Uncovering the link between a flexible exchange rate and fundamentals: the case of Central and Eastern European economies

ABSTRACT

This article examines the link between a nominal exchange rate and macrofundamentals in Central and Eastern European (CEE) countries. We use the model based on the monetary policy rule as a theoretical framework that explains the relations between the exchange rate and price level, risk premium, output gap, and expected inflation. It allows for endogeneity of the monetary policy – the issue ignored in the widely used monetary model. The sample covers the period January 2000 – December 2014, so the data are not plagued by high-inflation differentials characteristic for the early transition period and include countries with relatively flexible exchange rates. Our empirical strategy employs the panel error correction model that allows for cross-sectional dependence and a series of panel causality tests. The main finding is that the nominal exchange rates in CEE countries are not disconnected from macrofundamentals implied by the Taylor rule-based model. More specifically, we find that there is a strong cross-sectional dependence among CEE countries, exchange rates Granger-cause macrofundamentals and tend to revert to the long-run relation, and that the results are robust to the ‘extraordinary circumstances’ argument, i.e. do not rest on the dynamics during the global financial crisis.

KEYWORDS:

• Floating exchange rate
• interest rate rule
• Central and Eastern European countries
• cross-sectional dependence
• panel cointegration

JEL CLASSIFICATION:

• F36
• F41
• E44
• C33

Acknowledgements

The authors thank participants of EBES conference in Venice, participations of Macromodels 2015 conference in Trzebieszowice and participants of Macroeconometric Workshop in Berlin for their helpful comments on the previous versions of this paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ A similar point can be made about the PPP theory: Drine and Rault (2008) demonstrate that it is more easily accepted in countries with high inflation.

² For a similar argument see Angeloni, Flad, and Mongelli (2007), who claim that the reaction to inflation is partially reflected in the coefficient on the foreign interest rate (which is included in their equation).

³ We do not belittle this line of research. It is simply oriented at a different issue, i.e. the estimation of an equilibrium exchange rate (EER). See the seminal paper on EERs in CEE countries by Halpern and Wyplosz (1997). See also Égert, Halpern, and MacDonald (2006), Maeso-Fernandez, Osbat, and Schnatz (2006), Candelon et al. (2007), and Dąbrowski (2009).

⁴ This is in line with the fact that in the early transition period central banks in many CEE countries used the exchange rate as a nominal anchor to build their reputation (Sachs 1996; Markiewicz 2006).

⁵ The term in the real exchange rate can be written as $s_t-{\bar{s}}_t$, where ${\bar{s}}_t$ is the target nominal exchange rate. The target is set to $\left(\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{t}-p_t\right)$, i.e. the level implied by the PPP theory. Thus, omitting a constant for simplicity, one gets $q_t=s_t-{\bar{s}}_t$ (a constant is included in the empirical part).

⁶ The omission of u_t does not change the conclusion about the cointegration that we derive below.

⁷ The SUR method was also used by Beckmann, Belke, and Dobnik (2012) to estimate a dynamic panel ECM for the nominal exchange rate and its monetary fundamentals for 18 OECD countries.

⁸ We follow the IMF and include Turkey in the group of the CEE economies (see, e.g. IMF 2004), all the more it has had the floating exchange rate since 2001 (see below) and is a candidate country for the European Union (EC 2017).

⁹ A potential non-stationarity of the common factors is taken into account during estimation of cointegration relation.

¹⁰ The importance of cross-sectional correlations between exchange rates has recently been modelled by Greenaway-McGrevy et al. (2017). See also De Truchis, Dell’Eva, and Keddad (2017), who demonstrated that if two economies are economically linked, similar Taylor rules lead to long-run co-movement between exchange rates. In a related strand of literature Beckmann and Czudaj (2017) used common factor approach to examine exchange rate and current account dynamics across countries.

¹¹ The interpretation of the results of panel unit root tests is a difficult task. The null hypothesis of the CIPS test is that there are unit roots for all cross-sectional units and the alternative is that the series is stationary for at least one cross-sectional unit (Pesaran 2007).

¹² See Pedroni (2007) for another example: the within-sample behaviour of the saving rate per capita was interpreted as a ‘local’, not global property.

¹³ See also Pesaran (2007), who examined 17 real exchange rates of advanced OECD economies using the CIPS test and found that the null of a unit root test cannot be rejected at the 5% level.

¹⁴ In principle, one can claim that the nominal exchange rate and the relative price level are both stationary. It, however, is hardly appealing in the light of the literature referenced to above and our panel unit root test results depicted in Table 3.

¹⁵ We have experimented with the co-integrating equation that includes the nominal exchange rate and the relative price level only. The coefficient for the price level is −0.906 when CCE MG estimator is used and −1.057 when the AMG estimator is used. In both cases it is not statistically different from −1 at the 1% level.

¹⁶ The empirical evidence for CEE countries, however, is far from being unambiguous: mixed results have been reported by Acaravci and Ozturk (2010), Grabowski and Welfe (2016), Koukouritakis (2009), Sadoveanu and Ghiba (2012), Sideris (2006), Žďárek (2010), and evidence against the PPP can be found in Bekö and Boršič (2007), Drine and Rault (2008).

¹⁷ See also Takagi and Ciubotaru (2013).

¹⁸ Mackiewicz-Łyziak (2016) identified empirically two monetary policy regimes, i.e. ‘passive’ and ‘active’. In the former there is ‘a strong smoothing of the interest rate path and little response to inflation and output gap developments’ and in the latter ‘the smoothing parameter decreases and the reaction to inflation and/or output gap is stronger’. Moreover, she investigated monetary policy rules adopted by central banks in the Czech Republic, Hungary, and Poland.

¹⁹ The example provided by Frankel and Rose (1995) was hyperinflation.

²⁰ The correlation coefficient for levels is 0.53 with p-value less than 1%. It should, however, be interpreted with caution since both variables are likely to be non-stationary.

²¹ Instead of a median ratio, a mean ratio was also tried. The implied γ were even greater: 1.60 and 2.04, respectively.

Additional information

Funding

This work was supported by the Polish Ministry of Science and Higher Education and Cracow University of Economics [062/WE-KMA/01/2017/S/7062, 044/WZ-KS/03/2016/S/6044]. At its early stage this work was also supported by the National Science Centre in Poland [DEC-2013/11/B/HS4/00489].