Policy Research Working Paper 8893 Dollarization Dilemma Price Stability at the Cost of External Competitiveness in Cambodia Sovannroeun Samreth Miguel Eduardo Sánchez-Martín Sodeth Ly Macroeconomics, Trade and Investment Global Practice June 2019 Policy Research Working Paper 8893 Abstract Cambodia has recorded both rapid economic growth and shock has a negative impact on Cambodia’s trade balance macroeconomic stability in recent decades despite (or with the European Union, its main trading partner, as it thanks to) high levels of dollarization. Previous studies on leads to appreciation of the US dollar. Furthermore, this dollarization in Cambodia have largely focused on exam- shock also leads to a significant decrease in Cambodia’s ining its causes and estimating seigniorage losses. As an international reserve levels during the first two quarters attempt to further explore the effects of dollarization in following the shock. The surrendering of monetary and Cambodia, this paper examines its impact on the compet- exchange rate independence seems to affect the competitive- itiveness of the export sector. The main results, based on ness of the tradable sector negatively as well as exacerbate a vector autoregression estimation of quarterly data over financial sector vulnerability to solvency and liquidity risks. 1994Q4–2016Q4, indicate that a positive US interest rate This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at msanchezmartin@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Dollarization Dilemma: Price Stability at the Cost of External Competitiveness in Cambodia * Sovannroeun Samreth, † Miguel Eduardo Sánchez-Martín, ‡ and Sodeth Ly § JEL Classification: E5 Monetary Policy, Central Banking, and the Supply of Money and Credit F31 Foreign Exchange O24 Trade Policy • Factor Movement Policy • Foreign Exchange Policy * The authors are grateful to Charl Jooste, Naoko Kojo and Deepak Mishra for the insightful comments shared on previous versions of this paper, as well as the National Bank of Cambodia for providing necessary data for this research. Any error in this paper is the responsibility of the authors. † Graduate School of Humanities and Social Science, Saitama University, Japan. ‡ Senior Economist in the Macroeconomics, Trade and Investment global practice, the World Bank Group. § Senior Economist in the Macroeconomics, Trade and Investment global practice, the World Bank Group. 1. Introduction Over the past two decades, Cambodia has attained remarkable economic growth and macroeconomic stability. Growth has been driven by exports of goods and services (mainly garments and tourism), which grew by nearly 20 percent a year during 1994- 2015, among the top in the world and ahead of stellar performers such as Vietnam or Bangladesh. Cambodia’s impressive achievements have been built upon openness to trade and capital flows and driven by preferential access to the United States (US) and European Union (EU) markets. This has facilitated foreign direct investment (FDI) inflows and substantial official development assistance. In the late 1980s and early 1990s, Cambodia relied on monetization to fund its fiscal deficit, which resulted in hyperinflation at a time when the country was transitioning from a centrally planned to market economy. Macroeconomic instability coupled with political transition contributed to a rapid depreciation of the local currency, the Khmer riel (KHR), which encouraged Cambodians to start substituting other currencies as a store of value (Kem, 2001; Zamaróczy and Sa, 2002; Samreth, 2010). The arrival of the United Nations Transitional Authority in Cambodia (UNTAC) in 1992 and the opening to foreign aid and investors resulted in large inflows of US dollars (USD), which circulated freely as the authorities were not well prepared to handle the influx. The USD has since displaced the KHR as the primary currency. These days, the share of foreign currencies in broad money hovers around 80 percent, and the share of USD deposits to total deposits has remained above 90 percent during the past two decades. In Cambodia, the USD is used as a store of value (the so-called “financial dollarization”), a unit of account to determine prices and wages (“real dollarization”), and a means of exchange (“transaction dollarization”). Against this backdrop, Cambodian authorities adopted a managed float regime 1 in order to provide a nominal anchor for economic agents and to underpin the price stability of goods (and services) in local currency. Under a fixed exchange rate, the public sector cannot resort to money printing in excess by the central bank, since it would destroy the exchange rate arrangement (World Bank, 2015). Overall, under its exchange rate arrangement and in the context of underdeveloped capital markets, the country has been able to achieve macroeconomic stability. The indirect impact of dollarization on external competitiveness has not been well researched. In Cambodia, while eliminating exchange rate risk has been highly beneficial for an economy that was exporting intensely to the United States, its limitations have become more apparent since the EU became the main trading partner. As shown in Figure 1 The USD-KHR exchange rate has been contained between KHR 3,900 and 4,100 per USD since 2011. Thus, the KHR is de facto pegged to the USD (or at least fluctuates within a +/- 2 percent band). 2 1, episodes of USD appreciation vis-à-vis the Euro and other currencies seem to have negatively affected garment exports and have made trips to Cambodia more expensive for non-American tourists. The seeming decline in the competitiveness of the tradable sector may also partly explain the increasing share of credit flowing into construction and real estate that recent studies associate with higher financial risks (Ahmed et al., 2014; IMF, 2017). Figure 1: US dollar appreciation correlates with a decline in Cambodian garment exports to the EU 120% 10% 100% Change in the exchange rate 5% 80% Change in exports 60% 0% 40% 20% -5% 0% -10% 1999-Q2 2000-Q2 2001-Q2 2002-Q2 2003-Q2 2004-Q2 2005-Q2 2006-Q2 2007-Q2 2008-Q2 2009-Q2 2010-Q2 2011-Q2 2012-Q2 2013-Q2 2014-Q2 2015-Q2 2016-Q2 2017-Q2 -20% -40% -15% Exports to EU, yoy EU/USD, quarterly change Poly. (Exports to EU, yoy) Poly. (EU/USD, quarterly change) Source: European Central Bank, IMF Directions of Trade Database. Note: “Poly.” refers to a polynomial fit line to existing data points. In order to gain a more complete understanding of the limitations posed by the lack of independent monetary and exchange rate policy under dollarization, this paper assesses the hypothesis that US interest rate increases, resulting in USD appreciation, negatively affect Cambodian exports to the EU. The remainder of this paper is organized as follows. Section 2 presents a brief literature review. Section 3 discusses the theoretical framework and estimation methodology. Section 4 describes the data and presents estimation results. Section 5 discusses policy options, and Section 6 concludes. 3 2. Literature review Two different currents in literature serve as relevant background to the testing of the abovementioned hypothesis. One is on trilemma and exchange rate regime choice, and the other is on causes and effects of dollarization. According to the “Macroeconomic Trilemma” proposition, an economy cannot at the same time maintain monetary independence, exchange rate stability, and full integration in international financial markets (Aizenman, 2013). Over the past three decades, as most countries steadily increased their degree of integration in international financial markets, it has been argued that the “Trilemma” had become a “Dilemma,” and the currency regime options had been reduced to choosing between a fixed or a flexible exchange rate. After the Asian crisis, at the beginning of the past decade, a broad consensus started to emerge about the inappropriateness of opting for fixed exchange rates in a world of increasing financial integration (Kose et al., 2009). Thus, it was perceived that developing economies should either start floating (thus preserving monetary independence but sacrificing exchange rate stability) or commit to “hard peg” formulas such as monetary union or full dollarization. Reality, however, soon proved to be more complex, since most developing economies find it difficult to float in practice (Calvo and Reinhart, 2002). It has been argued that one of the main reasons behind this fear of floating is the lack of credibility of financial institutions, which usually is a problem in dollarized economies. Exchange rate depreciations tend to result in higher inflation in developing economies (passing through bilateral import prices). In addition, dollarization of liabilities is pervasive as, due to the “Original Sin,” agents in developing economies with incomplete financial markets are unable to borrow in domestic currency (Mckinnon and Schnabl, 2004). Invoicing in USD is also common practice in international trade (Boz et al., 2017). In another study, Casas et al. (2017) have argued that, while flexible exchange rates do provide valuable monetary policy independence, in a dollar-dominated trade environment, their ability to support full employment is severely limited, making the case for small open economies to opt for fixed exchange rates. Shi and Xu (2010) have argued that, under a fixed exchange rate, companies optimize their behavior by both borrowing and exporting in dollars. This so-called “twin dollarization” helps avoid domestic interest rate risk, which generates higher and more stable profits for export firms and, ultimately, higher consumption, partially compensating for other welfare losses caused by the fixed exchange rate regime. All these factors make the benefits of flexible exchange rates more elusive. As a result, in practice, many countries opted for softly pegging their currency to the USD, remaining in the middle ground of the trilemma. Notably, however, countries 4 sometimes find it hard to maintain intermediate regimes for long, since they are difficult to verify (Frankel, 1999; Frankel et al., 2001). Several benefits stemming from dollarization have been identified by the literature. For example, it constrains governments from resorting to excessive money printing by the central bank, which often results in inflation and heightened macroeconomic risks. The availability of deposits and financial instruments in foreign currency allows economic agents to protect the value of their savings and assets. Allowing foreign currency deposits in the domestic financial system also limits capital flight, reduces incentives to move savings abroad, and promotes overall financial deepening. Moreover, by mitigating exchange rate risks, dollarization supports international trade and investment. On the other hand, there are three commonly cited costs of dollarization. First, a dollarized country may face the loss of seigniorage or revenues stemming from the printing of local currency. Second, dollarization is alleged to limit the effectiveness of monetary policy and undermine monetary policy independence. Third, dollarization may heighten vulnerability to solvency risks, with the Central Bank being unable to serve as lender of last resort for domestic financial institutions in the case of a crisis (see Chang and Velasco, 2002; Cabral, 2010; World Bank, 2015). Policymakers tend prefer a stable exchange rate arrangement in the presence of high financial dollarization, as movements in the exchange rate could increase financial sector vulnerability to solvency risk. Overall, empirical evidence suggests that while financial dollarization has promoted a deeper financial system in countries with high inflation (Nicholo et al., 2005), the costs tend to outweigh the benefits, as dollarized economies are found to be more prone to crises. Meanwhile, the evidence on the relationship between dollarization and economic growth, effectiveness of monetary policy, and seigniorage is inconclusive. In Cambodia, in the context of shattered institutions and lack of credibility in the aftermath of the Khmer Rouge regime, economic agents started to hoard foreign currency (Zamaróczy and Sa, 2002). This was exacerbated by the country’s openness to aid, trade, and capital flows, which soon flooded with dollars the then small-sized Cambodian economy, which was operating under a weakened central bank. Lacking effective monetary policy independence, authorities were compelled to prioritize exchange rate stability over external competitiveness, which has ultimately resulted in self-fulfilling dollarization. Various studies on dollarization in Cambodia can be found. Some of them have focused on discussing its causes (Zamaróczy and Sa, 2002; Menon, 2008; Duma, 2011) and on analyzing the factors leading to the substitution between foreign and domestic 5 currencies that reinforce the dollarization process (Kem, 2001; Ra, 2008; Samreth, 2010, 2011; Siregar and Chan, 2014). Other studies discuss and analyze the negative impacts of dollarization by focusing on its negative effects on seigniorage (Zamaróczy and Sa, 2002; Kang, 2005; Samreth, 2010). However, other dimensions of costs and benefits of dollarization in Cambodia have not been addressed sufficiently. This paper aims to assess a consequence of high dollarization that is often overlooked by the literature: how US interest rate hikes (i.e. Fed decisions) might not only affect financial and capital flows but also indirectly hinder the competitiveness of the export sector of a highly dollarized country like Cambodia vis-à-vis third countries (in the EU, in particular) as the USD appreciates. To our knowledge, the closest study to this paper is Duma (2011), which assessed the impact of US interest rate increases on macroeconomic variables in Cambodia. Drawing from Duma (2011), this paper contributes to the literature in several ways. First, through a longer sample period, the vector autoregression (VAR) estimation is expected to provide more rigorous results. Second, it examines the impact of US interest rate increases not only on Cambodia’s total trade balance but also on its trade balances with its main trading partners (the EU and US). As a robustness check, it estimates Cambodia’s export function with the EU using a cointegration approach. Third, an extension of the analysis incorporating international reserves is introduced. Furthermore, to ensure the robustness of results, Bayesian VAR estimation is also conducted. 3. Analytical approach 3.1 Theoretical framework The Mundell-Fleming model, at the core of the “Macroeconomic Trilemma,” was widely applied in the analysis of transmission impacts of economic policies in an open economic framework. It was developed by Fleming (1962) and Mundell (1963, 1964) as an extension of the standard Keynesian IS-LM framework. The model was then extended by Dornbusch (1976) for the analysis of the exchange rate movement in the context of perfect foresight. Since then, this Mundell-Fleming-Dornbusch model has replaced the conventional Mundell-Fleming model as a tool for analyzing the effects of policy transmission. The model has been revisited by Obstfeld and Rogoff (1995) and Obstfeld and Rogoff (1996), who incorporated microfoundations to capture intertemporal behaviors of economic agents. 2 Our theoretical framework is essentially based on the standard Mundell-Fleming-Dornbusch model, given that the welfare effect examination 2 Duma (2011) also provides a brief background discussion of the Mundell-Fleming-Dornbusch model. 6 is beyond the scope of this paper. The first component of the model is the IS curve that helps capture the equilibrium in the market of goods and services. It can be expressed simply as follows: − ∗ = (, ( )), (1) where Y and Y* represent the actual output and the potential output (i.e. output under full employment), respectively; i is the domestic interest rate; NX(E) represents net exports (i.e. trade balance), which is a function of exchange rate E; F represents another function. An increase in the interest rate (i) reduces the demand for investment and induces savings, leading to a decrease in the output gap, Y-Y*. A depreciation of the exchange rate (i.e. the rise in E) leads to an increase in net exports. As a result, it increases the output gap. The second component is the LM curve illustrating the equilibrium in the money market. It can be expressed as a money demand function as follows: = (, ), (2) where M and P are money balance and domestic price level, respectively; L represents a function. An increase in the domestic interest rate (i) raises the opportunity cost of holding money, leading to a decrease in the real money balance, . An increase in output Y induces the demand for money for transactions. The third component is the Phillips curve capturing the movement of the domestic price level. The Phillips curve can be expressed as follows: = ∅( − ∗ ), (3) where represents inflation. A demand shock that increases the output gap induces inflation. Therefore, the parameter ∅ is positive. The fourth component is the uncovered interest rate parity capturing the relationship between domestic and foreign interest rates. The uncovered interest rate parity condition can be expressed as follows: = ∗ + , (4) 7 where i* and indicate the foreign interest rate and the rate of change of the exchange rate, respectively. An increase in z illustrates a depreciation of domestic currency, and its decrease represents an appreciation of domestic currency. Taking into account Equations (1) to (4), we can consider the impacts of an increase in foreign interest rate, i*. A policy shock raising i* can lead to an appreciation of foreign currency against domestic currency and an increase in domestic interest rate, i, from Equation (4). An increase in i decreases the output gap as discussed in equation (1), although it may be adjusted back due to the increase of the net exports resulting from the depreciation of domestic currency. From Equation (3), a decrease in output gap reduces inflation, and Equation (2) informs us that an increase in i leads the reduction of money balance. However, some of these impacts are likely to differ for Cambodia, a highly dollarized economy. In particular, exports to the EU and third countries are likely to be reduced (rather than increased), as US interest rate hikes result in USD appreciation, which affects exporting industry costs (via salaries and prices). Hence, to capture this, net exports NX in Equation (1) may need to incorporate i* (i.e. the US interest rate). 3.2 Estimation methodology In this study, we aim to examine the impact of the US policy interest rate on a highly dollarized country, Cambodia, by focusing on its effects on Cambodia’s external competitiveness. As mentioned above, the Mundell-Fleming-Dornbusch model provides a good framework for empirically examining the impacts of foreign policy change (i.e. policy interest rate) on the domestic economy. This examination can be done by the vector autoregression (VAR) estimation of the system including Equations (1) to (4). 3 The VAR estimation enables us to conduct the impulse response function and variance decomposition analyses of the variables in the system. The VAR model can be expressed as follows: = 0 + ∑ − + , =1 (5) where Y is the vector of the variables under consideration, m is the lag length, and u is an error term. As a first step, we need to conduct the unit root test for each variable in the VAR system. For variables integrating of order 0 (i.e. I(0) variables), they are included as 3 Duma (2011) also adopts the VAR estimation approach in an examination of the transmission impacts of the US policy interest rate on the key macroeconomic variables in Cambodia. 8 level variables. For variables integrating of order 1 (i.e. I(1) variables), their first difference needs to be used when including them in the estimation. The lag length, m, is usually determined by using the Akaike information criteria (AIC). 4 Motivated by the theoretical framework discussed above and taking into account the variables used in Duma (2011), the VAR system in our study consists of the following variables: the US federal fund rate (ffr), representing foreign policy interest rate; the growth rate of KHR to USD exchange rate (z) 5; trade balance as a share of GDP, money balance as a share of GDP (bmgdp) 6; domestic interest rate; output gap (gygap) as the difference between real GDP and real potential GDP (i.e. GDP under full employment), scaled by real potential GDP; and consumer price index (CPI) inflation (infl). In addition to including the total trade balance as a share of GDP (ttbgdp) as done by Duma (2011), the trade balances with main trading partners such as the EU and US are incorporated for the estimation in a separate case. They are represented by tbeugdp and tbusagdp, respectively. 7 Lacking Cambodia government bonds, the bank lending interest rate (lendint) is used as the domestic interest rate. 8 Duma (2011) incorporates lending and deposit interest rates into the VAR system simultaneously. However, we include only one domestic interest rate in the VAR estimation, as indicated in Equation (4) of the theoretical framework. Following Duma (2011), the fluctuation of the crude oil price represented by its change rate (goilp) is exogenously included in the VAR system as a control for external shocks. Furthermore, we add another exogenous dummy variable (dummy) to the estimation to capture the macroeconomic effects resulting from: political instability during the violent conflicts and political unrest in 1997; the 1997-98 Asian financial crisis; election periods in 1998, 2003, 2008, and 2013; and the 2008-09 global financial crisis. As an extension to the VAR system, another variable added is international reserves. Cambodian authorities may use the international reserve adjustment as a policy tool to, 4 In this study, the cointegration methodology based on the vector error correction model (VECM) proposed by Johansen and Juselius (1990) that requires that all variables are I(1) is not adopted. This is because the integration order of the variables under consideration in our estimation are mixed. We have both I(0) and I(1) variables in the VAR system, as discussed later. 5 z illustrates the depreciation of KHR if it increases and the appreciation of KHR if it decreases. 6 Broad money is incorporated as a share of GDP to take into account the scale effect. 7 Trade balances are incorporated as GDP shares, taking into account scale effect. 8 The VAR estimation by using deposit interest rate as domestic interest rate is also conducted. We obtain very similar results of impulse responses and variance decompositions. Data on the deposit interest rate is obtained from International Financial Statistics (IFS) published by the International Monetary Fund (IMF). The results can be provided upon request. 9 for example, stabilize the KHR to USD exchange rate in response to a US policy rate increase by the Fed. Therefore, we can expect a decrease in international reserves when there is a positive shock in the US federal fund rate. International reserves are included in the estimation as a share of GDP (intresvgdp). 4. Data and estimation results 4.1 Data Quarterly series between the fourth quarter of 1994 and the fourth quarter of 2016 for the variables mentioned above are obtained from various sources. 9 The data for US federal fund rate, KHR to USD exchange rate, lending interest rate, and CPI are obtained from the International Financial Statistics (IFS) published by the International Monetary Fund (IMF). Information on exports and imports is obtained from the Direction of Trade Statistics (DOTS) published by the IMF. Since quarterly GDP data for Cambodia is not available, it is generated by interpolation from the annual GDP series obtained from the World Development Indicators (WDI) published by the World Bank (WB). Broad money (M2) is the sum of money in circulation, demand deposits, and time and saving deposits of both KHR and foreign currencies. Data is obtained from the Economic and Monetary Statistics published by the National Bank of Cambodia (NBC). Data on crude oil price is retrieved from the Primary Commodity Prices published by the IMF. International reserve levels are obtained from the NBC. The dummy variable is 1 for 1997Q3-1998Q4, 2003Q3-2003Q4, 2008Q3-2009Q4, and 2013Q3-2013Q4, and zero elsewhere. Potential GDP in real values required for the calculation of output gap data is generated applying the Hodrick-Prescott filter on the real GDP series. Since our estimation uses quarterly data, adjustments are conducted to correct for seasonal fluctuations, except for the US federal fund rate and domestic lending interest rate. CPI is used to convert nominal GDP into real GDP. 4.2 Unit root tests As mentioned above, we need to apply unit root tests to the variables under consideration to identify their integration orders. Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests are employed. The results are provided in Table 1. From the table, while z, gygap, and infl are I(0) (i.e. variables having integration order of 0), ffr, bmgdp, ttbgdp, tbeugdp, tbusagdp, lendint, and intresvgdp are I(1) (i.e. variables having integration order of 1). 9 The sample periods are different by specification, depending on data availability. 10 Table 1: Unit root tests Variable Augmented Dickey-Fuller test Phillips-Perron test Level First difference Level First Difference ffr -1.8433 -3.5214*** -1.5620 -6.0814*** z -2.8946* -7.8303*** -5.6547*** -14.8536*** bmgdp 3.3403 -7.3074*** 2.5853 -7.3152*** gygap -2.9672** -4.3173*** -3.9751*** -7.4338*** infl -5.5818*** -8.1268*** -5.3785*** -23.3783*** ttbgdp -2.1970 -9.8562*** -2.1667 -9.8559*** tbeugdp -1.0499 -8.9135*** -1.0499 -8.9131*** tbusagdp -1.7918 -9.1621*** -1.8007 -9.1621*** lendint 0.2436 -8.1607*** 0.1383 -8.1201*** intresvgdp -0.0837 -3.1247** 0.9367 -5.3481*** 1. Null hypothesis: non-stationary 2. The asterisks ***, **, and * indicate 1%, 5%, and 10% of significance levels, respectively. 4.3 VAR estimation results Variables are included in the VAR estimation as their levels or first differences depending on whether they are I(0) or I(1) variables. In the baseline case, ffr, ttbgdp, bmgdp, and lendint are I(1). Hence, their first differences are included. For z, gygap, and infl, their levels are used as they are I(0) variables. The selected lag length for this VAR estimation is one, based on AIC. For the case in which the total trade balance is decomposed into trade balance with the EU (tbeugdp) and trade balance with the US (tbusagdp), the lag length of one is also used for the estimation from AIC selection. Both tbeugdp and tbusagdp are included in the estimation by using their first differences since they are I(1) variables from the unit root tests. 4.3.1 Impulse response results for the baseline case In examining the impact of a US policy interest rate shock on Cambodia’s highly dollarized economy, VAR-based impulse responses for the variables under consideration are analyzed. This is done by applying the Cholesky method, which requires that variables in the VAR system are ordered from the exogenous one to the most endogenous one. Following Duma (2011), the order in which variables are introduced in the baseline model is from D(ffr), to z, to D(ttbgdp), to D(bmgdp), to D(lendint), to gygap, and to infl, where “D” indicates the first difference. As indicated above, two exogenous variables are also included: goilp capturing the fluctuation of crude oil price and the dummy capturing political unrest and socio-economic instability. Figure 2 illustrates the results of impulse responses for the variables under 11 consideration to a one standard deviation (S.D.) shock in the federal fund rate. Except for the broad money balance (bmgdp), the impulse response results in the figure are generally as expected from a theoretical viewpoint and in line with Duma (2011): a US interest rate increase leads to higher interest rates in Cambodia, as well as lower inflation and a reduced output gap. 10 A first-quarter increase in the broad money balance as a share of GDP by approximately 0.86 percent in response to a one S.D. shock in the US federal fund rate (an increase of about 0.45 percent) is also observed. One possibility is that an increase in the US interest rate can cause currency depreciations with respect to the USD in other countries. Foreign capital may be mobilized from such countries to avoid exchange rate risk. Given its highly dollarized economy, Cambodia can be a short-term destination of such capital inflows. This can result in a temporary increase in foreign currency deposits in Cambodia, as a negative response of broad money balance is recorded in the second quarter. This may be due to the outflow of capital to a destination with increasing interest rates (i.e. the US). 11 In addition, the impulse response results show that a one S.D. shock in the US federal fund rate (an increase of about 0.45 percent) raises total trade balance as a share of GDP by about 0.19 percent in the first quarter. However, the response of total trade balance is negative in the second quarter. These mixed results may be a consequence of high dollarization in Cambodia. A positive US policy interest rate shock that leads to an appreciation of the USD may temporarily improve Cambodia’s trade balance with the US as exporters may try to export more to the US market, since appreciation of the USD makes their exports to non-US markets more expensive. At the same time, in making Cambodian export goods more expensive in the EU and third countries, the positive US policy interest rate shock worsens Cambodia’s trade balance with other major trading partners such as the EU, as discussed in the next section. 10 The VAR estimation is often criticized for including many variables that may be not important in the estimation and may cause biased results, especially for an estimation with a small sample size. As a robustness check, we also conduct the estimation based on the Bayesian vector autoregression (BVAR). Its impulse response results are similar to the VAR case. The results can be provided upon request. 11 There may be a need to explore this possibility with a detailed examination. We also try the VAR estimation including an additional variable, the net inflow of foreign direct investment (FDI) as a share of GDP, in the baseline case specification. The movement of the impulse response of this capital inflow with respect to one S.D. shock in the US federal fund rate is quite similar to that of broad money. 12 Figure 2: Impulse responses for the baseline case (“D” denotes the first difference) 4.3.2 Impulse response results for the case of trade balance decomposition In order to identify the impacts of the change in the US policy interest rate on Cambodia’s trade balance in more detail, we look at the impact on the balance with its two largest trading partners, the EU (tbeugdp) and the US (tbusagdp). Cambodian exports to these two partners amount to about two-thirds of total exports. The order of variables for the estimation is the same as the baseline case, but D(tbeugdp) and D(tbusagdp) are included to replace D(ttbgdp), where “D” is the first difference. 13 Figure 3 presents the impulse responses of this VAR estimation. Although a one S.D. shock in the US federal fund rate (an increase of about 0.45 percent) temporarily improves Cambodia’s trade balance with the US (as exporters may try to export more to the US market, given the fact that appreciation of the USD makes their exports to non-US markets more expensive), it decreases Cambodia’s trade balance with the EU by about 0.15 percent during the first quarter, and this impact seems to persist in the following quarters. 12 Furthermore, as indicated by the variance decomposition in Figure 4, the US policy interest rate shock contributes to about 2 percent of the variation in Cambodia’s trade balance with the EU. This negative impact is also confirmed when conducting a cointegration-based estimation of Cambodia’s export function with respect to the EU. The results, presented in Appendix A, confirm that an increase in the US federal fund rate does lead to a decrease in Cambodian exports to the EU. Other variable impulse responses are consistent with those of the baseline case. The impact of a US policy interest rate shock on non-dollarized economies is also examined. The VAR-based impulse response estimation for two countries in the region, Malaysia and Singapore, is provided in Appendix B. 13 The results indicate that a one S.D. shock in the US interest rate improves Malaysian and Singaporean trade balances with the EU, providing more evidence of how high dollarization affects Cambodia’s external competitiveness with respect to a positive US interest rate shock. 12 Similar results are obtained when using BVAR estimation. 13 An estimation for another non-dollarized economy in the region, Thailand, is also implemented. The sample period for the estimation is 1994Q4-2006Q4, based on data availability. However, the impulse response results seem inconsistent with theory. This may be partially due to a relatively small sample size for this case. 14 Figure 3: Impulse responses for the case of trade balance decomposition (“D” denotes the first difference) 15 Figure 4: Variance decomposition of D(tbeugdp) for the case of trade balance decomposition (“D” denotes the first difference) 4.3.3 Impulse response results for the case of trade balance decomposition and international reserve inclusion As an extension, an additional estimation including international reserves as a share of GDP (intresvgdp) is conducted. As Cambodia de facto operates under a peg to the USD, authorities often use international reserves and other instruments such as negotiable certificate of deposits to maintain the KHR to USD exchange rate within a +2/-2 percent fluctuation band. In the estimation of the VAR-based impulse response for this case, the order of variables runs from D(ffr), to z, to D(tbeugdp), to D(tbusagdp), to D(intresvgdp), to D(bmgdp), to D(lendint), to gygap, and to infl, where “D” indicates the first difference. Similar to previous estimations, goilp, capturing the fluctuation of crude oil price, and dummy, capturing years of exceptional political unrest and/or socio-economic instability, are incorporated exogenously. The sample period of this estimation is from 2003Q1 to 2016Q4, the period for which data on Cambodian international reserves is available. Figure 5 shows impulse response results. It indicates that a one S.D. shock in the US federal fund rate (an increase of about 0.44 percent) leads to a decrease in Cambodian international reserves as a share of GDP by about 0.02 percent in the first quarter. The negative impact in the second quarter is stronger. The decrease of intresvgdp is about 0.44 percent as a share of GDP. This is partly compensated in the following quarters, as the exchange rate readjusts. 16 Figure 5: Impulse responses for the case of trade balance decomposition and international reserve inclusion 5. Policy discussion This study provides new evidence on a consequence of dollarization in Cambodia: a positive US policy interest rate shock has a negative impact on Cambodia’s trade competitiveness with its main trading partner, the EU. As pointed out by Kojo (2015), contrary to the myth, (even partially) dollarized economies could be subject to Dutch Disease, and this could be a singular variety of it. Thus, policymakers need to consider that, aside from the commonly cited advantages and disadvantages of dollarization discussed above, the tradable sector will face a loss (gain) of competitiveness when the Fed increases (reduces) interest rates. 17 Given that Cambodia´s exports to the EU currently account for more than 40 percent of total exports while exports to the US market account for only about 25 percent, the current arrangement under dollarization may be hindering economic growth. For a small export-led economy such as Cambodia, already a lower middle-income economy which is becoming less dependent on the US market, this may be an excessive cost to pay. As is also discussed in the introduction (but not proven in this paper due to lack of available data), dollarization may also be behind rising prices in the non-tradable sector (and a potential construction bubble). To regain some monetary independence, the NBC has been adopting a series of measures to promote use of the national currency, including annually celebrating Khmer Riel Day to help increase public awareness, persuading public and private sector institutions to begin to pay salaries in KHR, and introducing Liquidity Providing Collateralized Operations (LPCO). This gradual market-based approach toward de- dollarization seems adequate to avoid shaking market confidence but is unlikely to reduce the levels of dollarization significantly (at least in the medium term). Policies aimed at promoting use of the national currency should take into account the causes of dollarization, as highlighted in the literature. In Cambodia, the crisis of confidence in monetary institutions experienced in the 1980s led to the widespread use of foreign currency (mostly USD), which has caused hysteresis, a status in which dollarization “self-feeds” (Menon, 2008; Samreth, 2011). In recent research, Samreth and Sok (2018) reconfirm this effect and further demonstrate that the currency substitution process toward foreign currency holdings when KHR depreciates is stronger than that toward KHR holdings when it appreciates. These factors suggest the need for administrative measures that directly regulate foreign currency use and holdings, as well as those that increase people’s incentives to hold KHR, such as introducing higher returns on its deposits. 14 In recent years, the NBC has already introduced differentiated reserve and liquidity requirements in USD and KHR, and the reasons why dollarization levels have not decreased significantly need to be studied in more detail to craft other suitable policies going forward. Prior to introducing further additional administrative/macroprudential measures, an in-depth assessment of their feasibility and adequacy is needed, since they could lead to financial disintermediation, reduce access to credit, and ultimately affect economic growth. Notably, in its pursuit of de-dollarization since the early 2000s, Peru introduced macroprudential measures such as provisioning requirements for foreign currency loans, 14 Samreth and Sok (2018) provide a more detailed discussion of this. 18 higher liquidity requirements for banking liabilities denominated in USD, and reductions in the maximum net open position in USD, 15 although not all of them may have been equally effective. 16 Cambodia has so far been able to achieve sustained macroeconomic stability under dollarization, a nominal anchor, and a prudent fiscal stance. However, as a lower middle- income economy, Cambodia may want to further build institutional trust and operating autonomy by articulating a medium-term macroeconomic policy framework and establishing a track record of delivering it. In particular, monetary authorities could consider introducing an inflation-targeting regime, as successfully done in Peru, where it helped reduce the exchange rate pass-through in the presence of dollarization (Armas and Grippa, 2005; Leiderman et al., 2006). This could be coupled with a progressive increase in exchange rate flexibility, which would enhance liquidity management and could increase the appeal of holding assets and liabilities in local currency, since they would be subject to less volatility than those in USD (Levy Yeyati and Ize, 2005). For this shift toward inflation targeting to be successful, institutional reforms aimed at significantly enhancing the autonomy, data collection ability, technical capacity, transparency, and accountability of the NBC, would first be needed. Finally, macroprudential measures and a clearer monetary policy framework need to be complemented by the development of capital markets in local currency. In particular, establishing a domestic bond market with titles denominated in KHR would provide investors with a more diversified set of investment opportunities, increase financing options for both local firms and the government, facilitate hedging against exchange rate swings, and decrease the cost of financing in local currency. In Peru, credit dollarization decreased significantly as private sector bond issuances in local currency increased (Garcia-Escribano, 2010). In summary, macroprudential measures per se may not be sufficient to de-dollarize, if not accompanied by a clear monetary policy framework (e.g., inflation targeting), institutional capacity and confidence building, and development of capital markets in local currency. 15 A company, financial institution, or individual has an open position in foreign currency when the liabilities in that currency exceed the assets, thus carrying foreign exchange risk. 16 See Garcia-Escribano (2010) for more detail. 19 6. Conclusion In conclusion, similar to Duma (2011) and over a longer time period, we find that US interest rate increases lead to higher lending rates in Cambodia, a reduced demand for money, and thus lower inflation and a smaller output gap. In addition, a novel result in this paper is that an increase of one standard deviation in the US federal fund rate (about 0.45 percent) has a negative impact on Cambodia’s trade balance with the EU (0.15 percent). It is also found that reserve levels decline significantly during the first two quarters following a US interest rate increase, as intervention by authorities helps mitigate exchange rate movements. The main policy implication of these findings is that, aside from the well-known advantages (price stability, discourages excess money printing) and disadvantages (lack of independent monetary and exchange rate policy, seigniorage loss, no lender of last resort in the event of a crisis) derived from dollarization, policymakers need to factor in that the tradable sector will face a loss of competitiveness when the Fed increases interest rates. Cambodia, a small export-led economy, is paying an important price, since its lack of independent monetary and exchange rate policy may be disfavoring the manufacturing sector while indirectly incentivizing investment in construction and other non-tradable sectors. At the same time, this is resulting in increased macroeconomic risks. As discussed earlier, to successfully pursue de-dollarization, Cambodian authorities need to not only introduce further macroprudential measures but also introduce and implement a clear monetary policy framework (e.g., inflation targeting), build institutional capacity and confidence in monetary institutions, and develop capital markets featuring instruments denominated in local currency. Areas for further research that were beyond the scope of this paper include estimating the net cost in terms of export losses Cambodia may have suffered so far, as well as assessing the channels through which US monetary policy decisions would affect FDI and credit flowing into the tradable and non-tradable sectors of the economy. The findings would be relevant not only to Cambodia but also to any other highly dollarized economies, and the policy implications would be pertinent for their policymakers. 20 Appendix A: Cambodia’s export function with respect to the EU A1. Estimation equation and estimation method In our estimation of the VAR-based impulse response, the results indicate that a positive US policy interest rate shock worsens Cambodian trade balances with the EU. As a robustness check of this negative impact, an estimation of Cambodia’s export function with respect to the EU is conducted. A simple function of Cambodian exports with respect to the EU can be expressed as follows: = ( , ), (A1) where EXEU, YEU, and ffr are Cambodian real exports to the EU, the EU’s real income, and the US federal fund rate, respectively. E represents a function. Notably, instead of the real exchange rate between Cambodia and the EU zone, ffr is used in equation (A1). This choice is motivated by the fact that Cambodia is a highly dollarized country where the economic transactions are widely conducted in USD. Hence, a change in the currency exchange rate between the US and the EU, possibly influenced by a change in the US policy interest rate (i.e. the US federal fund rate, ffr), can affect Cambodian exports to the EU, justifying the inclusion of ffr in Equation (A1). The following functional form of equation (A1) is considered: ln = 0 + 1 ln + 2 + 3 + , (A2) where is an error term, and ln and index t represent natural logarithm and time, respectively. Dummy variable, d, is added in the estimation equation to capture the macroeconomic effects resulting from political instability during the violent conflict in 1997; Asian financial crisis during 1997-1998; election periods in 1998, 2003, 2008, and 2013; and the global financial crisis during 2008-2009. As in VAR estimation, d is 1 for 1997Q3-1998Q4, 2003Q3-2003Q4, 2008Q3-2009Q4, and 2013Q3-2013Q4, and it is zero for other periods. For the estimation methodology, the autoregressive distributed lag (ARDL) approach to cointegration proposed by Pesaran et al. (2001) is adopted. 17 An error correction representation of the ARDL model of equation (A2) can be written as follows: 17 For more explanation of the ARDL approach to cointegration and its advantages, see Pesaran et al. (2001). 21 1 2 3 ∆ ln = 0 + � 1 ∆ ln − + � 2 ∆ ln − + � 3 ∆− =1 =0 =0 + 1 ln −1 + 2 ln −1 + 3 −1 + 4 + (A3) where is an error term, and ∆ denotes difference. The error correction representation of the selected ARDL model can be constructed by replacing the part of the lag of level variables in equation (A3) with an error correction term (EC). A2. Data and estimation results Quarterly data on Cambodian exports to the EU and the US federal fund rate are obtained from the DOTS and the IFS databases published by the IMF, respectively. The EU’s GDP is used for its income, and its real value is calculated by using the GDP deflator. Both GDP and GDP deflator data for the EU are taken from the Eurostat database. The sample period of the estimation is from 1995Q1 to 2016Q4, based on data availability. In the ARDL approach to cointegration, a statistical test to check the existence of the cointegration or long-run relationship among variables needs to be implemented in advance. Specifically, the F-test is used to test the null hypothesis of no cointegration, H0: λ1=λ2=λ3=0, against its null hypothesis, H0: λ1≠0, λ2≠0, λ3≠0. The computed F-statistic from this test is compared with its critical value available in Pesaran et al. (2001) for determining whether or not the null hypothesis should be rejected. Although the computed statistic does not support the existence of the cointegration relationship, its values varies significantly with different lag lengths. Following Bahmani-Oskooee and Nasir (2004) and Okada and Samreth (2013), the statistical significance of the error correction (EC) term in the next estimation step may be examined to evaluate the existence of the cointegration relationship. Equation (A3) in which the part of the lag of level variables is replaced by an error correction term (EC) is estimated, using the ARDL approach to cointegration. The maximum lag order is set to four for this estimation. 18 Following Narayan (2004) and Okada and Samreth (2013), Schwarz Bayesian information criterion (SBIC) is used for selecting the optimal lag order. Based on SBIC, ARDL(1,0,0) is obtained. Table A1 presents the estimation results of the error correction model. From the table, the coefficient of the error correction term (ECt-1) is statistically significant and negative. Its 18 Setting maximum lag order to four is common in an estimation using quarterly data. 22 absolute value is smaller than one, supporting the existence of the cointegration relationship among the variables under consideration. Table A1: Estimation results of the error correction representation for the selected ARDL model (dependent variable: Δ ln ) Variable Coefficient Standard Error Δ ln 0.0071 0.1294 Δ(%) -0.0242** 0.0134 Δ -0.0879 0.0528 ΔConstant 1.5567 3.4694 −1 -0.0865*** 0.0412 �2 0.0255 = ln − 17.9960 − 0.0816 ln + 0.2794 + 1.0159 The asterisk *** and ** indicates the 1% and 5% significance levels, respectively. Table A2 shows the long-run or cointegration results. The table indicates the statistical significances of the coefficients of the US federal fund rate, ffr, and the dummy variable, d, at conventional significance levels. The coefficient of ffr is -0.2794, implying that a one percentage point increase in the US federal fund rate leads to a 28 percent decrease in Cambodian exports to the EU. This negative impact of the US policy interest rate is in line with the impulse response result of our VAR-based estimation, which indicates that an increase in the US interest rate worsens Cambodia’s trade balance with the EU. However, the interpretation of this size impact of the US policy interest rate should be treated with caution, given our simple specification of the estimation equation. Table A2: Long-run estimation results (dependent variable: ln ) Variable Coefficient Standard Error ln 0.0816 1.4785 (%) -0.2794*** 0.1221 -1.0159** 0.6052 Constant 17.9960 43.0503 The asterisks *** and ** are 1% and 5% of significance levels, respectively. 23 Appendix B: VAR-based impulse response results for Malaysia and Singapore B1. Estimation framework and data For comparison, the impact of a US policy interest rate shock on two non-dollarized economies in the Southeast Asian region, Malaysia and Singapore, is examined. This is done by investigating VAR-based impulse responses of the variables under consideration with respect to a change in the US policy interest rate. Variables included in the VAR model are the same as the estimation for the Cambodian case. Specifically, they include the US federal fund rate (ffr), the growth rate of domestic currency to USD exchange rate (z), the trade balances with the EU (tbeugdp) and with the US (tbusagdp), money balance as a share of GDP (bmgdp), domestic interest rate (domint), output gap (gygap), and consumer price index (CPI) inflation (infl). Output gap is the difference between real GDP and real potential GDP (i.e. GDP under full employment), scaled by real potential GDP. Potential GDP in real values for the calculation of output gap data is generated by applying the Hodrick-Prescott filter on the real GDP data. Furthermore, the fluctuation of the crude oil price represented by its change rate (goilp) and a dummy variable (dummy) are exogenously included in the VAR system as a control for external shocks. The dummy variable is included to reflect effects resulting from the 1997-98 Asian financial crisis and the 2008-2009 global financial crisis. Like the Cambodian case, quarterly data are considered for the estimation. Due to the availability of data for domestic interest rate, the sample period for Malaysia is from 1994Q4 to 2015Q3 and for Singapore is from 1994Q4 to 2013Q2. The data for US federal fund rate, domestic currency to USD exchange rate, broad money (M2, a proxy for money balance), domestic interest rate, 19 GDP, and CPI are obtained from the International Financial Statistics (IFS) published by the International Monetary Fund (IMF). Data for exports and imports are obtained from the Direction of Trade Statistics (DOTS) published by the IMF. Data on crude oil price is taken from the Primary Commodity Prices published by the IMF. For the dummy variable, it is 1 for 1997Q3-1998Q4 and 2008Q3- 2009Q4, and zero elsewhere. Seasonal adjustments are applied on our quarterly data, except for the US federal fund rate and domestic interest rate, and CPI is used in converting nominal GDP into real GDP. 19 Domestic interest rate for both Malaysia and Singapore is the interest rate of government securities (i.e. treasury bills). Even if lending interest rate is used for both countries, we obtain similar impulse response results. 24 B2. Estimation results B2.1 Unit root tests The Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) based unit root tests for the variables under consideration are applied to identify their integration orders. Like the Cambodian case, for both Malaysia and Singapore, the unit root test results show that z, gygap, and infl are I(0) (i.e. integration order of 0), and ffr, tbeugdp, tbusagdp, bmgdp, and domint are I(1) (i.e. integration order of 1). These results can be provided upon request. B2.2 VAR-based impulse response results Since ffr, tbeugdp, tbusagdp, bmgdp, and domint are I(1), their first differences are included in the VAR estimation. For I(0) variables, z, gygap, and infl, their levels are included. The selected lag length for this VAR estimation is one for both Malaysia and Singapore, based on Akaike information criterion (AIC) and Schwarz information criterion (SIC). 20 VAR-based impulse responses for the variables under consideration are obtained by applying the Cholesky method. Variables in the VAR system are ordered from the exogenous one to the most endogenous one, following the Cambodian case. Specifically, the order is from D(ffr), to z, to D(tbeugdp), to D(tbusagdp), to D(bmgdp), to D(domint), to gygap, and to infl, where “D” indicates the first difference. Two exogenous variables are also considered. One is goilp, capturing the fluctuation of crude oil price, and another is a dummy variable, capturing impacts of the Asian financial crisis and global financial crisis. Impulse response results for the variables under consideration to a one standard deviation (S.D.) shock of the federal fund rate are shown in Figure B1 for Malaysia and Figure B2 for Singapore. As shown in Figure B1 and Figure B2, except for the trade balance with the EU, impulse response results for both Malaysia and Singapore are roughly as expected from a theoretical viewpoint and are consistent with the case of Cambodia. For trade balance with the EU, a one S.D. shock of the US federal fund rate temporarily improves it for both Malaysia and Singapore, in contrast to the result for Cambodia. This finding provides more evidence on the disadvantage of high dollarization in terms of Cambodia’s external competitiveness. 20 Both AIC and SIC indicate optimal lag length of one for Singapore. For Malaysia, while SIC always chooses optimal lag length of one, AIC indicates different optimal lag length, depending on maximum lags set for the estimation. Optimal lag length of one is applied for Malaysia. This also helps save the degree of freedom for the estimation, given our relatively small sample size for Malaysia. 25 Figure B1: Impulse responses for the case of Malaysia (“D” denotes the first difference) 26 Figure B2: Impulse responses for the case of Singapore (“D” denotes the first difference) 27 References Ahmed, F., Kim, Y., and Suphaphiphat, N. (2014). “What Can Go Wrong? Shifting Vulnerabilities as Asia Rebalances.” Chapter 5 in Cambodia Entering a New Phase of Growth. Ed. Unteroberdoerster, O. International Monetary Fund, Washington, Washington, D.C. Aizenman, J., (2013). “The Impossible Trinity — From The Policy Trilemma to the Policy Quadrilemma.” Global Journal of Economics 2, pp. 1-17. Armas, A., and Grippa, F. (2005). “Targeting Inflation in a Dollarized Economy: The Peruvian Experience.” IDB Publications (Working Papers) 1925. Inter-American Development Bank. Bahmani-Oskooee, M. and Nasir, A. (2004). “ARDL Approach to Test the Productivity Bias Hypothesis.” Review of Development Economics 8, pp. 483-488. Boz, E., Gopinath, G. and Plagborg-Moller, M. (2017). “Global Trade and the Dollar.” NBER Working Paper No. 23988. National Bureau of Economic Research. Cabral, R. (2010). “Why Dollarization Didn’t Succeed: Comparing Credibility and the Impact of Real Shocks on Small Open Economies.” North American Journal of Economics and Finance 21, 297-313. Calvo, G. and Reinhart, C. (2002). “Fear of Floating.” Quarterly Journal of Economics 117, pp. 379-408 Casas, C., Diez, F., Gopinath, G., and Gourinchas, P-O. (2017). “Dominant Currency Paradigm: A New Model for Small Open Economies.” IMF Working Paper No. 17/264. International Monetary Fund, Washington, D.C. Chang, R. and Velasco, A. (2002). “Dollarization: Analytical Issues.” NBER Working Paper No. 8838. National Bureau of Economic Research. Dornbusch, R., (1976). “The Theory of Flexible Exchange Rate Regimes and Macroeconomic Policy.” Scandinavian Journal of Economics 78(2), pp. 255-275. Duma, N. (2011). “Dollarization in Cambodia: Causes and Policy Implications.” IMF Working Paper No. 11/49. International Monetary Fund, Washington, D.C. Fleming, M. (1962). “Domestic Financial Policies under Fixed and under Floating Exchange Rates.” IMF Staff Papers 9, pp. 369-380. Frankel, J. (1999). “No Single Currency Regime is Right for All Countries or At All Times.” NBER Working Paper No. 7338. National Bureau of Economic Research. Frankel, J., Fajnzylber, E., Schmukler, S. and Serven, L. (2001). “Verifying Exchange Rate Regimes." Journal of Development Economics 66, pp. 351-386. Garcia-Escribano, M. (2010). “Peru: Drivers of De-dollarization.” IMF Working Paper No. 10/169. International Monetary Fund, Washington, D.C. Halvorsen, R. and Palmquist, R. (1980). “The Interpretation of Dummy Variables in 28 Semilogarithmic Equations.” American Economic Review 70, pp. 474-475. IMF (2017). “2017 Article IV Consultation.” IMF Country Report No. 17/325. International Monetary Fund, Washington, D.C. Ize, A. and Yeyati, E. L. (2005). “Financial De-Dollarization: Is it for Real?” IMF Working Paper No. 05/187. International Monetary Fund, Washington, D.C. Johansen, S. and Juselius, K. (1990). “Maximum Likelihood Estimation and Inference on cointegration: with applications to the demand for money.” Oxford Bulletin of Economics and Statistics 52, pp. 169-210 Kem, R. (2001). “Currency Substitution and Financial Sector Developments in Cambodia.” International and Development Economics Working Paper No. 01-4. Australian National University. Kojo, N. (2015). “Demystifying Dutch Disease.” Journal of International Commerce, Economics and Policy 6, pp. 1-23. Kose, A., Prasad, E., Rogoff, K. and Wei, S.-J. (2009). “Financial Globalization: A Reappraisal.” IMF Staff Papers 56, pp. 8-62. McKinnon, R. and Schnabl, G. (2004). “The East Asian Dollar Standard, Fear of Floating, and Original Sin.” Review of Development Economics 8, pp. 331-360. Menon, J. (2008). “Cambodia’s Persistent Dollarization: Causes and Policy Options.” ASEAN Economic Bulletin 25, pp. 228-237. Mundell, R. (1963). “Capital Mobility and Stabilization Policy under Fixed and Flexible Exchange Rates.” Canadian Journal of Economics and Political Science 29, pp. 475- 485. Mundell, R. (1964). “A Reply: Capital Mobility and Size.” Canadian Journal of Economics and Political Science 30, pp. 421-431. Narayan, P. (2004). “Fiji’s Tourism Demand: The ARDL Approach to Cointegration.” Tourism Economics 10, pp. 193-206. Nicoló, G. De, Honohan, P. and Ize, A. (2005). “Dollarization of Bank Deposits: Causes and Consequences.” Journal of Banking & Finance 29, pp. 1697-1727. Obstfeld, M. and Rogoff, K. (1995). “Exchange Rate Dynamics Redux.” Journal of Political Economy 103, pp. 624-660 Obstfeld, M. and Rogoff, K. (1996). Foundations of International Economics, Massachusetts: MIT Press. Pesaran, H., Shin, Y. and Smith, R. (2001). “Bounds Testing Approaches to the Analysis of Level Relationships.” Journal of Applied Econometrics 16, pp. 289-326. Ra, H.-R. (2008). “Modeling the Dollarization: A Case Study for Cambodia, Laos and Vietnam.” Global Economics Review 37, 157-169. 29 Reinhart C.M., Rogoff K. and Savastano, M.A. (2014). “Addicted to Dollars.” Annals of Economics and Finance 15, pp. 1-50. Samreth, S. (2010). “Currency Substitution and Seigniorage-Maximizing Inflation: The Case of Cambodia.” Applied Economics 42, pp. 1907-1916. Samreth, S. (2011). “An Empirical Study on the Hysteresis of Currency Substitution in Cambodia.” Journal of Asian Economics 22, pp. 518-527. Samreth, S. and Sok, P. (2018). “Revisiting the Impacts of Exchange Rate Movement on the Dollarization Process in Cambodia.” Paper presented at the 16th International Convention of the East Asian Economic Association held on October 27-28, 2018, Taiwan. Shi, K.and Xu, J. (2010). “Twin Dollarization and Exchange Rate Policy.” Journal of International Economics 81, pp. 109-121. Siregar, R. and Chan, N. (2014). “Factors behind Foreign Currency Holding by Household in Cambodia.” CAMA Working Paper No. 58/2014. Centre for Applied Macroeconomic Analysis, Crawford School of Public Policy, Australian National University. World Bank (2015). “Maintaining High Growth.” Cambodia Economic Update April 2015. World Bank, Phnom Penh. Zamaróczy, M., and Sa, S. (2002). “Macroeconomic Adjustment in a Highly Dollarized Economy: The Case of Cambodia.” IMF Working Paper No. 02/92. International Monetary Fund, Washington, D.C. 30