Currency Order Flow and Exchange Rate Determination: Empirical Evidence from the Malaysian Foreign Exchange Market

Abstract

This article presents empirical test results of Malaysian foreign exchange market microstructure assessment of exchange rate dynamics. We apply vector autoregressive (VAR) model to estimate the influential role of currency order flow in the determination of the currency exchange rate for the Malaysian ringgit (MYR) against the US dollar (USD). We investigate whether currency order flow captures the movements of exchange rate of MYR against USD, and how the long-term and short-term components impact the relative estimation of MYR in the international market. We, construct a measure of order flow in the Malaysian foreign exchange market to reflect the pressure of currency excess demand. Our focus is on the cumulative currency order flow and the exchange rate relationship of MYR and USD. A hybrid model of order flow and exchange rate dynamics proposed by Evans and Lyons (2002a) is applied to the Malaysian foreign exchange market (MYR/USD) to analyse a dataset of every 15-minute currency order flow and exchange rate movements from January 2010 to December 2015. Our dataset has unique features in terms of the quality of the data, extensive period and precise high frequency. Our results show that currency order flow explains an important portion of the movements in the MYR–USD exchange rate.

Keywords

Exchange rate currency order flow market microstructure foreign exchange market

Introduction

The existing macroeconomic models of currency exchange rate determination are deficient at frequencies over and above a period of a year. Undoubtedly, these models’ explanatory power is almost nil (Meese & Rogoff, 1983). Without missing words, the statement made by Frankel and Rose (1995) is that undesirable outcome has had a ‘pessimistic effect on the field of empirical exchange rate modelling in particular and international finance in general’. Many researchers have unearthed one of the variables that appear correlated with currency exchange rate changes at high frequencies, namely, order flow (Evans & Lyons, 2002a). These currency order flows are the driving force behind the turnover in the foreign exchange market (Evans & Lyons, 2002a).

O’Hara (1995, p. 1) defines the market microstructure approach as one that studies ‘the process and outcomes of exchanging assets under explicit trading rules’. In turn, the trading process is centred on order flow; hence, order flow becomes the focal point of the microstructure approach (Frankel & Rose, 1995). Although, existing research has been primarily concerned with matured economies and the world’s major currency pairs, very few studies have investigated the important role that currency order flow plays in the foreign currency markets in emerging market.

Indeed, among the high-performing economies in Association of Southeast Asian Nations (ASEAN) is Malaysia. Given this country diverse economic relationship with the USA, the economy of this nation ought to achieve a reasonable degree of exchange rate stability. However, it is unfortunate for this country to experience a continuous reduction in the foreign exchange reserves, which also led to currency depreciation in the international market, especially against USD. Meanwhile, the successful transition of this emerging economy to full development is important both to the world economy and as a model for other emerging economies. With this country’s rising importance in the world economy and the growing complexity of the economic and financial globalization, it is desirable yet challenging to achieve a superior appreciative of how the value of Malaysian ringgit (MYR) against the US dollar (USD) is determined in the international currency market both at the long run and short run.

A hybrid model of order flow and exchange rate dynamics proposed by Evans and Lyons (2002a) to test DM/USD and JPY/USD in the foreign exchange market is applied to the Malaysian foreign exchange market (MYR/USD) to analyse a new dataset of every 15-minute order flow and exchange rate movements on the Reuters and Bloomberg for the MYR–USD currency pair from January 2010 through December 2015. Mainly, our dataset has unique features in terms of quality of the data, its extensive period and its precise high frequency. We construct a measure of cumulative order flow in the Malaysian foreign exchange market context, which centred on 15-minute currency transaction data to reflect excess demand pressure in the foreign exchange market. With the application of vector autoregression (VAR) model, we examine the cointegrating relation between cumulative currency order flow and movements in currency exchange rate in the Malaysian currency exchange market. Fundamentally, we crave to resolve the accompanying inquiries:

Does currency order flow capture the movements of MYR exchange rate against the USD in the international currency market?

Does the long- and short-term components impact on the estimation of the MYR in the international currency market?

Our findings show that exchange rate Granger causes order flow and vice versa. In essence, there exists bidirectional causality. In testing the strength of the relationship at longer horizons, we consider 10 trading days as 2 weeks, 20 trading days as 4 weeks and 30 trading days as 6 weeks. Therefore, we test with Cholesky decomposition for a time horizon of 30 trading days, and we find that there is a strong relationship even at 6 weeks. Consequently, cumulative order flow has a positive and strong relationship with exchange rates even at longer the horizon.

The results from our sample indicate that, from every $1million USD/MYR purchase, currency order flow can explain up to 24 per cent of the currency exchange rate movements. Thus, important significant movements of fluctuations in the MYR/USD exchange rate can be explained by currency order flow in the Malaysian foreign exchange market. This study is motivated by the desire to throw more light on the currency exchange rate determination and dynamics in Malaysia with a market microstructure perspective. Essentially, we construct a measurement of currency order flow of 15-minute frequency in tick-by-tick trading data from the Malaysian foreign exchange currency market where our daily cumulative currency order flow is computed.

In effect, our research contributes to the market microstructure of the exchange rate theory by shedding more light on the appreciativeness of the international currency exchange rate in the emerging markets economy, concentrating on the currency order flow and determination of exchange rate in the international currency market. This research will help scholars to have profound grasp of currency order flow as a major microeconomic factor to be considered in the foreign exchange market, especially in the emerging markets economy. Also, the practitioners and policymakers will have a deeper understanding of the explanatory power of currency order flow on how it drives the exchange rate movements in the foreign exchange market.

This research article is structured as follows: the second section reviews literature in brief on exchange rate dynamics with reference to market microstructure. Third section discusses the data and methodology. Fourth section presents the empirical results and fifth section provides the conclusion.

Review of Literature in Brief

Recent exchange rate study, centred on microstructure thoughts, points out the part that currency trading performs in price formation through a concept termed order flow. Order flow is well defined to be the difference between buyer-initiated and seller-initiated currency trading importance in a given market and thus correspond largely to what practitioners might refer to as aggressive buying or selling pressure (Evans & Lyons, 2007). Evans and Lyons (2007), in their research work, made two basic classifications on the role currency order flow plays in the exchange rate models from the market microstructure perspective. More so, Evans (2011), in his research paper, considered order flow to be an important mechanism of information transmission that links price movements with information dispersion. Empirically, currency order flow can be considered to be a measure of the sum signed of the seller and that of the buyer initiated currency order flows.

In the models of Lyons (1995), Perraudin and Vitale (1996) and Evans and Lyons (2002a) currency order flow explain concomitant exchange rate fluctuations as it encompasses information, either about essential rudiments or long-run premia, which was earlier disseminated among market participants. More so, researchers’ interest to determine a consistent set of essential economic variables that do influence the behaviour of exchange rate as a result of its increasing wave of volatility and complexity is on the high (Effiong, 2014).

Thus, one of the fundamental variances between the microstructure-level analysis and traditional exchange rate framework is that similar information is not disclosed by all market participants in and/or is understood in a different way by participants. Influential research by Evans and Lyons (2002a) with the application of interdealer order flow for 4-month transaction data on the currency exchange rate from Reuter’s database analysed the daily changes of deutschmark and Japanese yen (JPY) with that of USD. Their results show that order flow can explain over 60 per cent of daily changes in the USD against deutschmark. Evans and Lyons (2002b), in another study, focused on British pound sterling (GBP), Belgian franc, French franc, Swiss franc (CHF), Dutch guilder, Italian lira and Swedish Krona, all these currencies are against USD. Their results show that an R² of 78 per cent can be generated daily through order flow. Berger, Chaboud, Chernenko, Howorka and Wright (2008) find that interdealer order flow has 0.65 correlations with the currency exchange rate of the EUR/USD.

The research work of Osler (2006) gives a summary of how currency order flow drives the movements of currency exchange rate with a basic explanation of inventory, information and liquidity effects. To have inventory effect, there must be a departure from the anticipated position which will eventually uncover the dealers to an unwarranted risk. Therefore, in order to guide against unwarranted risk, dealers in this market will try to increase or decrease their price, thereby attracting more buying or selling orders when there is a deviation from their inventory positions, which is different from their desired levels. In effect, temporary exchange rate fluctuations can be explained with the use of inventory models but not permanent exchange rate fluctuations. On the information models, market prices should be permanently affected by order flow. Therefore, cumulative currency order flow should be cointegrated with the exchange rate (Zhang, Chau & Zhang, 2013).

A simple linear VAR model was proposed in 1991 by Hasbrouck based on microstructure to examine New York stock exchange. This modelling strategy was applied by Payne (2003) to investigate for a period of 1-week, USD against Dutch mark between 6 October and 10 October 1997. The results show that there is an explanatory power on currency exchange returns with informed currency order flow up to 60 per cent fluctuations. Similarly, the VAR model was proposed by Froot and Ramadorai (2005) to examine the currency order flow as a major factor of exchange rate fluctuations in the foreign exchange market. The researchers study the interaction between permanent shock and transitory shock by dividing the exchange rate earnings. Their findings show that long-term and values can be easily and better explained with the use of macroeconomic fundamentals but for short-term currency returns, a micro-economic variable, that is, order flow, is appropriate. The results of these findings actually show that order flow as a microeconomic factor is of importance to research on, especially, in the foreign exchange market, by examining its role in the determination of currency exchange rate both in short-term and long-term dynamics.

It is noteworthy to say that most of the researchers in this field of study has focused on the major currency pairs. For example, Rime (2000) research on the following currencies: deutschmark, GBP, Canadian dollar (CND), CHF and JPY, all against USD, covering the period of 1995/7–1999/9. His results of cointegration test show that there is a cointegrating relationship between currency exchange rate and currency order flow for the following currency pairs: deutschmark/USD, GBP/USD and CHF/USD. This shows that there is an explanatory power of currency exchange rate fluctuations when order flow is lagged. Other empirical studies that investigate the explanatory power of order flow include Evans and Lyons (2005), Andersen et al. (2003), Berger et al. (2008), Martin (2001), Bjønnes and Rime (2005), Boyer and Norden (2006), Danielsson and Payne (2012), Evans (2002, 2010), Evans and Lyons (2006, 2008) and Rime et al. (2010).

Attention of researchers focusing on currency order flow and its explanatory power in the emerging markets is of importance. In the study of order flow and exchange rate in the emerging markets, De-Medeiros (2004) investigates the Brazilian foreign exchange market and includes a variable from the international finance field, country’s risk premium. His findings indicate that among the tested variables, the country’s risk premium, a variable from the international finance field appears more significant but order flow does not have any significant performance. That is, the performance of order flow is weak while comparing this result with developed economies. In the same manner, Wu (2010) did add to his model’s risk premium to investigate the interactions between the commercial customer order flow and financial customer order flow and divide these customers’ 4 years daily trading data into temporary period and permanent period, respectively. He finds that there exists a positive relationship between financial customer order flow and intervention flows in relation to exchange rate fluctuations. However, there exists a negative relationship between the commercial customer order flow and currency exchange rate. Duffuor, Marsh and Phylaktis (2012) investigate the foreign exchange market of the Ghana economy to explain currency order flow effect on the exchange rate fluctuations from the end user currency point of view. The researchers examine currency order flow by focusing on the unofficial exchange rate market (black market) and that of the official exchange rate market, respectively. Their results show that there is, in the expected currency order flow, a weak performance. Zhang et al. (2013) investigate foreign exchange market of China economy focusing on Chinese renminbi and the USD, to determine the extent to which order flow can influence exchange rate movements in the long-term and short-term between this currency pair. Their findings confirm that currency order flow as a microeconomic variable explains significantly major movements in the exchange rates between this currency pair.

The results of these research studies and their findings motivated us to investigate further the intensity of currency order flow to explain the fluctuations in the exchange rate of an evolving market currency against other major world currencies, especially, MYR against USD. On this note, it is imperative to have a vivid understanding of the Malaysian foreign exchange market and its trading mechanism on the strength of market microstructure in focus.

Theoretical Framework of the Study

In the past few decades, a set of macroeconomic variables, such as inflation rates, interest rates, countries current account balances, money supplies, gross domestic products and government budget deficits, are used to determine exchange rates from the view point of fundamental analysis.

Empirically, these macroeconomic variables, applying fundamental analysis approach, account for no appreciable movements of exchange rates as predicted. The classical research work of Meese and Rogoff (1983) reveal that macroeconomic fundamental models have failed to justify reason(s) for exchange rate movements. Furthermore, Meese (1990) encapsulates that ‘The proportion of (monthly or quarterly) exchange rate changes that current models can explain is essentially zero’. In view of these, it is not an exception to say that the macroeconomic fundamental models are weak in this direction.

A significant number of surveys on the various types of models and theories on the determination of exchange rates have been carried out by various authors (Backus, 1984; Eaton & Turnovsky, 1983; Evans & Lyons, 2002a; Fisher & Hillman, 2003; Frankel, 1993; Frankel & Rose, 1994; Galati, 2000; Isard, 1995; Ito et al., 1998; Lyons 2001a, 2001b; Macdonald & Taylor, 1992; Marsh & O’Rourke, 2005; O’Hara, 1995; Taylor, 1995) and we consider it unnecessary discussing these theories and models in detail in this article. Nevertheless, the most popular models of exchange rate determination include the flexible price model, the sticky price model and the portfolio balance model. The flexible price model is of the opinion that purchasing power parity exists and that domestic and foreign currencies demand are stable in both economies. By definition, the variability of real exchange rate is impossible.

The sticky price model provides variability of interest rate and exchange rate with the acceptance of deviations in the real exchange rate and nominal exchange rate, respectively, while the portfolio balance model of exchange rate determination states that there must be relative supplies of domestic bonds and foreign bonds for an effective exchange rate. In essence, the portfolio balance models assume imperfect substitutability of domestic bonds and foreign bonds. In a nutshell, portfolio balance models have risk premiums in the forward exchange rate, which is a function of relative asset supplies. With all these empirical difficulties of the traditional models of exchange rate determination, Evans and Lyons (2002a) propose a model ‘portfolio shifts model’ of exchange rate determination in an attempt to resolve the difficulties.

This model can be stated as follows:

Δ P_{t} = Δ m_{t} + λ Δ X_{t}

where ΔP_t represents changes in spot exchange rate; Δm_t represents macroeconomic information innovations (e.g., changes in interest rate differential); λ represents positive constant; ΔX_t is daily accumulated signed order flows.

Objectives of the Study

Existing research on microstructure approach to foreign exchange market has been primarily concerned with matured economies and the world’s major currency pairs, but very few studies have investigated the important role currency order flow plays in the foreign currency markets in emerging market. Meanwhile, the successful transition of emerging economies to full development is important to the world economy. With the emerging markets rising importance in the world economy, and the growing complexity of the economic and financial globalization, it is desirable yet challenging to achieve a superior appreciative of how the value of MYR against the USD is determined in the international currency market both at the long run and short run. Specifically, this study is carried out to:

Examine the extent to which currency order flow analysis can explain the short-term determination of the exchange rate value of the MYR against the USD.

Determine the interaction among exchange rate, currency order flow, international interest differential and country risk premium on the long-run and short-run dynamics of the Malaysian exchange rate.

Market Microstructure of Malaysian Foreign Exchange Market

Foreign exchange is the realm where a country’s currency is exchanged for that of another in the currency market. The leading financial market in the world is the foreign currency exchange market with daily trading volume of an equivalent of over 4 trillion USD, a transaction which can be described to be three times over and above the total aggregate amount of transactions on combination of US equity and Treasury market. This market operates through a global network of authorized banks and licensed corporations trading one currency for another. The foreign exchange market operates on a 24 hour bases, cut across all the major financial centres (i.e., the major trading sessions: New York, London, Tokyo, Sydney and Frankfurt).

The Bank Negara Malaysia (Central Bank of Malaysia) administered foreign exchange controls on behalf of the Malaysian Government with specific authorities delegated to the authorized banks. The Malaysian Government placed the effective rate for her currency on a controlled and fluctuating basis in June 1973. However, the Bank intervenes as the need arises to maintain and sustain orderly foreign exchange market conditions and to circumvent too many variations in the value of the ringgit in relations with Malaysia trading partners and other international currencies of settlements (Ariff, 1991). Meanwhile, ringgit pegged to the USD in 1997 was replaced with a managed float system in July 2005. The primary motivation for the policy shift according to the Central Bank of Malaysia is to better position Malaysia to respond and benefit from the structural changes happening in the region and in the international environment (Bank Negara Malaysia, 2016).

Malaysian foreign exchange trading starts at Monday morning and ends at Saturday morning, Malaysian time. Only authorized dealers are permitted to trade in the foreign exchange market in Malaysia pursuant to Section two of the Exchange Control Act of 1953. Spot trading are conducted on 27 currencies including the world major currencies: USD, European Euro (EUR), GBP, JPY, Australian dollar (AUD), CHF and CND. Spot trading in these currencies starts from 0900 to 1700 with four trading sessions (i.e., 0900, 1130, 1200 and 1700). The trading periods are in Malaysian time and usually open for business on Monday morning and closes on Saturday morning, excluding public holidays, and the settlement period for foreign exchange transaction is set at T+2 (i.e., two days after the transaction day).

Noticeably, the introduction of the large value payment system (LVPS) into the foreign exchange market by the Malaysian Government actually made the transaction of high value and real time easy to process. In addition, real-time electronic transfer of funds and securities (RENTAS) is the only LVPS for high value and time critical payments acceptable in the country and this operates under real-time gross settlements (RTGS). The main objective is to improve the overall efficiency of the LVPS. The RENTAS participants stand at 69 among which are Commercial Banks, Islamic Banks, Investment Banks and Development financial institutions classified as active players in the money market. In 2006, Central Bank of Malaysia collaborates with Hong Kong Monetary Authority to implement Payment versus Payment (PvP) infrastructure for the purpose of settling inter-bank ringgit–USD trade transactions during Malaysia business hours. The purpose is to eliminate foreign exchange settlement risk for ringgit and USD foreign exchange transactions. The Bank in March 2012 included renmimbi (RMB) settlements to improve and to enhance the capability of RENTAS in cross-border payments and settlements Figure 1 presents MYR/USD exchange rate as well as the correlation between the USD/MYR and currency order flow.

Figure 1.

The Daily Exchange Rate of USD/MYR MYR and Order Flow (04/01/2010 – 31/12/2015)

Data and Methodology

Data Sources

The data for this research study are sourced from Reuters and Bloomberg. These databases provide tick-by-tick trading prices data for spot transactions in foreign exchange rates. Our focus is on the Malaysian spot foreign exchange currency market and their trade transactions. We analyse a new dataset of every 15-minute currency order flow and exchange rate fluctuations on the Reuters and Bloomberg platform for the MYR against the USD currency pair from 4 January 2010 through 31 December 2015. Our dataset has a unique feature in terms of quality of the data, its extensive period and its precise high frequency, a total of 1,497 trading days excluding weekends and public holidays. We exclude Saturdays and Sundays from our sample in the sense that foreign exchange trading activity during these periods is minimal. In addition, we exclude general public holidays and Malaysian public holidays, for these days are of unusually light trading volume: New Year 1 January, Christmas day 25 December, Labour Day, Chinese New Year, National day, Malaysian day, Hari raya, Depavali, Awal Muharam and Maulidur rasul.

The opening time for spot foreign exchange trading in Malaysia starts from 0900 to 1700 (i.e., Malaysian Time GMT+8) with four trading sessions 0900, 1130, 1200 and 1700, respectively. The trading periods are in Malaysian time and usually open for business on Monday morning and closes on Saturday morning, excluding public holidays. The settlement period for foreign exchange transaction is set at T+2 (i.e., 2 days after the transaction day).

Measurement of Variables

Measurements of our variables are in this order: P_t represents the log of each working day closing exchange rate transaction price; X_t is daily accumulated order flow; (i_t – i_tf) represents the difference in interest rate for short-term period; (l_t – l_tf) represents the difference in interest rate for long-term period and (R_t – R_tf) represents the difference in the country’s risk premium. Evans and Lyons (2002a) indicate that the daily currency order flows X_t represent the net position between the buyer- and the seller-initiated currency order flows for the day-trading transactions. The difference in the interest rate for short-term period (i_t – i_tf) represents Malaysian interest rate daily overnight period minus the US interest rate daily overnight period. The difference in the interest rate for long-term period (l_t – l_tf) represents Malaysian inter-bank daily lending rate for 1 year minus the US inter-bank daily lending rate for 1 year. Country’s daily risk premium R_t represents the difference between the prime lending rate and 3 months Treasury bill rate. Therefore, the difference between the two countries the risk premium is given as (R_t – R_tf), the Malaysian risk premium minus that of the US risk premium. The interest rate data are expressed on an annual basis.

Trade direction and the sum of transaction volume are the two major important things from the definition of order flow. Thus, our major task is to determine the trade direction and sum up the tick trading direction of our 15-minute intraday data. Measure of spot currency order flow is constructed by assigning values to trade, that is, we assigned a value to every single buying and selling trade +1 and –1, respectively. Therefore, the summation of these trade signs is equal to 1-day spot order flow over the entire trading period.

Table 1 presents the summary of descriptive statistics then correlation matrix of the major items; P_t transaction price, X_t daily accumulated order flow, (i_t – i_tf) difference in interest rate for short-term period, (l_t – l_tf) difference in interest rate for long-term period and (R_t – R_tf) difference in the country’s risk premium. Our findings indicate that all the variables fail the Jarque-Bera test, meaning that all the variables depart from normality. The skewness for all the variables is less than two. Meanwhile, the correlation matrix results show that short-term interest and long-term interest have a negative relationship with the exchange rate. However, there is a positive relation between exchange rate, order flow and country’s risk difference. Therefore, the extent to which interaction exists among these variables needs further investigation.

Table 1.

Summary of Descriptive Statistics, then the Correlation Matrix

STRATUM A: SUMMARY STATISTICS	Exchange rate. (P_t)	Order flow (X_t)	Short-term IRD. (i_t – i_tf)	Long-term IRD. (l_t – l_tf)	Country’s risks premium. DIFF. (R_t – R_tf)
Observations	1497	1497	1497	1497	1497
Mean	0.307555	408.0541	2.784139	2.784139	0.442438
Median	0.314564	84.00000	2.849700	2.842358	0.435000
Maximum	0.340483	49123.00	3.164800	3.116288	0.790000
Minimum	0.224517	–21955.00	1.588750	1.780500	0.160000
Std. Dev.	0.025377	4507.573	0.317728	0.309980	0.091526
Skewness	–1.587282	0.752007	–1.484479	–1.492246	1.153289
Kurtosis	5.003744	16.53709	4.842390	4.743018	5.970604
JB Normality test	879.0415 (0.000000)***	11571.50 (0.000000)***	761.5436 (0.000000)***	745.0884 (0.000000)***	882.2813 (0.000000)***
STRATUM B: CORRELATION MATRIX
(P_t)	1.000000	0.121196	–0.339260	–0.229219	0.109375
(X_t)	0.121196	1.000000	–0.059674	–0.083287	0.038865
(i_t – i_tf)	–0.339260	–0.059674	1.000000	0.699348	–0.066652
(l_t – l_tf)	–0.229219	–0.083287	0.699348	1.000000	–0.075450
(R_t – R_tf)	0.109375	0.038865	–0.066652	–0.075450	1.000000

Source: Authors’ calculations.

Notes: The table presents the summary of descriptive statistics then correlation matrix of the major items; P_t Transaction price; X_t Daily accumulated order flow; (i_t – i_tf) difference in interest rate for short-term period; (l_t – l_tf) difference in interest rate for long-term period; and (R_t – R_tf) difference in the country’s risk premium. 10%, 5% and 1% levels denote *, ** and *** levels of statistical significance, respectively.

Transaction Price (P_t) and Cumulative Currency Order Flow (X_t)

Evans and Lyons (2002a) proposes a model based on a portfolio shift model. This model can be stated as

Δ P_{t} = Δ m_{t} + λ Δ X_{t}

(1)

Transaction Price (P_t) and Interest Rate (l_t – l_tf)

As a result of public information innovations Δm_t and the change in the log of the spot exchange rate ΔP_t, our Equation (1) needs modification to be comparable to the standard macroeconomic models. The estimation specification can be expressed as:

Δ P_{t} = α \cdot Δ (l_{t} - l_{t f}) + β \cdot Δ X_{t} + e_{t}

(2)

where ΔP_t represents change in log of the spot exchange rate; Δm_t in Equation (1) is the change in interest rate differential; that is, Δm_t = Δ(l_t – l_tf), we substitute Δm_t for change in long-term interest rate differential Δ(l_t – l_tf). Interest rate is considered to be an important variable that causes exchange rate movements in macroeconomic models, also available on a daily basis. Hence, it is considered suitable for experiential research. ΔX_t represents the daily cumulative order flow, while α and β represent regression parameters, and e_t is the error term.

Term Spread and Country’s Risk Premium (R_t – R_tf)

Country’s risk premium is a variable considered in the literature to have a positive and strong significance in the studies of emerging markets (De Medeiros, 2004; Duffuor et al., 2012; Wu, 2010; Zhang et al., 2013). Country’s daily risk premium R_t represents the difference between the prime lending rate and 3 months Treasury bill rate. Therefore, the difference between the two countries risk premiums is given as (R_t – R_tf), the Malaysian’s risk premium minus that of the US’s risk premium.

The research work of Evans (2011) states that currency transaction spot rate P_t of a pair currency with their interest rate short-term period is practically determined according to the standard of the monetary policy of the central banks concerned. Therefore, we consider Bank Negara Malaysia and the Federal Reserve as the central banks concerned in this study. Quote for all dealers is at a USD/MYR and is given as:

P_{t} = E P_{t} + (i_{t} - i_{t f}) - R_{t}

(3)

where P_t is the transaction price; (i_t – i_tf) represents difference in interest rate for short-term period; R_t represents country’s daily risk premium, that is, the difference between the prime lending rate and 3 months Treasury bill rate.

Figure 2.

Hodrick–Prescott Filter for the Major Items of Measurement

The long-term (l_t) and short-term (i_t) difference represent term spread, which is given as:

(l_{t} - l_{t f}) - (i_{t} - i_{t f}) = (R_{t} - R_{t f})

(4)

Therefore, we can equate country’s daily risk premium difference to the term spread. Figure 2 presents the Hodrick-Prescott Filter for the major items of measurement (exchange rate, currency order flow, short-term interest rate, long-term interest rate and country risk premium).

Methodology

The portfolio shift model (Evans & Lyons, 2002a) is extended in this article and we apply a VAR model proposed by Hasbrouck’s (1991) to examine the market microstructure elements of Malaysian currency exchange rate fluctuations. From the literature reviewed, VAR model is one of the preferred econometric methods to investigate both the long-run as well as short-run relation between currency order flow and exchange rate fluctuations with their feedback effects (Duffuor et al., 2012; Froot & Ramadorai, 2005; Wu, 2010; Zhang et al., 2013). Importantly, it considers the currency order flow coefficient on the ordinary least square regression with their likely feedback effect (Zhang et al., 2013).

Johansen’s (1995) cointegration is applied to run our analysis with particular reference to the setting of VAR. With cointegration analysis, we are more precise about the long-run relation between cumulative currency order flow and exchange rate movements. Cointegration is said to exist between two time series if they are individually non-stationary, even though there exists a linear combination of them with stationarity (Evans & Lyons, 2007). By interpretation, we can say that a stable long-run equilibrium relation exists. Therefore, VAR framework is extended in our analysis to calculate approximately the explanatory power of currency order flow on exchange rate movements in Malaysia.

The Vector Autoregression (VAR) Model.

The VAR model assumes that quotes from the market are immediately reflected based on public information available to the traders; hence, the informed traders take advantage of this to earn returns via their currency market orders.

Therefore, let H_t denotes attribute vector, D_t is the log of each transaction attribute, t is the time event.

The model:

H_{t} = B D_{t} + E_{t}

(5)

and

\underset{}{H_{t} = (\begin{matrix} P_{t} \\ X_{t} \\ (i_{t} - i_{t f}) \\ (l_{t} - l_{t f}) \\ (R_{t} - R_{t f}) \end{matrix})} 5 x 1; \underset{5 x 5 l}{B = [\begin{matrix} β_{1, 1} R β_{1, 5 l} \\ ⋮ ⋮ ⋮ \\ N O N \\ ⋮ ⋮ ⋮ \\ β_{5, 1} U β_{5, 5 l} \end{matrix}]}; \underset{5 l x 1}{D_{t} = (\begin{matrix} P_{t - 1} \\ ⋮ \\ N \\ ⋮ \\ R_{t - 1} \end{matrix})}; \underset{5 x 1}{E_{t} = (\begin{matrix} ε_{t} \\ ε_{t} \\ ε_{t} \\ ε_{t} \\ ε_{t} \end{matrix})}

(6)

where P_t represents transaction price, X_t represents daily accumulated order flow, (i_t – i_tf) represents difference in interest rate for short-term period, (l_t – l_tf) represents difference in interest rate for long-term period and (R_t – R_tf) represents the difference in the country’s risk premium. B represents matrices of coefficients to be estimated (β, R, N, O and U). Ordinary least square with heteroskedasticity robust standard errors is applied to estimate each VAR equation.

The VAR terms:

H_{t} = Γ D_{t - 1} + ε_{t}

(7)

hence,

{H^{'}}_{t} = f [P_{t}, X_{t}, (i_{t} - i_{t f}), (l_{t} - l_{t f}), (R_{t} - R_{t f})]

(8)

where H_t represent the transaction attributes vector, P_t represents the transaction price, X_t represents daily accumulated order flow, (i_t – i_tf) represents difference in interest rate for short-term period, (l_t – l_tf) represents differential in interest rate for long-term period and (R_t – R_tf) represents the difference in the country’s risk premium. The companion matrix Γ and variable P_t are let on uniform crosswise of the currencies and also the lags.

Presentation of Empirical Results

Table 2 reports the result of Johansen cointegration tests. The cointegration rank test (Trace and maximum eigenvalue statistics) analyse the propositions of maximum g number of cointegrating relations of the key variables. The subscript g denotes the number of significant cointegrating vectors. The results show that three cointegrating relationships exist, based on our full sample. At the 5 per cent significance level, the null hypothesis L₀: g ≤ III cannot be rejected. In addition, Table 3 shows the results of the uniqueness of the cointegrating relationships of our variable space tested in the VAR specification, that is, ${H^{'}}_{t} = f [P_{t}, X_{t}, (i_{t} - i_{t f}), (l_{t} - l_{t f}), R_{t}, R_{t f}, T r e n d]$ . Among the hypotheses tested, H₁ tests the cointegrating relationships if there exists any trend, but when we exclude the trend from the model, the p-value of 0.0606 is rejected. Also, we test the long-run cointegrating relationships between exchange rate and currency order flow, that is, P_t = –X_t, interest rate spread and country’s risk difference, that is, (i_t – i_tf) = – (l_t – l_tf), R_t = – R_tf using hypotheses H₂ to H₆. Hypothesis H₅ with p-value of 0.2887 is accepted from the test results. This result shows that there exists a relationship between exchange rate and country’s risk premium (difference). Our optimal lag length is of automatic specification based on the Schwarz information criterion (SC) and the Akaike information criterion (AIC) with maximum lag of 23.

Table 2.

Cointegration Analyses with Levels

Levels (ranks)	L₀: g ≤ NIL	L₀: g ≤ I	L₀: g ≤ II	L₀: g ≤ III	L₀: g ≤ IV
Eigenvalue	0.137864	0.021104	0.016107	0.004938	0.000298
Log likelihood	2694.781	2710.318	2726.230	2738.344	2742.037
Trace test	285.2089	63.88314	32.05888	7.831313	0.445313
Crit. Value (0.05)	69.81889	47.85613	29.79707	15.49471	3.841466
Probability	(0.0001)***	(0.0008)***	(0.0270)**	(0.4836)	(0.5046)
Max-Eigen	221.3258	31.82425	24.22757	7.386000	0.445313
Crit. Value (0.05)	33.87687	27.58434	21.13162	14.26460	3.841466
Probability	(0.0001)***	(0.0134)**	(0.0177)**	(0.4445)	(0.5046)

Source: Authors’ calculations.

Notes: The table reports the result of Johansen cointegration analyses. The cointegration rank test (trace and maximum eigenvalue statistics) analyse the propositions of maximum g number of cointegrating relations of the key variables. g denotes the cointegrating vectors number of significance. 10%, 5% and 1% denote the statistical levels of *, ** and ***, respectively.

Table 3.

Cointegrating Equations Restriction Tests

	P_t	X_t	(i_t – i_tf)	(l_t – l_tf)	R_t	R_tf	Trend
Unrestricted:
β ₁	0.5808	–0.0283	1.0000	6.6361	0.1405	–0.6056	–0.0041
β ₂	1.0000	–0.0249	0.0258	–0.2168	–0.3805	–0.0460	0.0015
H₁: Trend = 0, χ² (2) = 14.9265 [0.0606]*
β ₁	5.4624	–0.0419	1.0000	8.9061	–0.8289	–0.9066	0.00
β ₂	1.0000	–0.0350	–4.1204	–0.4613	–0.9645	3.6627	0.00
H₂: P_t = –X_t, χ² (2) = 77.9526 [0.0000]***
β ₁	–0.0283	0.0283	1.0000	9.4724	0.8978	0.3361	–0.0051
β ₂	1.0000	–1.0000	512.28	–27.969	–787.94	903.20	–0.0828
H₃: (i_t – i_tf) = – (l_t – l_tf), χ² (2) = 15.4072[0.0517]**
β ₁	–46.133	0.0481	1.0000	–1.0000	16.277	2.4902	–0.0734
β ₂	1.0000	0.0314	–279.69	279.69	22.798	73.463	–0.0920
H₄: (i_t – i_tf) = – (l_t – l_tf), Trend = 0 χ² (4) = 24.2126[0.0027]***
β ₁	3.0611	–0.0152	1.0000	–1.0000	–3.3618	– 0.6070	0.00
β ₂	1.0000	–0.0089	–4.1952	4.1952	0.0135	2.0234	0.00
H₅: R_t = – R_tf, χ² (2) = 9.6730[0.2887]
β ₁	2.9205	–0.0350	6.9462	35.3809	1.0000	–1.0000	–0.0585
β ₂	1.0000	–0.0283	–0.3274	0.4724	–0.4902	0.4902	0.0003
H₆:: R_t = – R_tf, Trend = 0 χ² (4) = 23.8440[0.0930]*
β ₁	–2.0920	0.0585	9.7631	0.4623	1.0000	–1.0000	0.00
β ₂	1.0000	–0.0283	0.4972	3.4519	–0.4248	0.4248	0.00

Source: Authors’ calculations.

Notes: This table reports the result of cointegrating relationships among key variables with and without trends. *, **, *** indicate statistical significance at the 10%, 5%, and 1% levels.

Table 4.

Granger Causality Test

Granger causality /long-run weak exogeneity test
	P_t	X_t	(i_t – i_tf)	(l_t – l_tf)	(R_t – R_tf)
X² (4)	50.04507	86.39550	24.21266	76.13401	23.84407
Probability	(0.0000)***	(0.0000)***	(0.0850)*	(0.0000)***	(0.0930)*

Source: Authors’ calculations.

Notes: This table presents the results of Granger causality tests and long-run weak exogeneity test of the key variables. 10%, 5% and 1% levels denote *, ** and *** levels of statistical significance.

Table 5.

Long-run Formation

	P_t	X_t	(i_t – i_tf)	(l_t – l_tf)	(R_t – R_tf)
Cointegrating vector. β	–2.402246 –1.0000	–0.05082 0.00547	1.0000 3.648690	–4.507371 –1.0000	–0.887760 1.0000
Feedback coefficients (α) with 2 ranks	0.002052 (0.00076) –0.000280 (0.00017)	30319.57 (2291.11) –2093.856 (525.556)	–0.001149 (0.01922) 0.010664 (0.00441)	0.017619 (0.01222) –0.001974 (0.00280)	–0.046737 (0.01431) –0.001324 (0.00328)

Source: Authors’ calculations.

Notes: The table reports the outcomes of hypotheses test on the cointegrating relationship among the variables. The cointegration coefficients are β and adjustment coefficients α with their standard errors in (). We consider 1 to 4 lag interval.

Table 4 presents the results of Granger causality tests and long-run weak exogeneity test of the key variables. The results show that exchange rate Granger causes order flow and vice versa. In essence, there exists bidirectional causality. Table 5 reports the results of hypotheses test on the cointegrating relationship among the variables with their cointegration coefficients β and adjustment coefficients α and their standard errors. However, from the results of the p-values for the long-run beta, none of the variable appears weak in the model.

From this results, level data can be formulated with the following cointegrating equations:

u_{i} = - P_{t} + 0.00547 * X_{t} + 3.64869 * (i_{t} - i_{t f}) - (i_{t} - i_{t f}) + (R_{t} - R_{t f});

(9)

u_{i i} = - 2.402246 * P_{t} - 0.05082 * X_{t} + (i_{t} - i_{t f}) - 4.507371 * (i_{t} - i_{t f}) - 0.88776 * (R_{t} - R_{t f})

(10)

The interest rate difference is significant and appropriately signed. Also, currency order flow is positively significant, meaning that there will be higher MYR price against the USD once there is a higher imbalance currency position in the net buying activities. Moreover, with a beta coefficient of 0.00547 in the USD/MYR exchange rate calculation, it implies that, for every currency order flow increasing at 1 per cent, there will be a corresponding increase within the day transactions, 54 basis points of the MYR price against the USD. Also, the country’s risk premium long-run coefficient β is likewise significant.

Δ P_{t} = u + α_{1} * Δ P_{t - 1} + β_{1} * Δ X_{t - 1} + β_{2} * Δ X_{t - 2} + θ * u_{i i t - 1} + ε_{p, t}

Δ X_{t} = u + α_{1} * Δ P_{t - 1} + α_{2} * Δ P_{t - 2} + θ * u_{i t - 1} + ε_{X, t}

\begin{array}{l} \begin{matrix} Δ (R_{t} - R_{t f}) = α_{1} * Δ P_{t - 1} + α_{3} * Δ P_{t - 3} + φ_{3} * Δ (i_{t - 3} - i_{t f - 3}) + λ_{1} * Δ (l_{t - 1} - l_{t f - 1}) \end{matrix} \\ \begin{matrix} + λ_{2} * Δ (l_{t - 2} - l_{t f - 2}) + λ_{3} * Δ (l_{t - 3} - l_{t f - 3}) + δ_{1} * Δ (R_{t - 1} - R_{t f - 1}) \end{matrix} \\ \begin{matrix} + δ_{2} * Δ (R_{t - 2} - R_{t f - 2}) + δ_{3} * Δ (R_{t - 3} - R_{t f - 3}) + ε_{R, t} \end{matrix} \end{array}

(11)

Table 6.

Error Correction Modelling Estimates

	ΔP_t	ΔX_t	Δ(R_t – R_tf)
Constant	0.0311 (0.0259)	–0.1419 (0.495)	–
α ₁	–0.0912*** (0.0409)	–0.1480** (0.0637)	–0.8780 (0.4389)
α ₂	–	0.1442** (0.0601)	–
α ₃	–	–	–0.8346 (0.3490)
β ₁	–0.5520 (0.1520)	–	–
β ₂	–1.0847** (0.3637)	–	–
θ	–0.0413 (0.0014)	–0.5215*** (0.0361)	–
φ ₃	–	–	0.9182*** (0.1534)
λ₁	–	–	–10.2205** (4.8070)
λ₂	–	–	15.6880*** (7.9193)
λ₃	–	–	–44.7526*** (10.6702)
δ₁	–	–	–0.4058*** (0.6661)
δ₂	–	–	–0.3860*** (0.0260)
δ₃	–	–	–0.1558*** (0.0301)
R ²	0.1669	0.3683	0.3773

Source: Authors’ calculations.

Notes: The table reports the estimating outcomes for ΔP_t , ΔX_t and Δ(R_t – R_tf) of the short-run vector error correction model. 10%, 5% and 1% levels denote *, ** and *** levels of statistical significance.

Table 6 shows the result of the short-run vector error correction model (VECM) estimates for ΔP_t, ΔX_t and Δ(R_t – R_tf). We removed from our model insignificant variables thereby reducing it to a partial VECM. The short-term correction result is significantly negative at 5 per cent level with a coefficient error correction term θ of –0.0413. This suggests that currency order flow in the Malaysian currency market Granger causes exchange rate fluctuation in the short-term. In addition, order flow speed of adjustment on the long-run relation is negative and significant. Therefore, this shows that an important factor influencing currency order flow is exchange rate fluctuations. Also, there is a negative relation between country’s risk premium and exchange rate movements. We compare our specification with that of Evans and Lyons (2002a), it shows that the coefficients of both studies are significant. Although, the R² obtained in our study (almost, 0.17) is low compared with Evans and Lyons 0.64 and 0.46, respectively. We realize that, in emerging markets, the frequency at which their currencies are being freely traded is relatively low compared with world major currencies of the developed markets. This implies that, there likely, in the emerging markets, more occurrences of currency interventions by the government. Thus, this may be one of the major reasons that might account for the difference in our results with that of Evans and Lyons (2002a). However, this result is harmonious with other results of emerging economies of a similar nature. For example, De-Medeiros (2004) obtained R² value of 0.06 in analysing order flow in the Brazilian foreign exchange market. Also, Zhang et al. (2013) obtained R² value of 0.13 in the analysis of order flow in China.

Table 7.

Variance Decomposition of Exchange Rate

Period	Standard error	P_t	X_t	(i_t – i_tf)	(l_t – l_tf)	(R_t – R_tf)
10	0.004547	96.1204	13.3680	0.06533	0.17900	5.9001
20	0.005744	94.2066	20.6703	0.29658	0.09297	5.4960
30	0.007843	91.4263	24.2662	0.32503	0.06819	5.3530

Source: Authors’ calculations.

Notes: The table reports the outcomes of decomposition of each item forecast error variance in our specification. We use Cholesky decomposition to test for a time period of 30 trading days.

In testing the strength of the relationship at longer horizons, we consider 10 trading days as 2 weeks, 20 trading days as 4 weeks and 30 trading days as 6 weeks. Therefore, we test with Cholesky decomposition for a time horizon of 30 trading days. Table 7 reports the results of decomposition of each item forecast error variance in our specification. That is, the variance decomposition of currency exchange rate fluctuations relative to other items in our specification, and it appears that order flow is the most exogenous variable. The results show that 24 per cent of changes in the exchange rate movements are caused by currency order flow. Hence, currency order flow may account for 24 per cent of exchange rate movements per day trading period. Also, the country’s risk premium difference explains approximately 5.4 per cent of exchange rate movements, while short-term interest and long-term interest differentials account for less than 1 per cent. Therefore, we can state that currency order flow variable with that of country’s risk variable appear important determinant factors of exchange rate fluctuations in the Malaysian foreign exchange market.

Conclusion

This study examines the potential role of cumulative currency order flow in the determination of MYR exchange rate against the USD in the long-term as well as short-term dynamics, respectively. We construct a measure of currency order flow in the Malaysian foreign currency exchange market context to reflect the pressure of currency excess demand. Also, we estimate the long- and short-run parameters with VAR framework and the results show that there exists cointegrating relation between cumulative currency order flow and exchange rate of USD against the MYR. There is a confirmation that order flow can actually explain the major movements in the exchange rate of the MYR/USD. The explanatory power of the currency order flow is strong and positive as the results revealed. Importantly, currency order flow with a positive beta coefficient of 0.00547 in the USD/MYR exchange rate implies that for every currency order flow increasing at 1 per cent, there will be a corresponding increase within the day transactions, 54 basis points of the MYR price against the USD. We compare our specification with that of Evans and Lyons (2002a), it shows that the coefficients of both studies are significant. Although, the R² obtained in our study (almost, 0.17) is low compared with Evans and Lyons 0.64 and 0.46, respectively. We realize that, in emerging markets, the frequency at which their currencies are being freely traded is relatively low compared with major currencies of the developed markets. This implies that, there likely, in the emerging markets, more occurrences of currency interventions by the government in the foreign exchange market. Thus, this may be one of the major reasons that might account for the difference in our results with that of Evans and Lyons (2002a). Nevertheless, this result is harmonious with other results of emerging economies of similar nature, such as that of Brazil and China.

Managerial Implications

This research study sheds more light on the appreciativeness of the international currency exchange rate in the emerging markets economy, thereby concentrating on the currency order flow and determination of exchange rate of MYR against USD by applying a market microstructure approach that is based on high-frequency dataset. Insomuch that the results show that currency order flow has significant explanatory power to capture the MYR exchange rate variability. Thus, having a deeper understanding of the explanatory power of currency order flow on how it drives the exchange rate movements in the foreign exchange market, it then brings to the attention of the Malaysian Monetary Authority the importance that should be attached to the market microstructure.

Footnotes

Acknowledgements

The authors are grateful to the anonymous referees of the journal for their extremely useful suggestions to improve the quality of the article. Usual disclaimers apply.

References

Andersen

T. G.

Bollerslev

Diebold

F. X.

& Vega

(2003). Micro effects of macro announcements: Real-time price discovery in foreign exchange. The American Economic Review, 93(1), 38–62.

Ariff

(1991). The Pacific economy: Growth and external stability. North Sydney: Allen & Unwin Pty Ltd.

Bank Negara Malaysia (2016). Retrieved 25 February 2016, from www.bnm.gov.my

Berger

D. W.

Chaboud

A. P.

Chernenko

S. V.

Howorka

& Wright

J. H.

(2008). Order flow and exchange rate dynamics in electronic brokerage system data. Journal of International Economics, 75(1), 93–109.

Bjønnes

G. H.

& Rime

(2005). Dealer behavior and trading systems in foreign exchange markets. Journal of Financial Economics, 75(3), 571–605.

Boyer

M. M.

& Norden

S. V.

(2006). Exchange rates and order flow in the long run. Finance Research Letters, 3(4), 235–243.

Danielsson

& Love

(2006). Feedback trading. International Journal of Finance & Economics, 11(1), 35–53.

Danielsson

Luo

& Payne

(2012). Exchange rate determination and inter-market order flow effects. The European Journal of Finance, 18(9), 823–840.

De Medeiros

O. R.

(2004). Order flow and exchange rate dynamics in Brazil. SSRN eLibrary.

10.

Duffuor

Marsh

I. W.

& Phylaktis

(2012). Order flow and exchange rate dynamics: An application to emerging markets. International Journal of Finance & Economics, 17(3), 290–304.

11.

Effiong

E. L.

(2014). Exchange rate dynamics and monetary fundamentals: A cointegrated SVAR approach for Nigeria. Global Business Review, 15(2), 205–221.

12.

Evans

(2002). FX trading and exchange rate dynamics. The Journal of Finance, 57(6), 2405–2447.

13.

Evans

(2010). Order flows and the exchange rate disconnect puzzle. Journal of International Economics, 80(1), 58–71.

14.

Evans

(2011). Exchange-rate dynamics. Princeton: Princeton University Press.

15.

Evans

& Lyons

(2002a). Informational integration and FX trading. Journal of International Money and Finance, 21(6), 807–831.

16.

Evans

& Lyons

(2002b). Order flow and exchange rate dynamics. Journal of Political Economy, 110(1), 170–180.

17.

Evans

& Lyons

(2005). Do currency markets absorb news quickly? Journal of International Money and Finance, 24(2), 197–217.

18.

Evans

& Lyons

(2006). Understanding order flow. International Journal of Finance & Economics, 11(1), 3–23.

19.

Evans

& Lyons

(2007). Exchange rate fundamentals and order flow (NBER Working Paper w13151). Georgetown University and University of Berkeley (mimeo).

20.

Evans

& Lyons

(2008). How is macro news transmitted to exchange rates? Journal of Financial Economics, 88(1), 26–50.

21.

Fisher

& Hillman

(2003). Comments on Richard K. Lyons, explaining and forecasting exchange rates with order flows. Economic Policy Forum, 96(4), 129–133.

22.

Frankel

J. A.

& Rose

A. K.

(1995). Empirical research on nominal exchange rates. In Handbook of international economics (No. 3, pp. 1689–1729). North-Holland, Amsterdam.

23.

Froot

K. A.

& Ramadorai

(2005). Currency returns, intrinsic value, and institutional-investor flows. The Journal of Finance, 60(3), 1535–1566.

24.

Galati

(2000). Trading volumes, volatility and spreads in foreign exchange markets: Evidence from emerging market countries. BIS, Monetary and Economic Department.

25.

Hasbrouck

(1991). Measuring the information content of stock trades. Journal of Finance, 46(1), 179–207.

26.

Ito

Lyons

R. K.

& Melvin

M. T.

(1998). Is there private information in the FX market? The Tokyo experiment. The Journal of Finance, 53(3), 1111–1130.

27.

Johansen

(1995). Likelihood-based inference in cointegrated vector autoregressive models, advanced texts in econometrics. Oxford and New York: Oxford University Press.

28.

Lyons

R. K.

(1995). Tests of microstructural hypotheses in the foreign exchange market. Journal of Financial Economics, 39(2), 321–351.

29.

Lyons

R. K.

(2001a). The microstructure approach to exchange rates. Cambridge, MA: MIT Press.

30.

Lyons

R. K.

(2001b). New perspective on FX markets: Order-flow analysis. International Finance, 4(2), 303–320.

31.

Marsh

I. W.

& O’Rourke

(2005). Customer order flow and exchange rate movements: Is there really information content? Cass Business School Research Paper.

32.

Martin

A. D.

(2001). Technical trading rules in the spot foreign exchange markets of developing countries. Journal of Multinational Financial Management, 11(2), 59–68.

33.

Meese

R. A.

& Rogoff

(1983). Empirical exchange rate models of the seventies: Do they fit out of sample? Journal of International Economics, 14(1), 3–24.

34.

O’Hara

(1995). Market microstructure theory (Vol. 108). Cambridge, MA: Blackwell.

35.

Osler

C. L.

(2006). Macro lessons from microstructure. International Journal of Finance & Economics, 11(1), 55–80.

36.

Payne

(2003). Informed trade in spot foreign exchange markets: An empirical investigation. Journal of International Economics, 61(2), 307–329.

37.

Perraudin

& Vitale

(1996). Interdealer trade and information flows in a decentralized foreign exchange market. In Frankel

J. A.

Galli

& Giovannini

(Eds), The microstructure of foreign exchange markets (pp. 73–106). Chicago: University of Chicago Press.

38.

Rime

(2000). Private or public information in foreign exchange markets? An empirical analysis (Working Paper No. 14/2000). Available at SSRN.

39.

Rime

Sarno

& Sojli

(2010). Exchange rate forecasting, order flow and macroeconomic information. Journal of International Economics, 80(1), 72–88.

40.

(2010). Order flow in the South: Anatomy of the Brazilian FX market. The North American Journal of Economics and Finance, 23(3), 310–324.

41.

Zhang

Chau

& Zhang

(2013). Exchange rate determination and dynamics in China: A market microstructure analysis. International Review of Financial Analysis, 29(2), 303–316.

Currency Order Flow and Exchange Rate Determination: Empirical Evidence from the Malaysian Foreign Exchange Market

Abstract

Keywords

Introduction

Review of Literature in Brief

Theoretical Framework of the Study

Objectives of the Study

Market Microstructure of Malaysian Foreign Exchange Market

The Daily Exchange Rate of USD/MYR MYR and Order Flow (04/01/2010 – 31/12/2015)

Data and Methodology

Data Sources

Measurement of Variables

Transaction Price (Pt) and Cumulative Currency Order Flow (Xt)

Transaction Price (Pt) and Interest Rate (lt – ltf)

Term Spread and Country’s Risk Premium (Rt – Rtf)

Hodrick–Prescott Filter for the Major Items of Measurement

Methodology

The Vector Autoregression (VAR) Model.

Presentation of Empirical Results

Conclusion

Managerial Implications

Footnotes

Acknowledgements

References

Transaction Price (P_t) and Cumulative Currency Order Flow (X_t)

Transaction Price (P_t) and Interest Rate (l_t – l_tf)

Term Spread and Country’s Risk Premium (R_t – R_tf)