Uncovered Interest-rate Parity and Risk Premium: Evidence from EUR/RSD Exchange Rate

ABSTRACT

This paper examines the uncovered interest-rate parity in a developing economy that implements inflation targeting. We study the exchange rate between the Euro and Serbian Dinar over different time horizons. We apply APARCH-in-mean to measure the impact and nature of a time-varying risk premium and capture the influence of higher-order moments on expected currency returns. We find a significant positive association between the returns and the interest-rate differential over shorter horizons when the risk premium is included. Asymmetries and fat tails are essential in explaining average returns over time horizons of up to one month.

KEYWORDS:

• Interest-rate differential
• time-varying premium
• exchange-rate fundamentals

JEL Classification:

• F31
• G15
• C58

Acknowledgments

The author acknowledges the financial support of the Ministry of Education, Science and Technological Development of the Republic of Serbia. I am thankful to the anonymous reviewers of this manuscript for their useful comments and suggestions.

Disclosure Statement

No potential conflict of interest was reported by the author.

Notes

1. Some notable exceptions are reported in a recent panel-data study by Lothian (2016).

2. Here, we focus on the exogenous aspects of the economic environment, setting aside the role of monetary policy (McCallum 1994) and related uncertainties (see, for example, Krol 2014; Della Corte, Riddiough, and Sarno 2016; Mueller, Tahbaz-Salehi, and Vedolin 2017; Berg and Mark 2018; Della Corte and Krecetovs 2019). Similarly, the issues related to market microstructure, such as buying and selling pressures related to carry trading and momentum strategies (Menkoff et al. 2012; Burnside, Cerrato, and Zhang 2020), foreign-exchange liquidity risk (Abankwa 2020), or bond liquidity premia (Lee and Jung 2020) are outside of the scope of this paper.

3. Engel and West (2005) show that exchange rates are unrelated to most alternative macroeconomic or financial fundamentals, including money supply, inflation, or real income. Rossi (2013) provides compelling arguments that it is not easy to reconcile data with most structural models. On the other hand, Molodtsova and Papell (2009) document that the inflation rate and the output gap have a strong out-of-sample forecasting power for international exchange rates against the U.S. dollar. These so-called Taylor-rule fundamentals seem to outperform conventional models. However, in a recent paper, Engel et al. (2019) find strong evidence of an in-sample anomaly that seems to drive the result: when the U.S. inflation is included in the UIP regression, it becomes significant, whereas the interest-rate differential does not. In any case, the economically intuitive fundamentals do not appear to be a robust alternative to a random walk, particularly in the short run (Cheung et al. 2019).

4. See, for example, Stoupos and Kiohos (2017) for a review of related work on exchange rates in transition countries.

5. Depending on the assumptions regarding the residual ${\epsilon }_{t+1}$, the parameter $\alpha$ can also contain Itô term.

6. At the same time, the correlation between exchange rates and consumption growth across countries is small and negative. This apparent contradiction with traditional international business cycle models is the so-called consumption-real exchange rate anomaly (see Chari, Kehoe, and McGrattan 2002).

7. An alternative would be to construct the interest-rate differential from zero-coupon rates of Treasury securities. Theoretically, they represent a ”purer” risk-free proxy, accessible to a broader set of non-financial investors. However, secondary trading of RSD-denominated Serbian Treasury instruments is very thin and illiquid, and data is not available for the entire sample period.

8. Note that the statistical moments shown in Table 1 are unconditional. Bollerslev (1986) and Bera and Higgins (1993) show that a GARCH-type model with conditional normality always implies heavier tails in returns than the corresponding unconditionally normal distribution. Nelson (1990) demonstrates that the continuous-time limit of a GARCH(1,1) process with conditionally normal residuals is equivalent to a continuous process whose distribution is an unconditional Student’s $t$. Nelson’s result indicates that the heavy-tailed distributions are more prevalent at higher frequencies of data. Here, we observe fat tails for daily data. Ding, Granger, and Engle (1993) show that an extremely high unconditional kurtosis of 25.42 in daily S&P 500 data can be successfully captured using the APARCH model with conditionally normal residuals, the assumption also used in this paper.

9. Note that the near-symmetrical shape of the exchange rate distribution by deciles of the interest-rate differential in Figure 3 does not contradict our interpretation. The asymmetry originates from the residuals ${\epsilon }_{t+1}$, which are unexpected and can be explained neither by the interest-rate differential $r_t-r_t^f$, nor the constant and the time-variable components of the risk premium, $\alpha$ and $\gamma {\sigma }_{t+1}$, respectively.

Additional information

Funding

This work was supported by the Ministarstvo Prosvete, Nauke i Tehnološkog Razvoja [OH 179005].