275 ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 | 275-308 INFLATION – THE HARROD-BALASSA- SAMUELSON EFFECT IN A DSGE MODEL SETTING Received: February 26, 2018 ČRT LENARČIČ 1 Accepted: September 15, 2018 ABSTRACT: This paper sets up a two-country two-sector dynamic stochastic general equilibrium model that introduces sector specific productivity shocks with quality improvement mechanism of goods. It provides a model-based theoretical background for the Harrod-Balassa-Samuelson phenomenon that describes the relationship between productivity and price inflation within different sectors in a particular economy. Both, the calibrated and the estimated model are able to show that the Harrod-Balassa-Samuelson effect is confirmed by inducing tradable sector productivity shocks as they drive the non- tradable sector price inflation higher than the tradable sector price inflation. By doing this, we overcome the problem that the tradable productivity increase in a typical open economy specification reduces the relative price of domestic tradable goods relative to the foreign ones. Key words: Harrod-Balassa-Samuelson effect, DSGE model, inflation, productivity, quality improvement JEL classification: C32, E31, E32 DOI: 10.15458/ebr.86 1 INTRODUCTION The relationship between productivity and price inflation is described by the theory of the Harrod-Balassa-Samuelson phenomenon (henceforth HBS). Harrod (1933), Balassa (1964) and Samuelson (1964) independently developed and formulated the HBS productivity approach in order to explain the purchasing power parity 2 . The HBS effect represents a tendency for countries that experience higher tradable-sector productivity growth compared to non-tradable sector productivity growth to have higher overall price levels (Obstfeld and Rogoff, 1996). In more detail, the basic idea behind it is that the growth in the productivity of a tradable sector influences the growth of wages in the tradable and later on in the non-tradable sector. W age growth in the tradable sector consequently affects the growth of prices in the non-tradable sector. Depending on the nominal exchange rate regime of a particular economy, it affects the real exchange rate as well. However, Betts and Kehoe (2008) studied the relationship between the real exchange rate and the relative price of non-tradable to tradable goods. Their conclusion is that the relation between the 1 Bank of Slovenia, Ljubljana, Slovenia, e-mail: crt.lenarcic@bsi.si 2 Baumol and Bowen (1967) developed a similar model that only describes the relationship between productivity and wages, and presents an important part of the HBS hypothesis, as discussed by W agner and Hlouskova (2004). ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 276 two variables is stronger in an intense trade environment. Therefore, the basic assumption is that the relationship between the relative growth in the productivities of the tradable to non-tradable sector and the relative price of non-tradable to tradable goods is relatively straightforward if we include sectoral data for European countries, for example. In addition to the close trade environment, the sole Euro area integration process suppresses the ability of economies to adjust through the nominal exchange rate channel, which could consequently put more pressure on non-tradable price inflation. The HBS hypothesis can be tested on different entities, which in general represent countries, regions, or in many cases, sectors. In our case, we divide these entities into a tradable sector and a non-tradable sector; we use a similar principle as the De Gregorio, Giovannini, and Wolf's (1994) methodology by using the ratio of exports to total production to define both sectors. In order to do that we include and combine the NACE Revision 2 10-sector breakdown statistical classification time series data of economic activities, which provides data on labour productivity and price levels across the two sectors, and the ratio of exports to total production data calculated from the input-output tables, which are available at the World Input-Output Database (WIOD) 3 . By obtaining the relevant tradable and non- tradable data for further analysis and adding other observable macroeconomic data, we estimate the constructed DSGE model. The problem of permanent tradable productivity increase in a typical dynamic open economy specification is reducing the relative price of domestic tradable goods relative to the foreign ones. This implies worsening the terms of trade for the domestic economy and consequently, its real exchange does not increase. These dynamics are not consistent with empirical evidence found for the new European Union member states. The main contribution of the paper is to overcome the typical dynamic open economy setting by constructing and estimating a two-country two-sector DSGE model with the quality improvement extension, proposed by Masten (2008), in a smaller calibrated version of a dynamic model. The basic assumption is therefore the separation of the economy into a tradable and a non-tradable sector. The tradable sector is open and allows domestic goods to be exported and foreign goods to be imported, whereas the non-tradable sector is closed to foreign markets (a similar structure was used by Masten, 2008; Rabanal, 2009; Micaleff and Cyrus, 2013). The assumption is that the tradable and non-tradable sectors are exposed to different productivity shocks; this means that non-stationary real variables can grow at a different pace, thus providing a case for the HBS effect. In specifying technology, we allow a quality improvement mechanism, which is needed to replicate the appreciation of prices, without resorting to the unrealistic assumption of perfect competition in the tradable sector (Masten, 2008). We find evidence for the HBS effect, based on an augmented technology process that considers a quality improvement mechanism, which affects marginal costs by requiring 3 In defining the tradable and the non-tradable sector we differ from the standard approach used in the literature by excluding the not distinctively tradable or non-tradable sectors from the analysis. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 277 the use of more advanced inputs in the production process. The quality improvement of goods overcomes the typical open economy theoretical specification that reduces the relative prices of domestic tradable goods relative to foreign prices, and consequently worsens the terms of trade for the domestic economy. By introducing a sector-specific domestic tradable technology shock, the modelled economy responds by increasing price differential of non-tradable relative to tradable prices and the overall domestic inflation. Doing this we are able to theoretically explain why the economies with higher economic and productivity growth during the catching-up phase experienced higher inflation. In Section 2, a review of the HBS related literature is presented and discussed. Section 3 provides a theoretical framework for the DSGE model. In section 4, the classification and definition of economic activities into a tradable and a non-tradable sector is presented, obtaining sectoral price indices and time series of sectoral labour productivity growths. The calibrated model is presented in Section 5, while the estimation results of the DSGE model are given and discussed in Section 6. Section 7 presents the conclusions. 2 LITERATURE REVIEW Despite treating the HBS theory as an old idea, in which the sectoral productivity differential is seen as the driver for price inflation in the non-tradable sector (Harrod, 1933; Balassa, 1964; and Samuelson, 1964), the empirical testing of the HBS effect only became more popular in recent years as econometric methods advanced and new (or additional) time series data became available. This availability was largely due to the establishment of the EU and later on its enlargement process, together with advances and convergence of methodologies in collecting data by the national statistical offices. At the same time, addressing the HBS issue became relevant from the economic policy perspective trying to identify the different sources of (structural) inflation. Betts and Kehoe (2008) show that a close trade environment lowers the significance of the nominal exchange rate adjustment. This was (and can still be) especially important for the future EU and euro area countries, which are obliged to satisfy the Maastricht criterion of low and stable inflation, as well as for other emerging economies in trying to stabilise their overall inflation. In their comprehensive survey, Tica and Družić (2006) gather empirical evidence regarding the HBS effect. They point out that most of the empirical work supports the HBS effect. Especially strong evidence comes from the work based on cross-section empirical studies, similar to Balassa's (1964) work. A large number of papers focus on studying the magnitude of the HBS effect in accession countries in the EU. Čihák and Holub (2001) for instance study the presence of the HBS effect in the Czech Republic vis-à-vis EU countries, while allowing for differences in structures of relative prices. Jazbec (2002) considers Slovenia as the HBS case of an accession country, while Dedu and Dumitrescu (2010) test the HBS effect using Romanian data. Papers by Cipriani (2000), Coricelli, and Jazbec (2004), Halpern and Wyplosz (2001), Arratibel, Rodríguez-Palenzuela, and Thimann (2002), Breuss (2003), Wagner and Hlouskova (2004), Mihaljek and Klau (2008) consider ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 278 a larger accession country panel. Some of the work focuses also on emerging economies. Jabeen, Malik, and Haider (2011) test the HBS hypothesis on Pakistani data, while Guo and Hall (2010) test the HBS effect on Chinese regional data. These empirical strands of the HBS effect related literature opened up new questions regarding data issues and were related mostly to availability in reliability of sectoral data. As databases, especially in Europe, became more complete, new available data enabled studying the HBS effect between individual tradable and non-tradable sectors of a particular economy. Since it is difficult to clearly divide tradable and non-tradable commodities in the real world, some of the early papers tried to identify the tradability/ non-tradability of commodities. Officer (1976) proposed that manufacturing and/or industry combine a tradable sector, while the services represent the non-tradable sector. De Gregorio, Giovannini, and Wolf (1994) used a ratio of exports to total production of each sector to define both sectors. In empirical studies, mostly total factor productivity (TFP) or average productivity of labour are used. Marston (1987), De Gregorio, Giovannini, and Wolf (1994), De Gregorio and Wolf (1994), Chinn and Johnston, (1997), Halikias, Swagel, and Allan (1999), Kakkar (2002), and Lojshová (2003) use total factor productivity as a productivity proxy, while due to the lack of data on TFP , many others, i.e., Coricelli and Jazbec (2004) and Žumer (2002), use average productivity of labour. Comparing total factor productivity and average productivity of labour, the argument against the use of average productivity of labour is that it is not completely clear if average labour productivity should be regarded as a reliable indicator for representing a sustainable productivity growth, which has a long-term effect on the economy (De Gregorio and Wolf, 1994). However, according to Canzoneri, Cumby, and Diba (1999), the argument against TFP is that TFP is a result of a possibly unreliable data collection of sectoral capital stocks comparing to data collection of sectoral employment and sectoral gross value added, especially in the case of the shorter-term series. Sargent and Rodriguez (2000) also conclude that if the intent of the research is to examine trends in the economy over a period of less than a decade or so, labour productivity would be a better measure than total factor productivity. According to Kovács (2002), another setback of using TFP is that during the catching-up phase the capital accumulation intensifies faster in the transition/accession countries than in the developed countries, due to the lower starting point in fundamentals of transition/accession countries. Therefore, the HBS effect might be overestimated. Listing some of the arguments against using TFP , we rather include the average labour productivity as a productivity proxy in the model. Comparing to the vast HBS literature in the 2000s in the accession process of the countries to the EU and the monetary union, less theoretical work was done with regards to the HBS effect in more structural and more complex models. Rogoff (1992) was the first to implement a general equilibrium framework, introducing the demand side of the economy within the HBS theory. This opened the possibility to further investigate the effects of relative productivities of production factors and the effects of the demand side of the economy on price levels. For instance, Mihaljek and Klau (2002) conclud that the Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 279 HBS effect could have important policy implications for the EU accession countries in order to satisfy the Maastricht inflation criterion. To further investigate Mihaljek's point, Masten (2008) constructs a two-sector DSGE model to see whether the HBS effect could represent an issue in satisfying the Maastricht inflation criterion. Further on, Natalucci and Ravenna (2002) compare the magnitude of the HBS effect within different exchange rate regimes in the general equilibrium model, while Restout (2009) allows for varying mark-ups in its general equilibrium framework. However, Asea and Mendoza (1994) conclude that the proof of the HBS theory within the framework of general equilibrium cannot reliably asses the relationship between output per capita and domestic relative prices. In other words, conclusions regarding the HBS theory from cross-country analyses can only be conditionally accepted since it is difficult to account for cross-country trend deviations from purchasing power parity (PPP). Even more, Bergin, Glick, and Taylor (2004) show that the relationship between output per capita and domestic relative prices had historically oscillated too much for the HBS theory to be proved by cross-section empirical studies. In order to test the HBS theory their suggestion is that it should be tested with a sector-specific analysis. Following the general equilibrium strand of the HBS related literature, Rabanal (2009) offers three explanations for studying sectoral inflation dynamics in Spain in a DSGE model structure. The first explanation relates to the role of productivity growth differentials, which directly brings the possibility to study the HBS effect. Altissimo et al. (2005) introduced a seminal paper on productivity growth differentials in a DSGE model setting. The second explanation adds the role of the demand-side effects in shaping the inflation dynamics (López-Salido et al., 2005). The third explanation suggests that, due to different product and labor market structures, there is heterogeneity of inflation dynamics processes in each country of the union (Angeloni and Ehrmann, 2007; Andrés et al., 2003). Rabanal (2009) concludes that even when economies are hit by symmetric external shocks, such as for example oil prices, world demand, or nominal exchange rate, the response of sectoral inflation will be different across countries. The Rabanal's model was adopted by Micaleff and Cyrus (2013) as well. They analyse the relative importance of the three main determinants of inflation differentials in Malta. Based on these considerations, a structured theoretical framework is presented in the following section. 3 MODEL In this section, we present the theoretical framework for the two-country two-sector DGSE model. The DSGE framework follows the Rabanal (2009) model, but the main contribution of the theoretical model is its extension for sectoral wage rigidites, thus making the model more realistic. Additionally, we introduce an augmented technology process with quality improvement (Masten, 2008). In order to investigate the HBS effect phenomenon, different sectoral productivity shocks have to be introduced providing assymetricity between sectors. The monetary union is made of two economies; a domestic and a foreign country with the common monetary policy rule. They are indexed on intervals [0,s] and [s,1], respectively, where s denotes the size of the domestic country with ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 280 respect to the two-country universe. In our case, we relate to Slovenia and the rest of the euro area. The following section only gives a structural domestic economy description since the foreign economy block is analogous to the domestic economy, which is in our case Slovenia. 3.1 Households The assumption is that the representative household maximizes its utility function, given by where C t (i) and L t (i) present consumption and quantity of work effort of a particular household. The parameter 0< β<1 is the discount factor of household. We assume that households value the current consumption more than the future one. The parameter 0<ϖ<∞ is the inverse of the elasticity of work effort with respect to the real wage (Frisch elasticity parameter). We assume consumption habits as well, which is represented by the parameter 01 presents the elasticity of substitution between tradable and non-tradable goods. Since the demand for tradable goods is not dependent only on domestic goods but foreign as well, the index of the tradable consumption good is written analogously to the equation (3) with which the aggregate consumption index is defined 4 We scale the variables in the model with so that the variables enter the model detrended, for example, . The scaling variable ensures a constant steady-state level of utility and is determined by productivity dynamics (Masten, 2008). Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 281 where ω HF represents the share of domestic tradable goods in the tradable consumption basket. The parameter ν HF >1 is therefore the elasticity of substitution between domestic tradable goods and tradable goods produced abroad. The indexes of individual goods are defined by the following equations and represent a continuum of differenced goods of the same type and The parameter ν>1 denotes the elasticity of substitution within one type of differentiated goods: c t H , c t F and c t N . The same principle can be applied to price indexes. The aggregate price index P t is then given by As above, the price index for tradable goods is given by Households have a set of contingent riskless euro area bonds B t EA at their disposal that pay one unit of currency in every possible state of nature in t+1. The assumption is that households can trade these bonds that pay a gross interest rate of R t EA . Since households are ex ante identical, they face the same budget constraint in each period: where W t represents the real wage, while ς t represents other income sources of households. As shown in Chari, Kehoe, and McGrattan (2002), the real exchange rate is given by ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 282 where the variables μ t and μ t * represent the marginal utilities of domestic and foreign consumption, respectively. Labour market is, in comparison to the Rabanal (2009) model, differentiated, thus provides a more realistic model assumption. Further on, the aggregation of work effort of both sectors (i.e., tradable and non-tradable) holds Against this backdrop, each household working in the tradable or the non-tradable sector sets its own wage (Erceg et al., 2000; Christiano et al., 2005). Firms aggregate the differentiated supply of labour by transforming it into a homogenous input of labour L t j , where j=N,T, in accordance with the Dixit-Stiglitz (1977) aggregator The parameter ν L,j is defined as the wage elasticity within different varieties of labour services in a particular sector, where j= N,T. Based on that the labour demand function for a particular is given by Combining equations (12) and (13) we get the aggregate wage, which is obtained from differentiated labour In order to introduce wage frictions in the model, we apply the Calvo (1983) principle. Each household has monopolistic power over the setting of its wage W t j (i), where j=N,T. Yet not all the households can set their optimal wage at any point of time, but only a fraction of households (1– α L,j ), where the Calvo parameter is defined on the interval 0<α L,j <1. The other part of households ( α L,j ) indexate their wage according to inflation target and current inflation. The wage inflation of a non-optimizing household is then given by Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 283 where the parameter 0< φ L,j <1 stands for the degree of wage indexation with respect to inflation target and current inflation, where j=N,T. When reoptimizing their wage in period t, workers of a particular sector choose an optimal wage W t j,opt in order to maximize household utility as opposed to their individual utility, where j=N,T. The utility is subject to a sequence of iso-elastic demand schedules for their labour type, and the usual sequence of household flow budget constraints. The first order condition associated with that problem can be written as where the expression Λ t,t+k =β k λ t+k / λ t represents the stochastic discount factor. The wage dynamics should therefore be where j=N,T. The average wage on an economy scale is then given by W t =(W t T ) ωTN (W t N ) 1-ωTN . 3.2 Firms On the supply side, there are three types of firms, producing two types of tradable goods (indexed by H,H * ) and domestic non-tradable goods (indexed by N). Each type of firm is facing price rigidities (Calvo, 1983). That means that only a fraction of firms (1– α i ), where i=N,H,H * , can set their optimal price. Other firms ( α i ), where i=N,H,H * , index their prices according to the inflation target and current inflation based on the parameter 0< φ i <1, where i=N,H,H * , which stands for the degree of price indexation with respect to inflation target and current inflation. Domestic and foreign economies are facing the same deterministic technology process, providing a case for output growth. This means that all the real variables entering the model are non-stationary in levels, but stationary in first differences. 3.2.1 Tradable sector In the tradable sector there are two types of firms. One type of firm produces tradable goods for the domestic market and tries to satisfy domestic consumption of tradable ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 284 goods, C t H . The other type of firm produces tradable goods meant for export and tries to satisfy the foreign consumption of domestic goods, C t H, * . Each firm in the tradable sector follows the Cobb-Douglas production function, where work effort is the only production factor and The variable A t T is a sector-specific productivity process that is characterised by quality improvement of higher-quality goods in the tradable sector index χ t =(Z t T ) θZ with quality improvement parameter θ Z >0 (Masten, 2008), so that The variable χ t represents a quality improvement of goods index that influences wages and marginal costs via positive productivity shocks. Masten (2008) finds that the problem of permanent tradable productivity improvement in a typical open economy specification reduces the relative price of domestic tradable goods relative to the foreign ones, thus worsens the terms of trade. Consequently, the real exchange does not increase and is not consistent with empirical evidence based on the new European Union member states. On the other hand, introducing quality improvement of higher-quality goods may require the use of more advanced inputs in the production process and will consequently increase the marginal costs and product prices. Sallekaris and Vijselaar (2004) introduce a similar mechanism, as they adjust capital with a simple quality correction. 5 The variable Z t T represents a tradable sector productivity shock, which is country-specific We assume that productivity shocks of both sectors can be different and that their growth rates could be different. We let the tradable productivity process Z t T to be affected by two different productivity innovations ε t Z,T , which are country and sector specific, and ε t Z , which represents a euro-area wide innovation. For the labour supply it holds L t T =L t T,H +L t T,H, * . 5 The idea of adjusting prices with quality improvements goes back into the 90s, as the study of Gordon (1990) tried to empirically document these biases. Later research focused on constructing quality-adjusted price indexes (Hulten, 1992; Greenwood et al., 1997; Cummins and Violante, 2002), production based estimates (Bahk and Gort, 1993), and capital model (Hobijn, 2000). Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 285 Tradable sector firms producing domestic goods for the domestic market maximize their profits according to subject to where the expression Λ t,t+k =β k λ t+k ⁄ λ t represents the stochastic discount factor, and is the tradable goods demand of a firm in time t+k. Y t H is the aggregate domestic- made tradable goods demand. Similarly , we can write the maximization profit function for tradable sector firms producing domestic goods for the foreign market subject to where the expression Λ t,t+k =β k λ t+k ⁄ λ t represents the stochastic discount factor, and (h ) is the tradable goods demand of a firm in time t+k. Y t H, * is the aggregate domestic tradable goods demand from abroad. Real marginal costs in the tradable sector for both types of firms are defined as MC t T . Marginal costs are defined as the real wage normalized for augmented productivity ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 286 Both types of tradable sector firms maximize their profit with respect to prices p t H (h) and p t H, * (f) and demands (h )and (f), respectively. The tradable price dynamics of domestic produced goods for the domestic market is where P t H, * ,opt is the optimal price and . The tradable price dynamics of domestic goods for the foreign market is where P t H, * ,opt is the optimal price and . 3.2.2 Non-tradable sector Analogously to the tradable sector, each non-tradable sector firm follows the Cobb- Douglas production function, where work effort is the only production factor The variable A t N is a sector-specific productivity process that is characterised by quality improvement index so that In this respect, we assume that the sector-specific productivity process A t N is affected by quality improvement of goods χ t in the tradable sector, while the variable Z t N represents a non-tradable sector productivity shock, which is again country-specific where we let the non-tradable productivity process Z t N to be affected by a sector-specific innovation, ε t Z,N . Non-tradable sector firms maximize their profits Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 287 subject to where the expression Λ t,t+k =β k λ t+k ⁄ λ t represents the stochastic discount factor, and is the non-tradable goods demand of a firm in time t+k. Y t N is the aggregate non-tradable goods demand. Real marginal costs in the non-tradable sector are defined as MC t N . From the cost-optimization perspective, the marginal costs are defined as the real wage normalized for productivity A non-tradable sector firm maximizes its profit with respect to price p t N (n) and demand . The non-tradable price dynamics should therefore be where P t N,opt is the optimal price and . 3.3 Monetary policy Monetary policy is modelled as a Taylor rule (Taylor, 1993) and is the same for both economies where ε t MP represents the monetary policy shock, while the interest rate R t EA responds to inflation and output gaps. The total output of the euro area is defined by Y t EA =(Y t ) s (Y t * ) 1-s , while the overall inflation in the euro area is defined by Π t EA =(Π t ) s (Π t * ) 1-s , where s is the size of the domestic country. The parameter is the weight parameter for the responsiveness of the past interest rate, while γ π and γ y are Taylor type paramaters for the response of the interest rate accordingly to both gaps. 3.4 Market clearing The clearing conditions are ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 288 and where variables G t T and G t N represent exogenous government spending shocks. Combining equations (37) and (38), the real GDP is What is left to do is to define the government sectoral spending process where i=N,T. 4 TRADABILITY OF SECTORS AND DATA As the theoretical model is divided into tradable and non-tradable sectors, some attention is needed for the specification and the sectoral definition of the data. The dataset consists of quarterly Slovene and euro area sectoral data, which is available from the Eurostat 6 website. The time series data spans from 1998Q4 to 2018Q1 and includes sectoral gross value added data and sectoral price indexes data. 4.1 Tradability of sectors To begin with, the tradability of the sectors has to be defined. Officer (1976) proposes the following sector division. Manufacturing and other industry activities represent the tradable sector, while the services represent the non-tradable sector. De Gregorio et al. (1994) use a ratio of exports to total production to define both sectors. Their division threshold is set to 10 percent, stating that the sector is defined as tradable if the ratio of exports exceeds the 10 percent threshold, and the sector is defined as non-tradable if the ratio of exports does not exceed the 10 percent threshold. Following the De Gregorio et al. (1994) sector division, we take a step further by strictly distinguishing between the tradable and the non-tradable sector. This means that we exclude those activities from the analysis that oscillate around the 10 percent threshold too much. We provide a more detailed specification below. 6 Available at the European Commmission's statistical database site http://epp.eurostat.ec.europa.eu/portal/ page/portal/eurostat/home/. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 289 First, data on the share of exports in total value added have to be extracted from the input- output tables available at the World Input-Output Database (WIOD). We use a standard ISIC/NACE Revision 2 aggregation category, which is used for reporting data from the System of National Accounts (SNA) for a wide range of countries. We present a 10-sector breakdown in Table 1. Table 1: NACE Revision 2 10-sector classification of economic activities Source: European Commission, author's calculations. *Note: Countries, such as Belgium, the Netherlands and Luxembourg, stand out with their ratio-of-export figures, thus driving up the average of ratio of exports in the agriculture sector. **Note: Countries, such as Ireland, the Netherlands and Luxembourg, stand out with their ratio-of-export figures, thus driving up the average of ratio of exports in the professional services sector. As mentioned above, to divide the 10 sectors into tradable and non-tradable sectors, we use a similar approach as De Gregorio et al. (1994). However, in the present paper we put emphasis only on strictly tradable and non-tradable sectors, meaning that the sectors which are not distinctively tradable or non-tradable are exluded from the analysis. A sector is then treated as tradable if its ratio of exports exceeds the 10 percent threshold for at least 75 percent of time using the WIOD data in the 2000-2011 period. The same principle is applied for the definition of a non-tradable sector. A sector is treated as non-tradable if its ratio of exports is under the 10 percent threshold for at least 75 percent of time using the WIOD data in the 2000-2011 period. Applying stricter conditions regarding the division of sector means that NACE Rev. 2 sectors, such as agriculture, forestry and fishing (A), information and communication (J), financial and insurance activities (K), professional, scientific, technical, administration and support services (M and N), are excluded from the analysis. These excluded sectors account for around 20 percent in total value added. Based on this threshold the manufacturing, mining, quarrying and other industries (B, C, NACE Revision 2 Sector description Ratio of exports (in %) Tradability A Agriculture, forestry and fishing 18.32* B, C, D, E Manufacturing, mining and quarrying and other industry 45.99 T F Construction 2.20 N G, H, I Wholesale and retail trade, transportation and storage, accommodation and food services 17.25 T J Information and communication 10.42 K Financial and insurance activities 12.63 L Real estate activities 0.56 N M, N Professional, scientific, technical, administrative and support services 16.39** O, P , Q Public administration, defence, education, human health and social work services 0.95 N R, S, T, U Other services 6.27 N ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 290 D and E), wholesale, retail, transportation, storage, accommodation and food services (G, H and I) are treated as tradable sectors, while construction (F), real estate activities (L), public administration, defence, education, human health, social work services (O, P and Q), and other services (R, S, T and U) are treated as non-tradable sectors. 4.2 Sectoral inflation and productivity Based on quarterly data available from the Eurostat website and consideration of the classification of economic activities into a tradable and a non-tradable sector (as defined in Table 1), supported by time-varying sectoral gross value added weights expressed in millions of euros in 2015, growth rate in prices for the tradable and the non-tradable sector are obtained. We use the same principle that was applied to divide economic activities into the tradable and non-tradable sectors to divide sectoral growth rate of value added for both sectors, based on the aggregation done for sectoral inflation. This way we get growth rates for the output on a quarterly frequency basis for a separate sector, i.e. tradable and non-tradable. 4.3 Data entering the model After defining and obtaining the sectoral data, we can provide a full description of the dataset entering the model in Table 2. There are 9 observable variables at a quarterly frequency in the period of 1998Q4-2018Q1, thus providing 78 observations. Tradable sector figures stand out the most and have the highest variability. Intuitively, this means that the tradable sector is more responsive to changes in different phases of business cycles. Additionally, Slovene data in comparison to the euro area data varies more, thus providing a case that small open economies are more vulnerable to macroeconomic imbalances. Table 2: Descriptive statistics (in p.p. deviations from the steady state) Variable description Data transformation Country Minimum Maximum Standard deviation Weighted tradable sector inflation demeaned log-differences SI -2.59 2.21 0.92 Weighted tradable sector inflation demeaned log-differences EA -1.08 1.31 0.39 Weighted tradable sector gross value added demeaned log-differences SI -10.17 3.02 1.64 Weighted tradable sector gross value added demeaned log-differences EA -6.36 1.23 1.07 Weighted non-tradable sector inflation demeaned log-differences SI -1.22 1.84 0.69 Weighted non-tradable sector inflation demeaned log-differences EA -0.76 0.80 0.30 Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 291 Source: Eurostat, author's calculations. 5 CALIBRATION OF THE MODEL We set the values of the calibrated parameters accordingly to known empirical facts from the existing literature and characteristics of the modelled economies, in our case Slovenia and the euro area. The discount factor β is set to 0.99, following Smets and Wouters’ (2003) paper. The degree of habit formation parameter h for Slovenia is set to 0.80 (as in Kilponen et al., 2015), while for the euro area it is set to 0.60 (as in Smets and Wouters, 2003), thus making Slovenia’s consumption slower to respond and more persistent. The Slovene economy size parameter s is set to 0.01. 7 The Frisch elasticity or the inverse of the elasticity of work effort for both economies has a typical parameter value of 2 (Smets and Wouters, 2003; Rabanal, 2009; Rabanal, 2012; Micallef and Cyrus, 2013). The elasticities of substitution between tradable and non-tradable goods for both, domestic ( ν TN ) and foreign (ν TN, * ), economies, take the value of 0.44, following the values set by Stockman and Tesar (1995). The elasticities of substitution between domestic produced and foreign produced goods for both, domestic ( ν HF ) and foreign (ν HF, * ) economies, take the value of 1.5, following Chari, Kehoe, and McGrattan (2002). Furthermore, the shares of important economic variables are calibrated as well. The share of government spending relative to GDP in Slovenia is set to 0.17 and for the euro area it is set to 0.20, while the average share of tradable goods in the consumption basket is set to 0.58 in Slovenia and 0.61 in the euro area. The Calvo wage parameters for both areass and both sectors are set to 0.81, while the price stickiness is set to 0.75, following the values set for Slovenia in Clancy, Jacquinot, and Lozej (2014) and Kilponen et al. (2015). The wage indexation parameters are set to 0.75, according to Rabanal (2012). The quality improvement parameters θ Z and θ Z,* for both economies are set to 0.25. The Taylor rule values inflation and output gap response parameters γ π = 1.5 and γ y = 0.1 take usual values when modelling the euro area monetary policy close to Fourçans and Vranceanu’s (2004) estimation of the euro area parameters. 7 In comparison to the euro area the size of the Slovene economy is even smaller. The reason behind a slightly bigger economy size parameter is that very small numbers of the parameters could represent numerical difficulties for the model. These are shown in a very slow convergence after shocking the model or even in the inability of computing the responses of the shocks. However, 0.01 economy size parameter does not significantly influences the universum of both economies, which would be the case for small open economies. Variable description Data transformation Country Minimum Maximum Standard deviation Weighted non-tradable sector gross value added demeaned log-differences SI -3.20 5.13 1.51 Weighted non-tradable sector gross value added demeaned log-differences EA -0.73 1.02 0.41 3-month Euribor Interest rate given by log(1+r t ⁄400), demeaned log-differences EA -0.55 0.78 0.42 ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 292 The calibrated model is able to produce the HBS type of productivity shock. The following figure shows the impulse responses of the main macroeconomic variables to a 1 p.p. domestic tradable sector productivity shock, based on the calibrated model. The productivity shock increases the production of both sectors, tradable and non-tradable. As the quality improvement mechanism takes place, firms are compelled to raise wages since more sophisticated labour force is needed with the productivity picking up. The pick-up in wages increases inflation and consumption in both sectors. What is noteworthy is that inflation in the non-tradable sector increases more than in the tradable sector, thus providing a case for the HBS effect. Figure 1: Impulse responses of the main variables to a 1 p.p. domestic tradable sector productivity shock (deviations from steady state, in p.p.) 6 ESTIMATION OF THE MODEL AND COMPARISON WITH THE CALIBRATED MODEL With the obtained dataset and the calibration parameters set, the two-country two- sector DSGE model is ready to be estimated. Doing that, we use the Bayesian inference methodology. We set the prior distribution of the estimated parameters, given in Table 3. The prior and the posterior distribution of the estimated parameters and the shocks is presented in Table 3, while the figures with comparisons between the prior and the posterior distribution of the parameters are presented in Appendix A, and in Appendix B the dynamics of the exogenous shocks is presented. The Metropolis-Hastings MCMC algorithm is used with 300,000 steps and two sequential chains with the acceptance rate per chain of around 30%. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 293 W e estimate the quality improvement parameters θ Z and θ Z,* for both economies. The priors of both parameters were set to 0.25, while the estimates of both parameters took the values of 0.1676 and 0.2127, respectively. The estimated values of both quality improvement parameters are below the calibrated value of the parameter for the domestic economy in Masten (2008). Since Slovenia was catching up the average of the euro area and experienced higher growth and inflation, the estimate of the quality improvement mechanism had to be stronger during this period. With respect to the other estimated parameters, the shock persistence parameters seem to suggest that the productivity persistence parameters show less persistence than the demand shocks entering both the non-tradable and the tradable sector. The parameter of the monetary policy rule is estimated as well and takes the value of 0.6250, suggesting a relatively high persistence of the past interest rate. In comparison to the calibrated model, the Calvo price and wage rigidity parameters ( α's) are estimated to be higher, meaning that the prices and wages respond slower to exogenous shocks. The values of the Calvo parameters are similar comparing the foreign or domestic economy. Table 3: Prior and posterior distribution of the estimated parameters and shocks Parameter Calibration model values Prior mode Posterior mode 90% HPD interval Prior distribution Prior distribution θ Z 0.250 0.250 0.1676 0.1061 0.2268 inv. gamma 0.100 θ Z,* 0.250 0.250 0.2127 0.1153 0.3393 inv. gamma 0.100 α H 0.750 0.750 0.6259 0.5828 0.6676 beta 0.150 α F 0.750 0.750 0.8955 0.8524 0.9355 beta 0.150 α H,* 0.750 0.750 0.8742 0.7620 0.9975 beta 0.150 α F,* 0.750 0.750 0.9200 0.8963 0.9412 beta 0.150 α N 0.750 0.750 0.8519 0.8250 0.8746 beta 0.150 α N,* 0.750 0.750 0.9550 0.9415 0.9686 beta 0.150 α W, T 0.810 0.810 0.9010 0.8395 0.9659 beta 0.070 α W ,T,* 0.750 0.750 0.8249 0.7279 0.9198 beta 0.070 α W, N 0.810 0.810 0.8889 0.8392 0.9367 beta 0.070 α W ,N,* 0.750 0.750 0.7920 0.6909 0.8920 beta 0.070 ν TN 0.440 0.500 0.5471 0.1888 0.8864 gamma 0.200 ν HF 1.500 1.500 1.1671 0.5602 1.7190 gamma 0.500 φ H 0.500 0.500 0.2600 0.0614 0.4389 beta 0.200 φ F 0.500 0.500 0.4762 0.2029 0.7278 beta 0.200 φ H,* 0.500 0.500 0.4643 0.3093 0.6329 beta 0.100 φ F,* 0.500 0.500 0.1014 0.0120 0.1864 beta 0.200 φ N 0.500 0.500 0.3958 0.2158 0.6011 beta 0.100 ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 294 Source: Author’s calculations. 6.1 Impulse response functions and the historical shock decomposition In this subsection, we present the historical shock decomposition and impulse response functions. The purpose of both is to provide a description of the severity of shocks that influence the macroeconomic variables. Figure 2 shows the contributions of the exogenous shocks onto the price differential between the non-tradable and tradable sectors through time. It is evident that the inflation differential between the non-tradable and tradable sectors has been influenced by productivity components. As the financial crisis lingered on in the second wave after 2010, the difference between the non-tradable and tradable dynamics turned to be negative, implying a slowdown in the tradable sector productivity. Only with the start of the recovery of the Slovene economy in 2015, the difference between the inflation of both sectors returned to positive figures and has continued the pattern from before the financial crisis in 2008 by being affected with positive tradable sector productivity shocks. Parameter Calibration model values Prior mode Posterior mode 90% HPD interval Prior distribution Prior distribution φ N,* 0.500 0.500 0.2926 0.1718 0.4120 beta 0.100 ρ Z,T 0.750 0.750 0.3940 0.2721 0.5210 beta 0.100 ρ Z,T,* 0.750 0.750 0.3958 0.2485 0.5370 beta 0.100 ρ Z,N 0.750 0.750 0.5840 0.4225 0.7360 beta 0.100 ρ Z,N,* 0.750 0.750 0.4358 0.2903 0.6000 beta 0.100 ρ G,T 0.750 0.750 0.8392 0.6415 0.9662 beta 0.100 ρ G,T,* 0.750 0.750 0.8468 0.7520 0.9511 beta 0.100 ρ G,N 0.750 0.750 0.9030 0.6564 0.9918 beta 0.100 ρ G,N,* 0.750 0.750 0.8062 0.6567 0.9330 beta 0.100 0.750 0.750 0.6250 0.4788 0.7758 beta 0.100 ε MP - 0.4000 0.1266 0.1113 0.1417 inv. gamma 0.1000 ε Z - 0.5000 0.1840 0.1509 0.2156 inv. gamma 0.2000 ε Z,T - 0.7000 0.2593 0.2211 0.2978 inv. gamma 0.2000 ε Z,T* - 0.5000 0.3710 0.2322 0.5064 inv. gamma 0.2000 ε Z,N - 0.7000 0.2472 0.2126 0.2803 inv. gamma 0.2000 ε Z,N,* - 0.5000 0.5192 0.2784 0.7470 inv. gamma 0.2000 ε G,T - 1.0000 0.5061 0.4420 0.5733 inv. gamma 0.2000 ε G,T,* - 1.0000 0.4088 0.3539 0.4595 inv. gamma 0.2000 ε G,N - 1.0000 0.6004 0.5109 0.6883 inv. gamma 0.2000 ε G,N,* - 1.0000 0.4032 0.3538 0.4484 inv. gamma 0.2000 Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 295 Figure 2: Historical shock decomposition in the inflation differential between the non- tradable and the tradable sector (deviations from steady state, in p.p.) *Note: Tradable productivity shocks are the sum of the contributions of the country-specific domestic tradable sector shocks ε Z,T and ε Z,T, * and the common productivity shock ε Z . The non-tradable sector productivity shocks ε Z,N and ε Z,N, * are depicted separately. Other shocks are the sum of the contributions of the government spending shocks ( ε G,T, ε G,T, * , ε G,N and ε G,N, * ) and the monetary policy shock ε MP . Source: Author’s calculations. It is more intuitive to look at the impulse response functions in order to understand the effects of productivity shocks. Figures (3-6) show the responses of the main macroeconomic variables to different exogenous shocks and depict a 20-period horizon. In studying the impulse responses, we will only consider the productivity shocks that hit the two economies. Figure 3 displays the impulse responses of the main variables to a 1 p.p. domestic tradable sector productivity shock ε Z,T . When a positive productivity shock hits the tradable sector, tradable and non-tradable inflation increases in Slovenia, causing the overall inflation to increase. This is due to a wage increase in the tradable sector via quality improvement mechanism that increases the need for more demanding inputs in the production process, thus increasing the marginal costs, as wages increase the marginal costs increase, causing the inflation to increase. The Harrod-Balassa-Samuelson type productivity shock causes the increase of output and consumption as well. Under the implementation of quality improvement mechanism and under the price and wage frictions the HBS effect seems to hold, based on the impulse responses, the effects on the euro area macroeconomic variables are small. ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 296 Figure 3: Impulse responses of the main variables to a 1 p.p. domestic tradable sector productivity shock (deviations from steady state, in p.p.) The same pattern is observed when we analyse a 1 p.p. common tradable sector technology shock ε Z , shown in Figure 4. Similar effects happen when a 1 p.p. foreign tradable sector productivity shock ε Z,T,* hits the rest of the euro area (Figure 5). The difference is that this time the quality mechanism works abroad, so that spillovers come with a lag and in smaller magnitude. As a consequence, marginal costs do not increase in the domestic country, but positive effects from the price increase abroad make the tradable sector more profitable, increasing production, consumption, price and wages in the domestic country. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 297 Figure 4: Impulse responses of the main variables to a 1 p.p. common euro area tradable sector productivity shock (deviations from steady state, in p.p.) Figure 5: Impulse responses of the main variables to a 1 p.p. foreign tradable sector productivity shock (in p.p. deviations from steady state) ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 298 We are left to study the effects of the non-tradable sector productivity shocks, as they are depicted in Figure 6. In contrast to the tradable sector productivity shocks, the domestic non-tradable sector productivity shock ε Z,N does not enter the quality improvement mechanism. Consequently, it acts more as a (classical productivity) shock that decreases marginal costs and lowers non-tradable sector inflation, while the tradable sector marginally increases since the labour supply moves from the non-tradable sector to the tradable sector. The sectoral and the overall output, as well as the consumption, increase. Figure 6: Impulse responses of the main variables to a 1 p.p. domestic non-tradable sector productivity shock (deviations from steady state, in p.p.) 6.2 Policy implications and way forward The HBS effect is typically used to explain inflation differentials for countries experiencing a catching-up process. As the relatively poorer countries adopt new technologies in those sectors that are open to international trade (i.e., the tradable sector), they will experience higher productivity growth in the tradable sector, increased wages via quality mechanism, and consequently a higher inflation in sectors that are not open to international trade, as is the non-tradable sector. Therefore, the HBS effect hypothesis could help to explain higher inflation rates in the non-tradable sector than in the tradable sector, hence leading to higher overall inflation. Another important issue to point out is that the HBS effect theory does not explain the possible sources of productivity differentials between different sectors and countries. As Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 299 the HBS is often associated with catching-up and convergence phases of less developed countries, there is a possibility that a catching-up process could take place without the HBS effect. This happens if productivity growth in both sectors (i.e., tradable and non- tradable) is equally high. Additionally, some countries that already experience high productivity levels may for various reasons (i.e., economic policies that are conducive to technological innovation) also experience relatively high productivity growth in the tradable sector. Importantly, in those countries structural rigidities and different degrees of competition 8 can affect productivity growth differentials between sectors and overall productivity growth in a way that favours either positive or negative inflation differentials. Despite wage setting being typical for the DSGE model setting, following Calvo (1983) and later on Christiano (2005) labour market frictions, some issues could still arise in that respect. The wage setting in the non-tradable sector could be to a large extent governed by the non-market forces and other structural rigidities since a large part of the non-tradeable sector is comprised of the public sector. In our case, the model does not structurally distinguish between the private and the public sector and would consequently not be able to consider various types of non-market forces. However, it does provide some distinction in a sense of having two different (estimated) rigidity parameters of the wage setting equation for the non-tradable and the tradable sector. Based on the estimation figures the non-tradable sector wages seem to be more rigid than those in the tradable sector. They are slower in responding to exogenous shocks, which would to some extent simulate the differences between the private and the public sector. This issue could go beyond the scope of the present paper, but it could represent an additional way forward to extend the model into a more complex one by additionally restricting and dividing the modelled labour market, as well as the government sector. Nonetheless, the continued process of convergence processes in the euro area should lead to a decline in inflation dispersion amongst the euro area countries due to a price level and income convergence in the long-run. On the other hand, other structural factors such as differences in the degrees of wage and price rigidities and divergent degree of competition in domestic markets may have also contributed to the observed inflation differentials and their persistence. In this respect, the relative degree of market competition seems to be an important parameter in explaining the size and volatility of relative price responses to symmetric shocks across euro area countries. 7 CONCLUSION This paper draws conclusions based on a construction of a theoretical two-country two- sector DSGE model with both economies operating in a common monetary union. We were able to produce and show the existence of the HBS effect in a calibrated and estimated structural dynamic setting of the DSGE model by introducing a quality improvement mechanism that helps to explain why prices grow when productivity increases, especially in 8 i.e., the private vs. the public sector. ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 300 catching-up economies like the new EU member states in 2000. The quality improvement mechanism affects marginal costs by requiring the use of more advanced inputs in the production process. Quality improvement of goods overcomes a typical open economy theoretical specification that reduces the relative prices of domestic tradable goods relative to the foreign prices, and consequently worsens the terms of trade for the domestic economy. Despite showing the presence of the HBS effect, the effect per se is not large enough to pose significant risks to central banks in their quest for price stability. REFERENCES Altissimo, F., Benigno, P., & Rodriguéz-Palenzuela, D. (2005). Long-Run Determinants of Inflation Differentials in a Monetary Union . National Bureau of Economic Research Working Paper 11473. Andrés, J., Ortega, E., & Vallés, J. (2003). Market Structure and Inflation differentials in the European Monetary Union. Bank of Spain Working Paper 0301. Angeloni, I., & Ehrmann, M. (2007). Euro Area Inflation Differentials. The B.E. Journal of Macroeconomics, 7(1), 1-36. Arratibel, O., Rodríguez-Palenzuela, D., & Thimann, C. (2002). Inflation Dynamics and Dual Inflation in Accession Countries: A New Keynesian Perspective. ECB Working Paper, 132. Asea, P. & Mendoza, E. (1994). The Balassa-Samuelson Model: A General Equilibrium Appraisal. Review of International Economics, 2 (3), 244-267. Bahk, B.H., & Gort, M. (1993). Decomposing Learning by Doing in New Plants. Journal of Political Economy, 101(4), 561-583. Balassa, B. (1964). The Purchasing Power Parity Doctrine: A Reappraisal. Journal of Political Economy, 72(6), 584-596. Bank of Slovenia. (2018). Economic and Financial Developments, January 2018. Bank of Slovenia. Baumol, W .J., & Bowen, W .G. (1967). Performing Arts - The Economic Dilemma. A Study of Problems Common to Theater, Opera, Music and Dance. College Music Symposium, 7, 127-142. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 301 Bergin, P ., Reuven, G., & Taylor, A.M. (2006). Productivity, Tradability, and the Long-Run Price Puzzle. National Bureau of Economic ResearchWorking Paper 10569. Betts, C.M., & Kehoe, T.J. (2008). Real Exchange Rate Movements and the Relative Price of Non-traded Goods. Federal Reserve Bank of Minneapolis. Breuss, F. (2003). Balassa-Samuelson Effects in the CEEC: Are they Obstacles for Joining the EMU?. Wirtschaftsuniversit at Wien IEF Working Paper, 52. Calvo, G.A. (1983). Staggered Prices in a Utility-Maximizing Framework. Journal of Monetary Economics, 12, 383-398. Canzoneri, M.B., Cumby, R.E., & Diba, B. (1999). Relative Labor Productivity and the Real Exchange Rate in the Long Run: Evidence for a panel of OECD countries. Journal of International Economics, 47 , 245-266. Chari, V .V ., Kehoe, P ., & McGrattan, E. (2002). Can Sticky Price Models Generate Volatile and Persistent Exchange Rates?. Review of Economic Studies, 69, 555-563. Chinn, M.D., & Johnston, L. (1997). Real Exchange Rate Levels, Productivity and Demand Shocks: Evidence from a Panel of 14 Countries. IMF Working Paper, WP/97/66. Christiano, L.J., Eichenbaum, M., & Evans, C.E. (2005). Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy. Journal of Political Economy, 113(1), 1-45. Cipriani, M. (2001). The Balassa-Samuelson Effect in Transition Economies. IMF. Clancy, D., Jacquinot, P ., & Lozej, M. (2014). The Effects of Government Spending in a Small Open Economy within a Monetary Union. Central Bank of Ireland Research Technical Paper 12/RT/14. Comin, D. (2008). Total Factor Productivity. The New Palgrave Dictionary of Economics, 2nd ed. Palgrave Macmillan. Coricelli, F., & Jazbec, B. (2004). Real Exchange Rate Dynamics in Transition Economies. Structural Change and Economic Studies, 15 , 83-100. Cummins, J., & Violante, G. (2002). Investment-Specific Technical Change in the United States (1947-2000): Measurement and Macroeconomic Consequences. Review of Economic Dynamics, 5(2), 243-284. ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 302 Čihák, M., & Holub, T. (2001). Convergence of Relative Prices and Inflation in Central and Eastern Europe. IMF Working Paper, WP/01/124. De Gregorio, J., Giovannini, A., & Wolf, H.C. (1994). International Evidence on Tradables and Nontradables Inflation. IMF Working Paper, WP/94/33. De Gregorio, J., & Wolf, H.C. (1994). Terms of Trade, Productivity, and the Real Exchange Rate . The National Bureau of Economic Research Working Paper 4807. Dedu, V ., & Dumitrescu, B.A. (2010). The Balassa-Samuelson Effect in Romania. Romanian Journal of Economic Forecasting, 4/2010. Dixit, A.K., & Stiglitz, J.E. (1977). A Monopolistic Competition and Optimum Product Diversity. American Economic Review, 67(3), 297-308. Erceg, C.J., Henderson, D.W ., & Levin, A.L. (2000). Optimal Policy with Staggered Wage and Price Contracts. Journal of Monetary Economics, 46, 281-313. Fourçans, A., & Vranceanu, R. (2004). The ECB Interest Rate Rule under the Duisenberg Presidency. European Journal of Political Economy, 20(3), 579-595 Gordon, R.J. (1990). The Measurement of Durable Goods Prices . Chicago: University of Chicago Press. Greenwood, J., Hercowitz, Z., & Krusell, P . (1997). Long-Run Implications of Investment- Specific Technological Change. American Economic Review, 87(3), 342-362. Guo, Q., & Hall, S.G. (2010). A Test of the Balassa-Samuelson Effect Applied to Chinese Regional Data. Romanian Journal of Economic Forecasting, 2/2010. Halpern, L., & Wyplosz, C. (2001). Economic transformation and real exchange rates in the 2000s: The Balassa-Samuelson Connection . Economic Survey of Europe, no 1. Halikias, I., Swagel, P., & Allan, W. (1999). Greece, Selected Issues. IMF Staff Country Report, 99/138. Harrod, R.F. (1933). International Economics. Nisbet and Cambridge University Press. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 303 Hobijn, B. (2001). Embodiment in U.S. Manufacturing. mimeo, Federal Reserve Bank of New Y ork. Hulten, C. (1992). Growth Accounting when Technical Change is Embodied in Capital. American Economic Review, 82(4), 964-980. Jazbec, B. (2002). Balassa-Samuelson Effect in Transition Economies – the Case of Slovenia. William Davidson Working Paper 507. Kakkar, V. (2002). Capital-Labor Ratios and Total Factor Productivity in the Balassa- Samuelson Model. Review of International Economics, 10 (1), 166-176. Kilponen, J., Pisani, M., Schmidt, S., Corbo, V ., Hledik, T., Hollmayr, J., Hurtado,. Júlio, P ., Kulikov, D., Lemoine, M., Lozej, M., Lundvall, H., Maria, J.R., Micallef, B., Papageorgiou, D., Rysanek, J., Sideris, D., Thomas, C., & De Walque, G. (2015). Comparing Fiscal Multipliers accross Models and Countries in Europe . ECB Working Paper Series no. 1760. Kovács, M.A. (2002). On the Estimated Size of the Balassa-Samuelson Effect in CE5 Countries . Prepared by the CE5 National Banks for the Basle Meeting of March 2002. Kutasi, G. (2013). The Reverse Balassa-Samuelson Effect in the Euro Zone . Köz-Gazdaság 2013/1. Lojschová, A. (2003). Estimating the Impact of the Balassa-Samuelson Effect in Transition Economies. Institute for Advanced Studies, Vienna. López-Salido, J.D., Restoy, F ., & V allés, J. (2005). Inflation Differentials in EMU: The Spanish Case. Bank of Spain Working Paper 0514. Masten, I. (2008). Optimal Monetary Policy with Balassa-Samuelson-Type Productivity Shocks. Journal of Comparative Economics, 36, 120-141. Micaleff, B., & Cyrus, L. (2005). Inflation Differentials in a Monetary Union: the Case of Malta. Central Bank of Malta, WP/05/2013. Mihaljek, D., & Klau, M. (2002). The Balassa-Samuelson Effect in Central Europe: a Disaggregated Analysis. Bank for International Settlements Working Paper no. 143. ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 304 Mihaljek, D., & Klau, M. (2008). Catching-up and Inflation in Transition Economies: the Balassa-Samuelson Effect Revisited. BIS Working Papers 270, Bank for International Settlements. Obstfel, M., & Rogoff, K. (1996). Foundations of International Macroeconomics . The MIT Press. Officer, L.H. (1976). The Productivity Bias in Purchasing Power Parity: An Econometric Investigation. IMF Staff Paper 23, 545-579. Peltonen, T.A., and Sager, M. (2009). Productivity Shocks and Real Exchange Rates: a Reappraisal. ECB Working Paper Series no. 1046. Rabanal, P. (2009). Inflation Differentials between Spain and the EMU: A DSGE Perspective. Journal of Money, Credit and Banking, 41(6), 1141-1166. Rabanal, P., & Tuesta, V. (2013). Nontradable Goods and the Real Exchange Rate. Open Economies Review, 24(3), 495-535. Restout, R. (2009). The Balassa-Samuelson Model in General Equilibrium with Markup Variations. EconomiX Working Paper 2009-39. Rogoff, K. (1992). Traded Goods Consumption Smoothing and the Random Walk Behavior of the Real Exchange Rate. BOJ Monetary and Economic Studies, 10 (2), 1-29. Sakellaris, P., & Vijselaar, F.W. (2004). Capital Quality Improvement and the Sources of Growth in the Euro Area. ECB Working Paper Series no. 368. Samuelson, P.A. (1964). Theoretical Notes on Trade Problems. Review of Economics and Statistics, 46 (2), 145-154. Smets, F., & Wouters, R. (2003). An Estimated Dynamic Stochastic General Equilibrium Model of the Euro Area. Journal of the European Economic Association, 1 (5), 1123–1175. Sargent, T.C., & Rodriguez, E.R. (2000). Labour or Total Factor Productivity: Do We Need to Choose?. International Productivity Monitor, Centre for the Study of Living Standards, 1 , 41-44. Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 305 Stockman, A.C., & Tesar, L.L., (1995). Tastes and Technology in a Two-country Model of the Business Cycle: Explaining International Comovements. American Economic Review, 85(1), 168–185. Taylor, J.B. (1993). Discretion versus Policy Rules in Practice. Carnegie-Rochester Series on Public Policy, 39(1993), 195-214. Tica, J., & Družic, I. (2006). The Harrod-Balassa-Samuelson Effect: A Survey of Empirical Evidence. FEB Working Paper Series 06-07/686. Wagner, M., & Hlouskova, J. (2004). What’s Really the Story with this Balassa- Samuelson Effect in the CEECs? . Universität Bern Diskussionsschriften 04-16. Žumer, T. (2002). Estimation of the Balassa-Samuelson Effect in Slovenia . Banka Slovenije, Prikazi in Analize X/1. ECONOMIC AND BUSINESS REVIEW | VOl. 21 | No. 2 | 2019 306 APPENDICES Appendix A: Prior and posterior distribution Figure A1: Prior (dashed line) and posterior distribution (solid line) of the estimated shocks Figure A2: Prior (dashed line) and posterior distribution (solid line) of the estimated parameters Č. LENARČIČ | INFLATION – THE HARROD-BALASSA-SAMUELSON EFFECT ... 307 Appendix B: Exogenous shocks Figure B1: Exogenous shocks