Domestic vs. external economic sectors and the political process: Insights from Greece
Abstract
EconStor is a publication server for scholarly economic literature, provided as a non-commercial public service by the ZBW.
Full text
Petrakos, Giōrgos; Rontos, Kostas; Salvatici, Luca; Vavoura, Chara; Baburas, Iōannēs S. Article Domestic vs. external economic sectors and the political process: Insights from Greece Economies Provided in Cooperation with: MDPI – Multidisciplinary Digital Publishing Institute, Basel Suggested Citation: Petrakos, Giōrgos; Rontos, Kostas; Salvatici, Luca; Vavoura, Chara; Baburas, Iōannēs S. (2022) : Domestic vs. external economic sectors and the political process: Insights from Greece, Economies, ISSN 2227-7099, MDPI, Basel, Vol. 10, Iss. 8, pp. 1-12, https://doi.org/10.3390/economies10080198 This Version is available at: https://hdl.handle.net/10419/328498 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/
Citation: Petrakos, George, Kostas Rontos, Luca Salvati, Chara Vavoura, and Ioannis Vavouras. 2022. Domestic vs. External Economic Sectors and the Political Process: Insights from Greece. Economies 10: 198. https://doi.org/10.3390/ economies10080198 Academic Editor: Franklin G. Mixon Received: 30 March 2022 Accepted: 10 August 2022 Published: 17 August 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). economies Case Report Domestic vs. External Economic Sectors and the Political Process: Insights from Greece George Petrakos 1, Kostas Rontos 2, Luca Salvati 3,*, Chara Vavoura 4and Ioannis Vavouras 1 1Department of Public Administration, Panteion University, 136 Syngrou Avenue, EL-17671 Athens, Greece 2Department of Sociology, University of the Aegean, University Hill, EL-81100 Mytilene, Greece 3 Department of Methods and Models for Economics, Territory and Finance, Sapienza University of Rome, Via del Castro Laurenziano 9, I-00161 Rome, Italy 4Department of Economics, National and Kapodistrian University of Athens, 1 Sofokleous Str., EL-10559 Athens, Greece *Correspondence: [email protected] Abstract: Building on the well-established relationship between economic dynamics and political processes, we focus on the most important element of the political process, namely, general (or national) elections, and look into their effects on public finance and total economic output. In this vein, the present study has three objectives: (i) to investigate political budget cycles in Greece during the period known as the ‘Third Hellenic Republic’ (in Greek, ‘Metapolitefsi’, hereafter THR) since 1974; (ii) to assess whether national elections affect total economic activity in a stabilizing or destabilizing way; and (iii) to examine the possible effects of the external sector of the economy on the budget balance. The empirical findings of our analysis document how the Greek economy was characterized by sharp political budget cycles in correspondence with the THR, exerting a destabilizing effect on the total output of the economy. Performances of the external sector of the economy have significantly affected budget balances in Greece. Keywords: political budget cycles; electoral cycles; economic destabilization; international trade JEL Classification: D72; E62; H62 1. Introduction Over the last decade, Greece has repeatedly faced bankruptcy risk as a result of its public debt crisis. Bankruptcy is still plausible nowadays, since the Greek government’s consolidated gross debt has surpassed 200% of GDP 1 , topping all European Union (EU) member states and amounting to more than twice the EU average. In this study, we document how the political process has been instrumental in the Greek debt crisis—mainly through the emergence of severe political budget cycles—and we explore the generating mechanism of these cycles and their impact on the stability of the aggregate economy and international trade. Political budget cycles (or electoral fiscal cycles) occur when governments in preelection periods pursue expansionary fiscal policies (e.g., public spending increases, tax cuts, budget deficit increases) followed by restrictive fiscal policies in post-election periods. In fact, excessive public spending and tax reductions are basic mechanisms generating political budget cycles. Each one is associated with different social welfare implications (Alesina 1987). Reducing taxes tends to be of practical concern to relatively wealthier people, while policies leveraging public spending—usually materializing through increases in transfer payments—favor the less well-off (Alesina 1988). Earlier studies support the hypothesis that budget cycles are stronger on the expenditure side (e.g., Rogoff 1990;Alt and Rose 2007), although there is some evidence on the existence of budget cycles on the revenue side (e.g., Katsimi and Sarantides 2012). The final outcome of this political and Economies 2022,10, 198. https://doi.org/10.3390/economies10080198 https://www.mdpi.com/journal/economies
Economies 2022,10, 198 2 of 13 economic behavior is the creation of budget cycles which are neither due to exogenous shocks nor due to policy failures. Instead, they are deliberately generated in the context of governments’ opportunistic efforts to maximize their re-election chances (Philips 2016). The success of this practice hinges on the voters’ myopic perspective and the ability of incumbent politicians to exploit their temporary information advantage2. Political budget cycles have been the subject of empirical research documenting how the ability of governments to create political-economic cycles becomes limited as (i) the level of socioeconomic development rises (Block 2002;Shi and Svensson 2006;Klomp and de Haan 2013), (ii) the quality of institutions improves, and (iii) the transparency of its political process increases (Persson and Tabellini 2005;De Haan and Klomp 2013;Veiga et al. 2017). In other words, this happens when the quality and duration (or maturity) of democracy in a given country increase (Shi and Svensson 2003;Brender and Drazen 2005;De Haan 2013). In particular, the quality of institutions, and especially the existence of effective mechanisms controlling the financial decisions of governments (checks, balances, and fiscal rules), seem to be the most important means of limiting the extent of political budget cycles, as institutions shape the choices of the electorate and determine the incentives as well as the opportunities for incumbents to resort to these cycles (Chang 2008;Streb and Torrens 2013;Gootjes et al. 2021)3. Therefore, the main conclusion of the existing works is that political budget cycles are more evident in developing economies (Brender and Drazen 2005), being negligible (or very limited) in wealthier countries (Andrikopoulos et al. 2004;Shi and Svensson 2006;Mandon and Cazals 2019). Our analysis focuses on Greece, a European country representing a remarkable deviation from the aforementioned rule. Chortareas et al. (2018) have documented the occurrence of opportunistic budgetary policies in this country, despite mixed evidence on their effectiveness in maximizing re-election prospects (Brender and Drazen 2008;Aidt et al. 2011;De Haan and Klomp 2013). In a more recent work, we have documented the significance, direction and size of political budget cycles in Greece (Petrakos et al. 2021a). This process is justified with the limited quality of the country’s institutions (Afonso et al. 2015) and lack of internal/external controls (Trantidis 2016), as well as with the fact that its governments often pursue clientele policies (Mitsopoulos and Pelagidis 2011) and resort to populism (Christodoulakis 2019). This paper serves three main objectives. The first is to investigate the existence of political budget cycles in Greece in the Third Hellenic Republic (in Greek, ‘Metapolitefsi’, hereafter THR) since 1974. For this purpose, we specified econometric models in which the actual budget balance (as the percent share of Gross Domestic product, GDP) is used as the dependent variable, with the aim of quantifying the effects of the electoral cycle on the formation of budget balances in Greece. The second objective is to examine whether general (or parliamentary) elections affect a country’s GDP, and whether the effects are stabilizing or destabilizing. More specifically, the question is whether general elections have statistically significant (stabilizing or destabilizing) effects on GDP. The third objective is to examine the possible impact of the external sector of the economy (i.e., external trade) on budget deficits. Based on these objectives, the structure of the paper is organized as follows. Section 2introduces empirical data and presents the empirical methodology. Section 3illustrates our findings in full detail and discusses the main results of this study, and Section 4concludes the paper with some remarks for future research. 2. Methodology 2.1. Study Area Considering Greece, a Southern European country, as the study area, our empirical analysis tests the intensity of political economic cycles and, particularly, political budget cycles over the THR since 1974, and more specifically, during a time interval encompassing more than four decades (1974–2020). This relatively long time horizon is considered representative of sequential economic downturns (Salvia et al. 2020), from the democratic recovery after the colonel’s dictatorship (1967–1973) in the hand of the center-right ‘Nea
Economies 2022,10, 198 3 of 13 Dimokratia’ party of K. Karamanlis and G. Rallis (1974–1981), to intense economic development under the government of the socialist party of A. Papandreou (1981–1989). After a relatively short dominance of a center-right government with K. Mitsotakis (1990–1993), the socialist party of A. Papandreou won the 1993 elections, taking the lead continuously up to 2004 with three governments of K. Simitis (1996–2004), after the premature death of A. Papandreou (1996). A moderate economic growth—mostly characterized by rising expenditure in public infrastructure expanding the public debt significantly—was the main trait of this political phase. The subsequent center-right government of K. Karamanlis (2004–2009) consolidated the intense economic expansion following the 2004 Olympic Games celebrated in Athens, introducing some reforms and trying to manage the early signs of the imminent financial crisis (Vinci et al. 2022). The short center-left government of G. Papandreou (2009–2010) was unable to manage the drastic impact of the recession on Greek society (Tomao et al. 2021), and a more evident instability characterized the political life of the country for some years (2011–2012). Alternated center-right (A. Samaras) and center-left (A. Tsipras) governments was representative of the 2010s political course of Greece. While reducing the impact of the crisis, the center-right government—answering the pressing requests of the European Commission and International Monetary Fund—was forced to cut salaries and pensions (Salvati 2016), and introduced a sort of austerity regime in public expenditures causing urban poverty, among others (Gkartzios 2013;Rontos et al. 2016; Panori et al. 2019). The subsequent center-left government tried to alleviate the economic consequences of Troika’s austerity, while fueling social spending, public infrastructure, and tourism (Salvati 2019). 2.2. Logical Framework and Scope of the Study The existence of political budget cycles has been documented for the Greek economy, with specific reference to a shorter time period, namely, 1980–2017 (Petrakos et al. 2021a). We have also demonstrated how these cycles have been generated primarily on the expenditure side and not on the revenue side (Petrakos et al. 2021c). The present analysis differs substantially from the existing research in many aspects. For the first time in the literature, the effects of the external sector of the economy (exports, imports, trade balance) were examined as explanatory variables of the public budget deficit in the context of political budget cycles and prove to be significant. These outcomes support the ‘twin deficit hypothesis’ (Miller and Russek 1989;Cavallo 2005;Corsetti and Müller 2006;Kim and Roubini 2008;Algieri 2013;Badinger et al. 2017;Afonso et al. 2022). Relating budget deficits to trade deficits, these outcomes had important policy implications for economic competitiveness (e.g., Rontos et al. 2016;Di Feliciantonio et al. 2018;Benassi et al. 2022). Moreover, we explicitly examined the effects of unemployment on the evolution of public budget deficits, regarded as a significant policy issue (Salvati and Benassi 2021). We studied a relatively long time series, incorporating predictors that may evaluate the influence of several election rounds from 1974 to 2019. Our second goal was to ascertain the possible existence of a relationship between election years and total real output (hereafter ‘real GDP’). In other words, we investigated whether the rate of change of total real GDP (hereafter ‘DTYGR1’) during the election years differs from the corresponding rate of change of non-election years (Petrakos et al. 2021b). Our final objective is to examine the possible effects of the external trade (exports of goods and services, imports of goods and services, and balance of goods and services) on the state budget balance, testing in this way the hypothesis of a direct linkage of budget balances with economic competitiveness at large. 2.3. Variables and Indicators To achieve our objectives, we constructed five econometric models using the following dependent (1) and explanatory (2–10) variables: (1) The actual budget balance (ABB) of a general government as a percent of GDP, as defined and measured by Eurostat. The minus ( − ) sign corresponds to a deficit, while the
Economies 2022,10, 198 4 of 13 plus (+) sign corresponds to a surplus. We consider the actual budget balance as a percent share of GDP (instead of computing absolute monetary terms) for three main reasons: because (i) percent metrics provide a more reliable indicator of the relative magnitude of the actual budget balance; (ii) percentages remove the long-term effect of inflation on fiscal aggregates; (iii) the main fiscal policy condition of the EU member states, in order to avoid entering the Excessive Deficit Procedure (EDP) of the Stability and Growth Pact, is not to exceed 3% of their GDP. (2) The one-year lag of the actual budget balance (ABB-1) of a general government as a percent of GDP, since the public balance might be ‘compounded’ in the sense that the budget deficit of the previous year might affect, to some extent, the deficit of the current year. In effect, the one-year lag of the dependent variable is used to control for the autoregressive, AR(1), component of the actual budget balance (Petrakos et al. 2021a). (3) The two-year lag of the actual budget balance (ABB-2) of a general government as a percent of GDP, since not only the budget deficit of the previous year but also the deficit of the year before might affect the deficit of the current year. In effect, the two year lag of the dependent variable is used to capture the autoregressive, AR(2), component of the actual budget balance. We also considered the impact of a three-year lag of the actual budget, but our preliminary analysis suggests that this variable is largely insignificant. (4) The growth rate of real total GDP (TYGR), as estimated by the World Bank. (5) The change in the growth rate of total real GDP (DTYGR1), i.e., the difference between the growth rate of total real GDP of the current year and the previous one, i.e., DTYGR1 i = TYGR1 i− TYGR1 i−1 . This variable is introduced as it might not be the rates of change in total real GDP that cause pressures for expansionary fiscal policy as much as it is the variations (or fluctuations) of these rates from year to year. (6) The unemployment rate (UNR) as it is measured by Eurostat and the Hellenic Statistical Authority (ELSTAT). (7) Exports of goods and services as a percent of GDP (EXPO) as they are estimated by the World Bank (World Development Indicators). (8) Imports of goods and services as a percent of GDP (IMPO) as they are estimated by the World Bank (World Development Indicators). (9) Balance of goods and services as a percent of GDP (BAGS), estimated as the difference between exports of goods and services and imports of goods and services as a percent of GDP (BAGS = EXPO −IMPO). (10) Election (ELE), a binary variable taking the value of 1 for (general) election years and 0 otherwise. 2.4. Statistical Analysis We have specified five linear regression models with the actual budget balance as the dependent variable. In what follows (Section 3), results of these models were presented and commented in the order they are listed in Table 1, reporting the regression coefficients of each model in separated tables. All models were checked against the basic assumptions of Ordinary Least Square (OLS) model fitting, i.e., testing for linearity, homoscedasticity, independence, and normality assumptions (Ciommi et al. 2019). Results of statistical checks were presented in a graphical form (Lamonica et al. 2020). Table 1. Specification of econometric models used in our analysis (Acronyms in Section 2.3). Model Dependent Variable Predictors 1 Actual Budget Balance ABB-1, TYGR, UNR, ELE 2 Actual Budget Balance ABB-1, TYGR, EXPO, IMPO, ELE 3 Actual Budget Balance ABB-1, ABB-2, TYGR, EXPO, IMPO, ELE 4 Actual Budget Balance ABB-1, ABB-2, TYGR, BAGS, ELE 5 Actual Budget Balance ABB-1, ABB-2, TYGR, BAGS, UNR, ELE
Economies 2022,10, 198 5 of 13 3. Results and Discussion 3.1. Outcomes of Model 1 Model 1 provided the simplest specification of the variability in the actual budget balances of Greece (1974–2020) using only four variables. Since the lagged dependent variable is included in the explanatory variables, residuals are independently distributed (Salvati 2022). Plotting the residuals versus the fit, we verify that the linearity and homoscedasticity assumptions hold (Figure 1). Finally, a QQ plot and histogram of the residuals reveal no serious departures from normality (Ciommi et al. 2018). ABB-1 and UNR are significant at α = 0.01, ELE is significant at α = 0.05, and TYGR is significant at α = 0.1. Low Variance Inflation Factors (VIF) values indicate no evidence of multicollinearity (Table 2). The proportion of explained variance is relatively high (adjusted-R2= 0.70). Economies2022,10,xFORPEERREVIEW5of14 Table1.Specificationofeconometricmodelsusedinouranalysis(AcronymsinSection2.3). ModelDependentVariablePredictors 1ActualBudgetBalanceABB‐1,TYGR,UNR,ELE 2ActualBudgetBalanceABB‐1,TYGR,EXPO,IMPO,ELE 3ActualBudgetBalance ΑΒΒ‐1,ΑΒΒ‐2,TYGR,EXPO,IMPO,ELE 4ActualBudgetBalance ΑΒΒ‐1,ΑΒΒ‐2,TYGR,BAGS,ELE 5ActualBudgetBalanceABB‐1,ABB‐2,TYGR,BAGS,UNR,ELE 3.ResultsandDiscussion 3.1.OutcomesofModel1 Model1providedthesimplestspecificationofthevariabilityintheactualbudget balancesofGreece(1974–2020)usingonlyfourvariables.Sincethelaggeddependentvar‐ iableisincludedintheexplanatoryvariables,residualsareindependentlydistributed (Salvati2022).Plottingtheresidualsversusthefit,weverifythatthelinearityandhomo‐ scedasticityassumptionshold(Figure1).Finally,aQQplotandhistogramoftheresiduals revealnoseriousdeparturesfromnormality(Ciommietal.2018).ABB‐1andUNRare significantatα=0.01,ELEissignificantatα=0.05,andTYGRissignificantatα=0.1.Low VarianceInflationFactors(VIF)valuesindicatenoevidenceofmulticollinearity(Table2). Theproportionofexplainedvarianceisrelativelyhigh(adjusted‐R 2 =0.70). Table2.RegressioncoefficientsofModel1(seeTable1andSection3formodel’sdescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant−3.181.170−2.710.010‐ BudgetBalance,previousyear0.750.0988.020.0001.22 GDPgrowth0.220.1221.810.0781.52 Unemployment0.160.0572.830.0071.26 Electionyear−1.840.755−2.440.0191.02 Figure1.ResidualplotsforModel1withActualBudgetBalanceasdependentvariable. 3.2.OutcomesofModel2 EXPOandIMPOwereintroducedinModel2togetherwiththepredictorsalready consideredinModel1.ResidualplotsinFigure2documenthowresultsofModel2fully adheretolinearity,homoscedasticity,independence,andnormalityassumptions.ABB‐1, TYGR,andEXPOaresignificantatα=0.01,whileELEandIMPOaresignificantatα= Figure 1. Residual plots for Model 1 with Actual Budget Balance as dependent variable. Table 2. Regression coefficients of Model 1 (see Table 1and Section 2for model’s description). Variable Coefficient Standard Error t-Value p-Value VIF Constant −3.18 1.170 −2.71 0.010 - Budget Balance, previous year 0.75 0.098 8.02 0.000 1.22 GDP growth 0.22 0.122 1.81 0.078 1.52 Unemployment 0.16 0.057 2.83 0.007 1.26 Election year −1.84 0.755 −2.44 0.019 1.02 3.2. Outcomes of Model 2 EXPO and IMPO were introduced in Model 2 together with the predictors already considered in Model 1. Residual plots in Figure 2document how results of Model 2 fully adhere to linearity, homoscedasticity, independence, and normality assumptions. ABB-1, TYGR, and EXPO are significant at α = 0.01, while ELE and IMPO are significant at α = 0.05. VIF values indicate evidence of a residual multi-collinearity in Model 2, mainly attributable to IMPO and EXPO variables (Table 3). The proportion of explained variance is rather high (adjusted-R2= 0.74).
Economies 2022,10, 198 6 of 13 Economies2022,10,xFORPEERREVIEW6of14 0.05.VIFvaluesindicateevidenceofaresidualmulti‐collinearityinModel2,mainlyat‐ tributabletoIMPOandEXPOvariables(Table3).Theproportionofexplainedvarianceis ratherhigh(adjusted‐R 2 =0.74). Table3.RegressioncoefficientsofModel2(seeTable1andSection3forthemodeldescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant−3.7102.360−1.570.125‐ BudgetBalance,previousyear0.5310.1084.930.0001.86 GDPgrowth0.3350.1222.740.0091.77 Exports0.4440.1223.630.0015.97 Imports−0.3120.140−2.240.0314.66 Electionyear−1.7330.708−2.450.0191.03 Figure2.ResidualplotsforModel2withActualBudgetBalanceasthedependentvariable. 3.3.OutcomesofModel3 ThelaggedvariableABB‐2wasintroducedasapredictorinModel3andaccounted foraparticularlyhighproportionoftheexplainedvariance(adjusted‐R 2 =0.75).There‐ sidualplotsinFigure3indicatenodeviationsfromthebasicregressionassumptions. EXPOwassignificantatα=0.01,ELE,IMPO,TYGR,andABB‐1wereallsignificantatα =0.05,andABB‐2wassignificantatα=0.1(Table4).Aweakmulti‐collinearityissue(VIF >5)wasadistinctivecharacteristicoftheestimationofModel3,inlinewithwhathas beenobservedfortheresultsofModel2. Table4.RegressioncoefficientsofModel3(seeTable1andSection3forthemodeldescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant−3.1002.410−1.280.207‐ BudgetBalance,previousyear0.3650.1442.530.0163.29 BudgetBalance,twoyearsbefore0.2280.1301.760.0872.52 GDPgrowth0.3130.1212.580.0141.71 Exports0.4580.1213.780.0015.88 Imports−0.3260.138−2.360.0234.51 Electionyear−1.7670.701−2.520.0161.03 Figure 2. Residual plots for Model 2 with Actual Budget Balance as the dependent variable. Table 3. Regression coefficients of Model 2 (see Table 1and Section 2for the model description). Variable Coefficient Standard Error t-Value p-Value VIF Constant −3.710 2.360 −1.57 0.125 - Budget Balance, previous year 0.531 0.108 4.93 0.000 1.86 GDP growth 0.335 0.122 2.74 0.009 1.77 Exports 0.444 0.122 3.63 0.001 5.97 Imports −0.312 0.140 −2.24 0.031 4.66 Election year −1.733 0.708 −2.45 0.019 1.03 3.3. Outcomes of Model 3 The lagged variable ABB-2 was introduced as a predictor in Model 3 and accounted for a particularly high proportion of the explained variance (adjusted-R 2 = 0.75). The residual plots in Figure 3indicate no deviations from the basic regression assumptions. EXPO was significant at α = 0.01, ELE, IMPO, TYGR, and ABB-1 were all significant at α = 0.05, and ABB-2 was significant at α = 0.1 (Table 4). A weak multi-collinearity issue (VIF > 5) was a distinctive characteristic of the estimation of Model 3, in line with what has been observed for the results of Model 2. Economies2022,10,xFORPEERREVIEW7of14 Figure3.ResidualplotsforModel3withActualBudgetBalanceasthedependentvariable. 3.4.OutcomesofModel4 Tocopewiththeweakmulti‐collinearityobservedinModel3,thevariablesIMPO andEXPOwerereplacedwiththeirdifference,namedBAGS,inModel4. Theresidual plotsinFigure4indicatethatthebasicassumptionswerenotviolatedwiththisspecifica‐ tion.Theproportionofexplainedvarianceremainedsatisfactory(adjusted‐R 2 =0.73). ABB‐1andBAGSaresignificantatα=0.01,ELEissignificantatα=0.05,TYGRismargin‐ allysignificantatα=0.05,andABB‐2ismarginallysignificantatα=0.1(Table5).Low VIFvaluesindicatenoevidenceofmulticollinearity. Table5.RegressioncoefficientsofModel4(seeTable1andSection3forthemodeldescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant1.0030.9791.020.312‐ BudgetBalance,previousyear0.4260.1452.930.0063.12 BudgetBalance,twoyearsbefore0.2240.1341.680.1022.52 GDPgrowth0.2360.1182.010.0521.51 Balanceofgoodsandservices0.4490.1253.600.0011.52 Electionyear−1.7240.723−2.390.0221.03 Figure4.ResidualplotsforModel4withActualBudgetBalanceasthedependentvariable. Figure 3. Residual plots for Model 3 with Actual Budget Balance as the dependent variable.
Economies 2022,10, 198 7 of 13 Table 4. Regression coefficients of Model 3 (see Table 1and Section 2for the model description). Variable Coefficient Standard Error t-Value p-Value VIF Constant −3.100 2.410 −1.28 0.207 - Budget Balance, previous year 0.365 0.144 2.53 0.016 3.29 Budget Balance, two years before 0.228 0.130 1.76 0.087 2.52 GDP growth 0.313 0.121 2.58 0.014 1.71 Exports 0.458 0.121 3.78 0.001 5.88 Imports −0.326 0.138 −2.36 0.023 4.51 Election year −1.767 0.701 −2.52 0.016 1.03 3.4. Outcomes of Model 4 To cope with the weak multi-collinearity observed in Model 3, the variables IMPO and EXPO were replaced with their difference, named BAGS, in Model 4. The residual plots in Figure 4indicate that the basic assumptions were not violated with this specification. The proportion of explained variance remained satisfactory (adjusted-R 2 = 0.73). ABB-1 and BAGS are significant at α = 0.01, ELE is significant at α = 0.05, TYGR is marginally significant at α = 0.05, and ABB-2 is marginally significant at α = 0.1 (Table 5). Low VIF values indicate no evidence of multicollinearity. Economies2022,10,xFORPEERREVIEW7of14 Figure3.ResidualplotsforModel3withActualBudgetBalanceasthedependentvariable. 3.4.OutcomesofModel4 Tocopewiththeweakmulti‐collinearityobservedinModel3,thevariablesIMPO andEXPOwerereplacedwiththeirdifference,namedBAGS,inModel4. Theresidual plotsinFigure4indicatethatthebasicassumptionswerenotviolatedwiththisspecifica‐ tion.Theproportionofexplainedvarianceremainedsatisfactory(adjusted‐R 2 =0.73). ABB‐1andBAGSaresignificantatα=0.01,ELEissignificantatα=0.05,TYGRismargin‐ allysignificantatα=0.05,andABB‐2ismarginallysignificantatα=0.1(Table5).Low VIFvaluesindicatenoevidenceofmulticollinearity. Table5.RegressioncoefficientsofModel4(seeTable1andSection3forthemodeldescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant1.0030.9791.020.312‐ BudgetBalance,previousyear0.4260.1452.930.0063.12 BudgetBalance,twoyearsbefore0.2240.1341.680.1022.52 GDPgrowth0.2360.1182.010.0521.51 Balanceofgoodsandservices0.4490.1253.600.0011.52 Electionyear−1.7240.723−2.390.0221.03 Figure4.ResidualplotsforModel4withActualBudgetBalanceasthedependentvariable. Figure 4. Residual plots for Model 4 with Actual Budget Balance as the dependent variable. Table 5. Regression coefficients of Model 4 (see Table 1and Section 2for the model description). Variable Coefficient Standard Error t-Value p-Value VIF Constant 1.003 0.979 1.02 0.312 - Budget Balance, previous year 0.426 0.145 2.93 0.006 3.12 Budget Balance, two years before 0.224 0.134 1.68 0.102 2.52 GDP growth 0.236 0.118 2.01 0.052 1.51 Balance of goods and services 0.449 0.125 3.60 0.001 1.52 Election year −1.724 0.723 −2.39 0.022 1.03 3.5. Outcomes of Model 5 Unemployment rate (UNR) was introduced as an additional predictor in Model 5. The proportion of explained variance was satisfactory (adjusted-R 2 = 0.75). Plots of residuals (Figure 5) indicate no violation of the basic econometric assumptions for Model 5. ABB-1 is significant at α = 0.01, ELE, ABB-2, TYGR, BAGS, are all significant at α = 0.05, and UNR is significant at α= 0.1. Low VIF values indicate no evidence of multicollinearity (Table 6).
Economies 2022,10, 198 8 of 13 Economies2022,10,xFORPEERREVIEW8of14 3.5.OutcomesofModel5 Unemploymentrate(UNR)wasintroducedasanadditionalpredictorinModel5. Theproportionofexplainedvariancewassatisfactory(adjusted‐R 2 =0.75).Plotsofresid‐ uals(Figure5)indicatenoviolationofthebasiceconometricassumptionsforModel5. ABB‐1issignificantatα=0.01,ELE,ABB‐2,TYGR,BAGS,areallsignificantatα=0.05, andUNRissignificantatα=0.1.LowVIFvaluesindicatenoevidenceofmulticollinearity (Table6). Table6.RegressioncoefficientsofModel5(seeTable1andSection3forthemodeldescription). VariableCoefficientStandardErrort‐Valuep‐ValueVIF Constant−0.9201.450−0.630.532‐ BudgetBalance,previousyear0.4170.1412.950.0053.12 BudgetBalance,twoyearsbefore0.2900.1362.140.0392.73 GDPgrowth0.2740.1172.350.0241.56 Balanceofgoodsandservices0.3040.1472.060.0462.23 Unemploymentrate0.1170.0671.750.0891.96 Electionyear−1.7440.704−2.480.0181.03 Figure5.ResidualplotsforModel5withActualBudgetBalanceasthedependentvariable. Inordertoinvestigatepossibledifferencesinthemeangrowthrate(%)oftotal(real) GDPbetweenthetwoELEgroups(i.e.,electionandnon‐electionyears),thenullhypoth‐ esisH 0 :μ 1 −μ 2 =0wastestedagainstthealternativeH 1 :μ 1 −μ 2 ≠0,whereμ 1 = Ε(DTYGR1|ELE=0)andμ 2 =Ε(DTYGR1|ELE=1).H 0 isrejectedatα=0.05,andthedif‐ ferencesinthemeansofGDPgrowtharesignificant.Theresultsofouranalysiswerepre‐ sentedinTable7andFigure6. Table7.Descriptivestatisticsofthegrowthrate(%)oftotal(real)GrossDomesticProduct (DTYGR1)byelectionyear. YearNo.YearsMeanStandardDeviation95%ConfidenceInterval Noelection310.8233.620(−0.338;1.983) Election14−1.2291.929(−2.956;0.499) Figure 5. Residual plots for Model 5 with Actual Budget Balance as the dependent variable. Table 6. Regression coefficients of Model 5 (see Table 1and Section 2for the model description). Variable Coefficient Standard Error t-Value p-Value VIF Constant −0.920 1.450 −0.63 0.532 - Budget Balance, previous year 0.417 0.141 2.95 0.005 3.12 Budget Balance, two years before 0.290 0.136 2.14 0.039 2.73 GDP growth 0.274 0.117 2.35 0.024 1.56 Balance of goods and services 0.304 0.147 2.06 0.046 2.23 Unemployment rate 0.117 0.067 1.75 0.089 1.96 Election year −1.744 0.704 −2.48 0.018 1.03 In order to investigate possible differences in the mean growth rate (%) of total (real) GDP between the two ELE groups (i.e., election and non-election years), the null hypothesis H 0 : µ1−µ2 = 0 was tested against the alternative H 1 : µ1−µ26= 0, where µ1 = E(DTYGR1|ELE = 0) and µ2 = E(DTYGR1|ELE = 1). H 0 is rejected at α = 0.05, and the differences in the means of GDP growth are significant. The results of our analysis were presented in Table 7and Figure 6. Economies2022,10,xFORPEERREVIEW9of14 Figure6.IntervalPlot(95%ConfidenceIntervalaroundthemean)ofchangesovertimeinthe growthrate(%)oftotal(real)GrossDomesticProduct(DTYGR1)againstelectionyear(1)ornon‐ electionyear(0);pooledstandarddeviation(3.20)wasusedtocalculatetheintervals. 3.6.DiscussingtheOutcomesofEconometricModels(1to5) ThemodelsillustratedabovehaveagoodfitastheadjustedR 2 wassystematically above0.7.Inallofthem,theactualbudgetbalances(whichcorrespondtopublicdeficits inthecaseofGreece)werestronglycorrelatedtotheirone‐yearlaggedvalues(thecoeffi‐ cientofABB‐1is0.75and0.53,respectively,inModels1and2).TheimpactofABB‐1was reducedslightlywhenconsideringthetwo‐yearlaggedvariableABB‐2(ABB‐1coefficient rangedbetween0.37inModel3and0.43inModel4,whileABB‐2coefficientrangedbe‐ tween0.22inModel4and0.29inModel5).Hence,weconcludethatpublicdeficitswere persistentinGreece,adjustingslowlyoverlongerperiodsoftimethatcouldlastuptotwo years. Regardingtheimpactoftheexternalsectoroftheeconomy(EXPO,IMPO,BAGS), exportsofgoodsandservicestendtoreducethebudgetdeficit,whileimportsseemto increaseit(Models2and3).Moreover,thebalanceofgoodsandservices(BAGS=EXPO −IMPO)tendstoreducethebudgetdeficit(Models4and5).Thisresultisastrongindi‐ cationforthepresenceof‘twindeficits’inthecaseoftheGreekeconomy.Inparticular, accordingtothe‘twindeficitshypothesis’,largeandgrowingbudgetdeficitsarereflected inthewideningofthecurrentaccountdeficit,leadingtodecliningeconomicpolicycred‐ ibility,macroeconomicimbalances,andaslowdownineconomicgrowth.Althoughmak‐ inguseofdifferentapproaches,thesefindingsareinlinewithearlierliteraturetestingthis hypothesisforGreece(Vamvoukas1999;Pantelidisetal.2009;KalouandPaleologou 2012;Magazzino2012;Piersanti2000;ForteandMagazzino2013;PanousisandKou‐ kouritakis2020;KatrakilidisandTrachanas2011;Paparasetal.2016;Kosteletou2013). Fewadditionalstudiesendedupwithmoremixed,andsometimescontrasting,results (Algieri2013;PapadogonasandStournaras2006). Theresultsofalleconometricmodelspresentedinourstudydemonstratehowthe growthrateofrealGDPpositivelyaffectstheactualbudgetbalance,reducingbudgetdef‐ icits.ThecoefficientofGDPgrowthratevariedbetween0.335(Model2)and0.221(Model 1).Thisfindingimpliesthatduringtimesofeconomicslowdown,Greekgovernments tendtorespondwithexpansionaryfiscalpolicies.Unemploymentalsoaffectedthede‐ pendentvariableinapositiveway,althoughwitharelativelylowcoefficient(0.162in Model1and0.117inModel5).Thiscorrelationappearscounterintuitiveatfirst(sensu Rontosetal.2019).However,economicdynamicsinthe2010smayjustifythisoutcomein thecaseofGreece,sincethecountrywasunderbudgetarysupervisionbytheEuropean Institutions(‘Troikamemoranda’).Inthatdecade,budgetdeficitswerestrictlycontrolled, andunemploymentskyrocketedtounprecedentedlevels,fuelingincomeinequalitiesand risingsocialsegregation,especiallyinurbanareas(Gavalasetal.2014;DiFeliciantonio andSalvati2015;Panorietal.2019;Rontosetal.2016;Salvati2016,2018). Figure 6. Interval Plot (95% Confidence Interval around the mean) of changes over time in the growth rate (%) of total (real) Gross Domestic Product (DTYGR1) against election year (1) or non-election year (0); pooled standard deviation (3.20) was used to calculate the intervals.