Empirical Economics

, Volume 51, Issue 2, pp 517–546 | Cite as

Are European fiscal rules that bad? Discretionary fiscal policies in New Member States

Article

Abstract

There is no clear-cut evidence on how the adoption of the European fiscal standards influences discretionary fiscal policies within the Member States. This study investigates that phenomenon on the example of the 2004 enlargement. The results show that the effects of the adoption of EU fiscal rules bring a statistically significant change toward more counter-cyclical behavior. The results are robust for different model specifications, including alternative time spans and correcting for the possible influence of the financial crises and political forces. Interestingly, the year 2004 did not have any significant impact on the change in fiscal policies in the Old Member States, suggesting that the EU entry might motivate new members to run more prudent budgetary policies.

Keywords

Discretionary fiscal policy New Member States Counter-cyclical policies Fiscal rules 

JEL Classification

E62 O52 

1 Introduction

Fiscal arrangements belong to the most criticized aspects of the European Union’s (EU) internal policy (Candelon et al. 2010). A widely cited reason for that concerns the inability of the EU governments to run stabilizing fiscal policies having the Maastricht Treaty (MT) and the stability and growth pact (SGP) in the background (Wyplosz 2006; Gali and Perotti 2003; Candelon et al. 2010). However, there has been no clear-cut evidence on how the EU fiscal rules affect discretionary fiscal policies nor what effects they pose on business cycles within the EU. Some studies prove a significant change in fiscal behavior after the adoption of the MT or the SGP fiscal rules, whereas other research shows the opposite.

This paper contributes to the topic by assessing the influence of the adoption of the EU fiscal standards on discretionary fiscal behavior empirically, on the example of the 2004 EU enlargement. Moreover, in order to determine whether the EU entry was a core event which influenced fiscal policies among the New Member States (NMS), the analysis compares the tendencies in the fiscal behavior between them, the Old Member States (OMS) and a group of control countries which did not belong to the EU at that time.

The study is carried out on all the NMS that joined the EU on May 1, 2004; Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia and Slovenia. Following the literature, different model specifications are applied to correct for possible biases. The first one is the simplest one and similar to the approach taken by Gali and Perotti (2003). It estimates the influence of the output gap on discretionary fiscal decisions using the fixed effects panel technique and assuming the existence of a structural break in 2004. To correct for the possible bias resulting from the financial crisis and the sovereign debt crisis in the euro zone, we carry out the robustness check by adding a proxy variable, reflecting the situation in the loan market, and by cutting the time span of the analysis in 2008. The fifth model combines the last two specifications. The time span of the study has been limited by data availability and quality and, therefore, covers years 2000–2011. Additionally, a possible policy change around the year of accession is checked by formal structural break tests.

To further investigate this phenomenon, the study focuses on the comparison of discretionary fiscal policies between the NMS and OMS. The OMS comprise 15 Member States which were in the EU before 2004 (Gali and Perotti 2003), i.e., Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and the UK. As another robustness check, we conducted the same experiment on a control group consisting of the so-called safe havens, represented by Australia and Canada, and two further central eastern European countries (CEEC) that joined the EU later on January 1, 2007, i.e., Bulgaria and Romania, in order to detect whether these economies experience a switch in policy behavior around 2004 as well.

Interestingly, a remarkable and significant switch toward counter-cyclical fiscal behavior is observed for the NMS, especially in the most recent years, with no similar effect on the OMS or the control group. Among all the models for the OMS, the only well-specified one suggests that the OMS were running acyclical fiscal policies for the entire period, driven mostly by external factors rather than complying with the EU fiscal rules.

The remainder of this paper is organized as follows. Section 2 brings forward the recent results on the topic and discusses the influence of fiscal policies on a business cycle. Section 3 investigates the changes in discretionary fiscal policies before and after the 2004 EU enlargement. Additionally, we carry out the robustness checks to assess the validity of the results. Section 4 summarizes and discusses some concerns regarding the results.

2 Literature review and motivation

Thinking of fiscal policy and its influence on the business cycle, one should distinguish between the effects of automatic stabilizers and discretionary actions. The former relates to the natural budget responses to the business cycle fluctuations (Candelon et al. 2010). For example, the revenues from taxes should increase during upswings so that the budget balance becomes larger. Likewise, when there is a recession, the budget balance should decrease as a consequence of, for instance, increased unemployment spending. In short, automatic stabilizers work as a natural mechanism to accumulate capital during booms and spend it during downswings. This all makes the business cycle less volatile and aims to provide solid economic growth (Wyplosz 2006).

However, fiscal authorities are able to stipulate the economy beyond the automatic stabilizers’ level. Such actions are called discretionary fiscal policy. Following Kydland and Prescott (1977), it may be driven by the time inconsistency of economic decisions or could be a straightforward consequence of political motivation (Pendleton 1938; Alesina et al. 2008). An example of a discretionary action could be a change in the tax levels or a revision in government investment policy.

A common view supposes that once a budget is balanced, discretionary actions should be avoided and the reaction to business cycle fluctuations should be left for automatic stabilizers only (European Comission 2001; Beetsma 2001; Artis and Buti 2000). Looking at the EU countries, the condition of a balanced budgetary position has not been always satisfied, mostly because when the economic and monetary union (EMU) was introduced, fiscal policy became the only macroeconomic management instrument. Therefore, it is unlikely that Member States would resign from it easily, violating the balanced budget plan (Wyplosz 2006).

Excessive discretionary policy could have dramatic consequences for the economy (Wyplosz 2006). It can not only drive the amount of public debt beyond the sustainable level but can also exacerbate the business cycle or/and foster uncertainty in the markets. The current situation in Europe, with many economies facing fiscal difficulties, reveals these problems. For about five years now, EU and especially euro zone members have been impaired by the financial and debt crises. Enormous financial rescue programs had to be arranged in order to support seriously debt-troubled economies, such as Greece for instance, and to restore financial stability. This assistance is tied to requirements to implement strict austerity measures (European Commission 2013). Some euro zone members, i.e., Ireland, Greece, Spain, Portugal and Cyprus, had to apply to receive financial support. Recently, Ireland, Portugal and Spain left the programs.

In order to induce sound fiscal performance and to limit the excessive discretionary actions within the EU, the MT implemented basic fiscal rules. They refer to the upper deficit and debt ceilings which each country has to respect, i.e., 3 % deficit-to-GDP and 60 % debt-to-GDP ratio. Since their introduction in 1992, they have been steadily modified in order to make them more accurate. In 1997 they were extended by the SGP and in 2005 by the decision of European Council entitled ‘Improving the Operation of the Stability and Growth Pact.’1 As a result, the performance of the EU fiscal rules has been improved; however, they still have some drawbacks (Buti et al. 2003; Wyplosz 2006).

The biggest critique of the EU fiscal arrangements comes from the fact that they might limit the flexibility of national fiscal authorities too much, especially during recessions (Gali and Perotti 2003), but also during recently observed financial crises. Consequently, governments would not be able to follow counter-cyclical policies and impose balanced budgetary plans because they would be absorbed by fighting external shocks instead or they would be facing internal political pressure.

Scholars have tried to explain empirically whether those threats are justified. The results are, however, ambiguous. Two basic concepts arise from the literature. The first one, proved by Buti et al. (1997), Gali and Perotti (2003) and Wyplosz (2006), states that discretionary fiscal policies in the EMU countries were pro-cyclical before 1992 and acyclical afterward.2 The second concept refers to studies of Von Hagen (2005) and Candelon et al. (2010) who also confirm that before 1992 the discretionary fiscal policies in the EMU were pro-cyclical; however, they remained pro-cyclical after the MT was signed. An interesting conclusion from Candelon et al. (2010) is that the empirical results are very susceptible to the time spans of the analysis. In fact, using similar methodology to Gali and Perotti (2003) but expanding the data set by two years only, Candelon et al. (2010) obtained the opposite results.

Moreover, there have been some key studies on fiscal policy and business cycle responses in NMS. For instance, Budina and Van Wijnbergen (1997) study fiscal policies in CEEC and former Soviet Union members and their impact on economic stabilization in the early phase of transition. They focus on sustainability of public finances and find empirical evidence that countries which quickly implemented ambitious and tight reforms have been more successful in terms of inflation stabilization and growth performance, both, with respect to the time of recovery and the size of the rate. Also, Lewis (2009) studies fiscal policies and their stabilizing effect in CEECs, as well as possible inertia of budget balances and the effect of accession on public finances. He finds some empirical evidence for counter-cyclical behavior and a less inertial response for the Eastern economies than in the Western EU countries. Moreover, he detects a fiscal loosening prior to accession. Staehr (2007) analyzes fiscal policy functioning as regards cyclical properties and fluctuation effects in an empirical comparative study for the euro zone members and CEECs. Even though more volatile, he finds more ‘agile’ and counter-cyclical policies for the NMS for several specification types with pro-cyclical tendencies in the euro zone members due to the revenue side. In a recent paper, Darvas (2009) studies the effect of the current crisis on fiscal policies in CEECs. Among other points, the results indicate some pro-cyclicality of fiscal policy.

Our paper differs from the contributions by Staehr (2007) and Lewis (2009) in several aspects. Whereas Staehr (2007) focuses on a comparison of fiscal policy between the two groups of NMS and euro zone members, we are interested in different behavior before and after EU accession, i.e., over time with focus on 2004, plus, comparing NMS and OMS. Also, the analysis done by Staehr (2007) ranges from 1995 until 2005. We extend the period to 2011, which allows us to take the early stage of the crisis into account. Lewis (2009), for instance, employs real-time data, whereas we rely on ‘conventional’ final data.

Several papers have studied fiscal policies and the business cycle response in the EU. Generally, the results are rather mixed. For instance, Beetsma and Giuliodori (2008) study fiscal policy in terms of deviations from plans due to new information with regard to the business cycle for OECD countries. They find differences for EU members (acyclical) and non-EU economies (counter-cyclical) ex ante, but as concerns the deviations the EU countries behave pro-cyclical, whereas the other countries reveal acyclical responses. Claeys (2008) analyzes fiscal policies, rules, business cycle effects and sustainability in Sweden. His results suggest a more counter-cyclical policy concerning the primary balance influence in Sweden. Additionally, he identifies regimes and shifts by incorporating structural breaks and switches and provides a comparison with other northern European economies. Like our study, he also uses GMM and quarterly data. Darby and Melitz (2011) study simultaneously effects of automatic and discretionary fiscal policy for OECD countries explicitly, by separating the components of the government budget they address the revenue and spending effects individually. Notably they do not confirm any effect of the MT and SGP for the output gap and fiscal policy relationship in Europe. Darby and Melitz (2008) study automatic stabilization in OECD countries and the response of the government balance and its components to the business cycle. Detailed estimations reveal a fairly destabilizing effect of taxes, whereas expenditures function in a stabilizing manner. Poplawski-Ribeiro (2009) analyzes the effectiveness of the fiscal framework in the euro zone to fiscal policy disciplining and distinguishes two sub-periods to the MT and SGP. He finds that the SGP is not an effective instrument to prompt fiscal discipline and it does not induce counter-cyclical fiscal policies in the euro zone. Moreover, he finds a relaxing of fiscal policy before the official accession.

For the EMU members incentives to implement structural reforms in view of the fiscal criteria may be a crucial aspect, both, for the NMS in their entry stage and for the OMS as many of them bear structural inefficiencies (Beetsma and Debrun 2004). Beetsma and Giuliodori (2010) mention that the literature indicates mixed effects, even though one of the reasoning pro EMU was perceived in enhancing structural reforms due to competitive pressure. Beetsma and Debrun (2004) analyze with a formal model the incentives for structural reforms in face of fiscal rules, like the SGP, and discuss the problems of the trade-off between stability and growth. They suggest a more flexible handling linked to progress in structural reforms and independent surveillance.

There are several commonalities of findings and proceeding in these mentioned studies and our contribution. The two studies of Lewis (2009) and Staehr (2007) are close to our proceeding and results, as they technically also use the instrumental variable (IV) GMM method. Moreover, they also support the finding of less inertia in NMS than in OMS and counter-cyclical behavior in NMS. Also, the loosening of the fiscal behavior prior to official entry, as found in Lewis (2009) and for euro zone members in Poplawski-Ribeiro (2009), may in part be related to our results.

Our study contributes to the discussion by expanding the scope of the analysis to the NMS and includes the most recent years. Additionally, our data set consists of quarterly data having more degrees of freedom and making the analysis more robust.

3 The 2004 enlargement

Fiscal policies in the NMS have been characterized by a strong heterogeneity among countries with some running pro-cyclical and others running counter-cyclical policies (Schneider and Zapal 2005). However, none of the studies treated the characteristics of fiscal policies in accordance with a structural break in 2004—a year when the NMS officially became a part of the EU and adopted its fiscal standards. Moreover, previous research took a very narrow view on the recent situation of fiscal policies in the NMS. In fact, none of the studies analyzes the NMS in isolation after 2006. Eventually, we may find an evidence in favor of the pro-cyclicality of fiscal policies in the NMS until 2006 with barely any insight on what happened afterward.

Even though the year 2004 can be considered as influential for the fiscal policies among the NMS, as it was the official date of adopting European standards, it is very likely that some of the countries had anticipated the accession to the EU and had changed their fiscal behavior before. Therefore, in this study we follow a two-step procedure. In the first step, we investigate whether there was an actual statistically significant break in fiscal policies across the NMS. We roll the breaking point quarter by quarter, from Q3 2000 until Q3 2007.3 We then run the Chow (1960) test on the fiscal policies’ coefficients, determining which of them were the most statistically different from each other between periods. We rank the breaking points by their statistical significance, and we determine the most likely candidates for the structural break. In the second step, we estimate the profile of fiscal policies in the NMS quantitatively, assuming the structural break point we have found in the first step. We adopt the methodology proposed by Gali and Perotti (2003) and extended by Candelon et al. (2010). The idea is to estimate the influence of the economy output gap on discretionary policy, which is approximated by the structural (primary) budget balance. In other words, fiscal policy is assessed from the perspective of a reference value that is not influenced by the cyclical component. In this study, the reference value is the potential GDP.

The workhorse panel model specification may be written as:
$$\begin{aligned} \begin{aligned} d_{i,t}&= \beta _0 + \beta ^{\mathrm{{BEU}}}_d d_{i,t-1} + \beta ^{\mathrm{{AEU}}}_d d_{i,t-1} \\&\quad + \beta ^{\mathrm{{BEU}}}_{\mathrm{{gap}}}E_{i,t-1}\left[ \mathrm{{gap}}^{\mathrm{{BEU}}}_{i,t}\right] + \beta ^{\mathrm{{AEU}}}_{\mathrm{{gap}}}E_{i,t-1}\left[ \mathrm{{gap}}^{\mathrm{{AEU}}}_{i,t}\right] \\&\quad + \beta ^{\mathrm{{BEU}}}_b b_{i,t-1} + \beta ^{\mathrm{{AEU}}}_b b_{i,t-1} + B^{\mathrm{{BEU}}} {\varOmega }^{\mathrm{{BEU}}}_{i,t} + B^{\mathrm{{AEU}}} {\varOmega }^{\mathrm{{AEU}}}_{i,t} + u_{i,t}, \end{aligned} \end{aligned}$$
(1)
where \(d_{i,t}\) is a proxy for structural budget balance, \(E_{i,t-1}[\mathrm{{gap}}_{i,t}]\) reflects the output gap expected at a previous period and \(b_{i,t}\) is the debt level. Matrix \({\varOmega }\) consists of control variables, described in detail later, and B is a corresponding vector of coefficients. Subscripts i and t refer to country and time dimensions, respectively. The error term consists of individual and time-specific effects (\(u_{i,t}=v_i+\varepsilon _{i,t}\)) . The model distinguishes between the influence of output gaps on the discretionary policy before (\(\beta ^{\mathrm{{BEU}}}_{\mathrm{{gap}}}\)) and after (\(\beta ^{\mathrm{{AEU}}}_{\mathrm{{gap}}}\)) the structural break in 2004. Also, the control variables are split before and after EU entry as well. In fact, since we are endowed with quarterly data, we cut the sample in the second quarter of 2004. To correct for the country size, all of the variables are taken relative to GDP.

Taking the expectation of the output gap at a previous period makes a clear sense. Any government decides on fiscal policy using the expected output gap variations (Gali and Perotti 2003). This study assumes rational expectations (\(E_{t-1}[\mathrm{{gap}}_t]=\mathrm{{gap}}_t\)) which is not far from reality taking into account that the fiscal authorities should know the response functions. In fact, perfect foresight assumption implies that our results should not differ from the ones estimated on a real-time data set, as the ex-post data corrections should be negligible. Due to scarcity of the real-time data in the majority of the EU countries, similar procedures are commonly applied in the literature (Gali and Perotti 2003; Candelon et al. 2010; Wyplosz 2006; Marinheiro 2005).

Positive values of \(\beta ^{\mathrm{{BEU}}}_{\mathrm{{gap}}}\) or \(\beta ^{\mathrm{{AEU}}}_{\mathrm{{gap}}}\) indicate counter-cyclical discretionary policy in a given period. In fact, this suggests that a fiscal authority was increasing structural budget balance during upswings which is exactly the definition of a counter-cyclical fiscal behavior. The \(\beta _d\) coefficient suggests that a fiscal authority is concerned about previous period conditions and tries to include them in the current fiscal policy. This makes a clear sense. If a government is worried about the deficit levels from the past, it should increase today’s balance to stabilize it. Additionally, more technically speaking \(\beta _d >0\) indicates extrapolative or a kind of trend-following while \(\beta _d <0\) reveals mean reversive behavior. The stationarity of the deficit time series is supported by \(|\beta _d|<1\) (Roodman 2006).

The specification of the model may lead to an endogeneity bias which arises between structural deficit and output gap. To correct for that, the IV technique is used with a standard fixed effects model.4 To determine the proper instruments, we follow closely the settings from Gali and Perotti (2003) and Candelon et al. (2010) and use up to the second lag of the output gap in the USA. Clearly, one would expect a high correlation between the lagged US and current EU production as a result of high trade flows and foreign direct investments. It is, however, doubtful to find a relationship between it and the structural budget deficits across the EU, which should make it a good instrument. Additionally, its validity is assessed statistically.

To further confirm our findings, we run several robustness checks, which include estimating the relationships on the annualized data and dynamic panel estimation technique.

3.1 Data

As we are interested in studying the effects of the adoption of EU fiscal rules on discretionary fiscal behavior on the example of the 2004 enlargement, for a first step, an illustration of the relevant variables is a good mean to get an idea of the economic circumstances and public finances in the NMS and the OMS. Here, a special focus may be set to second quarter in 2004, as the ten NMS officially joined on May 1, 2004. The data for the following figures comes from the European Central Bank (2013) and the International Monetary Fund (2013a). They show for each of our main variable series the averages of both groups over the considered time horizon.5

The variables we take into consideration in our analysis refer to the central public finance indicators as well as to instruments which governments may influence. As a proxy for measuring discretionary fiscal policy design, the dependent variable in our setting, we chose the seasonally adjusted primary balance to GDP ratio (Gali and Perotti 2003). The primary balance is a parameter which the administration can directly influence through its budget, tax or spending (in general: fiscal) policy. This is also in line with similar studies such as Claeys (2008) or in some cases Staehr (2007) for example. Moreover, the total budget balance is influenced by ‘inherited’ decisions from the past, as it differs to the extent of including interest payments on outstanding debt. This squeezes the available resources and limits the policy decisions. However, we do not neglect that influence in our regressions as we include both, the debt ratio variable and an interest rate gap regressor to capture their effects.6

The data of our seasonally adjusted primary balance ratios (averages) are depicted in Fig. 1 for the NMS and the OMS.
Fig. 1

Primary balance ratio. Source: ECB (2013), own calculations

What concerns the primary balance ratios for the NMS, there is a one-off effect in Hungary in the first quarter 2011 as a distinctive feature. This irregular budgetary surplus was due to revenues from a shift of pension assets (International Monetary Fund 2012).7 Besides that, the average primary balance ratio mainly increases over time until it drops with the crisis. For the OMS, generally a slight negative tendency of the primary surplus ratios may be observed with some improvement around 2006, recently there is again a pronounced decline, which is in part caused by the extreme values of Ireland in 2010, due to stabilization of the banking sector. Soon afterward, in November 2010, Ireland applied for financial assistance with a volume of 85 billion euros (European Financial Stability Facility (EFSF) EFSF 2013). Also for Greece there is a severe deterioration of the primary balance ratio in the crisis years, however, for both economies there appears to be some improvement with the recent observations.

Regarding the explanatory variables, our central regressor is the output gap ratio. It is constructed by calculating the difference between the observed real GDP value, denote it by Y, and the real GDP trend, \(Y^*\), relative to that trend: \((Y-Y^*)/Y^*\). More precisely, for the calculation of the GDP trend \(Y^*\), the real GDP series had been filtered with a Hodrick-Prescott method in \(R \ 2.9.0\). As it is quarterly frequency, we applied the commonly used smoothing parameter \(\mu =1600\).

Obviously, positive values indicate economic booms while negative values represent downswings. The data for the seasonally adjusted averages are depicted in Fig. 2 for both, the NMS and the OMS. The average output gap ratio of the NMS generally shows a slight positive trend with a small decline for the early 2000s followed by an almost steady increase until about 2007, which, however, then collapsed with the financial and debt crisis. Extreme negative values were conducted in Latvia and Estonia in 2009. Latvia also achieved the highest output gap ratio in the last quarter of 2006. Recently, some catching up tendency for most economies can be observed. For the OMS, the general pattern is similar: The output gap ratios mainly decrease toward the early 2000s; afterward, they improve until about 2007 with especially large positive values for Finland and Luxembourg. Certainly, the situation deteriorated with the outbreak of the crisis. Nevertheless, the recent observations indicate a slow recovery and upward trend in the output gap for many of the NMS and OMS.
Fig. 2

Output gap ratios. Source: ECB (2013), IMF (2013a), own calculations

Due to the recent developments with the financial and debt crisis afflicting Europe, the plots of the debt-to-GDP ratios and the interest rate gap (compared to Germany’s long-run bond yield) have been chosen to visualize the situation for the time span 2000–2011. Certainly, the public debt-to-GDP ratio is included as an explanatory variable, since it is one of the reference indicators from the MT which recently has become important as a result of the financial and debt crises. (For endogeneity and likely correlation effects, it is included in lagged terms.)

Without any doubt, the evolution of the seasonally adjusted average debt-to-GDP ratios in Fig. 3 has significantly been shaped by the crisis. Generally, the average debt ratio level of the OMS is higher than in the NMS. Moreover, in some OMS countries, for instance in Belgium, Greece and Italy, the initial values in 2000 were above 100 %. Moreover, for the group of the NMS Fig. 3 shows a slight decline of the debt ratio after the 2004 enlargement (our reference year for the behavior change due to the EU extension). This effect is hardly observable for the OMS. Thus, while for the OMS the distinctive feature here is an increase in public debt during the crisis, many of the NMS’s debt ratios additionally tend to be shaped by the 2004 enlargement.
Fig. 3

Debt ratios. Source: ECB (2013), own calculations

Further, in order to take into account possible crisis volatility, interest rate spreads are included in our model. They are determined by the difference between the individual long-run government bond yield compared to the reference value of the German long-run bond yield. Figure 4 plots the seasonally adjusted averages for the NMS and the OMS. The development of the spreads reveals a familiar pattern: Fig. 4 shows a characteristic almost u-shaped run for the NMS, which displays the dynamics of the proceeding of the European integration, too. On average NMS’s interest rate gaps were remarkably decreasing until about 2004 and then stayed low until about 2008. With the outbreak of the crisis, they began to rise again. For the OMS’s spread development in, again, the crisis years shape the characteristic run.
Fig. 4

Interest rate gaps. Source: IMF (2013a), own calculations

Moreover, to capture potential influence of political effects, a dummy variable for parliamentary elections is added to the model for the years when the election took place, similar to Lewis (2009) and Poplawski-Ribeiro (2009). Such a specification corrects for possible bias resulting from irresponsible budgetary spending in the election years (De Haan et al. 2003).

Usually, empirical analyses are limited by data quality and availability, which holds especially true for studies concerning NMS. Therefore, we constructed a data set combining data from different (but few) sources. It mainly consists of data from European Central Bank (2013) Statistical Data Warehouse for the governmental data (public finances and GDP) and International Monetary Fund (2013a) International Financial Statistics for the interest rates and the deflator. Concerning the demarcation of the public sector, our data set mainly refers to general government, i.e., including lower government tiers.8 The election variable reflects parliamentary election dates and has been taken from IFES (2013). Table 7 in ‘Appendix 1’ summarizes the data information and provides some descriptive statistics on the variables for the two groups of member states. Interestingly, on average the primary balance for both, NMS and OMS, is in deficit and for the OMS the average debt ratio is close to the MT reference value. Plus, on average interest rate gaps are higher in NMS than in OMS, and the average output gaps are close to balance. For both groups, the lowest debt ratio levels were achieved in 2008, with their maxima held at the end of the series in 2011 with the crisis. The output gap minimum values for both groups were in Q2 2009, the highest values for the NMS were performed at the end of 2007, and for the OMS at the beginning of 2008. The largest primary deficits for the NMS were also in Q2 2009, however, for the OMS in Q1 2010.9 Further, both groups benefited from lowest values in interest rate gap at the beginning of 2006, whereas the largest values are given by the recent observations.

Regarding the frequency, we employed quarterly data in order to enhance the time span. This is in line with Claeys (2008) or Darvas (2009). Even though fiscal and budget decisions are made on a yearly basis, our proceeding allows for a more profound empirical analysis and conveys more power to the results. Nevertheless, as a robustness check the estimations are also run with annual data. Thus, our data cover the years from 2000 (Q1) until 2011 (Q4). This yields 48 observations per country and a total size of \(N = 480\) for NMS and \(N = 720\) for OMS. Obtaining a continuous and consistent series required a thorough data set preparation. Seldom missing observations had been extrapolated in order to make the panel balanced.

As regards our control group of countries, we decided on two ‘safe havens,’ Australia and Canada, and two European countries that joined EU later (in 2007), Bulgaria and Romania. The safe havens data stems from OECD (2014a, b). For the two 2007 EU members the same data sources as for the NMS group has been used.10

Finally, and very importantly, the original data in our set were raw or crude (in terms of not being seasonally adjusted). Thus, at last all variables had been seasonally adjusted with X12 ARIMA methodology.

All but one variable are stationary at standard significance levels according to the Levin–Lin–Chu unit-root test (2002). The only non-stationary variable is the debt level. As a robustness check, we run all the regressions without the debt level, finding no qualitative difference from our main findings. Therefore, we decide to keep the debt variable in our regressions.

3.2 Results

In accordance with our two-step procedure, we first determine the breaking points in fiscal policies for the OMS and NMS. We run the model from Eq. 1 with a rolling breaking point and using the Chow test we determine which of them is the most likely candidate for the change. We rank the all the points by their p-values, where the higher the rank the lower the significance level of a change.11 The results for both the NMS and OMS are presented in Fig. 5.
Fig. 5

The rank of the statistical significance of various breaking points in fiscal policies among the NMS and OMS. The higher the rank, the lower the p value from the Chow test of no structural change

The results indicate that for the NMS the most likely breaking points fall within the period Q3 2003–Q2 2004, confirming our hypothesis that the NMS anticipated the EU accession and adjusted their fiscal standards before Q1 2004. The break in fiscal policies among the OMS was most likely to happen between Q3 2000 and Q3 2003, being somehow more in line with the consequences of the adoption of the common currency.

In our analysis, we decide to use the last quarter from the highlighted period to balance the number of observations between both time spans. Therefore, in the forthcoming analysis, we use Q2 2004 as a structural break point. We confirm the results by varying the break point from Q3 2003 until Q2 2004, finding no differences with the main results.

The regression outcomes for the NMS, estimated on the model from Eq. 1, are presented in Table 1.12 For clarity reasons, we present the results for all the model specifications. First, for the full sample and standard regressors (lagged budget balance, output gap ratio and lagged debt ratio—all before and after EU entry) see Model 1, presented in the first column. Taking into account the current crisis, we followed a twofold approach. On the one hand, we cut the sample and consider only the time period before the crisis (Q12001–Q42008), model type 2, second column. On the other hand, we include the interest rate gap, model type 3, third column. Finally, as an additional specification, we also include an election dummy variable, capturing the influence of politics, types 4 and 5, forth and fifth columns, respectively. The reason for cutting the sample in the last quarter of 2008 is twofold: On the one hand, this allows to equalize the number of the degrees of freedom, and on the other hand, it enables to observe the behavior before the crisis began.
Table 1

Discretionary fiscal rule in the New Member States before (BEU) and after (AEU) accession to the EU

Model

1

2

3

4

5

lag. deficit (BEU)

0.224**

0.216***

0.222**

0.225**

0.223**

(0.089)

(0.065)

(0.088)

(0.089)

(0.088)

lag. deficit (AEU)

0.314***

0.395***

0.265***

0.307***

0.260***

(0.05)

(0.08)

(0.05)

(0.05)

(0.05)

H0: BEU \(=\) AEU

0.362

0.063*

0.664

0.407

0.709

output gap (BEU)

\(-\)0.01

0.036

0.014

\(-\)0.014

0.009

(0.124)

(0.085)

(0.122)

(0.125)

(0.123)

output gap (AEU)

0.337***

0.178***

0.206***

0.349***

0.217***

(0.058)

(0.067)

(0.065)

(0.059)

(0.066)

H0: BEU \(=\) AEU

0.011**

0.198

0.167

0.009***

0.137

lag. debt (BEU)

\(-\)0.024

\(-\)0.053*

\(-\)0.003

\(-\)0.021

\(-\)0.001

(0.026)

(0.03)

(0.026)

(0.026)

(0.026)

lag. debt (AEU)

\(-\)0.001

\(-\)0.015

0.0333

0.001

0.035

(0.022)

(0.029)

(0.023)

(0.022)

(0.023)

H0: BEU \(=\) AEU

0.059*

0.000***

0.004***

0.06*

0.005***

int. gap (BEU)

  

\(-\)0.075

 

\(-\)0.073

  

(0.141)

 

(0.141)

int. gap (AEU)

  

\(-\)0.529***

 

\(-\)0.524***

  

(0.123)

 

(0.123)

H0: BEU \(=\) AEU

  

0.005***

 

0.005***

election (BEU)

   

\(-\)0.397

\(-\)0.485

   

(0.976)

(0.957)

election (AEU)

   

\(-\)0.808

\(-\)0.643

   

(0.698)

(0.685)

H0: BEU \(=\) AEU

   

0.731

0.893

Observations

460

350

460

460

460

\(R^{2}\)

0.236

0.252

0.267

0.238

0.268

Groups

10

10

10

10

10

Instruments

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

 

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

Sargan

0.069

0.86

0.067

0.064

0.063

Model 1: the benchmark model, i.e., full sample and standard regressors. Model 2: the same as 1 but on the truncated sample Q12001–Q42008. Model 3: the same as 1 but with the interest rate gap. Model 4: the same as 1 but with the election dummy. Model 5: the same as 1 but with the interest rate gap and the election dummy. Coefficients are reported as well as robust standard errors between brackets. *, ** and *** imply significance at the 10, 5, and 1 % significance level, respectively. Equality between coefficients is tested by a standard F test, whose p-value is indicated in the separate row H0: BEU \(=\) AEU. The adequacy of instrument space is tested with the Sargan test

The results show that the coefficients of interest \(\beta _\mathrm{{gap}}^{\mathrm{{BEU}}}\) and \(\beta _\mathrm{{gap}}^{\mathrm{{AEU}}}\), which correspond to the effect of the output gap ratio before and after EU enlargement in 2004 on the primary balance ratio, are statistically significant after EU entry only. Interestingly, all \(\beta _\mathrm{{gap}}^{\mathrm{{AEU}}}\) coefficients are positive, indicating counter-cyclical fiscal policy in NMS after joining the EU. This holds true for all model type specifications. The results suggest also that the discretionary fiscal policies before the 2004 entry were acyclical, in a sense that there was no statistically significant influence from the output gap on the structural balance. This is somehow confirming the results from Schneider and Zapal (2005) but on the aggregate level, with some countries running pro-cyclical and the other having counter-cyclical budgetary positions. These results might be interpreted in two ways proving that either the heterogeneity in fiscal behavior decreased and the fiscal change was relatively small, or the heterogeneity persisted with the 2004 effect being relatively larger. We investigate this phenomenon in detail in Sect. 3.5.

Concerning the other coefficients, only the lagged primary balance turns out to be statistically significant for all of the model specifications. As expected the coefficient is positive and lies on the unit circle, confirming the stationarity (Roodman 2006). In line with the comments from above concerning the \(\beta _d\) being smaller or larger than zero, the results suggest a trend-following behavior. This reflects objections of the administration about previous periods conditions. Further, the interest rate gap variable proves to affect the primary balance significantly and negatively after 2004 entry. This may display the influence of the current crisis. The higher the gap, i.e., the worse the economic condition of the economy, the higher the deficit in this period. The negative election effect indicates lower surpluses in times of elections (potentially ‘election gifts’), also in line with theory (De Haan et al. 2003); however, our estimation results show that those effects are not statistically significant.

Interpreting our results for the NMS from an economic policy perspective yields very interesting insights on the actual discretionary fiscal policies in a given time span. The outcomes reveal a statistically significant change toward the counter-cyclical fiscal behavior after 2004. The result is robust with respect to 2 out of 5 model specifications; however, the F-tests for all the remaining 3 specifications reject the null of no change at 0.2 significance level. When looking at the results, one should pay attention to the fact that the change was always from acyclical to counter-cyclical. Given the limited sample size and the data shortcomings, we consider those results as convincing, however, to underpin our findings we investigate the robustness of the results in Sect. 3.4.

Year 2004 brought a significant change to the debt and interest rate gap in all the model specifications. One may claim that this is a result of an increased number of degrees of freedom for the after-2004 specification. However, even if we correct for the sample size, the results remain similar. In fact, it seems that the EU entry did not only change the discretionary fiscal behavior among the NMS, but it also shaped the fiscal responses against the debt and interest rate gap factors.

One should pay attention to the possible over-identification problems, resulting from the choice of instruments. The number of lags in the instrument space was chosen in order to yield a maximum Sargan test p-value. Nevertheless, they might seem troubling for some of the readers. When considering the p-values for the standard Sargan test for over-identifying restrictions, one should keep in mind the possible heteroscedasticity of the errors which could influence the test performance (Roodman 2006). In fact, the magnitude of the p-values is similar to the one presented in Candelon et al. (2010) for the comparable model specifications, being acceptable at the 5 % significance level. Therefore, in order to draw conclusions, one should pay attention to the holistic perspective of our estimation results. The well-specified model, i.e., model with the truncated time span, confirms the main inference from the other models together with satisfying the over-identification restrictions. Having pointed this out, we reproduce models 3–5 from Table 1 on the truncated sample to limit the over-identification problems. The main results are confirmed: In all the settings, the NMS run acyclical fiscal policies before the accession to the EU and counter-cyclical afterward (see Table 8 in the ‘Appendix 1’). In Sect. 3.4, we apply the dynamic panel estimation technique, yielding a better instrument space and confirming our main conclusions.

Interestingly, one may think of the reason why there are over-identification problems with years after 2008. An explanation could lie in the increased heterogeneity in reactions to the financial crisis among the NMS. Clearly, different NMS got different exposure to foreign risks and were differently affected by the crisis. For instance, countries such as Lithuania, Estonia or Latvia suffered from a huge economic losses immediately after the US crisis. At the same time, the economic situation in Cyprus or Slovenia was mostly affected in the aftermath of the sovereign debt crisis which spread across Europe two years later. Countries such as Poland were pretty robust to the foreign shocks as a result of a strong internal consumption and a relatively low degree of internalization of the capital flows. Hungary, however, is a kind of a special case, as the situation may be reckoned to be strongly politically influenced. Concerning Slovakia and Czech Republic, the situation indicates to be relatively similar to Poland, while the behavior of Malta resembles Cyprus.

3.3 Comparison with the OMS and other countries

The results for the OMS are presented in Table 2. The model specifications remain the same as in the previous section.
Table 2

Discretionary fiscal rule in the Old Member States before (BEU) and after (AEU) the 2004 enlargement

Model

1

2

3

4

5

lag. deficit (BEU)

0.607***

0.435***

0.614***

0.609***

0.616***

(0.063)

(0.052)

(0.063)

(0.063)

(0.063)

lag. deficit (AEU)

0.631***

0.551***

0.616***

0.633***

0.619***

(0.038)

(0.056)

(0.039)

(0.038)

(0.039)

H0: BEU \(=\) AEU

0.747

0.077*

0.976

0.738

0.968

output gap (BEU)

0.04

0.252***

\(-\)0.001

0.038

\(-\)0.002

(0.131)

(0.095)

(0.133)

(0.131)

(0.133)

output gap (AEU)

0.227***

0.111

0.24***

0.223***

0.236***

(0.065)

(0.069)

(0.065)

(0.065)

(0.066)

H0: BEU \(=\) AEU

0.201

0.237

0.106

0.206

0.108

lag. debt (BEU)

\(-\)0.035***

0.014

\(-\)0.056***

\(-\)0.034***

\(-\)0.055***

(0.013)

(0.019)

(0.016)

(0.012)

(0.0160)

lag. debt (AEU)

\(-\)0.0422***

0.0155

\(-\)0.0620***

\(-\)0.0420***

\(-\)0.0617***

(0.0131)

(0.0201)

(0.0167)

(0.0131)

(0.017)

H0: BEU \(=\) AEU

0.049**

0.592

0.152

0.042**

0.142

int. gap (BEU)

  

1.133

 

1.145

  

(0.81)

 

(0.812)

int. gap (AEU)

  

0.189*

 

0.187*

  

(0.104)

 

(0.104)

H0: BEU \(=\) AEU

  

0.243

 

0.237

election (BEU)

   

0.032

0.074

   

(0.68)

(0.679)

election (AEU)

   

0.483

0.479

   

(0.501)

(0.499)

H0: BEU \(=\) AEU

   

0.592

0.629

Observations

688

522

688

688

688

\(R^{2}\)

0.552

0.286

0.556

0.553

0.556

Groups

15

15

15

15

15

Instruments

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

 

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

Sargan

0.000

0.018

0.000

0.000

0.000

Model 1: the benchmark model, i.e., full sample and standard regressors. Model 2: the same as 1 but on the truncated sample Q12001–Q42008. Model 3: the same as 1 but with the interest rate gap. Model 4: the same as 1 but with the election dummy. Model 5: the same as 1 but with the interest rate gap and the election dummy. Coefficients are reported as well as standard deviation between brackets. *, ** and *** imply significance at the 10, 5 and 1 % significance level, respectively. Equality between coefficients is tested by a standard F test, whose p-value is indicated in the separate row H0: BEU \(=\) AEU. The adequacy of instrument space is tested with the Sargan test

Since the results suffer from the over-identification bias, they should be taken with caution and should be treated as a comparison for the NMS rather than suiting to draw actual policy conclusions from. The only model specification which is not over-identified at the 1 % significance level, i.e., the model for the truncated time span, suggests, however, that there was no statistically significant change in fiscal behavior for the OMS in year 2004. Interestingly, the BEU coefficient in this model is counter-cyclical with the AEU coefficient being acyclical. The other model specifications show the opposite, however, the change is never statistically significant. This outcome may suggest counter-cyclical policy after 2004 being highly influenced by crisis effects (IMF 2013b). On the one hand, without the crisis years (the truncated sample), the coefficient of interest looses statistical significance, and on the other hand, including the crisis years yields significant positive effects of the output gap and negative effects of the debt ratio (unsustainable behavior).

The goal of this comparison is to check whether there were any other factors in 2004 which could have led to a switch in the discretionary fiscal policies among the member states. It seems that the 2004 effect could be clearly recognized for the NMS only, with no clear-cut evidence for a change in the other member states. Interestingly, one may get an impression that the fiscal policies in the OMS are shaped by different factors than those in the NMS. The over-identification problems and the counter-intuitive sign for the lagged debt variable are just symptoms of that phenomenon. One may also observe no political influence on the fiscal behavior, together with much higher \(R^{2}\) coefficients.

In fact, this might be the confirmation of previously mentioned literature, which pointed out the problem of lack of incentives to run more counter-cyclical fiscal policies and curb down the deficit and debt levels once in the EU (Wyplosz 2002, 2006). Our analysis confirms that the NMS were trying to satisfy the MT and SGP regulations and they actually changed their fiscal behavior entering the EU. It is difficult, however, to determine the fiscal stance of the OMS. Certainly, they did not change their fiscal policies in 2004. Moreover, their fiscal behavior was mostly driven by the crisis effects with barely any insight on other indicators. All this might be viewed as a contribution toward the ongoing discussion on the relevance and effectiveness of the EU fiscal standards.

In order to show whether the 2004 effects were the results of the EU accession or they could have been shaped by different factors, we run the same regressions on the sample of countries which did not belong to the EU at that time, i.e., Canada and Australia, together with countries which entered the EU later, i.e., Romania and Bulgaria. The results are presented in Table 3.
Table 3

Discretionary fiscal rule in Canada, Australia, Romania and Bulgaria before (BEU) and after (AEU) the 2004 enlargement

Model

1

2

3

4

5

lag. deficit (BEU)

0.199

0.257

0.026

0.202

0.022

(0.163)

(0.163)

(0.199)

(0.163)

(0.2)

lag. deficit (AEU)

0.521***

0.153

0.468***

0.519***

0.468***

(0.065)

(0.111)

(0.071)

(0.065)

(0.071)

H0: BEU \(=\) AEU

0.049**

0.581

0.03**

0.053*

0.029**

output gap (BEU)

0.063

0.075

0.096

0.062

0.088

(0.118)

(0.103)

(0.12)

(0.12)

(0.121)

output gap (AEU)

0.188**

0.052

0.148*

0.186**

0.147*

(0.083)

(0.109)

(0.084)

(0.083)

(0.084)

H0: BEU \(=\) AEU

0.397

0.882

0.73

0.409

0.699

lag. debt (BEU)

\(-\)0.004

\(-\)0.019

\(-\)0.018

\(-\)0.005

\(-\)0.019

(0.017)

(0.024)

(0.018)

(0.017)

(0.018)

lag. debt (AEU)

\(-\)0.035*

\(-\)0.027

\(-\)0.054**

\(-\)0.035*

\(-\)0.054**

(0.021)

(0.031)

(0.022)

(0.021)

(0.022)

H0: BEU \(=\) AEU

0.0012***

0.522

0.001***

0.002***

0.001***

int. gap (BEU)

  

\(-\)0.088**

 

\(-\)0.09**

  

(0.045)

 

(0.045)

int. gap (AEU)

  

\(-\)0.434**

 

\(-\)0.427**

  

(0.189)

 

(0.191)

H0: BEU \(=\) AEU

  

0.044**

 

0.051*

election (BEU)

   

0.083

0.422

   

(1.246)

(1.255)

election (AEU)

   

\(-\)0.513

\(-\)0.260

   

(0.871)

(0.865)

H0: BEU \(=\) AEU

   

0.694

0.653

Observations

184

140

184

184

184

\(R^{2}\)

0.397

0.060

0.417

0.398

0.418

Groups

4

4

4

4

4

Instruments

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

 

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

Sargan

0.007

0.351

0.008

0.006

0.008

Model 1: the benchmark model, i.e., full sample and standard regressors. Model 2: the same as 1 but on the truncated sample Q12001–Q42008. Model 3: the same as 1 but with the interest rate gap. Model 4: the same as 1 but with the election dummy. Model 5: the same as 1 but with the interest rate gap and the election dummy. Coefficients are reported as well as standard deviation between brackets. *, ** and *** imply significance at the 10, 5 and 1 % significance level, respectively. Equality between coefficients is tested by a standard F test, whose p-value is indicated in the separate row H0: BEU \(=\) AEU. The adequacy of instrument space is tested with the Sargan test

Interestingly, the output gap variable is weakly statistically significant after 2004 in all but the truncated-sample specification. The change from acyclical toward counter-cyclical behavior is, however, never statistically significant. Similar to the OMS, the other regressions suffer from the over-identification problems and unsuitable debt behavior and should therefore be treated through a prism of illustrative purposes. The only model which passes the Sargan test, i.e., model 2, suggests that there was no change in fiscal behavior in 2004. International Monetary Fund (2013b) delivers evidence that the advanced economies applied counter-cyclical fiscal policies as crisis-management measures in the aftermath of the global financial crisis. Our sample covers only two of the non-EU advanced economies, i.e., Australia and Canada, being non-representative and too small to confirm this result empirically. Nevertheless, this reflection builds an interesting scope for future research, including applications of fiscal policies as fire-fighting tools for the advanced and emerging economies.

3.4 Robustness of the results

We carry out two robustness checks for the NMS estimations. Firstly, in order to control for the quarterly fiscal effects, we run the regressions on the annualized data. Secondly, we apply the dynamic panel estimator, as proposed by Candelon et al. (2010), in order to check the results on a different instrument space. The results for the former are presented in Table 4 and for the later in Table 5.
Table 4

Discretionary fiscal rule in the New Member States on the annualized data before (BEU) and after (AEU) the 2004 enlargement

Model

1

2

3

4

5

lagged deficit (BEU)

0.351*

0.434***

0.331*

0.322*

0.294*

(0.182)

(0.129)

(0.174)

(0.172)

(0.165)

lagged deficit (AEU)

0.751***

0.475***

0.672***

0.713***

0.646***

(0.099)

(0.156)

(0.095)

(0.095)

(0.091)

H0: BEU \(=\) AEU

0.041**

0.817

0.07*

0.036**

0.047**

output gap (BEU)

\(-\)0.075

\(-\)0.094

\(-\)0.097

\(-\)0.148

\(-\)0.220

(0.227)

(0.183)

(0.220)

(0.224)

(0.217)

output gap (AEU)

0.317***

0.211**

0.145*

0.366***

0.182**

(0.07)

(0.107)

(0.082)

(0.069)

(0.082)

H0: BEU \(=\) AEU

0.1*

0.17

0.295

0.027**

0.074*

lagged debt (BEU)

0.181***

0.141***

0.185***

0.186***

0.188***

(0.037)

(0.05)

(0.035)

(0.035)

(0.033)

lagged debt (AEU)

0.213***

0.159***

0.227***

0.221***

0.232***

(0.035)

(0.047)

(0.033)

(0.033)

(0.031)

H0: BEU \(=\) AEU

0.113

0.284

0.03**

0.065*

0.017**

interest gap (BEU)

  

\(-\)0.257

 

\(-\)0.313

  

(0.275)

 

(0.261)

interest gap (AEU)

  

\(-\)0.522***

 

\(-\)0.521***

  

(0.15)

 

(0.145)

H0: BEU \(=\) AEU

  

0.303

 

0.397

election (BEU)

   

\(-\)1.28*

\(-\)1.637**

   

(0.71)

(0.672)

election (AEU)

   

\(-\)1.322***

\(-\)1.046**

   

(0.479)

(0.455)

H0: BEU \(=\) AEU

   

0.959

0.46

Observations

100

80

100

100

100

\(R^{2}\)

0.587

0.437

0.637

0.629

0.676

Groups

10

10

10

10

10

Instruments

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-1}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

 

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

\(\mathrm{{GAP}}_{\mathrm{{US}},t-2}\)

Sargan

0.155

0.152

0.23

0.14

0.272

Model 1: the benchmark model, i.e., full sample and standard regressors. Model 2: the same as 1 but on the truncated sample Q12001–Q42008. Model 3: the same as 1 but with the interest rate gap. Model 4: the same as 1 but with the election dummy. Model 5: the same as 1 but with the interest rate gap and the election dummy. Coefficients are reported as well as standard deviation between brackets. *, ** and *** imply significance at the 10, 5 and 1 % significance level, respectively. Equality between coefficients is tested by a standard F test, whose p-value is indicated in the separate row H0: BEU \(=\) AEU. The adequacy of instrument space is tested with the Sargan test

Table 5

Discretionary fiscal rule in the New Member States on the annualized data with the dynamic panel estimators before (BEU) and after (AEU) the 2004 enlargement

Model

1

2

3

4

5

lagged deficit (BEU)

0.648

2.515

\(-\)0.132

6.008

\(-\)16.87

(1.89)

(2.354)

(1.96)

(3.299)

(20.53)

lagged deficit (AEU)

0.908**

0.924*

4.433

\(-\)0.407

22.44

(0.379)

(0.502)

(4.153)

(0.639)

(20.57)

H0: BEU \(=\) AEU

0.907

0.585

0.398

0.130

0.362

output gap (BEU)

\(-\)0.791

\(-\)2.486

13.04

\(-\)8.815*

74.68

(2.545)

(3.908)

(11.17)

(4.775)

(73.70)

output gap (AEU)

0.447***

0.261

\(-\)3.928

0.647***

\(-\)19.37

(0.094)

(0.262)

(4.507)

(0.173)

(18.04)

H0: BEU \(=\) AEU

0.641

0.493

0.301

0.084*

0.332

lagged debt (BEU)

0.129

0.195

1.437

0.216

6.679

(0.117)

(0.134)

(1.462)

(0.213)

(5.9)

lagged debt (AEU)

0.162

0.061

1.93

\(-\)0.328

10.58

(0.135)

(0.177)

(1.991)

(0.214)

(9.875)

H0: BEU \(=\) AEU

0.856

0.548

0.398

0.135

0.354

interest gap (BEU)

  

\(-\)0.351

 

\(-\)3.474

  

(0.652)

 

(4.037)

interest gap (AEU)

  

\(-\)10.38

 

\(-\)46.08

  

(10.53)

 

(41.59)

H0: BEU \(=\) AEU

  

0.743

 

0.873

election (BEU)

   

\(-\)12.08*

43.82

   

(5.518)

(53.71)

election (AEU)

   

0.007

\(-\)11.89

   

(2.66)

(11.95)

H0: BEU \(=\) AEU

   

0.123

0.352

Observations

110

80

110

110

110

Groups

10

10

10

10

10

AR(1)

0.126

.

0.162

.

0.562

AR(2)

0.257

0.167

.

0.396

.

Sargan

0.257

0.0866

0.417

0.391

0.368

Variables are instrumented with the next lags of the dependent variables in the two-step procedure with the small-sample correction and orthogonal transformation. Model 1: the benchmark model, i.e., full sample and standard regressors. Model 2: the same as 1 but on the truncated sample Q12001–Q42008. Model 3: the same as 1 but with the interest rate gap. Model 4: the same as 1 but with the election dummy. Model 5: the same as 1 but with the interest rate gap and the election dummy. Coefficients are reported as well as standard deviation between brackets. *, ** and *** imply significance at the 10, 5 and 1 % significance level, respectively. Equality between coefficients is tested by a standard F-test, whose p-value is indicated in the separate row H0: BEU \(=\) AEU. The adequacy of instrument space is tested with the Sargan test. The autocorrelation of the error terms is tested with the standard AR test

The results obtained from the annualized data confirm the findings from the quarterly data, indicating that the fiscal policies after 2004 were always counter-cyclical and statistically significant, with their earlier counterparts being acyclical. The change toward acyclical behavior was statistically significant in 3 out of 5 model specifications. Importantly, none of the models suffer from the over-identification problems, not the unsustainable debt behavior. The coefficients on the interest gap are preserved from the quarterly data estimations. Interestingly, the annualized data show the statistical significance of the negative election effects, confirming the ‘election gifts’ hypothesis (De Haan et al. 2003).

We run the dynamic panel estimations on the annualized data to comply with the ‘large-N and small-T’ requirement suggested by Roodman (2006). We instrument the variables of interest with their first and second lags. We apply the standard two-step Arellano-Bond dynamic panel estimator (Arellano and Bond 1991), with the small-sample correction (Windmeijer 2005) and orthogonal transformation on the instrument space in order to yield a higher number of degrees of freedom (Roodman 2006).

Choosing a different instruments space yields better Sargan test results. The only possibly over-identified model is the truncated-sample specification. The other models show no over-identification problems, as expected, together with no autocorrelation in the first- and the second-lagged error terms. The results show much less statistical significance for all variables, which could result from much smaller number of degrees of freedom. The statistically significant coefficients in all the models preserve the findings from previous regressions. In 2 out of 5 models, we find the statistically significant counter-cyclical fiscal policies in the after-2004 period with acyclical or pro-cyclical policies before. We confirm also statistically significant change toward more counter-cyclical behavior in model 4, with also the negative election effects in the pre-EU period.

3.5 Heterogeneity analysis

In order to investigate the heterogeneity in fiscal policies between various NMS, we run a simple partition-clustering analysis in means as advertised by Everitt et al. (2011). We first run the ‘kmeans’ algorithm between deficit, debt and output gap, specifying the numbers of cluster we want to find. We vary the number of clusters from 2 to 5, managing at least 2 countries in each cluster. We then run the standard panel IV model in Eq. 1 with the standard specification on each cluster. We count how many clusters showed a statistically significant change toward counter-cyclical behavior after Q2 2004, and we list the countries which belong to that clusters. The results are presented in Table 6.13
Table 6

Cluster analysis of the fiscal policy changes among the New Member States after the accession to the EU in 2004

Clusters

Changes

Countries

2

1

Cz. Rep., Lithuania, Slovenia, Estonia, Latvia

3

2

Cz. Rep., Lithuania, Slovenia, Estonia, Latvia, Slovakia

4

2

Cz. Rep., Lithuania, Slovenia, Estonia, Latvia

5

2

Cz. Rep., Slovenia, Estonia, Slovakia

Clusters determines the number of clusters which are to be found across all the NMS. Changes shows how many of those clusters changed their fiscal behavior toward more counter-cyclical. Countries which changed their behavior are listed in a separate column

We can observe that the change in the NMS fiscal policies after the EU accession was mostly driven by Czech Republic, Lithuania, Slovenia, Estonia, Latvia and Slovakia. While some of the countries, such as Lithuania, Estonia or Latvia, were largely exposed to the global financial crisis, Slovenia was suffering from the sovereign debt crisis and Czech Republic and Slovakia were quite robust to those events. The common denominator for those countries can lie in their economic size. Relatively smaller countries tend to have more volatile public spending, suggesting that they find it easier to adjust their fiscal behavior (Furceri and Poplawski-Ribeiro 2008).

4 Summary

In line with the economic literature and recent considerations on the European integration and adoption of the fiscal rules, this paper continues and contributes to the discussion by addressing the aspect how fiscal rules affect discretionary policy behavior of governments in the EU. In times like these, with the financial and debt crisis affecting especially Europe, such an analysis certainly is of special relevance.

Our study includes 25 EU member economies, however, a special focus is set on ten NMS, which joined the EU in 2004. Based on quarterly data covering the years 2000 until 2011 and different specifications regarding the explanatory variables, a panel regression with the IV technique has been implemented. In order to study the influence of the adoption of fiscal rules on discretionary policy, the model estimates the relationship between the structural primary balance ratio and the central regressor: the output gap ratio, which represents the economic performance and situation.

The results indicate a statistically significant switch toward counter-cyclical fiscal policy behavior of the NMS after their EU entry. The findings are robust to different data frequency and estimation methods. The specification methods confirm previous findings of possible heterogeneity in fiscal behavior in our sample (Schneider and Zapal 2005). We find that changes in fiscal policies among the NMS were mostly driven by relatively smaller economies. On average, however, NMS were running stabilizing budgetary positions as of the Q2 2004.

For reasons of comparison, the same investigation has been implemented for the OMS and a set of non-member countries, including Canada and Australia, and future-member countries, such as Romania and Bulgaria. The results suggest that the fiscal behavior among the OMS and other countries was driven by different factors from the ones among the NMS. It seems that once in the EU, a member loses the incentives for prudent fiscal behavior. This phenomenon has been recognized in previous studies and is currently a topic of consideration of the European Commission, European Parliament and national governments in the EU (Schaechter et al. 2012). We believe that this study contributes to the ongoing discussion, yielding a mechanism toward more prudent fiscal performance once in the EU.

From a policy point of view, this study suggests that the EU fiscal rules proved to be sufficient to motivate counter-cyclical behavior within the NMS which joined the EU in 2004. Similarly, the counter-cyclical policies of the OMS appeared to be driven by the crisis period rather than EU fiscal framework (see also IMF 2013b). Nevertheless, as the sovereign debt crisis evolved and the macroeconomic performance have further deteriorated, the ongoing fiscal consolidation could have reduced the possibility to run counter-cyclical policies, regardless of whether in the NMS or OMS. With this respect, the EU fiscal framework seems to be ineffective to motivate counter-cyclical actions in the Member States. Further reforms are needed to provide the EU with enough fiscal effectiveness and flexibility to build a solid framework for sustainable growth. Those can include fiscal responsibility laws comprising medium- and long-term fiscal frameworks, fiscal rules, and independent fiscal councils (Darvas 2009).

The debt-overhang problem can also suggest that the fiscal behavior within the EU can be asymmetric (for instance, counter-cyclical during the boost and acyclical or pro-cyclical in the boom). Our policy implications depend largely on the availability of fiscal space in particular Member States. Because the NMS benefit from relatively lower levels of public debt (see Table 7), it is relatively easier for them to run counter-cyclical policies during the boosts. The question of fiscal asymmetries within the EU remains a valid point for policy agenda and deserves further investigation in order to provide the EU with an effective fiscal environment.

Certainly, these results should not be interpreted in isolation. The budgetary policies have been considered as an everlasting topic for discussion in the European history and might be affected by other factors than we included in our analysis. In order to get a comprehensive picture, one should consider country-specific research, treating the fiscal performance of each NMS individually. Our outcomes reveal just a top of an ice berg in the fiscal performance in the NMS after the EU entry.

Footnotes

  1. 1.

    For a detailed analysis of the evolution of fiscal rules in the EU, see Wyplosz (2006).

  2. 2.

    The term ’acyclical’ refers to Gali and Perotti (2003) and means that there was no significant impact of discretionary policy on the business cycle fluctuations.

  3. 3.

    In the analysis, we focus on the period from Q1 2000 until Q2 2008 in order to avoid possible spurious relations caused by the global financial crisis and the sovereign debt crisis.

  4. 4.

    Candelon et al. (2010) applies the dynamic panel method to correct for the endogeneity bias. Originally, the dynamic panel technique was designed for large-N and small-T samples, meaning for settings with many individuals and limited number of periods. In our data set, however, we observe the opposite. Therefore, we favor the panel IV technique over the dynamic panel methods. Nevertheless, we apply the dynamic panel estimation as a robustness check, confirming our main findings.

  5. 5.

    More detailed and disaggregated figures based on countrywise data are included in ‘Appendix 1.’

  6. 6.

    See also Darby and Melitz (2008) for that discussion.

  7. 7.

    This outlier does not bring any qualitative change to the estimation results so as to the final inference. Therefore, we decide to leave it in the data set in order to make it more methodology consistent.

  8. 8.

    Recently, many studies use real-time data (Beetsma and Giuliodori 2008; Lewis 2009). However, we resort to ‘conventional’ final data as availability may be a crucial issue here.

  9. 9.

    This number is strongly influenced by Ireland, as Fig. 7 reveals.

  10. 10.

    Due to data availability, the Romanian interest rate until 2005 is the treasury bill rate.

  11. 11.

    All the breaking points with rank 1 are considered to be the most likely candidates for the structural change.

  12. 12.

    All the estimations have been implemented in STATA 12.

  13. 13.

    The results are robust to all the other model specifications.

Notes

Acknowledgments

Authors would like to thank Cees Diks, the participants of the 11th Annual NBP-SNB Joint Seminar in Starawieś, two anonymous referees and the editor for valuable comments. Any views expressed are only those of authors and do not necessarily represent the views of the European Investment Bank.

References

  1. Alesina A, Compante FR, Tabellini G (2008) Why is fiscal policy often pro-cyclical? J Eur Econ Assoc 6(5):1006–1036CrossRefGoogle Scholar
  2. Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo Evidence and an application to employment equations. Rev Econ Stud 58(2):277–297CrossRefGoogle Scholar
  3. Artis M, Buti M (2000) Close to balance or in surplus: a policy maker’s guide to the implementation of the Stability and Growth Pact. CEPR Discussion Paper 2515:1–38Google Scholar
  4. Beetsma R (2001) Does EMU need a stability pact? In: Brunila A, Buti M, Franco D (eds) The Stability and Growth Pact, Palgrave, Basingstoke, pp 23–52Google Scholar
  5. Beetsma R, Debrun X (2004) Reconciling stability and growth: smart pacts and structural reforms. IMF Staff Papers 51(3)Google Scholar
  6. Beetsma R, Giuliodori M (2008) Fiscal adjustment to cyclical developments in the OECD: an empirical analysis based on real-time data. CEPR Discussion Paper 6692Google Scholar
  7. Beetsma R, Giuliodori M (2010) The macroeconomic costs and benefits of the EMU and other monetary unions: an overview of recent research. J Econ Lit 48:603–641CrossRefGoogle Scholar
  8. Budina N, Van Wijnbergen S (1997) Fiscal policies in Eastern Europe. Oxf Rev Econ Policy 13(2):47–64CrossRefGoogle Scholar
  9. Buti M, Franco D, Ongena H (1997) Budgetary policies during recessions—retrospective application of the Stability and Growth Pact to the post-war period. Recherches conomiques de Louvain/Louvain Econ Rev 63(4):321–366Google Scholar
  10. Buti M, Eijffinger S, Franco D (2003) Revisiting EMU’s Stability Pact: a pragmatic way forward. Oxf Rev Econ Policy 19(1):100–111CrossRefGoogle Scholar
  11. Candelon B, Muysken J, Vermeulen R (2010) Fiscal policy and monetary integration in Europe: an update. Oxf Econ Pap 62:323–349CrossRefGoogle Scholar
  12. Chow GC (1960) Tests of equality between sets of coefficients in two linear regressions. Econometrica 28(3):591–605CrossRefGoogle Scholar
  13. Claeys P (2008) Rules, and their effects on fiscal policy in Sweden. Swed Econ Policy Rev 15(1):7–47Google Scholar
  14. Darby J, Melitz J (2008) Social Spending and automatic stabilizers in the OECD. Econ Policy 23(56):715–756CrossRefGoogle Scholar
  15. Darby J, Melitz J (2011) Joint estimates of automatic and discretionary fiscal policy: the OECD 1981–2003. CEPR Discussion Paper 8342Google Scholar
  16. Darvas Z (2009) The impact of the crisis on budget policy in Central and Eastern Europe. Bruegel Working Paper 2009/05Google Scholar
  17. De Haan J, Berger H, Jansen D (2003) The end of the Stability and Growth Pact? CESifo Working Paper 1093Google Scholar
  18. EFSF (European Financial Stability Facility) (2013) Publications: Frequently Asked Questions, 21 January 2013. http://www.efsf.europa.eu/attachments/faq_en.pdf. Accessed 7 June 2013
  19. European Central Bank (2013) Statistical data warehouse. http://sdw.ecb.europa.eu/. Accessed 15 Feb 2013
  20. European Comission (2001) Public Finances in EMU, European Economy, No. 3/2001Google Scholar
  21. European Commission (2013) Economic and financial affairs, financial assistance in EU Member States. http://ec.europa.eu/economy_finance/assistance_eu_ms/index_en.htm. Accessed 24 April 2013
  22. Everitt BS, Landau S, Leese M, Stahl D (2011) Cluster analysis. Wiley, HobokenCrossRefGoogle Scholar
  23. Furceri D, Poplawski-Ribeiro M (2008) Government spending volatility and the size of nations. Working Paper Series 0924, European Central BankGoogle Scholar
  24. Gali J, Perotti R (2003) Fiscal policy and monetary integration in Europe. Econ. Policy 18(37):534–572CrossRefGoogle Scholar
  25. International Foundation for Electoral Systems (IFES) (2013) Election Guide, Parliamentary elections. http://www.electionguide.org/. Accessed 7 May 2013
  26. International Monetary Fund (2012) IMF Country Report Hungary 2011 Article IV Consultation and Second Post-Program Monitoring Discussions. IMF Country Report (12/13)Google Scholar
  27. International Monetary Fund (2013a) International Statistical Yearbook, IMF’s International Financial Statistics. via StatistikNetzde, DSI Data Service & InformationGoogle Scholar
  28. International Monetary Fund (2013b) Reassessing the role and modalities of fiscal policy in advanced economies. IMF Policy Paper (September), Washington, DC, pp 1–63Google Scholar
  29. Kydland FE, Prescott EC (1977) Rules rather than discretion: the inconsistency of optimal plans. J Polit Econ 85(3):473–492CrossRefGoogle Scholar
  30. Levin A, Lin CF, James Chu CS (2002) Unit root tests in panel data: asymptotic and finite-sample properties. J Econom 108(1):1–24CrossRefGoogle Scholar
  31. Lewis J (2009) Fiscal policy in Central and Eastern Europe with real time data: cyclicality, inertia and the role of EU accession. DNB Working Paper 214/2009Google Scholar
  32. Marinheiro CJF (2005) Has the Stability and Growth Pact stabilized? Evidence from a panel of 12 European countries and some implications for the reform of the Pact. CESifo Working Paper 1411Google Scholar
  33. OECD (2014a) Economic Outlook No. 95, OECD economic outlook: statistics and projections (database). doi:10.1787/data-00688-en. Accessed 24 July 2014
  34. OECD (2014b) OECD. Stat, quarterly public sectors debt (database). http://stats.oecd.org/. Accessed 24 July 2014
  35. Pendleton HE (1938) The politics of fiscal policy. Yale Law J 47(5):724–745CrossRefGoogle Scholar
  36. Poplawski-Ribeiro M (2009) New evidence on the effectiveness of Europe’s fiscal restrictions. CEPII Working Paper 2009-13Google Scholar
  37. Roodman D (2006) How to do xtabond2: an introduction to difference and system GMM in Stata. Tech. rep. Center for Global DevelopmentGoogle Scholar
  38. Schaechter A, Kinda T, Budina N, Weber A (2012) Fiscal rules in response to the crisis? Toward the next-generation rules. a new dataset. IMF Working Paper WP/12/187Google Scholar
  39. Schneider O, Zapal J (2005) Fiscal policy in new EU Member States. Go East, prudent man! CESifo Working Paper 1486:1–28Google Scholar
  40. Staehr K (2007) Fiscal policies and business cycles in an enlarged euro area. CESifo working paper 1933Google Scholar
  41. Von Hagen J (2005) Fiscal rules and fiscal performance in the EU and Japan. IMES Discussion Paper 2005-E-5:1–29Google Scholar
  42. Windmeijer F (2005) A finite sample correction for the variance of linear efficient two-step GMM estimators. J Econ 126(1):25–51CrossRefGoogle Scholar
  43. Wyplosz C (2002) Fiscal discipline in EMU: rules or institutions? Tech. rep. European ComissionGoogle Scholar
  44. Wyplosz C (2006) European Monetary Union: the dark sides of a major success. Econ Policy 21(46):207–261CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2015

Authors and Affiliations

  1. 1.Bielefeld UniversityBielefeldGermany
  2. 2.European Investment BankLuxembourgLuxembourg

Personalised recommendations