Determinants of inflation differentials in the EMU: an empirical assessment of the role of labour and product market institutions

(1)

1

Determinants of inflation differentials in the EMU: an empirical assessment of the role of labour and product market institutions H.J. (Henk Jan) Reinders

10261907

Faculty of Economics and Business University of Amsterdam

14-07-2014 Final version

Supervisor: Dr. D.J.M. Veestraeten Second reader: Prof. F.J.G.M. Klaassen

(2)

2

Abstract

This thesis empirically investigates the role of labour and product market institutions in the determination of inflation differentials in the Economic and Monetary Union (EMU) during the period after the adoption of the euro as a single currency (1999-2012). Five hypotheses are formulated that investigate the effects of product market regulation (PMR), employment protection regulation (EPR) and coordination of wage-setting (CO) on inflation differentials. The results of this thesis indicate that differences in employment protection regulation between countries in the EMU contribute towards inflation differentials between them. These results seem not to suffer much from the presence of the global financial and Eurozone crises in the sample (by controlling for crisis effects using a crisis variable). No significant results are found in relation to product market regulation and coordination of wage-setting. This may however be partly caused by several drawbacks which are inherent in the use of structural indicators.

(3)

3

1. Introduction

Since the adoption of the euro as a single currency in the Economic and Monetary Union (EMU) in 1999, inflation differentials have continued to be observed between the participating countries1 (figure 1). These differences in inflation rates between countries have multiple causes. Factors that have often received attention in the literature include business cycle differences, catch-up effects and asymmetric (external) supply and demand shocks (see amongst others: Honohan and Lane, 2003, Andersson et al., 2009). Most of these factors are considered to be benign in nature, since they reflect the result of convergence or equilibrating dynamics. In the long run, their effects on price levels are expected to either cancel out or diminish (de Haan, 2010). However, inflation differentials may also indicate distortions that arise because of inefficiencies in product and labour markets, such as wage and price rigidities. These distortions may amplify the impact of shocks on inflation or increase the persistence of inflation and may generally be considered undesirable (Jaumotte and Morsy, 2012).

Figure 1 – Yearly HICP inflation differentials, in percentage points difference from the EMU12 average inflation rate (source: Eurostat, own computations)

1_{The focus of this thesis is at the countries that that adopted the euro in or close to 1999. These are Austria (OE),} Belgium (BG), Finland (FN), France (FR), Germany (BD), Greece (GR), Ireland (IR), Italy (IT), Luxembourg (LX), the Netherlands (NL), Portugal (PT) and Spain (ES). This thesis will refer to this group of countries as the EMU12.

(6)

6 Inflation differentials are of particular interest in currency areas such as the EMU. There are two main reasons for this. The first reason is that inflation differentials do immediately lead to diverging real exchange rates between member economies, because nominal exchange rates cannot be adjusted as a consequence of using a common currency. In the EMU, persistent inflation differentials have been observed that cumulated into losses of competitiveness for high inflation countries. 2 In particular Ireland, Greece, Portugal, Luxembourg and Spain have seen their price levels increase more rapidly than the EMU12 average since the adoption of the euro in 1999 (figure 2). In the EMU imbalances in competitiveness are hard to correct in particular because, besides the single currency, the EMU is also characterized by relative low wage and price flexibility – for instance compared to the US (Biroli et al., 2010).

Figure 2 – HICP price level development relative to the EMU12 average; 1999=100 (source: Eurostat, own computations)

2_{Note that diverging price level indices may also be expected as a result of price level convergence. For example,} prices in the non-tradables sector are generally lower in countries with lower GDP/capita, as will be discussed further on in this thesis. Countries that are catching up in GDP/capita terms can therefore show higher inflation rates. Coudert et al. (2012) have investigated the underlying causes of the observed strong appreciation of real exchange rates of Greece, Ireland, Italy, Portugal and Spain towards the rest of the EMU. They conclude that the peripheral member countries of the EMU have indeed been suffering from losses in competitiveness since the mid-2000s, since their real appreciation was not fully due to an improvement in their fundamentals (such as labour productivity).

(7)

7 A second reason why inflation differentials within the EMU are problematic is because monetary policy is tailored to the average inflation rate in the Eurozone. When individual countries show inflation rates that differ substantially from the Eurozone average, monetary policy will not be optimal for all members (ECB, 2003). Member economies with higher inflation rates will experience lower short-term real interest rates, given the identical short-term nominal interest rates as set by the ECB. This could stimulate consumption and investment that could in turn cause additional inflation.

Recently, labour and product market institutions have received increased attention in empirical studies on inflation differentials in the EMU (Bulír and Hurník, 2008, Andersson et al., 2009, Biroli et al., 2010, Jaumotte and Morsy, 2012). Since labour and product market institutions differ between countries, these differences may result in heterogeneous wage and price dynamics and hence may yield inflation differentials. In particular, these studies have pointed towards the role of labour and product market institutions in determining country specific coefficients in the Phillips curve, such as for the response of inflation to the output gap and inflation persistence. However, these studies have been exploratory in nature and the results with respect to the effects of structural indicators on inflation are sometimes found to be opposing when comparing studies that use different samples. With that in mind it is important to still better understand structural differences and their effects on inflation differentials in the EMU.

The aim of this thesis is to investigate the role of labour and product market institutions in determining inflation differentials in 12 EMU countries after the adoption of the euro. To that end this thesis will empirically investigate the effects of labour and product market institutions on inflation.3 In particular, this thesis will focus on product market regulation (PMR), employment protection regulation (EPR) and coordination of wage-setting (CO). Five hypotheses are tested, covering the effects of these variables on inflation both directly (as independent variables) and indirectly (as moderating variables, or in other words as determinants of the size of for example the effect of the output gap on inflation).

3

Note that inflation and inflation differentials will sometimes be used interchangeably in this thesis. Factors that generate inflation also generate inflation differentials, as long as those factors differ between countries. This is also used in the empirical analysis, which takes the inflation rate (and not the inflation differential) as the dependent variable.

(8)

8 The contribution of this thesis lies in combining several hypotheses from the literature and extending the time period covered in earlier studies. Given that the sample period includes the global financial and Eurozone crises, some additional robustness checks are performed, such as including a crisis variable and using smaller samples (excluding the crisis years and excluding the countries that received formal EU/IMF support). Also, a novel tax variable is introduced.

The remainder of this thesis is structured as follows. Chapter 2 will review the factors that are thought to determine inflation differentials and formulates the hypotheses. Chapter 3 will then provide some stylized facts on inflation differentials and on the structural indicators that will be used to measure labour and product market institutions. Additionally this chapter discusses some features of the global financial and Eurozone crises. Chapter 4 will estimate several dynamic panel models with the aim to test the relevance of the institutional variables as determinants of inflation differentials in the EMU. Chapter 5 discusses the findings and concludes.

Regarding the empirical approach, as a first stage, a basic model will be set up that covers a selection of relevant control variables with respect to inflation differentials. To a large extent, this model will build on the approach as used by Honohan and Lane (2003). Following their method, the dependent variable in the analysis is taken to be the inflation rate (and not the inflation differential). By including time specific effects, however, the common inflation in a given year is accounted for and the remaining variables are those that contribute to inflation differentials.

Control variables will be included on the basis of the factors as identified in chapter 1. These include factors to capture the lagged inflation rate, the output gap, imported inflation, catch-up effects and non-market based prices (in particular taxes on products). Note here that it is not within the scope of this thesis to include all possible control variables (since this would mean including all possible factors that may lead to inflation). Hence only those control variables that are found most relevant in earlier empirical studies will be used, in order to focus on the institutional variables of interest.

As a second stage, this basic model will be used to test the effect of three institutional variables, each corresponding to one or two hypotheses, in their effect on inflation differentials in the EMU. One particular problem that is encountered here is that institutional factors may to a certain extent be correlated. To deal with these potential multicollinearity issues, the institutional variables will

(9)

9 be tested both individually and collectively as additional factors in the basic model. Differences in estimated coefficients and standard errors between these models can indicate correlation effects. More formally, multicollinearity is tested for using the variance inflation factor (VIF).

Finally, the obtained results are checked for robustness with respect to some econometric issues. In particular, findings may suffer from omitted variables bias due to crisis effects that are not adequately captured by the model. To investigate this possibility, an additional model mode is estimated that includes a crisis variable. Furthermore, due to its dynamic nature, the regression results may suffer from a dynamic panel bias. A model is estimated that uses an Least Squares Dummy Variable (LSDV) bias correction procedure (based on the Arellano-Bond estimator). This will shed some light on the magnitude of the dynamic panel bias. Finally, a last model is estimated using Feasible Generalized Least Squares (FGLS). This estimation method is generally more efficient than OLS/LSDV and may hence provide some additional indications about the relevance of the institutional variables. However, to obtain the FGLS results some additional assumptions have to be made.

(10)

10

2. Determinants of inflation differentials

Inflation is calculated by examining a basket of goods and services and comparing the changes in prices in that basket over time. Eurostat (2013) defines inflation as the ‘percentage change in the price index for a given period compared to that recorded in a previous period’. Usually when speaking about inflation these periods are taken to be years, even though some research also considers monthly or quarterly inflation. To calculate the inflation differential of a given country i, the inflation rate of that country is compared to a benchmark inflation rate . In general the benchmark inflation rate is taken to be the average inflation rate of a group of countries j, which may include country i.4 Formally, the inflation differential of country i and benchmark country (or group of countries) j in year t can be expressed as follows (Altissimo et al., 2011):

(1)

Inflation differentials are driven by (a subset of) the factors that drive inflation. To see this consider equation (1), where all right hand side factors include only the inflation rate of a country and/or of the average inflation rate of a group of countries (like the EMU). Hence, all factors that drive inflation can also drive inflation differentials. Some factors that drive inflation can cancel out though, when the factor is common to the compared entities. In the EMU this may for example be largely the case when considering the (nominal) monetary policy rate as set by the ECB which is equal for all countries using the euro as a currency. The same may go for harmonised parts of labour and product market institutions.

To answer the question of what drives inflation differentials, it thus is important to understand the determinants of inflation. The most important determinants of inflation are identified and discussed in the next two sections, covering the general determinants of inflation and catch-up effects in an economic and monetary union respectively. These factors will serve as the control variables in the regression analysis. A third section will specifically focus on the role of labour and product market institutions and will formulate five hypotheses.

(11)

11

2.1 General determinants of inflation

An extensive body of literature exists around the determinants of inflation, traditionally using the Phillips curve equation as a starting point. However, more recent investigations also take other approaches, thereby relaxing some of the quite stringent assumptions that are often used in the derivation of the Phillips curve. An example of this is the use of a lagged term for inflation as an explanatory variable, to account for (suboptimal) price setting based on past inflation experience behaviour by some firms. Next to that, several other factors can be considered as relevant additions to a framework that considers inflation differentials. Amongst others, these include imported inflation, non-market prices and catch-up effects (such as price convergence of both tradable and non-tradable goods).

2.1.1 The Phillips curve

During the last decades a standard New Keynesian framework has been developed that is based on a dynamic general macroeconomic model with imperfect competition (Wickens, 2008). A central equation in this framework is the New Keynesian Phillips curve, which in its basic form relates inflation to its expected future value and to a measure of real disequilibrium (Ólafsson, 2006). In addition to inflation expectations, inflation persistence is observed when looking at real world data. Hybrid, or amended, versions of the New Keynesian Phillips curve additionally include a factor to account for lagged inflation (Moccero et al., 2013). A general hybrid specification of the Phillips curve can be stated as follows, following Wickens (2008) and adding a term for lagged inflation:

( ̅) (2)

In this equation β is a traditional utility discount factor, is the expected future inflation rate, represents a share of past inflation and ( ̅ )represents a share of the output gap. Hence, according to this model, inflation differentials can occur as a result of differences in inflation expectations (partly based on past inflation) and as a result of differences in the output gap between countries. Furthermore it is important to note that, when the coefficients , and are country specific, this may also result in inflation differentials between countries.

(12)

12 The output gap in this thesis is defined as actual GDP minus potential GDP5. The output gap can lead to inflation differentials for example when the respective business cycles of individual countries are not fully synchronized or when asymmetric shocks between countries occur. Also, the response of inflation to the output gap, here represented by the coefficient , may be country specific. These country specific responses may also be (partly) caused by labour and product market institutions, which will be further discussed in the third section of this chapter.

2.1.2 Imported inflation

With respect to the Phillips curve as shown in equation (2) it is important to notice that not all prices are formed on the domestic market. Especially for small open economies, many prices have to be taken from the world market. This makes inflation dependent on world prices in for instance oil and other commodities. Moreover, exchange rates may influence the domestic price of these imported products. For most imported products it may be true that domestic prices rise in the same manner for all importing countries. However, the dynamics of imported inflation are complex and there are several reasons why changing world market prices or exchange rates may still lead to inflation differentials between countries.

A first reason is given by incomplete pass-troughs. Depending on market characteristics (which may differ between countries) firms may decide to transfer the full change in price the domestic consumer (maximum pass-through) or to bear part of the price increase or decrease themselves. Second, countries do differ in the share of consumption of a particular good or service. Given that the inflation rate is a weighted average of a basket of prices, the weight of a particular good or service in that basket will determine the impact on inflation rates. For example, countries with a relative high share of oil (or related products) in their consumption basket may show higher inflation rates when oil prices increase. Finally, a similar argument can be made for countries that import a relative large share of their domestic consumption from extra-EMU trading partners. When exchange rates change while import shares are different, this may contribute to inflation differentials (Égert, 2007).

5

Potential output is defined by the European Commission (EC) as ‘the level of output that is consistent with stable inflation’. The EC estimates potential output using a synthetic measure of the aggregate supply of the economy from a Cobb–Douglas production function estimate with trend total factor productivity. The output gap gives an indication of inflationary pressures as a result of the cyclical position of an economy (EC, 2001).

(13)

13 2.1.3 Non-market based prices and inflation

Finally, not all prices are fully shaped by market forces. In fact, the majority of goods and services may have a component of its price that is not determined by the market. Governments may for example impose taxes on products that increase the price of consumer products. When changes in government tax policies are not harmonised across countries, this may result in inflation differentials between these countries. In particular this effect may be visible for value added tax (VAT) rates, which form a substantial part of consumer prices and also apply to a large share of consumer products. In the EMU, VAT increases or decreases are still a matter of national discretion and are moreover not a rare phenomenon (see figure 6 on page 24). As a result, differences in national policies with respect to VAT rates are expected to result in one-off inflation differentials in the EMU after VAT rates are adjusted (Égert, 2007).

2.2 Catch-up effects in an economic and monetary union

For an economic and monetary union, the process of convergence of price levels is of special interest when considering inflation and inflation differentials. When not all members of an economic and monetary union share the same price levels and/or GDP per capita from the start, it is likely that the price level of lagging countries will converge towards that of the more advanced countries. This convergence in price levels may on the one hand arise in tradables prices because of further integration of goods and service markets. On the other hand, especially non-tradables prices may increase in lagging countries due to real income catch-up effects (Hofmann and Remsperger, 2005). This phenomenon is also observed as a positive correlation between GDP per capita and the price level of a given country (Bergstrand, 1991). Catch-up effects can therefore lead to inflation differentials because of higher inflation rates in countries that are catching up.

2.2.1 The Balassa-Samuelson effect

One often discussed mechanism that can explain the link between GDP per capita and the price level is the Balassa-Samuelson effect (Balassa, 1964, Samuelson, 1964). This effect arises due to productivity increases in the tradables sector. This increase in productivity increases real wages in the tradable sector. As a result, because labour flows between the tradables and non-tradables sectors, wages in the entire country rise. This drives up the relative price of non-traded goods as well (where productivity did not increase to the same extent). This may then cause additional

(14)

14 inflation in non-traded goods, and hence contribute towards overall inflation in the country. Compared to other members of the economic and monetary union, where productivity in the tradable sector did not increase to the same extent, this will then result in additional inflation for the countries that are converging.

2.2.2 Other effects

The Balassa-Samuelson effect, however, is not the only explanation for the correlation between GDP per capita and the price level of a country. Other effects may for example stem from increases in product quality and in pricing-to-market strategies. With respect to increases in product quality, Égert (2011) argues that part of the correlation between GDP per capita and the price level may stem from quality increases that are not accounted for in the official statistics. Increases in quality of a product warrant a higher price and hence may account for additional inflation. When these unaccounted for increases in quality occur only (or to a larger extent) in the subgroup of countries that are catching-up, this may lead to inflation differentials. With respect to pricing-to-market strategies, these strategies may result in lower prices in economies with lower GDP per capita. This is because firms may price their products in line with disposable income. A growing GDP per capita may result in additional inflation due to higher product pricing in the economies that are catching up.

2.3 Possible effects of labour and product market institutions on inflation

Besides the factors mentioned in the previous paragraphs, there has been increasing attention for the role of labour and product market institutions6 in determining inflation and inflation differentials between countries in the EMU (see amongst others: Campolmi and Faia, 2004; Andrés et al., 2007). This section discusses two ways by which labour and product market institutions may affect inflation differentials between countries: by providing different degrees of pressure on wages and markups across countries and by determining the country specific response of inflation to the output gap. Based on those mechanisms, five hypotheses are formulated that will be tested as part of the empirical analysis.

6_{Labour and product market institutions can refer to both regulations (binding rules set by the government) and other} implicit or explicit rules, such as wage setting (which is partly shaped by the social partners) and unemployment benefit systems (Biroli et al., 2010).

(15)

15 2.3.1 Level effects of labour and product market institutions

One way in which labour and product market institutions can contribute towards inflation and inflation differentials is by providing different degrees of pressure on wages (in the form of unit labour costs) and markups over marginal costs (Bulír and Hurník, 2008). Both unit labour costs and markups are components of consumer prices. When the development of wages and markups is not synchronized across the EMU this may lead to inflation differentials between countries.

Considering labour market institutions, trade unions in the more protected and union-oriented economies (those with for example higher levels of employment protection regulation or union membership7) may provide strong upwards wage pressures (Hancké, 2013). This is mainly due to high bargaining power of workers in these economies. This higher bargaining power can result in higher growth of unit labour costs compared to countries with relative low employment protection regulation and/or union membership. Higher unit labour costs growth in turn puts upwards pressure on wages and inflation. This leads to a first hypothesis:

 Hypothesis 1. A higher (lower) level of employment protection regulation leads to higher (lower) inflation. When countries do not have the same level of employment protection regulation this may result in an inflation differential between these countries.

With respect to product markets institutions, the regulatory environment may either hinder or foster competition in product markets (as a result of for example state control, barriers to entrepreneurship and barriers to entry). The level of competition may in turn influence the markup that firms can charge over marginal costs. In uncompetitive market environments there may be less downward pressure on consumer prices (Bulír and Hurník, 2008). When the degree of competition in product markets differs between countries, this may then potentially result in inflation differentials between countries. This leads to a second hypothesis.

7

In particular with respect to wage pressures, there are likely multiple factors that can contribute towards (excessive) growth in unit labour costs. Besides employment protection regulation (as put forward by Bulír and Hurník (2008)) other authors for example include union density (the percentage of workers that are member of a union) as an indicator of pressure on wages (Jaumotte and Morsy, 2012). This thesis will explicitly test the effect of the level of employment protection regulation (defined as the strictness of regulation on dismissals and the use of temporary contracts) on inflation and will not focus at union density. The reason for this is that union density is practically constant over time for individual countries in the EMU12 sample between 1999 and 2012 and this effect will therefore most likely be captured by the country specific effects in the regression analysis.

(16)

16

 Hypothesis 2. A higher (lower) level of competition in product markets leads to lower (higher) inflation. When countries do not have the same level of competition in product markets this may result in an inflation differential between these countries.

Furthermore, another way by which labour market institutions may contribute to inflation is because there may be different levels of coordination of wage-setting in different countries. Coordination of wage-setting is defined as the degree of centralization in wage-setting (where usually this centralization is provided by the government, as a result of monopolistic union confederations and/or by influential large firms). High coordination of wage-setting likely increases the consideration of macroeconomic consequences of possible wage adjustments (such as inflation).8 Therefore it may be easier, for example, to reach an outcome of wage moderation (Bowdler and Nunziata, 2007). This in turn decreases the pressure on firms’ marginal costs and hence prices and inflation. A higher level of coordination of wage-setting may therefore lead to lower inflation than would be obtained under low coordination of wage-setting.

 Hypothesis 3. A higher (lower) level of coordination of wage-setting leads to lower (higher) inflation. When countries do not have the same level of coordination of wage-setting this may result in an inflation differential between these countries.

2.3.2 Response of inflation to the output gap

A second way by which labour and product market institutions may influence inflation is by affecting the way that inflation responds the output gap (Correa-López et al., 2013). Labour and product market institutions are likely reflected in the coefficients that multiply the determinants of inflation in the Phillips curve as stated in equation (2). Countries with different product and labour market institutions may then show different inflation rates when confronted with common shocks.

With respect to product market institutions, the response of inflation to the output gap may depend on the speed with which markups over marginal costs are adjusted in response to

8_{This is in contrast to fragmented wage-setting, in which each individual wage-setting entity (such as a small labour} union) has fewer incentives to consider macroeconomic consequences given that the outcome will not have a large impact on the economy as a whole (given its small size).

(17)

17 economic up- and downturns (as measured by the output gap). In the literature, two contrasting arguments are proposed. On the one hand, prices may respond less to the output gap when the level of competition is high. This is because highly competitive firms may choose not to adjust their prices to consumers in response to economic conditions in fear of losing market share (Correa-López et al., 2010). On the other hand, prices may (contrary to the first argument) respond more to the output gap in a highly competitive environment. Andres, Ortega and Vallés (2008) argue for example that less competitive firms may change prices less often than more competitive firms and hence respond slower to economic up- and downturns. The consequence then is that in a more competitive market, inflation is more responsive to the output gap than in less competitive markets (due to slower price adjustment.) This leads to two (opposing) hypotheses:

 Hypothesis 4a. A lower (higher) level of product market regulation decreases (increases) the responsiveness of inflation to economic up- and downturns. When countries have different levels of product market regulation (and hence different levels of price rigidities) this may result in inflation differentials between these countries.

 Hypothesis 4b. A lower (higher) level of product market regulation increases (decreases) the responsiveness of inflation to economic up- and downturns. When countries have different levels of product market regulation (and hence different levels of price rigidities) this may result in inflation differentials between these countries.

With respect to labour market institutions, the response of inflation to the output gap may depend on the speed with which wages adjust to economic up- and downturns Under more strict employment protection legislation it is likely harder for firms to instantaneously alter wages in response to economic up- and downturns. This may result in a delayed adjustment of marginal costs and consumer prices in response to the output gap (Abbritti and Mueller, 2007). This results in a final hypothesis for this thesis:

 Hypothesis 5. A higher (lower) level of employment protection legislation reduces (increases) the responsiveness of inflation to economic up- and downturns. When countries have different levels of employment protection legislation this may result in inflation differentials between these countries.

(18)

18

3. Stylized facts for the EMU

This chapter first describes inflation differentials in the EMU and the observed persistence in these differentials. A second section describes the development of labour and product market institutions in the EMU over time from 1999 to 2012. Finally, a third section explores two variables of interest (the output gap and changes in taxes on products) in the context of the global financial and Eurozone crises.

3.1 Inflation differentials in the EMU

Within the EMU substantial inflation differentials exists when comparing individual inflation rates of its members (table 1). As discussed in the previous chapter, this can be caused by several factors that determine inflation and that moreover are not the same for all member economies of the EMU. Moreover, when averaging these inflation differentials over time they do not cancel out, which is shown by averages provided in the bottom three rows of the table. This indicates that inflation differentials in the EMU are not merely a cyclical phenomenon or the result of differences in price adjustment mechanisms (which may for example lead to a delayed response of inflation to shocks and hence to observed inflation differentials in the short run).

Table 1 – Yearly and average HICP inflation differentials (in percentage points ), and the EMU12 average HICP inflation rate (source: Eurostat, own computations)

BD FR OE FN BG NL IT IR PT LX ES GR EMU 1999 -0.5% -0.5% -0.6% 0.2% 0.0% 0.9% 0.6% 1.4% 1.1% -0.1% 1.1% 1.0% 1.1% 2000 -0.8% -0.4% -0.2% 0.7% 0.5% 0.1% 0.4% 3.1% 0.6% 1.6% 1.3% 0.7% 2.2% 2001 -0.5% -0.6% -0.1% 0.3% 0.0% 2.7% -0.1% 1.6% 2.0% 0.0% 0.4% 1.3% 2.4% 2002 -0.9% -0.4% -0.6% -0.3% -0.7% 1.6% 0.3% 2.4% 1.4% -0.2% 1.3% 1.6% 2.3% 2003 -1.0% 0.2% -0.7% -0.7% -0.5% 0.2% 0.8% 2.0% 1.3% 0.5% 1.1% 1.4% 2.0% 2004 -0.3% 0.2% -0.1% -2.0% -0.2% -0.7% 0.2% 0.2% 0.4% 1.1% 1.0% 0.9% 2.1% 2005 -0.2% -0.2% 0.0% -1.3% 0.4% -0.6% 0.1% 0.1% 0.0% 1.7% 1.3% 1.4% 2.1% 2006 -0.4% -0.3% -0.5% -0.9% 0.1% -0.5% 0.0% 0.5% 0.8% 0.8% 1.4% 1.1% 2.2% 2007 0.2% -0.5% 0.1% -0.5% -0.3% -0.5% -0.1% 0.8% 0.3% 0.6% 0.7% 0.9% 2.1% 2008 -0.4% 0.0% 0.0% 0.7% 1.3% -1.0% 0.3% -0.1% -0.5% 0.9% 0.9% 1.0% 3.2% 2009 -0.1% -0.2% 0.1% 1.3% -0.3% 0.7% 0.5% -2.0% -1.2% -0.3% -0.5% 1.0% 0.3% 2010 -0.3% 0.2% 0.2% 0.2% 0.8% -0.6% 0.1% -3.1% -0.1% 1.3% 0.5% 3.2% 1.5% 2011 -0.2% -0.4% 0.9% 0.6% 0.7% -0.2% 0.2% -1.5% 0.9% 1.0% 0.4% 0.4% 2.7% 2012 -0.4% -0.3% 0.1% 0.7% 0.1% 0.3% 0.8% -0.6% 0.3% 0.4% -0.1% -1.5% 2.5% 1999-2007 -0.5% -0.3% -0.3% -0.5% -0.1% 0.4% 0.2% 1.3% 0.9% 0.7% 1.1% 1.1% 2.0% 2008-2012 -0.3% -0.1% 0.3% 0.7% 0.5% -0.2% 0.4% -1.5% -0.1% 0.7% 0.2% 0.8% 2.0% 1999-2012 -0.4% -0.2% -0.1% -0.1% 0.1% 0.2% 0.3% 0.3% 0.5% 0.7% 0.8% 1.0% 2.0%

(19)

19 High average inflation differentials before the global financial crisis in 2008 are found in particular in the Southern European economies such as Greece, Spain and Portugal, although consistently high inflation differentials are also observed for Luxembourg and Ireland (the right hand side of table 2). Other countries, such as Germany, France and Austria, have shown consistently low inflation differentials during the same period (the left hand side of table 2). Some countries seem to have reversed course with respect to inflation differentials during the crisis period after 2007 (such as Ireland and Portugal). Others, such as Spain and Greece show lower, but nonetheless still positive, inflation differentials compared to before 2008.

3.2 Labour and product market institutions in the EMU

To measure labour and product market institutions, several organizations provide structural indicators on different aspects of labour and product market institutions. The table below presents the definitions of these structural indicators, after which the development over time is shown for the countries in the sample (EMU12).

Table 2 – Definitions of structural indicators as used in this thesis

PMR Correa-López et al. (2010) define product market regulation (PMR) as the ‘regulatory environment that either hinders or fosters competition in product markets’. According to the OECD measurement of the concept, PMR includes the level of state control, barriers to entrepreneurship and barriers to entry (Koske et al., 2014). PMR will be used as an indicator for the degree of competition in the product market (hypotheses 1 and 3). Note that a high score on the indicator means having a less competitive product market.

EPR The OECD indicator of employment protection regulation (EPR) measures the strictness of regulation on dismissals and the use of temporary contracts and includes both individual and collective dismissals. EPR will be used as an indicator of the strictness of employment protection legislation (hypotheses 2 and 4). A high score on the indicator means having more strict employment protection regulation.

CO The Amsterdam Institute of Advanced Labour Studies (AIAS) indicator for coordination of wage-setting (CO) measures the degree of centralised coordination of wage-setting on a five point scale. The highest value (5) corresponds to a highly centralized wage-setting (imposed by the government, as a result of monopolistic union confederations and/or influential large firms) while the lowest value (1) corresponds to fragmented wage-setting (confined largely to individual firms or plants). CO will be used as an indicator with respect to coordination of wage-setting (hypothesis 5).

(20)

20 3.2.1 Structural reforms

From 1999 to 2012 labour and product markets institutions have changed substantially in the EMU. The OECD indicator for product market regulation (PMR) has declined for all countries except Luxembourg over the sample period, reflecting efforts to reduce product market regulations and increase competition in product markets across the EMU (figure 3). Countries with the highest average level of product market regulation between 1999 and 2012 are Greece, Spain and Portugal, while the Netherlands, Germany and Austria show the lowest average levels. This seems broadly in line with the hypothesis that countries with low levels of product market competition (and hence high levels of product market regulation) have higher inflation rates (hypothesis 2).

Figure 3 – Development of the PMR indicator over time (source: OECD, own computations)

The OECD indicator for employment protection regulation (EPR) has in the EMU shown a converging pattern during the sample period, mainly due to reductions in the level of employment protection in Portugal while the level of employment protection regulation has increased in Ireland (figure 4). Countries with the highest average level of employment protection regulation between 1999 and 2012 include Portugal, Italy and Germany while countries with the lowest average level include Ireland, Finland and Austria. Some of these average levels are well in line with the expectation that a high level of employment protection regulation leads to higher inflation, including Portugal and Finland (hypothesis 1). For other countries, including Germany and Ireland, the existence of this effect is less visible at this point. Of course, to test whether this hypothesis holds, other factors must be controlled for which is part of the empirical analysis.

(21)

21 Figure 4 – Development of the EPR indicator over time (source: OECD, own computations)

3.2.2 Coordination of wage-setting

The AIAS indicator for coordination of wage-setting (CO) does not show clear increasing or decreasing trends from 1999 to 2012 (table 3). An exception is Greece, where coordination of wage-setting has sharply increased from 2010 onwards, likely as part of the austerity measures that were imposed. For Germany, France, Austria, Belgium, Italy and Ireland values for coordination of wage-setting have been mostly stable, with only one change in value during the sample period. Finland, the Netherlands, Portugal, Luxembourg and Spain show more frequent jumps in the level of coordination of wage-setting. In particular France, Portugal, Luxembourg and Greece (before 2010) show low levels of coordination of wage-setting.

Countries with the highest levels of coordination of wage-setting include Germany, Austria, Finland and Belgium. In general there seems to be a slight tendency towards a higher level of coordination in wage-setting in countries with a low average inflation differential between 1999 and 2012. This is also indicated by a negative correlation coefficient (-0.11) between coordination in wage-setting and the inflation differential for a given country in a given year. Exceptions in that respect seem to be France (low coordination and low inflation) and Spain (high coordination and high inflation). However, like with PMR and EPR, whether this relationship holds while controlling for other factors (such as catch-up effects) is a question for the empirical analysis.9

9

Note that for some countries (France, Austria and Belgium) the value of the indicator for coordination in wage-setting is constant. If this would be the case for all variables, the effect of the level of coordination in wage-wage-setting on inflation would likely be captured by the country specific effects. However, given that other countries do show some variation in the variable it seems possible to estimate the coefficient for the effect of the level of coordination…

(22)

22 BD FR OE FN BG NL IT IR PT LX ES GR 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

: mixed industry and firm-level bargaining, with no or little pattern bargaining and relatively weak elements of government coordination through the setting of basic pay rates (statutory minimum wage) or wage indexation.

: a) informal (intra-associational and/or inter-associational) centralization of industry and firm level bargaining by peak associations (one side, or only some unions) with or without government participation; b) industry-level bargaining with irregular and uncertain pattern setting and only moderate union concentration; c) government arbitration or intervention.

: a) centralized bargaining by peak associations with or without government involvement, and/or government imposition of wage schedule/freeze, without peace obligation; b) informal (intra-associational and/or inter-associational) centralization of industry and firm level bargaining by peak associations (both sides) ; c) extensive, regularized pattern setting coupled with high degree of union concentration.

: a) centralized bargaining by peak association(s), with or without government involvement, and/or government imposition of wage schedule/freeze, with peace obligation; b) informal centralization of industry-level bargaining by a powerful and monopolistic union confederation; c) extensive, regularized pattern setting and highly synchronized bargaining coupled with coordination of bargaining by influential large firms.

Table 3 – Values for the coordination of wage-setting indicator over time (source: AIAS)

3.3 The global financial and Eurozone crises

The sample used in this thesis includes the global financial crisis (which peaked in 2008) and the Eurozone crisis that followed (2009 onwards). This crisis period may have had an impact on several variables that are relevant in this thesis.

In particular the measure for the output gap reflects the effects of these crises. Figure 5 shows that, for all EMU members, a sharp drop in the output gap is observed in 2008 and 2009. A noticeable case is Greece, where the output gap has dropped substantially further than for other countries in the sample.Comparing to the output gap in other countries, the values for Greece in 2010 to 2012 represent clear outliers in the variable.

…in wage-setting nonetheless (for the countries where the variable is constant this effect would otherwise be fully captured by the country specific effects). However a lack of variation in coordination of wage-setting (which seems to be somewhat true for PMR and EPR as well) may make it more difficult to prove or disprove the related hypotheses (because a share of the effect may be captured by the country specific effects in the regression analysis).

(23)

23 Figure 5 - Gap between actual GDP and potential GDP as percentage of potential GDP (source: European Commission)

Furthermore, the Eurozone crisis has increased the need for most EMU governments to increase their income. This is reflected in increases in taxes on products (figure 6) which in turn are likely to affect inflation rates and inflation differentials across EMU member countries. In particular Portugal has increased its taxes on products substantially in the period from 2009 to 2012.10

Other countries that have increased their taxes on products by more than 0.5 percentage points in a single year include Finland, Italy, the Netherlands and Spain.11 The observed increase in tax rate in Germany in 2007 represents an increase in the general VAT rate from 16% to 19% on the 1st of January 2007 (Miki, 2011). This increase is also clearly visible in the data on inflation differentials (see the previous table 1) where 2007 represents the only positive differential (0.2%) for Germany during the 1999-2012 period.

One problem with the observation that multiple variables may be affected by the global financial and Eurozone crisis is that this may cause additional correlation between these variables. Also, it may be that especially during the crisis some unobserved variables play a role in determining inflation (such as expectations) or that extreme values of variables (such as the output gap for Greece) are not well captured by a linear model. These issues will, to the extent possible, be addressed as part of the empirical analysis in the next chapter.

10_{Taxes on products in the HICP at constant taxes include: VAT; excise duties and special taxes; consumption taxes;} car registration taxes; taxes on entertainment, insurance premiums and other specific services (such as hotels). 11_{No data is available for Ireland and Greece.}

(24)

24 Figure 6 – Theoretical percentage point increase in HICP due to changes in taxes on products (source: Eurostat; data available from 2004 except Ireland and Greece, own computations).

(25)

25

4. Empirical analysis

This chapter first provides an overview of the empirical literature on the determinants of inflation differentials and the possible role of labour and product market institutions therein. Second, the data that is used for the regression analysis is described, followed by the model and estimation methodology. The last section provides the results and interpretation of the regression analysis.

4.1 Overview of empirical literature

4.1.1 Determinants of inflation differentials

An extensive body of literature has evolved around the empirical investigation of inflation differentials. As one of the first to empirically investigate inflation differentials in the EMU after the adoption of the euro, Honohan and Lane (2003) use an equation for inflation that includes factors for the output gap, the fiscal stance, the nominal effective exchange rate (NEER) and the lagged price level. Their findings suggest the importance of price level convergence, the output gap and the NEER as determinants of inflation differentials. Subsequent research also includes terms for lagged inflation, to account for inflation persistence. For instance, both Hofmann and Remsperger (2005) and Angeloni and Ehrmann (2007) find empirical evidence for the output gap, the lagged inflation rate and the NEER as relevant determinants of inflation differentials. This thesis will use these three variables as control variables in the regression analysis.

In terms of catch-up effects in the EMU, an important question is how to model the effects of this process on inflation and the inflation differential. Honohan and Lane (2003) do so by assuming a common long run price level for countries within the EMU, and by furthermore assuming that the additional inflation caused by the convergence process is proportional to the difference between current and long run price levels. An alternative is to model for example the Balassa-Samuelson effect directly: Andersson et al. (2009) include the growth rate of relative labour productivity between the tradables and non-tradables sectors. This factor is however not found to be relevant in their panel estimation. Also others do not find much evidence for a substantial Balassa-Samuelson effect in the EMU after the adoption of the euro (Rabanal, 2009, Égert, 2011). To capture catch-up effects in a broad way (including, but not limited to, the Balassa-Samuelson effect) this thesis therefore includes the deviation of the price level from the EMU12 average.

(26)

26 To capture part of non-market based price setting, Égert (2007) takes a binary variable for VAT rate changes in a model of inflation, but does not find significant results. Since the effect of VAT rate changes on the inflation rate is intuitively highly plausible, this is likely due to a lack of information on the magnitude of the effect in that variable and the expectation that the effect on inflation is only short lived. One new approach here is to subtract the Eurostat series on HICP from the HICP at constant tax rates (TAX). This then yields an indicator that does take the magnitude of a change in tax rates on products into account (see also the previous figure 6).

4.1.2 The role of product and labour market institutions

In recent years, several authors have investigated empirically the effects of product and labour market institutions on inflation differentials the EMU. Most of these studies investigate the effect of structural indicators on product and labour market institutions as interaction variables with traditional determinants of inflation (Bowdler and Nunziata, 2007, Biroli et al., 2010, Correa-López et al., 2010, Jaumotte and Morsy, 2012) This thesis follows the approach of this strand of literature by including two interaction variables between the output gap and institutional variables (output gap * PMR and output gap * EPR) and will in this way test hypotheses 4 and 5.

Additionally, some studies investigate the more direct effects of product and labour market institutions (Bulír and Hurník, 2008, Andersson et al., 2009). These may include level effects (where the independent variable in a regression on inflation is the level of a structural indicator) or dynamic effects (where the independent variable in a regression on inflation is the change in a structural indicator). With respect to the latter, Andersson et al. (2009) find that changes in level of product market regulation contribute to inflation differentials between 1999 and 2006 for 12 countries in the EMU. This thesis however follows the approach of Bulír and Hurník (2008) by including the level of employment protection regulation (EPR) and product market regulation (PMR) as independent variables in the analysis (in this way hypotheses 1 and 2 are tested). These authors have previously found that higher levels of EPR or PMR result in higher inflation in a sample of 15 European countries between 1996 and 2005, which is in line with the hypotheses in this thesis. Additionally, this thesis will furthermore investigate the effect of the level of coordination of wage-setting on inflation (to test hypothesis 3).

(27)

27

4.2 Data

To investigate the effect of labour and product market institutions on inflation differentials in the EMU, a model is set up including a set of institutional variables and a set of control variables. The sample used consists of the 12 EMU member states which adopted the euro close to 1999. These members include Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal and Spain. The sample period runs from 1999 to 2012.

The dependent variable of interest is based on the inflation rate (INFL), which is measured by the growth rate in the Harmonized Index of Consumer Prices (HICP). The HICP covers the prices of consumer goods and services acquired by households and is measured consistently for all countries in the EMU. Structural indicators are taken from the OECD and the Amsterdam Institute for Advanced Labour Studies (AIAS) as described earlier in table 2. Five control variables are used: these are the lagged inflation rate, the output gap, the change in the nominal effective exchange rate index (NEER), the difference between the price level of an individual country from the EMU12 average price level (CPI gap) and the change in taxes on products (TAX). All variables are shortly described below, while a full definition and operationalization of the variables can be found in appendix I.

EPR – To measure the strictness of employment protection dimension of labour market institutions, the OECD summary indicators on employment protection are used. This thesis uses the second version of the strictness of employment protection (EPR) indicator, since this series is available from 1999 to 2012. The indicator measures the strictness of regulation on dismissals and the use of temporary contracts and includes both individual and collective dismissals. A higher value on the indicator indicates more stringent regulation and is expected to lead to higher inflation (hypothesis 1). Also, a higher value for EPR is expected to lead to a lower response of inflation to the output gap (hypothesis 5).

PMR – To measure product market institutions, the OECD indicators on product market regulation are used. This thesis uses the integrated product market regulation (PMR) indicator, which includes measures of state control, barriers to entrepreneurship and barriers to trade and investment. It is available for 1998, 2003, 2008 and 2013. To create a time series the values for missing years are linearly interpolated. A higher value on the indicator indicates more stringent

(28)

28 regulations. The moderating effect of this variable on the response of inflation to the output gap may be either positive or negative, depending on the strength of the effects as put forward in the theoretical section (high competition leads to more constant prices versus low competition leads to infrequent adjustment of prices; hypotheses 4a and 4b). Additionally it is expected that a higher value for PMR leads to higher inflation (hypothesis 2).

CO – To measure the coordination of wage-setting within labour market institutions, an indicator (CO) is used that is part of the ICTWSS: Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts as provided by the Amsterdam Institute for Advanced labour Studies (AIAS). It measures the degree of centralised coordination of wage-setting on a five point scale. The highest value (5) corresponds to a highly centralized wage-setting (imposed by the government, as a result of monopolistic union confederations and/or influential large firms) while the lowest value (1) corresponds to fragmented wage-setting (confined largely to individual firms or plants). This indicator is available for all countries in the sample from 1999 to 2011. For 2012, the values are assumed to be the same as in 2011. A high level of coordination in wage-setting is expected to lead to lower inflation (hypothesis 3).

Output gap – To control for cyclical inflation in the economy an estimation of the output gap (output gap) is used, taken from the European Commission’s Ameco database. It is defined as the gap between actual (real) GDP and potential GDP as a percentage of potential GDP, and is available for the entire sample. A higher value of the output gap is expected to increase inflation.

CPI gap – To control for catch-up effects, the difference between the price level of an individual country and the EMU12 average price level is used (CPI gap). These price levels are based on the price level index for actual individual consumption (Eurostat) which is available for the entire sample. A higher value for CPIGAP is expected to decrease inflation.

NEER – To control for imported inflation, a weighted version of the rate of change in the nominal effective exchange rate index is used (NEER). This variable is calculated by taking the annual rate of change in the nominal effective exchange rate index (41 trading partners) multiplied by the imports of goods and services as percentage of GDP. Data is obtained from Eurostat and is available for the whole sample. A rise in the nominal effective exchange rate

(29)

29 index means a strengthening of the currency and hence it is expected that a higher value for NEER will lead to lower inflation.

TAX – To control for a part of non-market based prices, an indicator is constructed that captures changes in taxes on products for countries in the EMU. It is defined as the annual average rate of change in the HICP minus the annual average rate of change in the HICP at constant taxes. Taxes include excise duties and special taxes; consumption taxes; car registration taxes; taxes on entertainment, insurance premiums and other specific services (such as hotels). Data is obtained from Eurostat and is available from 2004. No data is available for Ireland and Greece. When no data is available, a zero change in taxes on products is assumed. An increase in taxes on products is expected to lead to higher inflation.

4.3 Regression model

To get to an empirical model that can test the hypothesized effects posed in the first chapter, two steps are taken. First, an equation is set up that includes the most relevant factors in determining inflation in a given country. This equation is based on a backwards-looking version of the Phillips curve, with additional control factors for imported inflation, catch-up effects, taxes on products and factors that capture the potential effects of labour and product market institutions. Second, using the fact that the average inflation in the EMU12 is the same for all individual countries, the empirical model will be tailored to analyse inflation differentials specifically. This is accomplished including time specific effects that capture the common area wide effect for each year.

Considering the first step, table 4 shows the factors that are included as determinants of inflation in a given country i and at time t. Based on the literature review in chapter 1, this includes factors from the Phillips curve, some additional control factors (such as the NEER) and the possible direct effects that stem from labour and product market institutions (hypotheses 1, 2 and 3). Country fixed effects are allowed for to capture additional unobserved differences between countries. 12 Note that the coefficients of the Phillips curve ( ) are taken to be both

12_{An additional possibility is that, unlike the fixed effects (LSDV) model, the variation across countries can be} modeled to be random and uncorrelated with the independent variables in the model. In such case a random effects model would be appropriate. However, a Breusch and Pagan Lagrangian multiplier test for random effects (based on

(30)

30 country and time dependent. This is to allow for labour and product market institutions (that vary between countries and over time) to partially determine these coefficients. In other words, labour and product market institutions are thought to moderate the effect of the output gap on inflation (relating to hypotheses 4 and 5).

Table 4 – Step 1: theoretical determinants of inflation

( ̅ )  backwards-looking Phillips curve ( ̇)  additional control factors

 direct effects of institutions

 country fixed effects

In the table above, is the inflation rate, is real output, ̅ is potential output and the rate of change in the trade-weighted nominal effective exchange rate index. Furthermore, is the country and time specific degree of inflation persistence, the price level and ̇ the EMU12 average price level. represents the country and time specific coefficient for the response of inflation to the output gap, while , , and represent coefficients that are assumed to be constant. furthermore is the difference in taxes on products compared to the previous year and the direct effects of labour and product market institutions. Finally, represents the error term.

Considering the second step, a time dummy variable is added to capture the common area-wide inflation in a given year. Even though the dependent variable continues to be the inflation rate (and not the inflation differential) the remaining factors will now effectively only capture the contribution towards inflation differentials. 13 Finally, the country specific coefficients for effects of the output gap are taken to depend in part on the value of the structural indicators. Hence the empirical model is supplemented with interaction terms for the structural indicators with the output gap.

a random-effects GLS regression of model (8) as stated in table 6) does not reject the null hypothesis of zero variances across entities (p=0.25). Random-effects GLS regression is therefore not preferred over simple OLS or fixed effects (LSDV) regression.

(31)

31 Table 5 – Step 2: empirical model to analyse inflation differentials

inflationi,t = β1inflationi,t-1 + β2output gapi,t  backwards-looking Phillips curve

+ β3CPI gapi,t-1 + β4NEERi,t + β5TAXi,t  additional control factors

+ β6PMRi,t*output gapi,t + β7EPRi,t*output gapi,t  moderating effects of institutions

+ β8PMRi,t + β9EPRi,t + β10COi,t  direct effects of institutions

+  time dummy variables

+  country fixed effects

 error term

In this equation, the variable names are the same as those used in the previous paragraph: CPI gap stands for the deviation of CPI from the EMU12 average, NEER stands for the rate of change in the trade weighted nominal effective exchange rate, TAX for the tax variable, EPR represents the strictness of employment protection indicator, PMR represents the product market regulation indicator and CO the coordination of wage-setting indicator. In this equation represents a set of time dummy variables, represents a set of country dummy variables and represents the error term.

Finally, to make valid and reliable estimations, several conditions have to be met. Since homoscedasticity and the absence of autocorrelation cannot be assumed for most models, heteroskedasticity and autocorrelation consistent (HAC) standard errors are used where appropriate.14 Furthermore, to check for relevant dependencies between variables, the next section will estimate multiple models that include the institutional variables both individually and collectively. Differences in estimated coefficients and standard errors may then indicate multicollinearity issues. 15 A further test will be applied by calculating the variance inflation factor (VIF) for the variables in each model. VIFs lower than five are generally considered to indicate unproblematic multicollinearity (Dormann et al., 2012).

14_{A modified Wald test for groupwise heteroskedasticity does reject the null hypothesis of a constant variance} (p=0.000) and a Wooldridge test for autocorrelation in panel data rejects the null hypothesis of no first order autocorrelation (p=0.001). Both are based on model (8) in table 6.

15_{From a correlation table (provided in appendix II) it can be seen that the correlation between the independent} variables is rather low (the maximum correlation between is found between EPR and CO with a value of -0.35).

(32)

4.4 Empirical results

4.4.1 Main results

Table 6 – determinants of inflation: main results (EMU12, 1999-2012, LSDV estimation, dependent variable: HICP inflation rate)

Model (1) (2) (3) (4) (5) (6) (7) (8) Lagged inflation 0.502 *** 0.497 *** 0.496 *** 0.484 *** 0.487 *** 0.502 *** 0.489 *** 0.464 *** (0.078) (0.077) (0.077) (0.075) (0.074) (0.072) (0.080) (0.074) Output gap _{0.188 ***} _{0.189 ***} _{0.191 ***} _{0.154 ***} _{0.188 ***} _{0.172 ***} _{0.196 ***} _{0.129 ***} (0.038) (0.038) (0.041) (0.037) (0.04) (0.041) (0.051) (0.048) CPI gap -0.007 - - - - (0.016) - - - - NEER _0.000 _-0.001 _-0.001 _0.000 _0.000 _0.000 _-0.001 _0.001 (0.004) (0.005) (0.005) (0.004) (0.003) (0.003) (0.005) (0.003) TAX _{0.741 ***} _{0.747 ***} _{0.758 ***} _{0.649 ***} _{0.754 ***} _{0.880 ***} _{0.751 ***} _{0.836 ***} (0.190) (0.189) (0.197) (0.205) (0.24) (0.248) (0.193) (0.254) PMR * output gap -0.014 -0.010 (0.051) (0.038)

EPR * output gap _-0.051 _{-0.061 **}

(0.041) (0.026) PMR 0.171 0.245 (0.234) (0.241) EPR _{0.495 *} _{0.691 **} (0.275) (0.284) CO _0.055 _0.043 (0.181) (0.115) Observations ₁₅₆ ₁₅₆ ₁₅₆ ₁₅₆ ₁₅₆ ₁₅₆ ₁₅₆ ₁₅₆ R-squared _0.72 _0.72 _0.72 _0.73 _0.72 _0.73 _0.72 _0.74