On the predictive power of monetary exchange rate model: the case of the Malaysian ringgit/US dollar rate

Abstract

The predictive power of the monetary model for the Malaysian ringgit/US dollar (RM/USD) rate is analysed using quarterly data ending in 2006:Q3. We find compelling evidence of a long-run relationship between exchange rates and the economic fundamental determinant. Macroeconomic factors systematically affect the long-run movement of the RM/USD rate. Additionally, the RM/USD rate was overvalued by about 10% several quarters before the 1997 crisis; after the crisis, rates fluctuated close to the equilibrium value. The out-of-sample forecasts demonstrate that the monetary model outperforms the naïve random walk model. The monetary and Purchasing Power Parity (PPP) models do well at the four to eight quarters horizon.

Acknowledgements

Part of this article was written when Ahmad Zubaidi Baharumshah was visiting Washington State University under the Fulbright program. Financial support from ACU and SCIENCE FUND [Project No: 04-01-04-SF0414] is gratefully acknowledged. Sung K. Ahn's work was supported by the Korea Research Foundation Grant (KRF-2005-070-C00022) funded by the Korean Government (MOEHRD).

Notes

¹ MacDonald and Taylor (1994) were among the first to show robust evidence on the long-run relationship between exchange rate and monetary fundamental. In the similar way, Mark (1995) found that monetary factors affect the movement of exchange rates and outperform a random walk at horizons of three and more. The reasons for the poor performance of some of these models could be either due to incorrectly specified models, or to inappropriate empirical techniques. Apart from that, Purchasing Power Parity (PPP) and Uncovered Interest Rate Parity (UIP), the building blocks of the monetary model perform poorly. Various attempts have been made recently in the literature to improve the specification of the models (see for example Mark and Sul, 2001; Morley, 2007).

² We thank an anonymous referee for pointing out this point.

³ We note that Chinn (2000) relied on pure time-series methods with the dollar-based rates, while Husted and MacDonald (1999) employed a panel framework with the yen-based rates to model the Asian currencies. More recently, Chin et al. (2007) applied several versions of the monetary model to the RM/USD rate. They demonstrated that the sticky price model outperforms the in-sample forecasts of the flexible price monetary model. The authors also showed that a sizeable fraction of the change in exchange rates can be attributed to monetary factors.

⁴ During the mid and late 1990s, we observed a large departure from fundamentals. Some authors (e.g. Kilian and Taylor, 2003) ascribe the poor performance of out-of-sample forecast to a short data span and the rarity of large departures from fundamentals. Our data covers the currency crisis period where we may expect the ringgit to depart from its fundamental values.

⁵ Malaysia's experience in the past three decades or so reflects its preference for exchange rate stability. The US dollar plays a major role in the exchange rate arrangement. Recent estimates indicate that the weights given to the US dollar rise from 0.40 in 1973–1977 to more than 0.85 from 1989 onwards (Yap, 2002). See also Ibrahim (2007) for a detailed documentation of the exchange rate regime in Malaysia.

⁶ The sticky-price model is well known and so we have not included the discussion on the theory behind it for brevity. The model can be interpreted as an extension of the PPP model with the price variable replaced by macroeconomic variables that capture money demand and overshooting effects.

⁷ There is no consensus among economist on the statistical property of interest rate or interest rate differentials. Some authors (e.g. Fedderke and Liu, 2002) have argued that interest rate differential might not be I(0) if risk premium change overtime. In our case, we find strong evidence that the interest rate differential variable is stationary as such we treat it as an exogenous variable.

⁸ Malaysia absorbed inflows of amounting to 10% of GDP in 1991, 15% in 1992, and more than 20% in 1993, averaging 12% for the entire boom period of 1989–1996. Foreign Direct Investment (FDI) accounted for the overwhelming share of net flows to Malaysia. A high share of the Malaysia's FDI inflows was in traded-goods sector, mostly in exported manufacturing (Ito, 2000). The degree of real exchange rate appreciation during the capital inflow episode was estimated at 6.9. Additionally, test based on recursive residuals also suggests that the parameters of the estimated VAR equations are associated with outliers in the same period.

⁹ Some authors have pointed out that the use of information criteria may not be adequate to select the optimum lag length in the presence of moving average terms. To overcome this shortcoming, we also ensure that the VAR residual are serially uncorrelated. For this purpose, we relied on the Box–Pierce statistic and the Lagrange Multiplier (LM) test for serial correlation. The optimum lag is five in our case.

¹⁰ The degree of freedom correction suggested by Reinsel and Ahn (1992) multiplies the test statistics by (T − pk)/T, where T is the sample size, p the number of variables, and k is the lag order of the estimated VAR system. In the analysis that follows, we applied this adjustment factor to check for the significance and the robustness of our results.

¹¹ To test which variable(s) need to be included in the cointegrating space, Johansen suggested the following test statistic: , where r here is the number of cointegrating vectors, λ* is the eigenvalue of the vector from the restricted space, and λ is the eigenvalue from the unrestricted cointegrating space. The test is distributed as χ² with r(p − s) degree of freedom where p is the dimension of unrestricted cointegrated space and s is the dimension of the restricted space.

¹² Following earlier studies, we analyse the VAR model where the interest rate differential is omitted. The results (not reported) yield incorrect sign for the money and income coefficients. Moreover, we found that exchange rate is not weakly exogenous with respect to the cointegrating vector. This means that it is the monetary fundamentals (not the exchange rate) that adjust to the long-run equilibrium.

¹³ Microfit 4.1 was used to carry out most of the empirical analysis.

¹⁴ It is worth pointing out here that an incorrect but simple model may outperform a correct model in forecasting; see Clements and Hendry (2001) on this issue.

¹⁵ For a detailed discussion, see Goldfajn and Baig (1998).

¹⁶ This is the general pattern in the Asian countries (including China) where they maintained an overvalued currency in the 1980s and early 1990s. The pattern then changed to undervaluation in the late and early 2000s. For a recent discussion on misalignment in the Asian and the G20, see Bénassy-Quéré et al. (2008).

¹⁷ Based on panel estimates, Husted and MacDonald (1999) also showed that the ringgit–yen rates were overvalued from the early 1990s.

¹⁸ The ratio of RMSE and MAPE are computed as following:

where k = 1, 4, 8 and 12 denotes the forecasting horizons; N_k the total number of forecasts in the projection period for which is actual A(t) is known and F(t) is the predicted value from the models.

¹⁹ We note that Kilian and Taylor (2003) find that forecast models based on Exponential Smooth Transition Autoregressive (ESTAR) PPP fundamentals perform better than the random walk at horizons of 2 to 3 years. The authors found no such evidence for shorter horizons.

²⁰ In this context, our findings differ from that of Mark (1995) in that the strongest out-of-sample evidence of predictability is at the highest forecasting horizons (12 and 16 quarters) for the major currencies. Meanwhile, it appears to be consistent with the work of Kilian (1999) and Groen (1999) who found monetary fundamentals predict foreign exchange rates, but no evidence of more forecastibility at longer horizons.