1
The impact of foreign direct investment, portfolio investment and other
investment on real effective exchange rates in East Asia and Pacific.
Abstract
This paper studies the effect of different capital flows on the real effective exchange rate. The three types of capital flows distinguished are foreign direct investment, portfolio investment and other
investment. An autoregressive distributed lag model has been used to study annual data from countries in East Asia and Pacific from the year 1982 to 2014. Total net capital flows, portfolio investment and other investment have a positive and significant impact on the REER in the short run,
while no long run impact is measured. Foreign direct investment appears to be unable to explain REER movements in either the short- or the long run.
JEL Classification: F21 International Investment; Long-Term Capital Movements, F31 Foreign Exchange Keywords: International Capital Flows, Real Exchange Rates, East Asia and Pacific
Name: Annemarije Santman Student number: 11111844 Programme: MSc ECO Date: 21 January 2017 Supervisor: Boe Thio
2
Statement of Originality
This document is written by Student Annemarije Santman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 1. Introduction
The real effective exchange rate (REER) can be seen as a measure of international competitiveness of a country. An appreciation of the REER is equivalent to a more expensive domestic currency of a country which could worsen the county’s international competitiveness. This could lead to lower demand for the country’s exports, which could stagnate economic growth. Large REER fluctuations thus increase uncertainty and are a threat to sustainable economic growth. A potential trigger for REER fluctuations are capital flows. Extensive research has been done on the causal effect of aggregate capital flows on the REER, but relatively few papers have differentiated among different types of capital flows in their empirical analysis. Capital controls implemented by policy makers are generally based on aggregate capital flows (Combes et al, 2011), while a number of studies (Combes et al, 2011; Wiboonchutikala et al, 2011; Bakardzhieva et al, 2010; Shen & Wang, 2001; Ahlquist, 2006; Athukorala & Rajapatirana, 2003) provide evidence that macroeconomic impact differs among capital flows.
This paper will add to economic understanding of the impact of different capital flows on the REER. The three categories that are most often distinguished and that will be used in this paper as well are foreign direct investment (FDI), portfolio investment and other investment. These capital flows combined with reserve assets comprise the financial account of the balance of payments. Based on existing literature it is expected that short-run effects are similar for the three types of capital flows as all indicate an increase in demand for the domestic currency which would thus lead to an appreciation. Long run effects however are expected to differ significantly as a result of the different drivers of the different capital flows which could thus be accompanied by different macroeconomic implications (Moore & Pentecost 2006). As there have not been many studies on the causal effect of different types of capital flows on the REER and as results of the studies that have been done on this relationship feature different results there is room for further research on this topic.
According to the World Bank Group, the region East Asia and Pacific has been, and still is, one of the main drivers of economic growth in the global economy. Capital inflows are generally large in fast growing economies as fast growth is generally accompanied by opportunities to gain high rewards on investment. These capital inflows encourage further development of the recipient country and could give economic growth a further boost. A negative side effect is loss of international competitiveness. This could reduce demand for tradable goods and thus limit growth. The region East Asia and Pacific has experienced large fluctuations in capital flows making it an interesting region to study. Large capital flows have been pointed out as main cause of the Asian crisis that erupted in 1995 by for instance Wiboonchutikala et al (2011), Khan et al (2005), Khan (2004) and Radelet & Sachs (1998). Does this mean that large capital flows are a threat to economic stability? Based on research by for instance Combes et al (2010), Bakardzhieva et al (2011) and Athukorala & Rajapatirana (2003) it
4
can be said that this is highly dependent on the composition of these capital flows. The impact of the composition of capital flows on the REER is what will be studied in this paper, leading to the research question: how do foreign direct investment, portfolio investment and other investment affect the REER in East Asia and Pacific? The method that will be used for this study is the fundamentals approach, which will be explained in section 2.1 and 3.1.
The structure of this paper will be as follows. The next section presents an overview of literature related to the research question in this paper. In section three the empirical model will be specified and in section four the methodology is explained. Chapter five provides an overview of the data used and the results will be discussed in chapter six. Implications of the results and limitations of this paper will be discussed in section seven and finally a conclusion is given in section eight.
2. Literature
2.1 Capital flows and the REER
Capital inflows lead to an appreciation of the domestic currency and, by the definition used in this paper, an increase of the REER. This argument is generally accepted in the field of economics and is supported by several macroeconomic models, such as the Dornbusch (1976) and Frenkel & Rodrigues (1982) models of exchange rate overshooting and the Portfolio Balance Model. An increase of capital inflows can be seen as an increase in demand for the domestic currency and thus the currency appreciates. This is however only the short-run effect, the long run effect may be different as other factors play an important role as well. Moore & Pentecost (2006) argue that if the nominal currency appreciation reduces international competitiveness and leads to a lower export level, this may cause the domestic price level to fall which would return the real exchange rate to its old level. On the other hand they argue that if the capital inflows were induced by financial liberalization, the liberalized interest rates could be accompanied by liberalized prices which could actually lead to an increase in the domestic price level and thus a further increase of the real exchange rate. Financial liberalization refers to the elimination of restrictions on financial markets and -institutions, which stimulates economic growth and improves the investment climate.
FDI, portfolio investment and other capital flows have different determinants. Dunning’s eclectic OLI theory for instance states that a multinational enterprise will engage in FDI if it faces ownership-, location-, and internalization advantages by doing so (Faeth, 2009). Portfolio investment on the other hand is mainly triggered by return on equity and is much more volatile (Taylor & Sarno 1997). Furthermore, John Ahlquist (2006) states that while political stability is an important FDI determinant, portfolio investment is more sensitive to the government’s economic policy. Moore & Pentecost (2006) pointed out that different drivers of capital flows can lead to different behaviour of
5
the real exchange rate, which supports the hypothesis that different capital flows could indeed have significantly different effects on the real exchange rate.
An approach to explain REER movements that has been used by for instance Clark & MacDonald (1998), Athukorala & Rajapatirana (2003), Combes et al (2011) and Bakardzhieva et al (2010) is the fundamentals approach. This approach estimates the equilibrium REER as a function of a country’s macroeconomic fundamentals. Combes et al (2011) use the fundamentals approach to estimate the effect of different capital flows on the REER in 42 developing countries. They found that portfolio investment has a much more severe impact on the REER than direct investment and that direct investment in turn had a greater impact than bank loans. They argue that FDI is mostly directed towards productivity increases and is a relatively stable capital flow, explaining its minimal impact on the REER. Portfolio investment on the other hand is more volatile and speculative, which they argue is a trigger of macroeconomic instability. Hence, portfolio investment leads to relatively large REER fluctuations. Combes et al (2011) use the pooled mean group estimator by Pesaran et al (1999) to estimate an ARDL model, with the assumption that long run effects of the capital flows and fundamentals on the REER are homogeneous. Their results provide evidence that this is indeed the case. A similar approach will be used in this paper to determine the extent to which REER movements can be explained by different capital flows in the region East Asia and Pacific.
Bakardzhieva et al (2010) studied the impact of different capital flows, aid and remittances on the REER of 57 countries across six different regions, including South East Asia, using the Generalised Method of Moments (GMM) estimator. Their results show a significant positive relationship between net capital flows and the REER in most regions, with the exception of Central and Eastern Europe. Furthermore, the authors found evidence for a significant positive relationship between FDI and the REER in Africa. However, no significant impact of FDI on the REER was found for the other regions. This leads the authors to conclude that FDI does not affect competitiveness of a country. The causal relationship between portfolio investment and the REER appears to differ per region. It is significant and positive in South East Asia, Latin America, the Gulf Cooperation Council and Central and Eastern Europe, but is significant and negative in the Middle East and North Africa while it is not significant at all in Africa. The argument provided by Bakardzhieva et al (2010) to explain the insignificant impact of portfolio investment on the REER is that the level of portfolio investment in the region is low. The negative relationship in the Middle East and North Africa is explained by the relatively young capital markets in the countries included in the studies. They explain that portfolio investment in the region mainly comprises the privatisation of public enterprises, which closely resembles FDI.
Athukorala & Rajapatirana (2003) performed a comparative analysis of capital flows in Latin America and Asia and found that the impact of total capital flows on the real exchange rate (RER) in Latin America was much larger than in Asia. The method used by the authors is a two stage least
6
squares instrumental analysis, using macroeconomic indicators as control variables. They found that capital flows other than FDI where jointly determined with the RER and used a range of instruments to correct for this endogeneity problem. Athukorala & Rajapatirana (2003) found that non-FDI capital flows lead to a RER appreciation, while FDI leads to a RER depreciation. They argue that FDI is mainly directed towards export oriented production, limiting pressure on prices in the non-traded goods sector compared to other types of capital flows and thereby leading to a depreciation of the RER. Capital flows in Asia comprise a larger share of FDI than capital flows in Latin America, which explains that capital flows in Asia have less impact on the RER than those in Latin America.
Moore & Pentecost (2006) used a vector auto regression (VAR) model to find the determinants of the real exchange rate appreciation after the liberalization of financial markets in 1997 in India. In VAR models endogenous variables are specified with respect to their lags, in order to avoid endogeneity. Moore & Pentecost (2006) focused on the difference between real and nominal shocks and found that real shocks were the main determinants of fluctuations in the real interest rate. This refers to the second case mentioned earlier, where the increase in the real exchange rate due to the nominal currency appreciation is offset by a decrease in the domestic price level. These results show that when looking at the effect of capital flows on the real exchange rate, the determinants of these capital flows have a significant effect on the long run real exchange rate changes.
Shen & Wang (2001) distinguish between FDI, portfolio investment and other investment in their empirical analysis where they look at the impact of central bank policy on the real exchange rate. Although the focus of their study lies on the role of the central bank and the implications of different inflation regimes, their results are interesting to look at for the purposes of this paper. Shen & Wang (2001) use Tsay’s arranged auto regression to study the effect of capital flows on the exchange rate under different inflation regimes by the central banks in selected Asian economies, looking at actual inflation. They found significant differences in the effects of FDI, portfolio investment and other investment within and among countries and under different inflation regimes. However, they do not elaborate on the causes and relationships of these differences as the aim of their study is to determine the impact of intervention of the central bank on real exchange rates.
Wiboonchutikala et al (2011) study the effect of net capital flows on several macroeconomic indicators including the REER in Thailand. In their study they also recognize that different capital flows might have different effects, so they also run a regression in which they distinguish between FDI, portfolio investment in equity, portfolio investment in debt investments, long-term loans, and short-term loans. Like Moore & Pentecost (2006) the authors use a VAR model to deal with endogeneity. They use the Akaike information criterion to determine the optimal lag length, which is twelve periods for their model. They study the effect of the different capital flows per gross domestic product (GDP) on the nominal and the real bilateral exchange rate, the real effective exchange rate, equity prices,
7
housing prices, domestic foreign exchange reserves, money supply, and the inflation rate. Each of these variables is incorporated in the model one at a time in order to estimate the effect of each of the capital flows on the selected macroeconomic indicators separately. The short run effects of FDI and portfolio investment in equity on the real exchange rate are similar, both lead to an immediate appreciation. In the long run however, the real exchange rate stabilizes after FDI inflows while it fluctuates afters portfolio investment in equity inflows. Portfolio debt inflows trigger a short-run depreciation after which the currency appreciates with persistence to future periods. An increase in both long- and short-term loans induces a steady and persistent appreciation of the currency. By using several dependent variables and entering them in the model one by one, the model of Wiboonchutikala et al (2011) lacks control variables. The effects estimated by the authors might thus be unreliable due to omitted variable bias. Singling out one dependent variable and using the others as control variables might generate a more accurate estimation.
The papers discussed in this section show that the impact of capital flows on the REER differs per region or country. Studies by Shen & Wang (2001), Combes et al (2011), Bakardzhieva et al (2010) and Athukorala & Rajapatirana (2003) distinguish capital flows similar to the ones distinguished in this paper, but provide deviating results and conclusions. For instance, Shen & Wang (2001) and Combes et al (2011) find a positive relationship between all types of capital flows and the REER. Wiboonchutikala et al (2011) find that both FDI and portfolio investment lead to an immediate appreciation of the REER, but that the REER returns to its old level in the case of FDI and portfolio investment in equity and leads to a long run depreciation in the case of portfolio investment in debt. Bakardzhieva et al (2010) compare the relationship between capital flows and the REER in six different regions and find that while a certain capital flow may lead to a REER appreciation in one region, it may lead to a depreciation in another region. Athukorala & Rajapatirana (2003) compare the impact of capital flows on the RER in Latin America and Asia and also find the results to differ per region. Furthermore, they find FDI inflows to lead to a depreciation of the RER, while non-FDI flows lead to an appreciation. Although the methodology of Bakardzhieva et al (2010) is different from the methodology used in this paper, the authors study the impact of capital flows on the REER in several regions including South East Asia. They use the same group of East Asian countries studied in this paper and use annual data over a similar timespan making it reasonable to expect the results found in this study to be in line with the results found by Bakardzhieva et al (2010). Total capital flows and portfolio investment are thus expected to have a positive and significant relationship with the REER, which implies that net inflows of total capital or portfolio investment lead to an appreciation of the REER. FDI is expected to have an insignificant impact on the REER. Bakardzhieva et al (2010) define other investment as debt and find a positive significant impact on the REER. This relationship is also expected
8
for other investment in this paper, implying that net inflows of other investment lead to an appreciation of the REER.
2.2 The Asian crisis
The period taken into account in this study includes the Asian Crisis. As a result of rapid economic growth and financial liberalization in the early nineties many Asian economies experienced large capital inflows at the time. The Asian crisis erupted in Thailand when economic growth slowed down in 1995. Roll-over of loans was largely discontinued and extensive capital outflows were experienced. This led to the collapse of the peg of the Thai Baht to the US Dollar in 1997 when the Thai government was forced to devaluate the Baht (Khan et al, 2005). The value of foreign debt increased significantly as a result of the devaluation which triggered the Thai currency and financial crisis. The crisis spread to other Asian countries in a similar fashion and led to the Asian crisis. Large capital inflows followed by large capital outflows have been named as a prime cause for the Asian crisis by for instance Wiboonchutikala et al (2011), Khan et al (2005), Khan (2004) and Sach & Radelet (1998). Wiboonchutikala et al (2011) explain that until the early nineties the only capital flows that entered Thailand were related to direct investment and these inflows increased at a substantial pace in accordance with an improvement of the terms of trade. As a result of this prosperity the Thai government decided to allow for other types of investment as well and due to the attractive investment climate the country experienced large surges of capital inflows. These capital inflow surges led to a real appreciation of the currency which had a negative effect on the terms of trade leading to a current account deficit of eight percent. A substantial share of the capital inflows was made up out of short terms loans and portfolio investment, two types of capital flows that can change of owner quickly and are thus easily reversible, Wiboonchutikala et al (2011) explain. This is exactly what happened in 1997 when the large current account deficit was accompanied by increased inflation. Investors could easily refuse to roll-over loans and direct portfolio investment out of Asia towards other economies. These volatile types of capital flows are thus accompanied by much greater macroeconomic risk than foreign direct investment that is generally more difficult to reverse. This leads to the conclusion that if a government were to impose capital controls it would be important to distinguish between long term and short term types of investment. A distinction that is also emphasized by Desai et al (2006), although they mention that this is difficult to implement.
Not all research papers distinguish between different capital flows in determining the causes of the Asian crisis. Khan et al (2005) explain that the peg of Asian currencies to the US Dollar that was appreciated at the time, led to an appreciation of these currencies as well that negatively affected the export levels leading to large current account deficits. The result were over valuated currencies which led to a collapse of the pegs to the US Dollar. While the large outflows of short term investment did
9
play an important role in the Asian crisis, the cause was the peg of several Asian currencies to a single major currency like the US Dollar Khan et al (2005) explain. In order to prevent a similar situation from occurring again capital controls would be unnecessary, as long as the exchange rate would be able to adjust in accordance to the domestic market and not be bound by a single large economy’s behaviour. As the Asian crisis is assumed to have impacted REER movements significantly, a dummy variable is included in this research to prevent biased results. The dummy will take a value of one if the time period falls within the Asian Crisis and zero otherwise. Irregular behaviour of the REER during the crisis will then be captured by the dummy variable, without distorting the relationship of the REER with the control- and determining variables.
3. Model specification
3.1 Fundamentals approach
The long run relationship between the REER and the macroeconomic fundamentals is specified as follows:
𝑅𝐸𝐸𝑅𝑖𝑡 = 𝜃0+ 𝜃1𝑇𝑂𝑇𝑖𝑡+ 𝜃2𝑃𝑅𝑂𝐷𝑖𝑡+ 𝜃3𝑇𝑅𝐴𝐷𝐸𝑖𝑡+ 𝜃4𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖𝑡+ 𝜈𝑖𝑡 (1)
𝑖 = 1,2, … , 𝑁 ; 𝑡 = 1,2, … , 𝑇
The different countries are denoted by i and t indicates the time period. 𝑅𝐸𝐸𝑅𝑖𝑡 represents the real
effective exchange rate, 𝑇𝑂𝑇𝑖𝑡 the terms of trade, 𝑃𝑅𝑂𝐷𝑖𝑡 the productivity gap, 𝑇𝑅𝐴𝐷𝐸𝑖𝑡 trade
openness, and 𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖𝑡 the aggregate capital flows. The 𝜃s represent the long-run coefficients of
the fundamentals. The derivation of this model will be explained in the methodology section. The choice to use the terms of trade, the productivity gap and trade openness as control variables has been made following the example of Combes et al (2011), as the empirical estimation method used in this paper closely resembles their method. The REER used in this paper is the weighted average of a country’s consumer price index (CPI) based real exchange rate and is defined as:
𝑅𝐸𝐸𝑅𝑖𝑡 = 𝑁𝐸𝐸𝑅𝑖𝑡 ∗ ∑ ( 𝐶𝑃𝐼𝑖𝑡 𝐶𝑃𝐼𝑗𝑡) 𝑤𝑗 𝑁 𝑗=1 (2) 𝑁𝐸𝐸𝑅𝑖𝑡 = ∑(𝑁𝐵𝐸𝑅𝑗𝑡) 𝑤𝑗 𝑁 𝑗=1
10
𝑅𝐸𝐸𝑅𝑖𝑡 is the real effective exchange rate of country i at time t, 𝑁𝐸𝐸𝑅𝑖𝑡 represents the nominal
effective exchange rate, 𝐶𝑃𝐼𝑖𝑡 and 𝐶𝑃𝐼𝑗𝑡 represent the consumer price index of country i and j
respectively, 𝑤𝑗 is the weight of country j which is based on the share of trade and 𝑁𝐵𝐸𝑅𝑗𝑡 is the
nominal bilateral exchange rate. An increase of the REER index reflects an appreciation of the domestic currency.
Terms of trade is an indicator of the value of exports as a percentage of the value of imports. A deterioration of the terms of trade affects the real exchange rate through the substitution- and income effect (Ostry, 1988). The substitution effect refers to the shift from expensive imported goods to less expensive non-traded goods which leads to an increase in demand for the domestic currency and thus an appreciation. The income effect on the other hand reflects that consumers can buy less of everything as a result of the terms of trade deterioration. This indicates a decrease in demand for the domestic currency and ceteris paribus this leads to a real depreciation. It is assumed that the income effect dominates the substitution effect which means that a deterioration of the terms of trade leads to a real depreciation of the domestic currency (Combes et al, 2011). A positive relationship is expected to be found between the terms of trade and the REER.
The productivity gap refers to the Harrod-Balassa-Samuelson effect. The effect is based on the assumption that productivity grows faster in the tradable sectors that in non-tradable sectors. The productivity increase in the tradable sectors results in higher wages in these sectors while relative prices do not change. Workers in non-tradable sectors will demand higher wages as well, but as these wage increases are not accompanied by a proportional productivity increase relative prices of goods in the non-tradable sector increase. The CPI increases relative to other countries and this is reflected by an appreciation of the REER (Lothian & Taylor, 2008) and thus, the productivity gap is expected to be positively related with the REER.
Trade openness reflects changes in a country’s trade policy. An increase of trade liberalization leads to an increase in the total amount of trade and is especially associated with an increased amount of imports at a lower relative price due to for instance abolition of tariffs. The lower price level is reflected in the consumer price index, which leads to a depreciation of the REER. This implies a negative relationship between trade openness and the REER. Trade openness is measured as the ratio of exports and imports to GDP.
Aggregate capital flows reflect the financial account of the balance of payments, excluding reserves. An increase of net capital flows generally results in an appreciation of the REER. Capital inflows are usually associated with an appreciation of the REER, as a result of increased demand for the currency (Calvo et al, 1993; Shen & Wang, 2001; Wiboonchutikala et al, 2011). A positive relationship between capital flows and the REER is thus expected to be found. The capital flows used in this research are in accordance with the financial account of the balance of payments. The capital
11
flows are recorded according to the sixth edition of the International Monetary Fund’s (IMF) balance of payments and international investment position manual (BPM6). All transactions between domestic entities and entities abroad involving financial assets are recorded on the financial account. The transactions are divided into three categories; foreign direct investment, portfolio investment and other investment. Reserves also make up part of the financial account, but for the purpose of this research they are not included in the model.
3.2 Composition of capital flows
The balance of payments is a method to record all international monetary transactions and theoretically it should be equal to zero. The balance of payments is divided into three categories; the current account, capital account and financial account. The current account records the inflows and outflows of goods and services, the capital account records international capital transactions of non-financial assets and the non-financial account records transactions of non-financial assets and liabilities that take place between domestic residents and foreigners. The capital flows distinguished in this paper are FDI, portfolio investment and other investment. As the data used in this paper has been extracted from the Balance of Payments Statistics database of the IMF, the definitions of the different capital flows will be in accordance with definitions used in BPM6. The sum of these capital flows comprises the financial account of the balance of payments, excluding reserves.
3.2.1 Foreign direct investment
Foreign direct investment is defined as investment in which a foreign firm establishes a subsidiary in a foreign country or acquires a significant interest of a foreign firm with control over management. This control can be established via immediate direct investment where at least ten percent of the voting power is obtained or via indirect direct investment where voting power is obtained in an enterprise that has voting power in another enterprise.
3.2.2 Portfolio investment
Portfolio investment is defined as investment in debt or equity securities across countries, other than the transactions included in foreign direct investment or reserve assets. Securities are assets or liabilities that are negotiable and change owner relatively easy. Acquisition of stocks of an enterprise worth less than ten percent of the total company value would for instance be reported as portfolio investment. Equity in another form than securities however, is not included in portfolio investment but is recorded as foreign direct investment or other investment.
12
3.2.3 Other investment
Other investment comprises of financial positions or transactions that are not included in foreign direct investment, portfolio investment and financial derivatives and employee stock options. Financial derivatives and employee stock options are recognised as a separate category of the financial account by the IMF, but due to the low volume of these capital flows they have been added to other investment in this paper. The residual transactions are categorised into seven categories: other equity in other forms than securities; currency and deposits; loans; nonlife insurance technical reserves, life insurance and annuities entitlements, pension entitlements and provisions for calls under standardized guarantees; trade credit and advances; other accounts receivable or payable and special drawing rights (SDR) allocations.
4. Methodology
Three approaches are commonly used to estimate dynamic panel data models (Combes et al, 2011). The first is the mean group (MG) estimator approach that averages the coefficients of parameters based on the individual coefficients of each cross-section unit. Pesaran & Smith (1995) showed that this method gives consistent estimates, however, homogeneity between the different units is not considered. The second approach assumes that variables in different cross-section units are homogeneous and estimates the coefficients by pooling the data. Some models that use this approach are the dynamic fixed effects model and the random effects model (Hill et al, 2008, pp. 391-406). Pesaran et al (1999) introduced a third intermediate approach based on the mean group estimator and the dynamic fixed effects approach called the pooled mean group estimation method for dynamic heterogeneous models which allows variables to be heterogeneous in the short-run, while it assumes long-term coefficients to be identical for each cross-section unit. This is the model that has been used by Combes et al (2011) to determine the effect of different capital flows on the REER. For all three estimation methods the model will take the form of an autoregressive distributed lag (ARDL) (p,q) model in order to deal with endogeneity and the possibility of stationary and non-stationary variables in a single equation.
This section describes the model as it has been developed by Combes et al (2011). Movements of the REER and macroeconomic fundamentals are assumed to differ across countries in the short-run, while they converge in the long-run. The pooled mean group estimation allows for the REER and REER determinants’ behaviour to be heterogeneous across countries in the short-run, while the coefficients are assumed to converge to a homogeneous equilibrium in the long-run. Whether this is indeed the case can be verified by performing a Hausman test which compares the results of the pooled mean group estimator with results of the mean group estimator. For the dynamic fixed effects estimator as
13
well as the pooled mean group estimator and the mean group estimator the model takes the form of an ARDL (p,q) model: 𝛥𝑦𝑖𝑡 = 𝛼𝑖𝑦𝑖,𝑡−1+ 𝛽𝑖𝑥𝑖,𝑡−1+ ∑ 𝜆𝑖𝑗𝛥𝑦𝑖,𝑡−𝑗 𝑝−1 𝑗=1 + ∑ 𝛿𝑖𝑗𝛥𝑥𝑖,𝑡−𝑗+ 𝜇𝑖+ 𝜀𝑖𝑡 𝑞−1 𝑗=0 (3)
𝑦𝑖𝑡 is the dependent variable, 𝑥𝑖𝑡 is the matrix of regressors, 𝜇𝑖 represents the fixed effects, 𝛼𝑖 is the
coefficient on the lagged dependent variable, 𝛽𝑖 is the vector of coefficients of the lagged regressors,
𝜆𝑖𝑗 the vector of coefficients of the lagged first differences of the dependent variable, 𝛿𝑖𝑗 the vector of
coefficients of the lagged first differences of the regressors. 𝑃 and 𝑞 represent the number of lags of the dependent variable and the explanatory variables respectively. The error term, 𝜀𝑖𝑡, is assumed to
be normally and independently distributed across i and t with a mean equal to zero, variances greater than zero and finite fourth moments. Under the assumption that 𝛼𝑖 is smaller than zero, indicating
that the model is stable as shocks die out over time, a long term relationship between 𝑦𝑖𝑡 and 𝑥𝑖𝑡 exist
that looks like:
𝑦𝑖𝑡 = 𝜃𝑖𝑥𝑖𝑡+ 𝜂𝑖𝑡 (4)
𝜃𝑖= −
𝛽𝑖
𝛼𝑖
𝜃𝑖 represents the long-run relationship between the regressors and the dependent variable. The error
term, 𝜂𝑖𝑡 is assumed to be stationary. Based on equation (4), equation (1) can be rewritten for the
long-run as: 𝛥𝑦𝑖𝑡= 𝛼𝑖𝜂𝑖,𝑡−1+ ∑ 𝜆𝑖𝑗𝛥𝑦𝑖,𝑡−𝑗+ ∑ 𝛿𝑖𝑗𝛥𝑥𝑖,𝑡−𝑗+ 𝜇𝑖+ 𝜀𝑖𝑡 𝑞−1 𝑗=0 𝑝−1 𝑗=1 (5)
𝜂𝑖,𝑡−1 has been estimated using equation (4), and 𝛼𝑖 represents the speed at which the dependent
variable converges to its long-run equilibrium. The three estimators use the maximum likelihood technique to derive the parameters. The parameters are defined as 𝛼̂ , 𝛽𝑖 ̂ , 𝜆𝑖 ̂ , 𝛿𝑖𝑗 ̂ , and 𝜃𝑖𝑗 ̂ , the 𝑖
estimators are defined as:
𝛼̂ =∑ 𝛼̂𝑖 𝑁 𝑖=1 𝑁 , 𝛽̂ = ∑𝑁𝑖=1𝛽̂𝑖 𝑁 , 𝜃̂ = ∑𝑁𝑖=1𝜃̂𝑖 𝑁 (6) 𝜆̂𝑗= ∑𝑁𝑖=1𝜆̂𝑖𝑗 𝑁 , 𝑗 = 1, … , 𝑝 − 1, 𝛿̂𝑗 = ∑𝑁𝑖=1𝛿̂𝑖𝑗 𝑁 , 𝑗 = 0, … , 𝑞 − 1 (7) This system of equations is used to establish the long-run relationship between the macroeconomic fundamentals and the REER given in equation (1). The long run coefficients are derived by taking the
14
average of the coefficients for each of the cross-section units. In case of the fixed effects estimator and the pooled mean group estimator, homogeneity among countries occurs for all- or the long term parameters respectively and thus the weighted average of these parameters is equal to each of the parameters itself (∑𝑁𝑖=1𝛼̂𝑖
𝑁 = 𝛼̂𝑖 = 𝛼̂).
Before constructing the model, the variables are tested on stationarity and cointegration and the number of lags to be included is determined. A variable is stationary if it has a constant mean and variance over time and if the covariance between two values depends only on the time between the observation moments, but not on the moment of observation (Hill et al, 2008). If any of these conditions is violated a variable is said to have a unit root. Multiple unit root tests should be performed for the results to be reliable. Many tests can be used to determine whether a variable contains a unit root, the tests used in this paper are the Im et al (2003) and Levin et al (2002) tests for a model with constant, but without trend. The following test is performed:
𝛥𝑦𝑡 = 𝛼 + 𝛾𝑦𝑡−1+ 𝜈𝑡 (8)
𝐻0: 𝛾 = 0
𝐻1: 𝛾 < 0
To test the hypothesis, equation (8) is estimated by ordinary least squares (OLS) and the τ-statistic for the hypothesis that 𝛾 = 0 is analysed. The τ-statistic is the t-statistic with an adjusted distribution to take into account changes in the variance is the null hypothesis is true and the variable is nonstationary.
Stationary variables and nonstationary variables that are integrated of order 1, namely I(1), can both be used in the ARDL model, but have to be treated differently. A variable is I(1) if its level values are nonstationary, but its first differences are stationary. If nonstationary variables are treated as stationary in a regression the estimated coefficients of unrelated variables could appear to be significant which is called a spurious regression. An exception to this problem are nonstationary variables that are cointegrated which implies that the variables share a similar stochastic trend. Cointegration can be tested by testing the least square residuals (𝑒̂𝑡 = 𝑦𝑡− 𝑏1− 𝑏2𝑥𝑡) for a unit root.
An endogeneity problem is expected to occur with FDI, portfolio investment and other capital flows as determining variables of the REER. These capital flows do not only affect the real exchange rate but they are also affected by changes in the real exchange rate causing reverse causality which would lead to biased results if not dealt with properly. The ARDL model corrects for endogeneity, including reverse causality, by taking lags of the different variables. When determining the optimal lag length for a model a trade-off is made between adding more information to the model and losing
15
efficiency. The lag length in most existing literature where the REER is estimated lies between zero and two (Combes et al 2011, Clark & MacDonald 1998). The lag length to be used in this paper will therefore be determined using the Akaike Information Criterion with a maximum lag length of two.
Next a separate ARDL(p, q, q, q, q) for each country must be run in order to establish whether a long-run relationship between the REER and the macroeconomic fundamentals, including the capital flows, exists. For this to be the case it is required that 𝛼𝑖≠ 0. Next, the estimation can be performed.
Under the assumption that the variables are I(1) and cointegrated the ARDL(p, q, q, q, q) model looks like: 𝑅𝐸𝐸𝑅𝑖𝑡 = 𝜇𝑖𝑡 + 𝛿10𝑇𝑂𝑇𝑖𝑡+ ∑ 𝛿11 𝑞 𝑗=1 𝑇𝑂𝑇𝑖,𝑡−𝑗+ 𝛿20𝑃𝑅𝑂𝐷𝑖𝑡+ ∑ 𝛿21 𝑞 𝑗=1 𝑃𝑅𝑂𝐷𝑖𝑡,𝑡−𝑗 (9) + 𝛿30𝑇𝑅𝐴𝐷𝐸𝑖𝑡+ ∑ 𝛿31 𝑞 𝑗=1 𝑇𝑅𝐴𝐷𝐸𝑖,𝑡−𝑗+ 𝛿40𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖𝑡 + ∑ 𝛿41 𝑞 𝑗=1 𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖,𝑡−𝑗+ 𝜀𝑖𝑡
And the equilibrium error correction representation of an ARDL (p, q, q, q, q) model for the relationship between the REER and the macroeconomic fundamentals is:
𝛥𝑅𝐸𝐸𝑅𝑖𝑡 = 𝛼𝑖[𝑅𝐸𝐸𝑅𝑖,𝑡−1− 𝜃0𝑖− 𝜃1𝑖𝑇𝑂𝑇𝑖,𝑡−1− 𝜃2𝑖𝑃𝑅𝑂𝐷𝑖,𝑡−1− 𝜃3𝑖𝑇𝑅𝐴𝐷𝐸𝑖,𝑡−1 (10) − 𝜃4𝑖𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖,𝑡−1] − ∑ 𝛿0,𝑗−1,𝑖 𝑝 𝑗=2 𝛥𝑗𝑅𝐸𝐸𝑅𝑖𝑡 − ∑ 𝛿1𝑗𝑖 𝑞 𝑗=1 𝛥𝑗𝑇𝑂𝑇𝑖𝑡 − ∑ 𝛿2𝑗𝑖 𝑞 𝑗=1 𝛥𝑗𝑃𝑅𝑂𝐷𝑖𝑡− ∑ 𝛿3𝑗𝑖 𝑞 𝑗=1 𝛥𝑗𝑇𝑅𝐴𝐷𝐸𝑖𝑡− ∑ 𝛿4𝑗𝑖 𝑞 𝑗=1 𝛥𝑗𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖𝑡+ 𝜀𝑖𝑡 With 𝜃0𝑖= 𝜇𝑖 1 − 𝜆𝑖 , 𝜃1𝑖 = ∑𝑞𝑗=0𝛿2𝑗𝑖 1 − 𝜆𝑖 , 𝜃2𝑖 = ∑𝑞𝑗=0𝛿3𝑗𝑖 1 − 𝜆𝑖 , 𝜃3𝑖= ∑𝑞𝑗=0𝛿4𝑗𝑖 1 − 𝜆𝑖 , 𝜃4𝑖= ∑𝑞𝑗=0𝛿5𝑗𝑖 1 − 𝜆𝑖 , 𝛼𝑖 = −(1 − 𝜆𝑖)
Where the 𝜃s and 𝛿s represent the long-run coefficients and 𝛼𝑖 the short-run coefficient. The error
correction component in this equation is 𝑅𝐸𝐸𝑅𝑖,𝑡−1. 𝐶𝐴𝑃𝐼𝑇𝐴𝐿𝑖𝑡 will be disaggregated into the before
mentioned three categories in order to determine the long-run relationship between the different capital flows and the REER.
16
5. Data
Annual data from 1983 to 2014 is used for seven countries in the region East Asia and Pacific: China, Indonesia, the Republic of Korea, Malaysia, the Philippines, Singapore and Thailand. This timespan and the countries have been selected based on data availability of different capital flows.
Data for the REER has been extracted from a Bruegel (2016) database. Data for the terms of trade, trade openness and the productivity gap have been extracted from the World Development Indicator Databank of the World Bank Group. The terms of trade have been calculated as the percentage of the export value index to the import value index. Trade openness has been defined as the sum of imports and exports as a share of GDP. The productivity gap is measured as domestic GDP per capita to the weighted average GDP per capita of the country’s ten largest trade partners. Data for the trade partners has been extracted from the Direction of Trade Statistics database of the IMF. The weights per country are included in Table 2 in the appendix. Data for the capital flows has been extracted from the Balance of Payments Statistics database of the IMF. In this research capital flows are weighted against GDP in order to make the flows in each country comparable to flows in other countries.
The data of the REER, terms of trade, trade openness and productivity feature kurtosis significantly different from the normal distribution kurtosis of three, as is shown in Table 3 and Figures 1 to 4 in the appendix. To impose normality the data is transformed by taking natural logarithms which will be used in the regression. The development of the log of these variables can be observed in Figures 5 to 8 in the appendix, which shows the average of each variable for the countries included in this study. Development of the average of the different capital flows is depicted in Figures 9 to 11 and development of average aggregate capital flows is shown in Figure 12. If the figures are positive, the region experiences net inflows and if the figures are negative the region experiences net outflows. Comparing development of the REER and aggregate capital flows shows that the REER features depreciation in periods where capital inflows are negative while it appreciates when capital inflows are positive. This is consistent with existing economic models (Dornbusch 1976, Frenkel and Rodrigues 1982). FDI and portfolio investment follow almost identical paths, which is not what would typically be expected. Portfolio investment is much more volatile than FDI and development of these flows would therefore be expected to feature more fluctuations than FDI flows.
6. Results
First the Im et al (2003) and Levin et al (2002) unit root tests are applied to the variables in order to verify that none of the variables are integrated of order 2 as this would invalidate the model. The results are depicted in Table 4 in the appendix and show that indeed all variables are either stationary or integrated of order 1. This confirms that an ARDL model is the appropriate model to use as this type
17
of econometric model features the ability to regress stationary and non-stationary variables in the same regression. A prerequisite is however, that the non-stationary variables are cointegrated. Running a regression with non-stationary variables that are not cointegrated could lead to a spurious regression where a significant causal relationship is estimated between variables that are in fact unrelated or vice versa. The four cointegration tests for panel data by Westerlund (2007) are used to determine whether the non-stationary variables the REER, terms of trade and productivity gap are cointegrated. The results of these tests are included in Table 5 of the appendix. Ga and Gt test the null hypothesis of no cointegration for each individual cross-section unit. A rejection of the null suggests that the variables are cointegrated for at least one of the cross-section units. The Pa and Pt tests test whether the variables for the panel as a whole are cointegrated. At a five percent level of significance the Gt and Pt tests reject the null hypothesis of no cointegration, while the Ga and Pa tests fail to reject. The evidence for cointegration is thus not very strong, however, following the example of Persyn & Westerlund (2008) and Fedeli (2012) the rejection of the null hypothesis by two of the four tests is interpreted as evidence for the existence of cointegration.
The optimal number of lags to use is estimated using the Akaike Information Criterion (AIC) with a maximum of two lags. The results are shown in Table 6 of the appendix. Figure 5 shows the development of the REER with a clear structural break from 1997 when the peg of the Thai Baht to the US Dollar collapsed to 2000. In order to capture the effect of the Asian crisis on the REER, the observation time span is divided into three periods: pre-crisis, crisis and post-crisis. The reasoning behind differentiating between pre- and post-crisis years is that before the crisis erupted the Asian currencies were pegged to the US Dollar, while this peg collapsed at the beginning of the crisis. Behaviour of the REER can therefore be expected to be different after the crisis compared to the pre-crisis years. Two dummies are included for the pre-pre-crisis and pre-crisis periods, the third is omitted to avoid multicollinearity. The pre-crisis dummy takes a value of one from 1982 to 1996 and zero otherwise, the crisis dummy takes a value of one from 1997 to 1999 and zero otherwise. In addition to these constant dummies, interactive dummies can be included if there are reasons to expect that one or more variable(s) will have affected the REER differently depending on the period. For instance, one of the channels through which capital flows affect an economy is via policy makers who may choose to mitigate fluctuations of the REER by adjusting monetary policy (Calvo et al, 1993). If policy makers would have altered their responses to capital flows after the crisis indicating that a capital flow of one pre-crisis would have a different effect on the REER than a capital flow of one post-crisis, ceteris paribus, this would be a reason to include interactive dummies in the regression. However, Wiboonchutikala et al (2011) explain that Asian policy makers have adjusted capital controls after the crisis which will have an effect on the volume of capital flows, but not on the way that these flows affect the REER. As the difference in exchange rate policy is captured by the pre- during- and post-crisis
18
dummies and there is no reason to expect such changes in the relationship between the REER and the other variables, no interactive dummies are used. This gives the ARDL (2,2,1,2,1,1,1) UECM model:
𝛥𝑅𝐸𝐸𝑅𝑖𝑡 = 𝛼𝑖(𝑅𝐸𝐸𝑅𝑖,𝑡−1− 𝜃0𝑖− 𝜃1𝑖𝑇𝑂𝑇𝑖,𝑡−1− 𝜃2𝑖𝑇𝑅𝐴𝐷𝐸𝑖,𝑡−1− 𝜃3𝑖𝑃𝑅𝑂𝐷𝑖,𝑡−1 (11)
− 𝜃4𝑖𝐹𝐷𝐼𝑖,𝑡−1− 𝜃5𝑖𝑃𝐼𝑖,𝑡−1− 𝜃6𝑖𝑂𝐼𝑖,𝑡−1) − 𝛿01𝑖𝛥2𝑅𝐸𝐸𝑅𝑖𝑡− 𝛿11𝑖𝛥𝑇𝑂𝑇𝑖𝑡
− 𝛿12𝑖𝛥2𝑇𝑂𝑇𝑖𝑡− 𝛿21𝑖𝛥𝑇𝑅𝐴𝐷𝐸𝑖𝑡− 𝛿31𝑖𝛥𝑃𝑅𝑂𝐷𝑖𝑡− 𝛿32𝑖𝛥2𝑃𝑅𝑂𝐷𝑖𝑡− 𝛿41𝑖𝛥𝐹𝐷𝐼𝑖𝑡
− 𝛿51𝑖𝛥𝑃𝐼𝑖𝑡− 𝛿61𝑖𝛥𝑂𝐼𝑖𝑡+ 𝛽1𝑝𝑟𝑒 + 𝛽2𝑐𝑟𝑖𝑠𝑖𝑠 + 𝜀𝑖𝑡
Estimating this model with the pooled mean group estimator, the mean group estimator and the dynamic fixed effects approach provides the results shown in Table 7 in the appendix. The first column of each estimator shows the results of the regression where total capital flows are recognised, while the regression in the second column distinguishes between the different capital flows. A Hausman test is used to determine which of the three approaches is the consistent one to use. The mean group estimator is assumed to be consistent in any case, so the results of the other two estimators are compared to the results of the mean group estimator. If the differences are significant, the mean group estimator should be used. If not, the pooled mean group estimator or dynamic fixed effects estimators are consistent. Contrary to Combes et al (2011), for both estimators the null hypothesis that the differences are insignificant is rejected at a five percent level of significance indicating that the mean group estimator is the best estimator to use for this regression. Both the long- and short-term effects are thus heterogeneous across the countries included in this sample. The adjustment speed is negative indicating that there is no omitted variable bias (Combes et al, 2011). The results of the mean group estimation are depicted in Table 1. The results indicate that total capital flows, portfolio investment and other investment have a positive impact on the REER in the short run, which makes sense. Net inflows of these types of investment trigger an immediate appreciation of the REER. FDI appears to be insignificant when looking at the impact on the REER in the short run. None of the capital flows appears to be able to explain REER movements in the long run. The terms of trade are insignificant in the short- and long run, however, the second lag appears to have a significant negative impact on the REER. Trade openness has a negative significant impact on the REER at a one percent level of significance in the short run, while this relationship is insignificant in the long run. The productivity gap has a positive significant effect on the REER in the short run, while the lag of the productivity gap is also significant but has a negative impact on the REER. In the long run the productivity gap affects the REER significantly negative. The Asian crisis has a significant negative impact on the REER in the regression where the different capital flows are distinguished, but appears to be insignificant if only total capital flows are taken into account. The difference between pre- and post-crisis estimation results appears to be insignificant.
19
Table 1: Composition of capital flows and the REER
1 2 Long run Terms of trade -6.208 (5.594) -0.274 (1.017) Trade openness -0.780 (1.047) -0.228 (0.441) Productivity gap -1.194 (2.095) -1.105* (0.641) Total capital 7.570 (11.187) FDI 0.923 (6.757) Portfolio investment -2.192 (2.561) Other investment -1.945 (2.549) Short run Ec -0.031 (0.085) -0.025 (0.066) D2 REER 0.402*** (0.059) 0.370*** (0.052) D terms of trade 0.048 (0.056) -0.054 (0.106) D2 terms of trade -0.075*** (0.027) -0.034 (0.043) D trade openness -0.179*** (0.042) -0.186*** (0.063) D productivity gap 0.544*** (0.093) 0.500*** (0.088) D2 productivity gap -0.254*** (0.075) -0.212*** (0.074) D total capital 0.330*** (0.118) D FDI -0.147 (0.733) D portfolio investment 0.394* (0.221) D other investment 0.284* (0.150) pre -0.003 (0.022) -0.016 (0.021 crisis -0.009 (0.016) -0.023** (0.010) Constant 0.227 (0.643) 0.035 (0.637) Observations 217 217
Standard errors in parentheses *p<0.10, **p<0.05, ***p<0.01
20
Table 8 in the appendix shows the results of the mean group estimation with results for each country where in the first column total capital flows are taken into account while the different capital flows are recognised in the second column. In the long run significant coefficients are neither found for the different capital flows nor for the total capital flows indicating long run movements of the REER cannot be explained by capital flows. In the short run the similar results are found for most of the countries. Only movements of the REER in the Republic of Korea can be partly explained by portfolio investment and the REERs in Malaysia and Thailand by FDI flows. Other investment appears to be insignificant for all countries. In addition to these results, it also appears that the Asian crisis has had no significant effect on the REER in any of the countries included in the study. The terms of trade cannot explain behaviour of the REER in any of the seven countries, trade openness only in China and Thailand and the productivity gap only in Thailand.
The results imply that the REER in most countries included in this study moves independently of capital flows and of the macroeconomic fundamentals used in this model. Implications of these results are discussed in the next section.
7. Discussion
The results per country in Table 8 show that FDI has a positive and significant impact on the REER in Malaysia and Thailand in the short run, while no significant relationship is found in the long run. This implies that FDI inflows lead to an immediate appreciation of the domestic currency, but that this appreciation is deferred in the long run. Potential explanations are that the appreciation of the local currency reduces the country’s international competitiveness which could trigger a decrease of exports (Moore & Pentecost, 2006) or that the capital inflows stimulate productive growth which lead to capital outflows by an increase of the level of capital goods imports (Bakardzhieva et al, 2010), both triggering the REER to depreciate back to its old level. This is consistent with the expectations formed at the beginning of this research. The other five countries included in the study and the region as a whole feature no causal relationship between FDI and the REER in either the short- or the long run. The absence of an immediate effect could be a result of FDI being directed towards productivity enhancement in the tradable goods sector. This would require import of capital goods, offsetting the appreciating effect of increased exports.
A positive and significant causal relationship of portfolio investment on the REER is found in the region East Asia and Pacific as a whole and in the Republic of Korea in the short run, but contrary to what was expected this relationship is insignificant in the long run. For none of the other six countries a significant causal relationship between portfolio investment and the REER is found. As these countries face a relatively high level of centralisation, it could be that portfolio investment is a result of financial liberalisation and is mainly driven by the privatisation of public companies. This type
21
of portfolio investment features similar behaviour as FDI (Bakardzhieva et al, 2010), explaining the absence of impact on the REER. The same arguments that were used to explain the diminishing of an initial currency appreciation as a result of FDI flows can be applied to portfolio investment as well.
The final type of capital flow recognised in this paper, other investment, appears to be insignificant both in the short- and the long run for all the individual countries studied in this paper. In the region East Asia and Pacific other investment has a positive and significant effect on the REER in the short run, but no significant impact in the long run. Similarly to portfolio investment, this is contrary to what was expected based on the results from Bakardzhieva et al (2010). This category comprises all capital flows not included in FDI and portfolio investment. If the major share of other investment would comprise financial transactions directed towards productivity enhancement, other investment shows similar behaviour as FDI. As was the case with portfolio investment, this can explain the insignificant relationship between other investment and the REER.
Total capital flows have a positive and significant impact on the REER in the region East Asia and Pacific in the short run. A one unit increase of total capital flows leads to a 0.33 unit increase of the REER. In the long run the impact of total capital flows appears to be insignificant. This is not what was expected, but as the long run impact of all the different capital flows on the REER is insignificant as well it makes sense that this also holds for total capital flows.
The Asian crisis has a significant negative impact on the REER in the regression where the distinction between the different capital flows is made, but is insignificant in the regression featuring total capital flows. The negative significant estimate implies that the crisis led to a depreciation of the REER which is a confirmation of what was observed in that period. The absence of a significant relationship between the crisis and the REER in the regression that only takes into account total capital flows suggests that the impact of the crisis is reflected in the net capital flows and control variables.
In the short run the terms of trade and trade openness both have a negative and significant impact on the REER, while the long run effects are insignificant. For both this implies that the substitution- and income effect neutralise one another in the long run. The first lag of the productivity gap is significant and positive in the short run and the second lag is significant and negative in the short run. The long run impact of the productivity gap is significant and negative, indicating that an increase in the productivity gap initially leads to an appreciation of the REER while it leads to a depreciation in the long run.
Some of the results found in this paper are surprising as studies by for instance Combes et al (2011), Wiboonchutikala et al (2011), Bakardzhieva et al (2010), Shen & Wang (2015) and Khan (2004) show that capital flows do have a significant impact on the REER in the long run. Combes et al (2011) take into account the exchange rate regime in their regression, while this variable has not been taken into account in this paper. This could explain the differences in results. The absence of a long run causal
22
relationship between capital flows and the REER found in this paper for the individual countries and for the region as a whole could also be related to the Asian region in specific. Athukorala & Rajapatirana (2003) studied the impact of FDI and other capital flows in Latin America and East Asia and found that the impact of other capital flows on the REER was much smaller in East Asia than in Latin America. Bakardzhieva et al (2010) study the impact of capital flows in, among other regions, South and East Asia and do not find a causal relationship between FDI and the REER. These results combined with the results found in this paper indicate that capital flows in the Asian region might have different implications in this region than in the rest of the world. This would be interesting to study in further research.
The results of this research should be interpreted with caution. A limitation of this research is for instance the relatively small dataset. Only seven countries have been taken into account, while the region East Asia and Pacific comprises of many other economies for which insufficient data was available for the purpose of this paper. The same holds for the relatively short time span of the data that was chosen based on data availability. This time span has been reduced further by the use of an ARDL model in which degrees of freedom are lost as a result of the use of lags. The ARDL model has been selected for its ability to deal with endogeneity and the use of I(1) and I(0) variables simultaneously. However, the loss of degrees of freedom for a dataset that was already small is a large disadvantage and should be considered when interpreting the results.
Another limitation of this paper lies in the transformation of the data. The natural logarithms are taken of the terms of trade, trade openness and productivity gap, but not of the different capital flows. The reason to take the natural logarithms of the control variables is clear; the distribution of all variables features skewness and/or kurtosis. The capital flows however are also featured with kurtosis, but this data is not transformed as some values of the capital flows are negative the natural logarithm cannot be taken of the original values. A possible solution would be to add a constant to all values equal to the highest negative value in order to make all values positive. However, this would alter the data significantly and could make the regression unreliable so it has been decided to follow the example of Wiboonchutikala et al (2011), Combes et al (2011) and Bakardzhieva et al (2010) not to transform the data of the capital flows even though the distribution is similar to the distribution of the other variables.
8. Conclusion
The purpose of this paper has been to study the effect of different capital flows on the REER in the region East Asia and Pacific. Capital control measures are often based on aggregate capital flows, while the composition of these flows might have a significant impact on how the capital flows affect the
23
REER (Combes et al, 2011; Bakardzhieva et al, 2010). Several studies (Wiboonchutikala et al, 2011; Shen and Wang, 2001; Ahlquist, 2006) have been performed on the effect of capital flows on the REER and a few have differentiated between different capital flows. The studies have had different research questions and have focused on different geographical regions, but they have in common that almost all have found a significant relationship between capital flows and the REER. The exception being Bakardzhieva et al (2010) who found no significant relationship between FDI and the REER.
The final dataset used for empirical analysis in this paper comprises of seven countries from the region East Asia & Pacific. An ARDL model is used to regress the REER on several macroeconomic fundamentals and the different capital flows following the example of Combes et al (2011). FDI appears to be insignificant when it comes to the explanation of REER movements in the short- as well as the long run. This implies that policy makers do not have to impose capital controls directed towards FDI, as this type of capital flow leads to sustainable growth and does not harm competitiveness of the country. Total capital flows, portfolio investment and other investment do have a significant positive effect on the REER in the short run, but appear to have no significant impact in the long run either. One could therefore argue that capital controls are unnecessary as the loss of competitiveness affects the country only in the short run while this effect dies out in the long run. However, evidence suggests that the Asian crisis has been caused by excessive capital flows (Wiboonchutikala et al, 2011; Khan et al, 2005; Khan, 2004; Sach & Radelet, 1998) so this argument is not very strong. Short run effects could have long run consequences if the shock is large enough, even if the direct impact appears to be insignificant in the long run. The Asian crisis erupted as a result of sudden large capital outflows, which are more likely to occur in the form of portfolio investment than FDI. Capital restrictions are therefore not redundant, but should be directed towards the more volatile capital flows. FDI should be encouraged as it leads to productivity enhancement which leads to sustainable growth.
For further research it would be interesting to study the causes for the absence of a long run impact of capital flows on the REER in the region East Asia and Pacific, as this relationship does appear to be present in other regions (Combes et al, 2011; Bakardzhieva et al, 2010; Athukorala & Rajapatirana, 2003). This paper focused on the impact of the different capital flows on the REER, but the explanation for the results has been based on existing literature and has not been studied for the region in specific. An interesting variable to take into account is the exchange rate regime of each of the Asian countries included in this study, as Combes et al (2011) find a significant negative relationship between flexible exchange rates and REER appreciation. Another study that would be interesting would be to look at portfolio investment in more detail. In some countries portfolio investment appears to be highly volatile with severe impact on the REER, while in other countries portfolio investment closely resembles FDI. This suggests that some types of portfolio investment contain a larger trade-off between economic growth and stability than others. Policy makers would be able to
24
impose capital controls more effectively if more knowledge would be available on the nature of portfolio investment.
9. References
Ahlquist, J.S. (2006) “Economic Policy, Institutions, and Capital Flows: Portfolio and Direct Investment Flows in Developing Countries” , International Studies Quarterly, Vol. 50, No. 3, pp. 681-704 Athukorala, P.C., Rajapatirana, S. (2003) “Capital Flows and the REER: A Comparative Study of Asia and
Latin America”, The World Economy, Vol 26, No. 4, pp. 613-637
Bakardzhieva, D., Ben Naceur, S. and Kamar, B. (2010) “The Impact of Capital and Foreign Exchange Flows on the Competitiveness of Developing Countries”, IMF Working Paper, Vol. 10, No. 154 Berument, H., Nergiz Dincer, N. (2004) “Do Capital Flows Improve Macroeconomic Performance in
Emerging Markets? The Turkish Experience”, Emerging Markets Finance & Trade, Vol. 40, No. 4, pp. 20-32
Branson, W.H. (1981) “Macroeconomic Determinants of Real Exchange Rates”, National Bureau of Economic Research Working Paper No. 801. Retrieved from http://www.nber.org /papers/w0801
Bruegel (2016) “Real effective exchange rates for 178 countries: a new database”, Retrieved from http://bruegel.org/publications/datasets/real-effective-exchange-rates-for-178-countries-a-new-database/
Calvo, G.A., Leiderman, L., Reinhart, C.M. (1993) "Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors", IMF Working Paper, Vol. 40, No. 1, pp. 108-151 Clark, P.B., & MacDonald, R. (1998) “Exchange rates and economic fundamentals: a methodological
comparison of BEERs and FEERS”, IMF Working Paper, Vol. 98, No. 67
Combes, J.L., Kinda, T., & Plane, P. (2011) “Capital Flows, Exchange Rate Flexibility, and the Real Exchange Rate”, IMF Working Paper, Vol. 11, No. 9
Corbo, V., Hernández, L. (1996) "Macroeconomic Adjustment to Capital Inflows: Lessons from Recent Latin American and East Asian Experience", The World Bank Research Observer, Vol. 11, No. 1, pp. 61-85
Desai, M.A., Foley, C.F., Hines, J.R. Jr. (2006) “Capital Controls, Liberalizations, and Foreign Direct Investment”, The Review of Financial Studies, Vol. 19, No. 4, pp. 1433-1464
Dornbusch, R. (1976) “Expectations and exchange rate dynamics”, Journal of political economy, Vol. 84, pp. 1161-1176
25
Edwards, S. (1989) “Real Echange Rates, Devaluation and Adjustment: Exchange Rate Policy in Developing Countries”, Cambridge, Massachusetts: MIT Press
Edwards, S. (1994) “Real and Monetary Determinants of Real Exchange Rate Behavior: Theory and Evidence from developing countries”, Published in Williamson, J. (1994), Chapter 4
Faeth, I. (2009) “Determinants of Foreign Direct Investment – A Tale of Nine Theoretical Models”, Journal of Economic Surveys, Vol. 23, No. 1, pp. 165-196
Fedeli, S. (2012) “The impact of GDP on health care expenditure: the case of Italy (1982-2009)”, Working paper: Dipartimento di Economia Pubblica
Frenkel, J.A., Rodriguez, C. (1982) “Exchange rate dynamics and the overshooting hypothesis”, IMF Working Paper, Vol. 29, pp. 1-30
Hill, R.C., Griffiths, W.E., & Lim, G.C. (2008) “Principles of econometrics”, Third edition, John Wiley & Sons, Inc
Im, K.S., Pesaran, M.H., Shin, Y. (2003) “Testing for Unit Roots in Heterogeneous Panels”, Journal of Econometrics, Vol 115, pp 53-74
International Monetary Fund (2009) “Balance of Payments and International Investment Position Manual”, Sixth edition, BPM6
Khan, S. (2004) “Contagious Asian Crisis: Bank Lending and Capital Inflows”, Journal of Economic Integration, Vol. 19, No. 3, pp. 519-535
Khan, S., Islam, F., Ahmed, S. (2005) “The Asian Crisis: An Economic Analysis of the Causes”, The Journal of Developing Areas, Vol. 39, No. 1, pp. 169-190
Levin, A., Lin, C.F., Chu, J.C.S. (2002) “Unit root tests in panel data: Asymptotic and finite-sample properties”, Journal of Econometrics, Vol 108, No 1, pp 1-24
Lothian, J.R., & Taylor, M.P. (2008) “Real Exchange Rates over the past Two Centuries: How Important is the Harrod-Balassa-Samuelson Effect?”, The Economic Journal, Vol. 118, No. 532, pp. 1742-1763
Moore, T., Pentecost, E.J. (2006) "The Sources of Real Exchange Rate Fluctuations in India", Indian Economic Review, New Series, Vol. 41, No. 1, pp. 9-23
Ostry, J.D. (1988) “The Balance of Trade, Terms of Trade, and Real Exchange Rate: An Intertemporal Optimizing Framework”, Palgrave Maxmillan Journals, on behalf of the IMF, Vol. 35, No. 4, pp. 541-573
Persyn, D., Westerlund, J. (2008) “Error-correction-based cointegration tests for panel data”, The Stata Journal, Vol 8, No. 2, pp. 232-241
Pesaran, M.H., Shin, Y., & Smith, R.P. (1999). “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels”, Journal of the American Statistical Association, Vol. 94, No. 446, pp. 621-634
26
Pesaran, M.H., Smith, R. (1995) “Estimating long-run relationships from dynamic heterogeneous panels”, Journal of econometrics, Vol. 68, No. 1, pp. 79-113
Radelet, S., Sachs, J. (1998) “The Onset of the East Asian Financial Crisis”, NBER Working Paper, No. 6680
Shen, C.H., Wang, L.R. (2001) “To Intervene or Not to Intervene: Exchange Rate Responses to Capital Flows in Selected Asian Economies”, ASEAN Economic Bulletin, Vol. 18, No. 1, pp. 63-82 Taylor, M.P., Sarno, L. (1997) “Capital Flows to Developing Countries: Long- and Short-Term
Determinants”, The World Bank Economic Review, Vol. 11, No. 3, pp. 451-470
Westerlund, J. (2007) “Testing for Error Correction in Panel Data”, Oxford Bulletin of Economics and Statistics, Vol 69, pp 709–748
Wiboonchutikula, P., Kotrajaras, P., Chaivichayachat, B. "An Analysis of Thailand's Net Capital Inflows Surges After the 1997 Crises", ASEAN Economic Bulletin, Vol.28, No.3 (December 2011), pp. 281-314
