Currency substitution, policy rule and pass-through: evidence from Turkey

Abstract

In this article we analyse empirically currency substitution and exchange rate pass-through in the Turkish Economy, where their ongoing presence could undermine the implementation of a successful monetary policy, especially in a flexible exchange rate regime. Even though a considerable time has passed after the implementation of a flexible exchange rate regime in Turkey, by using Vector Error Correction model for the period from 1987 to 2004, we find that the currency substitution and exchange rate pass-through still have importance in the Turkish Economy and the monetary policy stance has been considerably strong, possibly, as a response of ongoing presence of them. If this is the case, to avoid the undesired consequences of this strong monetary policy, Turkey should consider some policy measures to reduce the degree of pass-through and currency substitution.

Notes

¹ Turkey adopted a flexible exchange rate regime from February 2001 and has started to declare its inflation targets since January 2002.

² In fact such an outcome can be expected if inflation indeed decreases after the adoption inflation targeting. Taylor (2000a, b) states that a low inflationary environment leads to a low ERPT to domestic prices. Devereux and Yetman (2002) and Choudhri and Hakura (2006), by using large datasets consisting of several countries, show that estimated ERPT tends to vary systematically with the mean inflation rate. These results constitute some reasons to expect a reduction in the pass-through after the drop in inflation rate.

³ Following the definition of Calvo and Vegh (1992) we use the term ‘currency substitution' as the holding of foreign currency rather than domestic currency for some of the monetary services. Currency substitution is just one instance of a more general phenomenon called dollarization, that is, widespread holdings of financial assets denominated in foreign exchange (Calvo, 1998, p. 150).

⁴ Reinhart et al. (2003) challenge the notion that dollarization limits the scope for an independent monetary policy.

⁵ Moreover, this ineffectiveness in monetary policy causes authorities to become over sensitive to exchange rate volatility leading them to accumulate excess international reserves, intervene frequently in exchange rate markets and adjust interest rates often (Hausmann et al., 2001). These events exacerbate sudden stops, cause output volatility and ultimately result in costly self-fulfilling macroeconomic crises (Eichengreen and Hausmann, 2005).

⁶ They do not report specific numbers, their assessment are carried out on basis of the qualitative comparison of impulse response functions.

⁷ Magnitudes for currency substitution are reported for these countries.

⁸ This terminology belongs to Svensson (2003).

⁹ Main references of this literature can be found in Tekin and Yazgan (2007).

¹⁰ However, the degree of pass-through can be as low as 13% when it relates to the euro not the US dollar.

¹¹ As is well known, for the high inflation periods, aggregate money demand is more appropriately described by including inflation as an opportunity cost of money. Since, our estimation period is dominated by high and persistent inflation, we think inflation should be an important variable in explaining the money demand in Turkey.

¹² Note also, based on the theory, the expected depreciation rate should be used in money demand equation. However, since there is no such data available, one solution is to use ex-post rates to proxy the expected ones, as is done in this article and others. The other solution is using different proxies for the expected rates. Chaisrisawatsuk et al. (2004) uses the ratio of the difference between forward and spot exchange rate to spot rate as a proxy for expected depreciation of the exchange rate. Akcay et al. (1997) use expected depreciation rate derived from an EGARCH-M model.

¹³ Exchange rate is defined as the Turkish Lira per US dollars.

¹⁴ Unit-root tests, based on univariate regressions, such as Augmented Dickey Fuller (ADF) and Dickey-Fuller Generalized Least Square (DF-GLS) are also calculated up to 5th order. Although, the evidence provided by them predominantly support the conclusion that output and interest rate spread are I(1); money, exchange rate, and prices are I(2), there is some ambiguity around exchange rate variable. However, we think the multivariate analysis given below is more appropriate to determine the time-series properties of these variables. These and subsequent not-reported results can be found in working paper version of this article.

¹⁵ The details of these dummies are explained below.

¹⁶ FIML procedure is found to have better size properties than the two-step estimation procedure in Johansen (1995), see Nielsen and Rahbek (2003).

¹⁷ This and all subsequent calculations are carried out in Version 2.D of CATS for RATS.

¹⁸ Here, note that, the I(2) trend with r = 2 is borderline accepted. Given the fact that the dummy variables are included into the model, and since the dummies could affect the distribution of the test statistics, it can be argued that tests should be carried out at higher significance levels. However, even if we follow this logic the result does not change in terms of the I(2) trend. Only, the rank is determined as 3 instead of 2 with a p-value of 0.6329.

¹⁹ Note also that the condition for the long-run price homogeneity is , i.e. the nominal shocks u _1t affect nominal money and prices in the same way.

²⁰ Again, as in the previous footnote, we do not want to rely on this ‘low’ p-value, the result does not change in terms of the I(2) trend, but we need to accept cointegration rank as 4 with a p-value of 0.5494. This is not important in terms of our analysis, since by using the I(1) variables in Equation 6, we are able to determine the rank as 3 in the following section.

²¹ The lag order of this VEC model is determined as one according to results of a sequence of lag deletion LR tests and different information criteria performed by using the underlying vector autoregressive representation of the VEC model.

²² For a similar usage of shift dummies in cointegration space, see Lütkepohl et al. (1998). They model the demand for M3 in the Unified Germany by taking into account the structural break at 1990 because of German unification. Felmingham et al. (2001) use also a shift dummy to capture the 1991 deep recession in Australia.

²³ As suggested by extensive simulation studies in Lütkepohl et al. (2001), trace test appears to be superior in terms of power than maximum eigenvalue test, hence, we do not consider the maximum results of eigenvalue test.

²⁴ To check the constancy of these three cointegration relations, the recursively estimated trace tests are also calculated. These tests clearly indicate that the three cointegration relations are constant.

²⁵ To explain the differences among those, one can argue that the extent of the estimated coefficient of the depreciation rate is decreased, when the period is extended and/or the narrow definition of money is used instead of a broad one.

²⁶ It should also be noted that our results are not consistent with those of Kara et al. (2007) suggesting that there is a considerable drop in the ERPT during the float. As can be seen in the following section our constancy tests refute the presence of a break in the coefficients of ERBT equation. In this respect, our findings are more consistent with those of Selcuk (2005). He concludes that traditional measures of currency substitution indicate that there was not much of a reverse currency substitution during the float.

²⁷ See Juselius (2007) on these tests.