r+ pe + 1 + m − − & v m w w m p p m b const b 0,6ln 1,0(ln 0,6ln ) 0,8ln / ln = + −1+ − −1 + k v m p q p y y m const m 0,6ln 1,0(ln 0,6ln ) 0,3ln / 0,4ln ln = + −1+ − −1 − +
Where: c: private consumption, yb: disposable income, r: long-term interest rate, :
inflationary expectations, m2: m2 demand of money (exogenously determined by money market), b: level of exports, m
e
p&
w: world trade index, pm: index of import prices,
pv: expenditure price index, m: level of imports, y: national product, qk: utilization
rate of capital stock.
to more realistic fitness of the data in economic equations (as argued by all of the authors of the previously mentioned models, e.g. look Musila 2002).
Moreover, specially for Greece there have been several macroeconomic models developed, and here we mention the 4 sector model of Christodoulakis & Kalyvitis (1998) and the expenditure side model of Sakellariou & Howland (1993) as the most profound and relevant ones to the current project. Especially Sakellariou & Howland (1993), make usage of differentials in their equations, a method which will also be considered in the methodology part that follows. It serves furthermore as good example of estimating fundamental equations (e.g. consumption function) separately while introducing specialized dummy variables, a pattern which will be extensively followed in the procedure of the current thesis.
Incorporating political regime effects in macroeconomic equations, although has never been broadly discussed and presented for the case of Greece, has been however made for other countries. A very interesting example is the work of J. Svejnar (1981), where the effects of Hitler’s dictatorship on the relative wage of unions in Germany, are captured in the form of dummy variables, in a model which follows the
logarithmic pattern as well.
3. METHODOLOGY
Any discussion on the economic system in Greece requires first that a theoretical model is developed, tested and estimated and on a second stage that the results of the estimation are presented, analysed and interpreted economically. In this section we will review the methodology that has qualified for application to the model as well as the details of the data and the variables. The model will follow the pattern of the reviewed literature, and its justification and step-by-step construction will also be presented in the current part.
Step 1: Data and Limitations of the Study
The data sample selected to be examined and applied to the econometric
methodology includes 41 annual observations for basic macroeconomic variables of the Greek economy for the years from 1960 to 2000. I have decided not to include the observations for the years after 2000, since in 2001 Greece joined EMU and the macroeconomic policy (mainly the monetary policy) is exogenously controlled by the European Central Bank (ECB) and therefore key indexes, such as the interest or the exchange rate, that are endogenously examined in the theoretical model, are
transformed into exogenous variables and their values are harmonized throughout the EMU region. Furthermore, there is a huge lack of formal measuring and archiving for the decades preceding the 60’s, so the best possible timeframe that includes data both justified and consistent for the selected modelling is the period 1960-2000.
collected have been checked for compatibility between them and have been transformed in order to be able to be compared on a single measurement basis.
Step 2: Choosing Functions and Variables
The second step in building the model is to define the functions that our model will contain, as well as the variables that will serve as dependent and independent in each one of them. Based on the theoretical background, the demand side of the economy (in the form of expenditure) is described by four behavioural functions, namely the consumption function, the private investments function, the import demand and the export demand. Following Sakellariou & Howland (1993), we will add to the model a fifth function, the government expenditure function, which for the purposes of the current project is very interesting to be estimated, since it is of major importance to relate the governmental (public) subsidies with the different political regimes in Greece. Interesting conclusions, regarding the budget balance, can be reached by examining such a relationship. We will furthermore take a closer look in each one of these functions and their variables.
The Consumption Function:
The consumption function represents the relationship between the levels of private consumption and several other important macroeconomic variables. It is more commonly related to the levels of disposable income and interest rate. Thus, the dependent variable in this function is:
cons: the level of private consumption in real prices (billions of Greek Drachmas, GRD).
Additionally, following Den Butter (1991), Bucevska (2003) and the majority of the papers reviewed, we choose to relate the levels of private consumption with the following independent variables:
cons-1: the levels of private consumption of the previous year. It implies that the
levels of consumption made in the previous year, influence the decision of private individuals for consumption expenditures in the current year.
GDP (Gross Domestic Product) after the deduction of the income for the government from direct taxation (regarded as exogenous).
rp: a variable named “real interest rate”. It is calculated by Den Butter (1991) as following: ) 100 ln( / ) 100 ln( + + = rs pe rp & Where:
rs: the short-term annual interest rate provided by the Central Bank of Greece and the monetary authorities. It is important to make a distinction between this general rate and the nominal bank deposit interest rate of each year, which reflects the yield on current bank savings. We select the annual rate because the model is based on annual and not quarterly data, effects and observations.
1 −
= p
p&e & : inflationary expectations, where p&: the inflation rate as published by the
Greek ministry of National Economy. p& denotes the inflation rate of the previous −1
period.
md: the levels of money demand. It is the value of M4 money supply minus the value
of M1 money supply, in order to reflect the amount of money held in the form of
annual bonds, deposit notes, REPOS and other kinds of similar financial investments. The levels of money demand are considered to be exogenously determined in the current model.
The Private Investments Function:
The private investments function represents the relationship between the levels of private investment and other macroeconomic variables. It is more commonly related to the general levels of income and interest rate. According to economic theory, an increase in income will encourage higher investment, whereas a higher interest rate may discourage investment as it becomes costlier to borrow money. The dependent variable is thus:
Following economic theory and partially based on the formulation of Den Butter (1991) and Sakellariou & Howland (1993), private investment is related to the following independent variables:
rp, md plus:
inv-1: the levels of private investment of the previous year. It implies that the levels of
investment produced in the previous year, influence the decision of private individuals for investing in the current year.
y: General Income. It is taken as the Gross Domestic Product (GDP) of Greece in real prices (billions of GRD)
The Exports Function:
The exports function relates the amount of exports of a country, with several other macroeconomic variables, in order to explain their value and initiative. Thus the dependent variable is:
expo: the value of exports in real prices (billions of GRD)
In the current model it is related to the following independent variables:
expo-1: the value of exports of the previous year. It implies that the levels of exports
made in the previous year, influence the decision (and thus value) of exporting in the current year.
e
s&
: the expected exchange rate change, it is calculated by Den Butter (1991) as follows:1 −
= s
s&e & and
1 1)/ ( 100 − − − = s s s s&
Where s: the exchange rate of GRD versus the US dollar, provided by the International Monetary Fund (IMF). (GRD per USD).
s&,s&−1and s-1 notate the exchange rate change, the exchange rate change of the
yg: a variable named “income of foreign country”. According to economic theory, the income of the foreign country affects the volume and the value of exports of the home country (Leventakis, 2003). As “foreign country” we define the main export partners of Greece during the examined period 1960-2000. Their income, is an additive relative weight of their incomes, according to the importance of each one of the export partners. The main export partners of Greece and their respective ratios are the following:
Germany 12,6%, Italy 10,5%, UK 7%, France 4,2%, Rest of the World (OECD) 65,7% (Source: The Heritage Foundation)
The variable yg is then calculated as follows:
Ygt=0,126(GermanGDP)t+0,105(ItalianGDP)t+0,07(UKGDP)t+0,042(FrenchGDP)t
+0,657(OECDGDP)t.
Where t denotes each one certain year t (between 1960 and 2000) and the GDP of the rest of the world is calculated as the gross income of OECD countries (following Sakellariou & Howland 1993), minus the income of the countries already included in the equation. So, we create a time-series for the variable yg for the period 1960-2000. Value of GDP is in millions of USD and is provided by OECD.
The Imports Function:
The imports function relates the amount of imports made by a country, with several other macroeconomic variables, in order to explain their value and initiative. Thus the dependent variable is:
impo: the value of imports in real prices (billions of GRD)
In the current model it is related to the following independent variables:
e
s&
, y plus:impo-1: the value of imports of the previous year. It implies that the levels of exports
It is implied by economic theory that the value and volume of imports is related to the income of the home country (in this example Greece), unlike the case of the exports. (Leventakis, 2003)
The Government Expenditure Function:
The government expenditure function is a relationship between the value of the expenditure or public investments made by the state, and other macroeconomic variables which we will furthermore define. The dependent variable of the function is thus:
gov: the level of public investments originated by the Greek government in real prices (billions of GRD).
In the current model, it is related to the following independent variables: gov-1: the level of public investment of the previous year.
t: the level of direct taxes, as published by the Greek statistical authorities (billions of GRD). It is mainly regarded as exogenous variable in the literature, and it is assumed to have an effect on the level of governmental expenditure, since it is considered as the main source of income for the government.
ctpn: the level of transfer payments (billions of GRD), it also considered to be a secondary source of income for the government (Demopoulos, 1996).
General Note for all variables mentioned: The variables are given in real prices (base year 1995), considering the relevant deflators published by the Greek Ministry of National Economy (following Christou, 2001)
The Dummy Variables:
KING: a dummy variable representing the parliamentary monarchic regime that was predominant in Greece for the period 1960-1966. The dummy takes the value 1 for this specific period of time, and the value 0 for all other years after 1966.
DICT: a dummy variable representing the military junta regime imposed to Greece during the so called “dark seven-year” period of Modern Greek history. It takes the value 1 for the years 1967-1974, and the value 0 for all other years.
CONSERV: a dummy variable representing the period of time when the
conservatives (New Democracy) are in power, under a parliamentary democratic regime. It takes the value 1 for the years 1975-1980 and 1990-1993, and the value 0 for all other years.
SOCIAL: a dummy variable representing the period of time when the
social-democrats (PA.SO.K) are in power, under a parliamentary democratic regime. It takes the value 1 for the years 1981-1989 and 1994-2000 and the value 0 for all other years.
EU: a dummy variable representing the period of time when Greece has been a member of the European Union. It takes the value 1 for the years 1981-2000 and the value 0 for all other years.
Step 3: Testing for Stationarity and Cointegration
Dealing with time series data needs a more careful approach than dealing with any other random variables. Regressions based on time series, can produce spurious results, with no actual meaning or interpretation. This danger of obtaining apparently significant regression results while in reality they are spurious, appears when using non-stationary series in regression analysis (Hill, Griffiths & Judge, 2001).
pattern. Hill, Griffiths & Judge (HGJ, 2001), provide us with various examples of random walks from non-stationary time series.
Usually, the first indicator of non-stationarity is the very slowly declining
autocorrelations between the observations within a time series. The stationarity or not of a time series can also be tested with a formal unit root test using an econometrics software package. Let’s consider a regression of the type: yt =a+ρyt−1+vt, where yt
is a random variable, and vt is a disturbance term. We say that the variable yt has a
“unit root”, when the coefficient ρ is equal to one. Thus, when
ρ=1⇒ yt =a+yt−1 +vt , and the random walk process yt is non-stationary. The proof
is presented in HGJ (2001). Using EViews 5.1, we can easily perform unit root tests to our selected time series. The most commonly used unit root test is the formal test created by statisticians Dickey and Fuller, who have also developed critical values for the presence of a unit root (random walk process). Formal algebraic depiction of a Dickey-Fuller test can be found in HGJ (2001).
The null hypothesis of a Dickey-Fuller test is the following: H0: the time series has a unit root (is non-stationary).
Against the alternative: H1: the time series is stationary
Performing the test, we obtain ADF (Augmented Dickey-Fuller) values for each one of our time series. If the ADF statistic is within the critical values set (lower), we reject hypothesis H0 and we conclude that the time series is stationary. Moreover,
spurious regressions exhibit a low value of the Durbin-Watson (DW) statistic and a high R2. When R2>DW, the regression has a high probability of being spurious (HGJ, 2001).
As we can see from the tables, the natural logarithms (ln) and the differentials of the natural logarithms (dln) of the variables, have also been tested for stationarity. Note that the differentials are calculated using the econometrics software as follows:
) ln( ) ln( ) ln( ) ln( ) ln( ) ( ) ( 1 1 1 − − − = Δ = − = − = Δ = t t t t t t t t y y y y y y d y y y y d
Where t denotes time and y is a random variable or time series. (Source: Cordina, 1997)
A first conclusion is that, by applying logarithms and logarithmic differentials, the image we get for stationarity is clearly improved. Variables in logarithmic form, are more probable to be stationary in our model, than the original variables. This is one of the reasons why in literature, the usage of variables in logarithmic forms is more common. (Sakellariou & Howland, 1993).
Furthermore, even if non-stationary time series should generally not be used in regression models, there is a single exception to this rule. This is the case when the variables of the regression are cointegrated. The meaning of cointegration is that when two or more time series share similar stochastic trends, but in fact their difference is stationary, they actually never diverge too far from each other (HGJ, 2001). Thus, even when two or more variables are non-stationary, when their
difference (residual) is stationary (they are therefore cointegrated), there is a possible regression, without the danger of producing spurious results.
Again, there are formal procedures to test for cointegration between time series. The most common way, is to run the regression between the selected variables and then perform a DF unit root test on the residuals of the regression. Note that the
relationship of the non-stationary but cointegrated variables which is estimated via a least squares regression is referring to the long-run period (HGJ, 2001). This
distinction between the long and the short run is better displayed in the macroeconomic model of Musila (2002).
residuals of the regressions. The results are presented in the Appendix 2. We perform a simple OLS (as suggested by HGJ, 2001) regression including a constant term between the following non-stationary (at 5% significance level) variables for the following functions:
Consumption function: ln(cons) ln(cons-1) ln(yd) rp ln(md)
Investments function: ln(inv) ln(inv-1) ln(y) rp ln(md)
Exports function: ln(expo) ln(expo-1) s&eln(yg)
Imports function: ln(impo) ln(impo-1) s&e ln(y)
Government Expenditure function: ln(gov) ln(gov-1) ln(t) ln(ctpn)
Note: The variables with the subscript (-1), produce the same stationarity results as the original ones. The variable rp is already in logarithmic form as we have seen.
The null hypothesis of the test is: H0: the least square residuals are non-stationary, the
variables are not cointegrated.
Against the alternative: H1: least square residuals are stationary, the variables are
cointegrated.
We reject the H0 when the ADF statistic is lower than the critical values (HGJ, 2001).
Examining the results of the unit root tests, we conclude that the variables in all the functions are cointegrated, since in all cases (look Tables 7-11), the ADF value we obtain is below the border of 10% significance. In all but one (consumption function) cases it exceeds the borders of 5% and 1% significance as well. We therefore
conclude that there is indeed a way to perform a regression between the selected variables for each one of the functions, without the danger of producing spurious results.
Relationships between variables that are proved to be non-stationary but
Step 4: Choosing an Equational Form
Reviewing the existing literature, the selected variables and functions and the results of the formal tests applied to them, we are now in position to give an equational form to the functions we will furthermore estimate and examine.
First of all, it should be clear by now that a linear relationship between the variables for all functions is not applicable and the main reason for that is the fact that such an approach has already been tested as the basic IS-LM model visited in the literature review, and it has failed to produce noteworthy and economically significant results (Vercelli, 1999). Furthermore, as we have seen, modern macroeconomic modelling is based on logarithmic relationships between the variables. It is generally accepted that the log-log model, which produces non-linear curves as plots, fits better the data used for economic models, and mainly because of its “constant elasticity” property (HGJ 2001, Christou 2001).
Generally, the log-log model is described as follows:
t t t x e y )= + ln( )+ ln( β1 β2 , with a slope of t t x y 2
β and a constant elasticity equal to
β2, which is very convenient for demand determined economic models (Source:
Christou, 2001)
Where: yt and xt are random variables, β1 and β2 are the regression coefficients and et
is the error term.
In the context of the above described log-log model and following Den Butter (1991) and Sakellariou & Howland (1993), our functions get the following form:
The consumption function:
The private investments function: ) ln( ) ln( ) ln( ) ln( ) ln( )} 100 ln( / ) 100 {ln( ) ln( ) ln( ) ln( 5 1 4 3 1 2 1 5 1 4 3 1 2 1 md c rp c y c inv c c inv md c p rs c y c inv c c inv e + + + + = ⇒ ⇒ + + + + + + = − − − − & Identities: rp=ln(rs+100)/ln(p&e +100) p&e = p&−1
The exports function:
e s c yg c c c1 2ln(expo-1) 3ln( ) 4& ) expo ln( = + + +
Identities: s&e = s&−1
s&=100(s−s−1)/s−1
The imports function:
e s c y c mpo c c mpo) 1 2ln(i -1) 3ln( ) 4& i ln( = + + +
Identities: s&e = s&−1
s&=100(s−s−1)/s−1
The government expenditure function: ) ln( ) ln( ) ln( ) ln(gov =c1+c2 gov−1 +c3 t +c4 ctpn
Where, ci (i=1,2,3,4,5), the relevant coefficients of the variables.
Now that we have a formal description of the functions, we proceed by describing the econometric techniques that will be used for the estimation of the model.
Step 5: Applying an Econometric Technique
the fundamental functions separately. We thus reject the alternative methodology of estimating the system as a whole, for the obvious reasons described in the literature review and above.
Taking the consumption function for example, and following Svejnar (1981) and Sakellariou & Howland (1993), we apply the dummy variables as follows:
dDUMMY md c rp c yd c cons c c cons)= + ln( − )+ ln( )+ − + ln( )+ ln( 1 2 1 3 4 1 5
Where d denotes the dummy variable coefficient and DUMMY represents each one of the dummy variables created (KING, DICT, SOCIAL, CONSERV, EU).
A joint estimation of all 5 dummies in one function, is in this case not applicable mainly due to the fact that we are interested in the effects of each one dummy separately. Moreover, the dummies refer to different time periods and it thus
irrelevant trying to estimate joint effects for the same timeframe. Finally, according to Kennedy (1986) and Suits (1984), such forms of equations with multiple dummy variables produce an undesirable effect of multicollinearity which has to be treated alternatively. We therefore estimate each one of the functions 5 times separately, applying each time a different dummy variable.
However, the estimation of the fundamental functions of our model is not sufficient to be conducted by applying standard regression techniques, because as we have seen we are dealing with non-stationary and cointegrated time series and this approach may produce spurious results. The functions should be therefore estimated on the pattern of the error correction model.
The error correction model (more in detail presented in HGJ, 2001), can be described as: t t t t a a y x v y = + − − + Δ 1 2( −1 β1 β2 −1)
Where Δyt represents the changes on a non-stationary variable yt, t denotes time, xt is a
non-stationary variable and vt is again a disturbance term (shock). This change of the
variable yt is proved to be stationary. The error correction model can be estimated as
We use ordinary least squares to estimate the cointegrating relationship:
t
t x
y =β1+β2 and then we use the lagged (by one period) residuals of the regression: as right-hand side variable in the model, estimating it with a second least squares regression. We thus “correct” the model by incorporating the stationary changes Δy
1 2 1 1 1 ˆt− = yt− −b −b xt− e
t in the form of the lagged residuals to the function.
To display empirically how the “error correction model” method solves the problem of spurious results, we proceed with an estimation of one of the functions of our model. We take the consumption function:
) ln( ) ln( ) ln( )
ln(cons =c1 +c2 cons−1 +c3 yd +c4rp−1 +c5 md , and we run an OLS (Ordinary Least Squares) regression , as suggested by HGJ (2001) in EViews 5.1. The idea that the OLS estimator is the preferable one for small samples (less than 50), is also supported by Judge (1999) on his Monte Carlo studies. The results of the regression are presented in Appendix 3. On Table 12, we observe the results of an ordinary regression, without using the error correction model provisions. All of the variables are significant on a 5% level, since the p-values we obtain are less than 0,05. We can however clearly see that the R2 value is very high and even higher than the estimated Durbin-Watson statistic. We get: R2=0,99992>DW=0,646190. Thus, the results we get from the regression have a great probability of being spurious. The corellogram we obtain for the residuals is presented on Table 13. We can observe very slowly declining autocorrelations, which is an indicator of non-stationarity.
The next step is to create a series for the residuals of the initial regression and then a series for the lagged (for one period) residuals, as the theory of the error correction model suggests. We thus construct the series RESIDCONS and RESIDCONS1 for the residuals and the lagged residuals respectively. Then, we run an OLS regression on the following corrected equation:
1 6 5 1 4 3 1 2 1 ln( ) ln( ) ln( ) ˆ ) ln(cons =c +c cons− +c yd +c rp− +c md +c et−
Where the lagged residuals eˆt−1 are represented by series RESIDCONS1.
On Table 14 we observe the results of the new, corrected regression. All the variables are again significant at a 5% level. The R2 value is now 0,999956 and the DW statistic is DW=1,784372. Thus: R2<DW and there is no clear sign that the regression
autocorrelation (thus stationarity) and that the series is cointegrated at 1% level of significance.
We now conduct the same corrected regression, including a dummy variable. We select for example the dummy DICT. The equation we estimate is the following:
dDICT e c md c rp c yd c cons c c cons)= 1 + 2ln( −1)+ 3ln( )+ 4 −1 + 5ln( )+ 6ˆt−1+ ln(
The results of the regression are presented on Table 17. We once again observe a high R2 value (R2=0,999958) but an even higher DW statistic (DW=1,8411).
We get R2<DW and thus there is again no clear sign that the regression has produced spurious results even after the insertion of the dummy to the model. The corellogram of the corrected regression with dummy is presented on Table 18, and we see no signs of autocorrelation, thus no initial sign of non-stationarity. The unit root test of the residuals of the corrected regression with dummy, are presented on Table 19. We once again observe existence of cointegration between the variables at a 1% level of significance. Note that the dummy variable DICT is significant at a 10% level.
We therefore conclude that the results have been drastically improved with the application of the error correction model in estimating the consumption function.
Had our time series been non-stationary and non-cointegrated at the same time, we would have worked with an alternative method of estimation, and this is by running regressions on the actual differentials of the variables, following the dln(yt) or d(yt)
pattern we have earlier discussed (based on HGJ, 2001). However, for the current conditions of our time series, such an approach is finally inapplicable.
We will proceed by mentioning some further formal econometric tests that will be applied to our functions in order to ensure that the chosen model is able to fit the data and to explain the relationship between the variables. As we have seen, the R2 value of the regression is an important measurement of determining the “fitness” of the data to the selected functional form, however it is not the determining factor of the
regression equation but as discussed, it should always be compared to the DW statistic. A low R2 value does not necessarily mean a failure of the model.
All the variables of the model and the dummies as well, will be tested for
significance on a 5 and 10% level. Any conclusions on significance of the variables will be based on the p-value that will be obtained by EViews. E.g. a p-value p<0.05 for a variable’s coefficient, suggests that the examined variable is statistically significant on a 5% level. Here, we should note the difference between the statistical and the economic significance of the variable. A variable could be statistically significant, but the effects that it produces are of minor importance in the economic function under examination. (E.g. an independent variable x with a p-value 0,00003 and a coefficient 0,000004 is statistically significant, but economically unimportant since it produces minor changes in the value of the dependent variable y in the following regression: y =c+0,000004x).
Furthermore, the residuals of each one of the regressions will be tested for normality. A formal test for normality is included in EViews and its results are obtained by reviewing the Jarque-Bera statistic and its relevant p-value. This method, tests the hypothesis H0: the errors are normally distributed, against the alternative H1: the
errors are not normally distributed. Obtaining a p-value for the Jarque-Bera statistic that is greater than the 5% level of significance (p>0,05), suggests that we fail to reject the null Hypothesis, and then we conclude that the residuals are normally distributed. Additionally, for a normal distribution the kurtosis value is 3 and the skewness is around 0. But why is it important that the residuals are normally
distributed? This is mainly because the hypothesis tests and interval estimates for the coefficients we obtain, rely on the assumption that the errors (residuals) of the regression are normally distributed. Thus, when choosing an equational form, it is desirable to create a model in which the errors are normally distributed (HGJ, 2001).
squares estimator are incorrect. The existence of heteroskedasticity can be formally checked via a Goldfeld-Quandt test and EViews can provide us with White-corrected coefficients in case the response of the test is positive. The completion of the formal testing of the model will lead us to the core part of the master thesis, which is the derivation, the interpretation and the analysis of the results obtained by following the methodology reviewed.
4. RESEARCH ANALYSIS & EXPECTATIONS
The second and most important part of the master thesis project will be conducted after all regressions and estimations will have been completed and the results will have been obtained. The significance of the dummy variables set, will mainly guide the course of the analysis on the effects. Based on the main research question
presented in the beginning of the thesis, the formal hypothesis testing in each one of the fundamental macroeconomic functions examined has the following format:
H0: The dummy variable is statistically significant
H1: The dummy variable is statistically not significant
(Where the term “significant” has been defined in the previous section according to the theory)
After the regression results are obtained, we are mainly interested in the significance and interpretation of the dummy and not of the other independent variables, since as we have seen, the dummies reflect the effects we intend to measure. The
interpretation of the coefficients we obtain for the other independent variables in each one of the fundamental functions of the model has a secondary role, and mainly serves the verification of the relationship we have chosen for the variables.
Specifically for the independent non-dummy variables: Let’s assume for example that we accept the null hypothesis H0: the independent variable inv-1 is statistically significant
We therefore conclude that the selected variable inv-1 has a certain effect on the
will also be dependent on the magnitude of the relative coefficient obtain. This analysis and the formal certification of the interdependence between the selected variables, is of great economic importance and will be presented in the analysis of the results for each one of the functions. The formal conclusions however of the main research question of the thesis, will be based on the formal testing of the dummy variables.
As an example, we reject the null hypothesis:
H0: the dummy variable KING statistically significant in the consumption function
We reach the conclusion that the parliamentary monarchic regime that was
predominant in Greece during the period 1960-1966, had no significant effect on the private consumption levels of Greek individuals.
We have briefly reviewed the pattern on which our analysis will be based, after the concrete results of the regressions will have been obtained. The expectations of such a research project can never be solid and one should be very careful trying to foresee the results of a formal estimation testing procedure. We can generally argue that we expect significant results regarding the coefficients of the non-dummy independent variables, since their selection and formal representation has been based on already tested relationships from the literature. We can however obtain different results for the special case of Greece.
Regarding the dummy variables, and based on the theoretical review of the diverse political and economic system of Greece during the period 1960-2000, we can generally argue that we expect negative effects on the fundamental functions of the expenditure side of the economy during the “era of political instability and insecurity” 1960-1974. These results are mostly expected for the private consumption and
way. Furthermore, the examination of the economic effects of each one of the different governmental schemes, i.e. conservatives or socialists, under the general regime of parliamentary democracy is expected to provide us with different results. Our research and formal testing intends to provide us with evidence in the direction of supporting or rejecting the claims above. The research results are extensively
5. RESEARCH RESULTS
Following the methodology and applying the econometric techniques reviewed in the previous sections, we are now in position to present and discuss the results obtained. First we will examine the outcomes of the estimations for each one of the fundamental macroeconomic functions of the expenditure side of the Greek economy for the years 1960-2000 and then we will attempt a connection of these functions with the respective political regimes, which were predominant in each one of the examined periods. The presentation of the political regime effects on dependent variables representing the expenditure side of the Greek economy will thus be conducted in time periods, each one reflecting a certain dummy variable – political regime.
Note: For performing the estimations, the software package Eviews 5.1 by QMS has been thoroughly used in the context of the reviewed methodology. The formal outcomes of the estimations are presented in sections 2 and 3 of the Appendix.
The Consumption Function
The formal representation of the Greek consumption function for the period 1960-2000, is estimated as following: 1 1 1) 0,549ln( ) 1,155 0,07ln( ) 0,688ˆ ln( 548 , 0 869 , 0 ) ln(cons =− + cons− + yd − rp− − md + et− (0,0063) (0,0000) (0,0000) (0,0003) (0,0040) (0,0000) (p-values of t-statistics in brackets)
As we observe on Table 1 in Appendix 2.1, all of the selected independent variables are statistically significant on a 1% and 5% level. We moreover observe a high R2 value and a normal distribution of the residuals, indications that the model we have chosen fits well the data. (Note that the common usage of R2 statistic is just an
indication and not a rule of thumb).
The sign and the magnitude of the effect of each one of the independent variables, is measured by their respective coefficients estimated. A closer look on the consumption function reveals an “incompatibility” between textbook economic theory and the empirical evidence provided by Greek data. We can see that the level of the lagged real interest rate (rp-1), has a significant negative effect on the levels of private
consumption, although the theory suggests a positive relationship (e.g. see
Demopoulos 1996). This can be explained if we consider the fact that the real interest rate in Greece has been exogenously controlled by the Greek authorities and has not been subject to the market powers of demand and supply, while the levels of
consumption have been constantly growing during the examined time period, positively related to the growth of real GDP. There have therefore existed certain periods in Modern Greek economic history where the controlled real interest rate has followed a declining course, contrasting the steadily growing course of private consumption (Sakellaropoulos et. al. 2004). This effect is mainly observed for the years prior to 1975 and has reached its peak during the period of military dictatorship (1967-1974). The general relationship between the real interest rate and private consumption is thus estimated to be negative. (Referring to the literature review, Den Butter (1991) similarly finds a significant negative relationship between ln(cons) and rp-1 for the Netherlands).
We are moreover pleased to empirically support the argument of a positive relationship between the levels of disposable income and prior levels of private consumption with these of current levels of private consumption, results that are in line with the model estimations of Sakellariou & Howland (1998) for the period 1960-1985 of the Greek economy.
The relationship between the money demand and the levels of consumption is also negative, but this negative effect is not strong. Both levels of M4-M1 money demand
Carassava 2002) . This can be clearly observed on Table 3. We thus have a slight different course of the logarithms of these two variables.
The Investments Function
The investments function we acquire from our estimations for Greece is formally represented by the following function:
1 1 1) 0,148ln( ) 0,723 0,198ln( ) 0,289ˆ ln( 575 , 0 935 , 0 ) ln(inv = + inv− + y − rp− + md + et− (0,549) (0,006) (0,499) (0,681) (0,098) (0,252)
In this case, we observe that most of the variables we have chosen as independent for our function are proven to be statistically insignificant. We obtain however a high R2 value and a normal distribution of the residuals.
We thus have empirical evidence, according to the chosen model, that there is indeed a statistically significant positive relationship between the levels of money demand and prior investment, and these of current private investment on a 10% and 1% level respectively. The private investment that Greek individuals generate each year seems to have a positive effect on the amount of investment they generate the following year. Moreover, the level of private investments in logarithmic form seems to follow the same trend with the logarithm of M4-M1 money demand, and mainly after the
beginning of the 80’s as we observe on Table 8.
Private investments generally follow a growing trend in real prices, despite the fact that during some specific time periods they have declined. As we will more in detail later discuss, especially after the restoration of Modern Greek parliamentary
Another remarkable result is that although we estimate a positive relationship between Greek GDP and the level of private investments (given by coefficient c3 =
0,148), this relationship is statistically not significant for the examined period. Both trends of Greek GDP and private investments (in logarithmic forms) are plotted as group on Table 7.
The Exports Function
Our model, estimates the formal representation of the export function of Greece for the period 1960-2000 as follows:
ln(expo)=−1,993+0,859ln(expo-1)+0,439ln(yg)+0,004s&e+0,153 eˆt−1
(0,0003) (0,0000) (0,0001) (0,044) (0,385)
We can see that all of the independent variables we have selected to incorporate to the function, are proven to be statistically significant in a 1% level, except from
(the expected exchange rate change) which is statistically significant on a 5% level. This is an expected and pleasant result, since it indicates that the choice of variables while building the model has been reasonable. In combination with the positive results of the formal Jarque-Bera normality test for the distribution of the residuals and the high R
e
s&
2 value, we can assume a decent choice for the functional mode of
estimation.
following. Prior to 1973, Greek exports have included intermediate and capital goods (such as cement, aluminum and nickel) but after 1974 and the restoration of the Greek democracy, the exports have been mainly based on traditional industrial goods such as clothes, leather goods and beverages (Sakellaropoulos, 2004). Later on we will
examine if this transformation of the political regime had any significant effect on the Greek export levels. The course of Greek exports (in logarithmic form) throughout time can be observed on Table 13.
A noticeable result from our estimations is the significance of the variable Yg, which we have calculated specially for the case of Greece and have named “Income of Foreign Country”. It is thus empirically proven by our model that this variable has a significant positive connection to the levels of Greek exports. The more the income (GDP) of the exporting partners of Greece grows, the more their demand for imports from Greece becomes greater and the more the Greek level of exports grow, just as expected according to the theory. This relationship is depicted on Table 12.
We moreover have a significant positive relationship between the export levels and the levels of the expected exchange rate growth, which means that whenever
depreciation (rise of the exchange rate) is expected, the level of exports is also rising. This effect however is of minor economic importance since the obtained coefficient is very low. This relationship, which is in line with prior researches for other countries (e.g. Beenstock et. al. 1994), certainly holds for Greece for the examined period
The Imports Function
Following the reviewed methodology, we receive the following estimation for the Greek imports function for the examined period:
98 , 0 002 , 0 ) ln( 127 , 1 ) ln( 04 , 0 234 , 2 )
ln(impo =− − impo−1 + y + s&e+
1
ˆt− e (0,0000) (0,784) (0,0000) (0,0248) (0,0000)
independent variables are statistically significant on a 1% level and only the variable is significant on a 5% level, without however representing an economically significant result. The choice of variables can once again be regarded as successful.
e
s&
Unlike to our prior expectations, it seems that the variable impo-1 has no significant
effect on the Greek import levels. This means that the level of imports of the current year does not depend on the level of imports of the previous year, throughout the examined time period. The level of imports seems thus to follow a more independent pattern from its previous levels as we expected it to follow.
We furthermore confirm the positive relationship between the logarithms of Greek income (GDP) and the level of imports. This relationship is plotted on Table 17. As the domestic product and thus the income in Greece rises, we observe a rise in the levels of the amount and value of products that Greeks import from their trade partners. The late 1970’s find Greece amidst a general international crisis and the import levels are also affected. Furthermore, the contribution of European Union membership in early 1980s to the level of Greek imports is a hypothesis that will be tested on the section that follows. Generally, it can be argued that after the entrance of Greece to the unified market of European Union the trade barriers in imports and exports have been gradually reduced and finally diminished.
The Government Expenditure Function
We have finally come to the point to present the final fundamental macroeconomic function for Greece, as estimated by following the selected methodology. The estimated government expenditure function is the following:
146 , 1 ) ln( 63 , 0 ) ln( 307 , 0 ) ln( 054 , 0 ) ln( 635 , 0 525 , 0 ) ln(gov =− − gov−1 + y + t + ctpn + e ˆt−1 (0,0004) (0,0867) (0,897) (0,057) (0,0004) (0,0095)
and economically significant, referring to the magnitude of their respective coefficients. The only variable that has no significant effect on the level of government expenditure is the variable Y, which represents the Greek GDP. This means that the level of income in Greece does not significantly affect the expenditure made on behalf of the government. This relationship (in logarithmic values) is
presented on Table 22. We observe that the logarithm of government expenditure is behaving more independently than this of the logarithm of income, which follows a smooth growth trend.
Furthermore, we empirically confirm the theoretical argument of positive connection between the income from taxation, the general transfer payments and the government expenditure for the case of Greece. With relevant significance of 10% and 1%
respectively, these independent variables both affect positively the expenditure of the Greek government. This means that the expenditure deriving from the government rises, when its income from taxation and transfer payments rises as well.
It is furthermore remarkable, the fact that we observe a significant (on 10% level) negative relationship between the government expenditure of the current and the previous year. If the government has raised its expenses one year, it is probable that it will reduce its expenses during the following year. We will return to this issue more in detail, while examining the effects of each one of the political regimes, on the
expenses deriving from the state.
Incorporating the Dummy Variables
After completing the necessary review of the estimations on each one of the
Note for all estimations: The estimated effect is regarded as the magnitude of the obtained coefficient for each one of the dummy variables on each one of the
dependent variables. The effect thus applies to the logarithmic value of the variable and it is presented in decimal form. It is expected that effects on logarithmic values maintain their sign and significance on real values as well (Kennedy, 1986).
1960-1966: Greece under a constitutional monarchic regime
“I would like to refer to the officers. We are united by God himself. I belong to you and you belong to me. I hope that everyone will work towards this direction. You have
become soldiers of our homeland. Soldiers of the King. My soldiers…” King Paul I of Greece, Thessaloniki 1962
The year 1960, the time point where this research begins, finds Greece under a constitutional monarchic regime with King Paul I as the head of state. By definition, in representative democracies that are parliamentary monarchies, the monarch may be regarded as the head of state but the Prime Minister, whose power derives directly or indirectly from elections, is head of government. The case of Greece however was quite different. Although there existed an elected parliamentary body and a Prime Minister, which was indeed regarded as the head of government, the monarch has in reality always possessed influential powers in selecting members of the government, or even affecting the decisions of the respective Prime Ministers. The extent of democratic practices followed during the election procedure of the parliament during this period is also a matter in question.
left-winged Left Democratic Union – L.D.U. (Ε.Δ.Α.) which represented the leftist and communist movements in Greece.
The tight bonds between the National Radical Union and the royal court are generally admitted by historians and have been a matter of controversy between the conservative and progressive political powers of the time. Furthermore, the supreme power and authority of the king has been regarded as the alternative antipode to communism, an argument which gave enough support to the populist actions of the monarch in the eyes of common view. So, Greek political life had to face a double-sided illusion: The fear of the royally supported conservatives that their power will be lost in favour of the leftists through public demand, and the certainty of the leftists that they can transform their growing potential into greater parliamentary power (Vournas, 2004).
On March 6th 1964 King Paul I dies, and his 24-year-old son Konstantinos inherited the crown. The main course of interference that the royal court has followed
throughout the post-war years, has remained however unchanged. Karamanlis had already fled to Paris, Georgios Papandreou was the head of the Government and the institution of the democratically elected Greek parliament was once again in fact powerless. One Prime Minister after another was handing his resignation in disposal of the king, in many cases following the monarch’s pressure, and the legal political powers of the country were publicly declaring their disappointment from these developments.
is of no surprise that under unstable and mainly monarch-led political conditions, the economy was neglected by the state and private consumption was the only obvious way of directing money into the market. Although Greek banking system was quite powerful in comparison with the primitive money market alternatives, it has however remained underdeveloped throughout the examined period. Saving or investing, was thus a limited and unattractive alternative to consumption for the average Greek individual.
The greatest economic problems that Greek economy had to face are these of high unemployment in the city centers and of low productivity rates. The 1960’s was a period of a high migration wave from Greece towards western European countries such as Germany or Belgium. Private initiative in investment, although not forbidden such as in communist states, couldn’t easily find its way to the market since doing business in Greece had become a highly complicated procedure. Greek businessmen however, excelled and deployed their creativity abroad (e.g. Onassis or Niarchos). Moreover, industrial growth has been motivated in the direction of exporting companies, in order for Greece to gain sufficient foreign currency reserves and overcome the long lasting problem of deficits (Sakellaropoulos et al, 2004).
In the mid-60’s and after the successful test of application of Keynesian theory in the USA under President Kennedy, Keynesianism has gained ground as an alternative framework for achieving economic growth, especially in emerging economies (Stein, 1969). The news has reached Greece as well, and this required a fundamental change on how economic policy and institutions were conceived. The dynamics of such a change would cause a transformation to the social and economic environment of the country. The lack of a climate of political collaboration and the total inexistence of a Greek state of providence however, led Greek economists of this generation to turn over Keynesianism to the highlight of a political struggle and confrontation and not of economic analysis (Sakellaropoulos et al, 2004).
