The effect of exchange rate regimes on economic growth in developing countries Bachelor thesis Name: Tijmen de Graaf Student number: 10446273 Specialization: Economics & Finance Field: International Economics Number of credits thesis: 12 EC Title of your research: will either fixed or floating exchange rates lead middle income countries to higher economic growth in the long run? Supervisor: Lucyna Górnicka Abstract This paper examined the effect of exchange rate regimes on economic growth measured as annual change in GDP per capita. The exchange rate regimes are separated into a floating and a fixed currency. Research was conducted on middle income countries which are divided into two groups: upper and lower middle income countries based on their GNI per capita. Regression analysis was conducted based on the Solow growth model. First, the annual change in the U.S. dollar exchange rate and annual change in foreign currency reserves were researched. To investigate the influence of the Great Recession in 2007‐09, a crisis dummy was added. Thereafter, a dummy variable for fixed exchange rate regimes was created and, at last, the performances of lower and upper middle income countries were examined. The results did not show any statistically significant outcomes where either a fixed or floating exchange rate regime outperformed the other.
Table of contents Page 1. Introduction ……… 4 2. Exchange rate regimes ……… 5 2.1 Fixed exchange rate regime ……….. 5 2.2 Floating exchange rate regime ……… 6 2.3 Intermediate exchange rate regime ……… 8 3. Literature review ………... 8 3.1 Foreign direct investment ...……….. 8 3.2 Economic growth ………..……… 9 3.3 Price level ……… 10 3.4 Trade ……….………. 10 4. Empirical analysis ……… 11 4.1 Country identification ………. 11 4.2 Countries ………….……… 11 4.3 Variables .….……… 12 4.4 Panel data ……….. 15 5. Regressions and results ………..……….. 16 5.1 Regression on changes in exchange rate and foreign currency reserves ..……… 16 5.2 Regression with a crisis dummy ……….. 18 5.3 Regression with dummy variable ……….. 19 5.4 Validity threats ……… 22 6. Conclusion ……… 23 7. References ……… 24
Appendix 1 Countries in data set ………. 26 Appendix 2 Hausman tests ……… 29 Appendix 3 Robustness check: regression fixed versus floating ……….. 31
1. Introduction In 1947 the world set up the system of Bretton Woods to encourage international trade. This implied that every country adopted a fixed exchange rate against the U.S. dollar. Accordingly, to strengthen belief in the workings of the system, the U.S. dollar was pegged to gold. However, in the following years some renowned economists like Milton Friedman, Egon Sohmen and Harry Johnson promoted the use of a floating exchange rate regime (Pilbeam, 2013). In 1971 Bretton Woods collapsed because the U.S. dollar to gold peg became unsustainable. Since then, nations are free in adopting an exchange rate regime. Gosh et al. (2002) showed that in the period 1970‐1999 countries moved away from a pegged exchange rate. They examined a total of 167 IMF member countries based on which exchange rate regime they claim to have adopted. The percentage of floating exchange rate regimes increased from 4.3 percent of all countries in the 1970s to 27.0 percent in the 1990s. In the same interval, the percentage of nations with intermediate exchange rate regimes went from 11.0 percent to 26.4 percent, while countries with fixed exchange rate regimes declined from 84.8 percent to 46.6 percent. The 1990s were dominated by a series of crises in Mexico, Thailand, Russia and Brazil. One similarity of these crises was that each country had implemented a soft‐pegged exchange rate regime. This led economists to a bipolar view where only freely floating and fully fixed exchange rate regimes were feasible and intermediate exchange rate regimes should be abandoned (Fischer, 2001). In the beginning of the 21st century twelve countries established the Eurozone where each nation adopted the same currency, the euro. Even though the euro is left to float freely, the original currencies of the euro countries are fixed at a predetermined exchange rate. This can be seen as one currency pegged against a basket consisting of the other eleven currencies. The set‐up of the European Monetary Union (EMU) induced countries to remove the European trade barriers in order to allow increased capital movement. The adoption of a single currency also led to less exchange rate risk and transaction costs across the euro nations, but these benefits came at the cost of a loss of monetary independence. Now the national monetary policy is determined by the European Central Bank (ECB). This has led to major problems for countries regarding their government debt, most prominently Greece. There are many debates whether Greece should stay in the Eurozone or leave and implement a floating exchange rate regime, which would allow a devaluation of its original currency. One question that raises interest is whether or not one type of exchange rate leads to better economic performance in the long run. This is of considerable relevance for developing economies, because their choice affects their future economic growth. This paper examines the long run perspective of the economy as nations tend to stick to their choice for a long time and desire to have
perpetually well‐performing economies. Hence the research question of this paper is stated as follows: will either fixed or floating exchange rates lead middle income countries to higher economic growth in the long run? Part two of this paper examines the different exchange rate regimes and their advantages. Subsequently, there is a recapitulation on previous work on this topic. It focuses on exchange rate regimes and their effect on foreign direct investment, economic growth, price level and trade. The fourth section discusses which countries are selected in the data set, which variables are created and the rationale for using panel data. The fifth part contains the empirical research and its outcomes. At last, this paper ends with a conclusion that provides an answer to the research question. 2. Exchange rate regimes This section examines the reasons for choosing either a fixed or floating exchange rate regime. The final part discusses an intermediate exchange rate regime which combines elements of these extremes. 2.1 Fixed exchange rate regime Pilbeam (2013) identifies three arguments in favor of pegged exchange rates. Firstly, fixed exchange rates bring about more international trade and investment. A stable and fixed exchange rate leads to less uncertainty about the future value of the exchange rate and decreases the exchange rate risk. This induces risk‐averse economic agents to start trading as well. However, some caution is required, because when a country fixes its exchange rate against one or several countries it will only enjoy these benefits when trading among each other and not when trading with countries its currency still fluctuates to. Besides, a nation can only peg its nominal exchange rate and not its real exchange rate, which measures competitiveness across countries. Although flexible exchange rates have to cope with higher risk and more fluctuations, participants in the foreign exchange market can avoid these problems by using forward exchange rates. This allows economic agents to buy a certain amount of currency in the future at a predetermined exchange rate. However, if a risk premium exists, these forward contracts come at a cost. Secondly, fixed exchange rates avoid the conduct of reckless macroeconomic policies. Whereas a floating exchange rate is able to fluctuate freely, a pegged exchange rate has to be kept at the same level. When the economic circumstances change, the government has to intervene in the foreign exchange market to ensure the exchange rate remains fixed. This process works as follows. The authorities have no direct control over the demand side of the foreign exchange market, but they do have control over the supply side. The government can buy the domestic currency via open market operations by selling their foreign currency reserves. This reduces the supply of the national
currency in the foreign exchange market and increases the exchange rate. By selling its own currency the opposite mechanism will take place and the exchange rate decreases. If a country pursues irrational macroeconomic policies it takes a lot of effort to keep the exchange rate in line with the fixed rate. In case of a major increase in the money supply, the government has to buy its own currency by reducing its amount of foreign currency reserves. This can lead to the problem that a country runs out of foreign reserves and can no longer hold the exchange rate fixed. This can lead to a speculative attack where investors expect a nation to devaluate its currency and therefore sell the currency before it lowers in value. In the opposite case, the pressure for an appreciation induces a nation to buy foreign reserves. Large foreign reserves of one particular currency are vulnerable to negative economic shocks which will lower that currency’s value and thus the value of the reserves. A final argument put forward is that countries which peg their currency require more international coordination and cooperation. When two nations set up a fixed exchange rate they are both responsible for it. They have to avoid that the exchange rate peg comes under pressure and therefore have to align their policies with the policies of the country they pegged their currency to. This is due to the fact that the economies are mutually dependent. A devaluation of one currency leads to a revaluation of another currency and the other way around. Likewise, one country’s import is another country’s export and vice versa. This fosters the exchange of information between nations which reduces uncertainty and avoids conflicts. 2.2 Floating exchange rate regime Pilbeam (2013) provides four reasons to adopt a floating exchange rate regime. The first argument in favor of flexible exchange rates is that they equilibrate the balance of payments. The exchange rate adjusts in order to equal the supply of and demand for the currency. This equilibrating process works as follows. A balance of payments surplus induces an appreciation of the currency which makes domestic products more expensive for foreigners, resulting in a reduction of export. Import will increase because the higher value of the home currency makes foreign products cheaper. These events have a negative effect on the current account and hence the balance of payments will fall into equilibrium again. In case of a deficit on the balance of payments the reverse is true. Fixed exchange rates are not allowed to revaluate or devaluate and will subsequently face difficulties when the economic circumstances change. Hence, these countries have to buy or sell foreign currency reserves to bring the balance of payments into equilibrium. Another reason for choosing floating exchange rates is that countries have control over their own monetary policy. This is explained by means of the impossible trinity. In an open economy countries have three different instruments at their disposal. These equally desirable instruments include a fixed exchange rate, independent monetary policy and free movement of capital.
Independent monetary policy means that countries have control over the money supply which has direct influence on the inflation rate and, as a result of the Fisher equation, also on the interest rate. If there are no capital controls, money can flow to places where it is needed most. This will lead to efficient economies and more trade with the rest of the world. The issue of the trinity is that it is impossible to handle all instruments. A government can operate only two of the instruments at the same time. The rationale behind the trilemma is as follows. A fixed exchange rate regime requires a country to buy or sell foreign currency reserves in exchange for its national currency. In this way, it alters the money supply in order to ensure the exchange rate remains stable. The money supply, in turn, has to be in equilibrium with the demand for money. The interest rate will adjust the demand side until the equilibrium is restored. However, if the domestic interest rate differs from the world interest rate, capital will flow in or out of the country. These flows are directly related to the capital account in the balance of payments. The fact that the balance of payments has to be zero requires the buying or selling of foreign currency reserves in order to keep the exchange rate fixed. A nation is able to conduct independent monetary policy if it imposes capital controls. In this case it can change the interest rate without money crossing the border. If these capital restrictions are absent, countries have to make sure the balance of payments equilibrates by giving up independent monetary policy. Countries do not have to make a choice between these two instruments when they opt for a floating exchange rate. Thirdly, a floating exchange rate prevents a country from importing foreign inflation or deflation. Purchasing power parity (PPP) states that the exchange rate responds to differences in the price levels of the domestic and foreign nation. This means that if foreign prices go up, national prices must be raised to make sure that PPP holds, otherwise the fixed exchange rate will be violated. Due to the foreign price increase, the domestic goods are relatively cheap and this positively affects the balance of payments due to higher demand. This puts downward pressure on the exchange rate and induces the government to buy foreign reserves to avert an appreciation. The money stock will be expanded by this buying process and domestic prices will react upwards. The result is that the domestic country has imported the inflation from abroad. Nations with floating exchange rates isolate themselves from these foreign price shocks. At last, it may be easier to adjust the exchange rate than to adjust other economic variables. Exchange rates can go up or down immediately, whereas other variables, for example prices, only change in the long‐run and refuse to fall. This leads to more economic stability, because exchange rates can act as a shock absorber.
2.3 Intermediate exchange rate regimes Both fixed exchange rates and floating exchange rates are two extremes of a continuum. This means that countries can also adopt a hybrid version where they employ a combination of both sides. Williamson (2000) researched this so‐called intermediate exchange rate regime. There are certain advantages for countries that implement a band wherein the exchange rate can move. Contrary to a pegged exchange rate, it allows them to pursue a somewhat independent monetary policy by influencing the money supply and, in turn, the inflation rate. Secondly, Pilbeam (2013) notes that it allows a nation to align the exchange rate with the fundamental rate. A country can fix its exchange rate in accordance with the fundamental value, however when the economic conditions change misalignment will occur. Likewise, this can happen to floaters. Speculators may employ the wrong exchange rate determination model or misinterpret economic news spread by the government. This permits a country to intervene in the foreign exchange market and help realign the market exchange rate with its economic fundamental value. Motivation behind this is that the authorities have more information about their future path of policy, something the speculators lack. Frankel et al. (2000) point out that one of the disadvantages of an intermediate exchange rate regime is verifiability. It is more difficult to monitor whether the central bank keeps its promises. When an exchange rate deviates from its pegged value, economic agents know that the central bank did not comply with its announced policy of fixed exchange rates. In case of floating exchange rates, the market can check when the central bank intervened in the market by conducting open market operations to alter the foreign currency reserve level. Frankel et al. (2000) claim it is necessary to gather data over a long time period to verify whether the central bank sticks to its announced policy. 3. Literature review The question of the exchange rate regime matter is a much discussed topic in economic literature. Research was conducted on the influence of the exchange rate regime regarding four variables, namely its effect on foreign direct investment, economic growth, price levels and trade. 3.1 Foreign direct investment Aizenman (1992) studied the effects of exchange rate regimes on domestic investment and foreign direct investment (FDI). The importance of this subject originated from the increased internationalization of world markets after the Second World War. Another reason was that after the break‐up of the Bretton Woods system in 1971 countries adopted different exchange rate regimes. Aizenman (1992) differentiated between fixed and floating exchange rate regimes and his results showed that both domestic investment and foreign direct investment are higher under the former regime in cases of monetary and productivity shocks. Countries with a fixed exchange rate regime
are better able to insulate real wages and production from shocks in the money supply. Hence, they will have lower volatility and higher income than floaters. The latter induces higher investment, both domestic and foreign. In case of a productivity shock, a flexible exchange rate country will let its currency appreciate. However, a nation with a fixed exchange rate regime is unable to do that so it will raise employment and thereby increase productivity, domestic investment and FDI. Although Aizenman (1992) made no distinction between developed and developing nations, this is of particular interest for the latter, because they have lower marginal productivity. Hence, they will benefit from access to new technologies and knowledge which can lead to higher economic growth. Research by Abbott et al. (2012) focused on the choice of exchange rate regime and its effect on foreign direct investment in developing nations. They introduced a third, intermediate category of exchange rate regime. The conclusions of their work were in compliance with those of Aizenman (1992). They found that de facto fixed or intermediate exchange rates provoke higher foreign direct investment flows than flexible regimes. This is consistent with the fact that pegged exchange rates exhibit less exchange rate risk and less transaction costs. Thus, it is expected that pegging a currency will lead to more international trade and investment. 3.2 Economic growth Levy‐Yeyati and Sturzenegger (2003) analyzed the relationship between exchange rate regimes and economic growth for both developing and developed nations. Their work showed that for developing countries floating exchange rates are associated with higher output growth and less output volatility than fixed exchange rates. This result was absent for developed countries. These conclusions were contrary to previous literature which was ambiguous about the effect of a pegged currency on growth. Mundell (1997) argued that fixed exchange rates led to higher growth and lower volatility. As described before, pegging reduces uncertainty and exchange rate risk which will attract more foreign investment. In addition, a fixed exchange rate regime prevents outrageous monetary policy. Both phenomena promote economic growth and lower volatility. Conversely, Broda (2001) argued that if a pegged exchange rate regime was hit by a shock it was unable to modify either the exchange rate or the price level. The latter was not easily changed in the short run, because wages and product prices are rigid due to contracts. This meant that no adjustment mechanism was present and thus more output volatility and, in case of a negative shock, lower economic growth was provoked. Levy‐Yeyati and Sturzenegger (2003) concluded that a longer duration of shocks and vulnerability to speculative attacks were prevalent in fixed exchange rate countries. Even if nations adopted a high‐credibility peg, they did not perform better than floaters. A study by Husain et al. (2005) concluded that fixed exchange rates are favorable to poor countries with limited capital mobility. They argued that these countries imported low inflation and
high credibility from the country they were pegging to. However, if the capital controls would be loosened these nations would lose their ability to conduct fully independent monetary policy. When they grew and became more developed a switch to a floating exchange rate regime would benefit them. Developed countries had higher growth levels without higher inflation rates. There was no preferred exchange rate regime for developing countries, although nations with a pegged currency were more prone to exchange rate and banking crises. 3.3 Price level Broda (2006) researched the relationship between price levels and exchange rate regimes. The importance of this subject lies in the fact that prices influence the level of competitiveness of a country. Broda (2006) found that the price level of nations that adopted a fully flexible exchange rate is twenty percent lower than countries with a fixed currency. This made those countries more competitive and induced a rise in export which increased the level of GDP. The rationale behind this is that countries with fixed exchange rates import inflation from the nation it is pegged to. Due to inflation inertia the price level in nations with a fixed exchange rate regime is higher. 3.4 Trade One of the arguments for choosing a pegged exchange rate regime is its promotion of international trade. Klein and Shambaugh (2006) found empirical proof for this statement. They stated that trade increased between a developing country that adopted the fixed exchange rate and a developed nation which its currency is pegged to. This effect was absent between two industrialized countries. Frankel and Rose (2002) studied the same relationship, but they looked at currency unions instead of a single fixed currency. They proved that adoption of a common currency led to more international trade with the other participants of the union. Trade would increase by a factor three between members of the currency union. Furthermore, it did not impede trade with nonmember countries. Another finding was the fact that the increased trade was accompanied by rising income. A one percent increase in total trade led to a one‐third percent improvement of income per capita over the long term. Even though these results were based on currency unions which this paper does not take into consideration, the findings are useful. The similarity of currency unions with pegged exchange rate regimes is that both have a fixed exchange rate regime and lost their ability to conduct independent monetary policy. The difference is that nations within a currency union fix their exchange rate against a basket of countries, while other countries peg their currency against only one other currency.
4. Empirical analysis This section shows how empirical research was conducted, how the dataset was composed, which variables were used and the rationale for choosing these variables. The final section discusses the motivation for using panel data. 4.1 Country identification This paper takes only middle income countries into consideration. The rationale behind this is a research conducted by Rogoff et al. (2003). Their results showed that low income nations should opt for a fixed exchange rate regime. These countries had poorly performing financial institutions and had to cope with high inflation. A pegged currency would enable them to lower their inflation rate. In addition, these nations could control their own money supply, because capital was restricted from crossing the border. Contrarily, developed nations had well‐working financial institutions and money was free to flow in to or out of the country. Their credible national bank provided low inflation, so it was unnecessary to adopt a fixed exchange rate. Hence, a floating exchange rate was the preferred exchange rate regime for these countries. The class of middle income nations was in between these two sides. Middle income countries have great potential to develop themselves to high income nations by enhancing their financial institutions and opening up their borders even more for international trade. Economic growth is one measure of their progress. The choice for the right exchange rate regime may fasten this process. The rest of the world will benefit as well due to more global trade and less poverty. To identify which nations belong to the group of middle income countries, data was used from the Worldbank. This institution classifies nations according to their gross national income (GNI) per capita in dollars. Countries with a GNI per capita lower than 1035 dollar are viewed as low income. A GNI per capita higher than 12.616 dollar is classified as high‐income. Middle income countries have a GNI per capita between 1036 dollar and 12.615 dollar. The Worldbank also distinguishes between lower middle income and upper middle income nations where the latter starts at a GNI per capita of 4086 dollar. 4.2. Countries in data set For the empirical analysis panel data was used. Data was collected on 50 lower middle income countries and 55 upper middle income countries for the time period 1995 until 2013. This gave us a total of 1995 observations. The list of all 105 countries in the data set can be found in appendix 1.
4.3 Variables For all these countries data was collected concerning the GDP growth per capita, population growth, gross savings, technological progress, foreign currency reserves and the exchange rate. The Worldbank provided this information. This led to the creation of the following variables: Variable name Variable label GDPg GDP growth per capita, annual POPg population growth, annual Sav gross savings in % of GDP Tech annual change in scientific and technical journal articles published Rev foreign currency reserves in dollars Ex_dollar exchange rate domestic per dollar, annual average Ex_euro exchange rate domestic per euro, annual average Year year Country country Revloc foreign currency reserves in local currency Change_Rev annual change in foreign currency reserves Change_dollar annual change in exchange rate domestic per dollar Change_euro annual change in exchange rate domestic per euro UMI (dummy) 1=upper middle income countries 0=lower middle income countries FIX (dummy) 1=fixed exchange rate regime 0=floating exchange rate regime Crisis (dummy) 1=data from years 2007‐09 0=data from years 1995‐2013, except 2007‐09 ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ The rationale for choosing these variables is provided by the Solow growth model. Mankiw (2013) explains this model on the basis of the production function which states that GDP (Y) depends on the amount of capital (K), the amount of labor (L) and technological growth (E). Y = F(K, LE) (1)
The model states that the output per worker (y) is positively related to the capital per worker (k) and the level of technological growth (E). Y/L = F(K/L, E) (2) Factors that influence the capital per worker are investments (s), depreciation (δ), population growth (n) and technological progress (g). k = sy – (δ + n + g)k (3) Workers will either spend their money on consumption or they will invest it in new capital. While investments have a positive effect on capital per worker, the reverse is true for depreciation. When the population is growing the amount of capital has to be divided over a larger amount of people which reduces capital per worker. Finally, technological advancements induce the labor force to work more efficiently. Mankiw (2013) notes that this means that workers can produce more output and hence fewer workers are needed to produce the same amount of output. However, these redundant workers need to be supplied with new capital. Hence, technological progress negatively affects the capital per worker. The total effect of technological growth is ambiguous. In formula 2 there is a positive effect of technological progress to GDP per capita, whereas in formula 3 there is a negative correlation between technological development and capital per worker. GDP per capita (GDPg) is used to measure economic growth. Investments and savings (Sav) are in equilibrium on the financial market, so by measuring the latter it is possible to find out the former. Changes in the amount of people in a country are controlled by the population growth rate (POPg). Technological progress (Tech) is measured by annual change in the amount of scientific and technical journal articles published. The motivation for using this variable is that these articles contain information about new research so that other researchers can use it to proceed with it. Hence, the more articles published, the more technical development there will be. There is no variable present to measure depreciation. However, it is assumed that the depreciation rate changes over time due to technological advancements but does not change across countries, because the nations in the data set belong to the same class of development. Panel data allows us to make this assumption. To examine the influence of the exchange rate regime on the economic growth rate, two new variables were created. The first of those variables is the exchange rate written as domestic currency per U.S. dollar (Ex_dollar). This is a direct method to monitor if a country has either a fixed
or a floating exchange rate. However there is one pitfall here. Data is only available for exchange rates against the U.S. dollar. This means that a country can float its currency against the U.S. dollar, but can peg the exchange rate to another currency. To solve this issue, the exchange rate in domestic currency per euro (Ex_euro) was added. This value is calculated by dividing the domestic currency per U.S. dollar exchange rate by the euro per U.S. dollar exchange rate. Only these two large currencies were used, because most nations either peg its currency to one of these or they do not peg at all. However, use of this method to compute the domestic currency per euro exchange rate must be approached with some caution. It is an indirect way and thus may deviate by small amounts from the actual exchange rate. One other indicator of exchange rate regimes is the level of foreign reserve currency (Rev). As outlined in section 2, a country that adopted a fixed exchange rate regime has to buy or sell foreign currency to maintain the peg. A nation that let its currency float does not have to perform these transactions. Hence, its level of foreign currency reserves remains stable. The data collected provides the level of foreign currency reserves in U.S. dollars. However a change in value of reserves can have two meanings. Either the level of foreign currency reserves changed or the exchange rate to the U.S. dollar changed. To make sure that only the changes in foreign currency reserves are accounted for, the variable Revloc was created. By multiplying the foreign currency reserves by the exchange rate, the foreign currency reserves are measured in the local currency, just like they are measured in the balance of payments. This paper examines the effect of exchange rate regimes on economic growth in two ways. The first method is to look at the annual change in the exchange rates and the foreign currency reserves. Therefore the variables Change_Rev, Change_dollar and Change_euro were constructed. These variables measure the yearly differences in percentages for the level of foreign currency reserves, the domestic currency per U.S. dollar and domestic currency per euro exchange rates. It does not matter if these changes are positive or negative, that is why these variables were squared. However, the newly created variables cannot be properly interpreted. By taking the square root of all squared changes, it is possible to get the changes in percentages again without any negative values. The second method to examine the influence of exchange rates on economic growth is by creating dummy variables. The variables FIX_dollar and FIX_euro will only take the values one or zero. The variable FIX_dollar equals one if the yearly change in domestic currency per U.S. dollar is exactly zero and equals zero otherwise. The dummy FIX_euro equals one if the yearly change in domestic currency per euro is equal or smaller than 0.25 percent, and equals zero in every other case. By using an indirect method to compute the domestic currency per euro exchange rate none of the yearly changes in this exchange rate is exactly zero. To account for these mathematical errors, the value of 0.25 percent was chosen arbitrarily. The problem with this way of identifying which
nations adopted a fixed exchange rate is that floaters not necessarily need a high level of annual change in exchange rates. The reverse is true for a country with a fixed exchange rate which may have to change its exchange rate in bad economic circumstances. Besides, there are also nations with an intermediate exchange rate regime which combine characteristics of both exchange rate regimes. These two dummies were used to construct a new variable called FIX. This paper distinguishes between fixed and floating exchange rates and it does not matter if the exchange rate is either fixed against the U.S. dollar or the euro. This dummy equals one if the variables FIX_dollar and FIX_euro add to one or higher. If this is not the case, it will have a value of zero. This binary variable was computed for each country observed for each year of the time series. At last, to distinguish between lower middle income and upper middle income countries the variable UMI was created. This dummy variable equals one if the country belongs to the upper middle income class and zero otherwise. However, the boundaries used to divide countries in various classes based on level of GNI per capita were arbitrarily chosen. 4.4. Panel data This paper used panel data to examine the research question. Panel data combines cross‐ sectional data with time series data where the 105 middle income countries are the entities used and 1995 until 2013 are the years considered. Stock and Watson (2012) note that the advantage of using panel data is that changes in dependent variables of many countries can be seen over many years. Hence, there are more observations available. Another advantage is that panel data allows us to leave out omitted variables if they are constant over the years but change across the countries observed. This is called the fixed effects model. Stock and Watson (2012) state that the fixed effects model constructs a dummy variable for each nation in the data set to account for omitted variables. To avoid the dummy variable trap and the problem of perfect multicollinearity, one dummy is omitted. Hence, 104 binary variables were constructed. In the regression function this looks like this: γ1calbania + … + γ104cyemen where γ is the parameter for the binary variables for each nation and c is the country it corresponds to. In this way a different intercept for each country is created. This allows us to examine the pure effect of the exchange rate on economic growth, while controlling for all country‐specific effects. Stock and Watson (2012) note that one assumption of the fixed effects model is that the omitted variables can be correlated within a country but not with the other nations. On the other hand, the random effects model is a special case of the fixed effects model. It accounts for error terms that are correlated across countries. However, these error terms cannot be correlated with the regressors of the model. To decide which model suits the regression better a Hausman test can be
conducted. This is a chi‐squared test with a null hypothesis that the error terms are not correlated with the regressors. If the chi‐squared test has a probability higher than 0.05 the null hypothesis is rejected and the Hausman test recommends using the random effects model. The results of the Hausman tests for each of the regressions conducted in the following section can be found in appendix 2. A second test will be conducted to control for time fixed effects. The time fixed effects model controls for omitted variables that are constant across the nations observed, but change over the years. Depreciation is an example of such an omitted factor. It is part of the Solow growth model, but is difficult to measure. It is assumed that depreciation changes over the years due to technological improvements but is constant across the countries observed, because they belong to the same class of economic development. Just as the fixed effects model, the time fixed effects model constructs a dummy variable for each year. There are 19 years of data collected, but due to the dummy variable trap only 18 dummies were created. In the regression function this looks like this: δ1t1995 + … + δ18t2012 where δ is the parameter for the binary time variables and t is the year it corresponds to. The test for time fixed effects is an F‐test with a null hypothesis that the time coefficients are jointly zero. If the probability that F is lower than 0.05 the null hypothesis is rejected and the time fixed effects model should be employed. In this way, both country‐fixed and time fixed effects are controlled for. 5. Regression and results This section presents the regressions that were conducted, what the results were and how these were interpreted. First, the correlation between fluctuation of the U.S. dollar exchange rate and foreign currency reserves with economic growth was examined. To research the influence of the Great Recession from 2007 until 2009, a crisis dummy was constructed. Thereafter, binary variables were created to account for countries with a fully fixed exchange rate and to distinguish between lower and upper middle income countries. The last paragraph discusses validity threats. 5.1 Regression on changes in exchange rate and foreign currency reserves First, the effect of the annual change in the U.S. dollar exchange rate on economic growth was examined. No distinction between lower and middle income countries was made yet. The expectation is that there is a negative correlation between these two variables. The higher the annual change in the exchange rate, the more uncertainty there is regarding the future value of the exchange rate. This impedes trade and thus the economic growth per capita will be diminished. The annual change in the foreign currency reserves is predicted to be positively correlated with the GDP
growth per capita. The level of foreign currency reserves will fluctuate when a country has a fixed exchange rate. As said above, a fixed rate leads to higher economic growth. Hence, the more the foreign currency reserves fluctuate, the higher economic growth should be. The following regression was conducted:
GDPg = β0 + β2POPgct + β3Techct + β4Savct + β5Change_dollarct + β6Change_Revct + γ1calbania
+ … + γ104cyemen + δ1t1995 + … + δ18t2012 + uct (1) where GDPg is the annual change in GDP per capita, POPg is the annual change in population, Tech is the technological progress measured by the annual change in scientific and technical research articles published, Sav is gross domestic savings as a percentage of GDP, Change_dollar is the annual change in exchange rate with respect to the U.S. dollar, Change_Rev is the annual change in foreign currency reserves including gold, γ is the coefficient for the binary variables for each nation, c is the country it corresponds to, δ is the coefficient for the binary time variables, t is the year it corresponds to and u is an error term with mean zero and constant variance. The results of the country and time fixed effects using regression (1) are as follows: Country and time fixed effects GDPg Coef. Std. Err. t POPg ‐.4834669 .2590654 ‐1.87*** Sav .0934766 .015528 6.02* Tech ‐.0001831 .0012902 ‐0.14 Change_dollar ‐.0003894 .0017711 ‐0.22 Change_Rev ‐.0001795 .0006863 ‐0.26 _cons 1.863205 .637467 2.92*
*=1 percent significance level **=5 percent significance level ***=10 percent significance level The results demonstrate that both the population growth and savings are significant at the 10 and 1 percent level respectively. There is a negative relationship between the former and economic growth and a positive correlation between the latter and economic growth. Technological progress is also negatively related to economic growth, although, it is not significant at the 10 percent level. The latter is also the case for the parameters of the annual change in the U.S. dollar exchange rate and the foreign currency reserves. The fluctuation of the U.S. dollar exchange rate has a negative sign just as the fluctuation of the foreign currency reserves. The former outcome is in
compliance with what Mundell (1997) wrote. He stated that a fixed exchange rate regime reduces uncertainty and exchange rate risk which will attract more foreign investment. In addition, outrageous monetary policy is prevented. These effects promote economic growth. An argument in favor of the insignificance of the fluctuation in U.S. dollar exchange rate is that a country with a fixed exchange rate regime not necessarily has a low fluctuation. When a country is hit by a crisis it has to devalue its currency. This is exactly what happened during the Latin‐ American debt crisis in the 1990s. Argentina, Brazil, Mexico and Venezuela were countries that had to cope with these economical bad times and are also part of the data set used. A reason for the insignificance of the fluctuation of foreign currency reserves might be the data that was used. According to the literature the level of foreign currency reserves should be stable for countries with floating exchange rate regimes. However, the data shows this never happened to any of the countries in the timespan from 1995 until 2013. That means that countries change the level of foreign currency reserves irrespective of the exchange rate regime they use. 5.2 Regression with a crisis dummy Broda (2001) noted that a fixed exchange rate regime will be heavily impacted by an adverse economic shock. This is due to the fact that these countries are unable to alter either the exchange rate or the price level and this should lead to lower economic growth. Hence, the prediction is a higher value for the parameter Change_dollar and a lower value for the parameter Change_Rev. To see if the results are affected by the Great Recession, which took place from 2007 until 2009, a binary variable was created. This dummy will equal to one if the data concerns a crisis year and will be zero otherwise. The previous regression was altered to:
GDPg = β0 + β2POPgct + β3Techct + β4Savct + β5Change_dollarct + β6Change_Revct + β7Crisisct
+ γ1calbania + … + γ104cyemen + δ1t1995 + … + δ18t2012 + uct (2)
where GDPg is the annual change in GDP per capita, POPg is the annual change in population, Tech is the technological progress measured by the annual change in scientific and technical research articles published, Sav is gross domestic savings as a percentage of GDP, Change_dollar is the annual change in exchange rate with respect to the U.S. dollar, Change_Rev is the annual change in foreign currency reserves including gold, Crisis is a dummy which equals one if the data concerns the years 2007 until 2009, γ is the coefficient for the binary variables for each nation, c is the country it corresponds to, δ is the coefficient for the binary time variables, t is the year it corresponds to and u is an error term with mean zero and constant variance. The outcomes for the country and time fixed effects are as follows:
Country and time fixed effects GDPg Coef. Std. Err. t POPg ‐.4834669 .2590654 ‐1.87*** Sav .0934766 .015528 6.02* Tech ‐.0001831 .0012902 ‐0.14 Change_dollar ‐.0003894 .0017711 ‐0.22 Change_Rev ‐.0001795 .0006863 ‐0.26 Crisis ‐4.156288 .5953817 ‐6.98* _cons 1.863205 .637467 2.92*
*=1 percent significance level **=5 percent significance level ***=10 percent significance level The newly created variable Crisis is clearly negative and significant at the 1 percent level. This means that economic growth in the crisis years is lower than in normal years. However, the values and signs of the other parameters are exactly the same as in the previous regression. Hence, the parameters of interest, Change_dollar and Change_Rev, are not influenced by the crisis, which is in contrast with what Broda (2001) stated in his work. However, some caution is advised. It is possible that a country coped with a financial crisis in the years other than 2007 until 2009. This paper did not control for this possibility, but this would influence the results. 5.3 Regression with dummy for fixed exchange rate The previous section looked at the relationship between the fluctuation of the U.S. dollar exchange rate and economic growth. However the fluctuation of the exchange rate does not necessarily corresponds to a specific exchange rate regime. A country with a floating currency can still have very low fluctuation in its exchange rate when there are no economic problems and economic growth is constant. On the other hand, a country with a fixed exchange rate regime can experience a lot of fluctuation when it has to cope with a financial crisis and has to devalue its currency. Hence instead of looking at the fluctuation, this time a dummy variable FIX was created that equals to unity if the exchange rate is fixed to either the U.S. dollar or the euro. The dummy equals zero if a country let its exchange rate float against the other currencies. This section looks at the differences in performance of using a floating or fixed exchange rate regime. According to the same reasoning as in the previous paragraph, it is expected that fixed exchange rate regimes will lead to higher economic growth. The regression employed is:
GDPg = β0 + β2POPgct + β3Techct + β4Savct + β5FIXct + γ1calbania + … + γ104cyemen + δ1t1995 + … + δ18t2012 + uct (3) where GDPg is the annual change in GDP per capita, POPg is the annual change in population, Tech is the technological progress measured by the annual change in scientific and technical research articles published, Sav is gross domestic savings as a percentage of GDP, FIX is a dummy variable that equals one if a country has a fixed exchange rate regime, γ is the coefficient for the binary variables for each nation, c is the country it corresponds to, δ is the coefficient for the binary time variables, t is the year it corresponds to and u is an error term with mean zero and constant variance. The outcomes for the country and time fixed effects model are as follows: Country and time fixed effects GDPg Coef. Std. Err. t POPg ‐.4611707 .258989 ‐1.78*** Sav .0893245 .014944 5.98* Tech ‐.0001634 .0012878 ‐0.13 FIX .264292 .4245109 0.62 _cons 1.812372 .6415154 2.83*
*=1 percent significance level **=5 percent significance level ***=10 percent significance level The outcome looks the same as the previous regressions regarding the three variables from the Solow growth model. The variable FIX is positively related with the regressand just as predicted, which means that nations with a fixed exchange rate regime have a higher GDP growth rate per capita. However, it is not significant at the 1, 5 or the 10 percent level. The result is in accordance with the study by Klein and Shambaugh (2006). A developing country which fixed its exchange rate can raise its economic growth by increased trade with a developed nation where it pegged its currency to. This is exactly what the dummy FIX represents. In addition, to account for any effects of the Great Recession, another regression was conducted. This time a dummy variable Crisis was added, just like in the previous paragraph. The regression function looks like this:
GDPg = β0 + β2POPgct + β3Techct + β4Savct + β5FIXct + β6Crisisct + γ1calbania + … + γ104cyemen
+ δ1t1995 + … + δ18t2012 + uct (4)
where GDPg is the annual change in GDP per capita, POPg is the annual change in population, Tech is the technological progress measured by the annual change in scientific and technical research articles published, Sav is gross domestic savings as a percentage of GDP, FIX is a dummy variable that equals one if a country has a fixed exchange rate regime, Crisis is a dummy which equals one if the data concerns the years 2007 until 2009, γ is the coefficient for the binary variables for each nation, c is the country it corresponds to, δ is the coefficient for the binary time variables, t is the year it corresponds to and u is an error term with mean zero and constant variance. The results of regression (4) using the country and time fixed effects model are the following: Country and time fixed effects GDPg Coef. Std. Err. t POPg ‐.4611707 .258989 ‐1.78*** Sav .0893245 .014944 5.98* Tech ‐.0001634 .0012878 ‐0.13 FIX .264292 .4245109 0.62 Crisis ‐4.136638 .5892353 ‐7.02* _cons 1.812372 .6415154 2.83*
*=1 percent significance level **=5 percent significance level ***=10 percent significance level The outcome is not much different from regression (3). The signs and values of the parameters are not influenced. A fixed exchange rate regime leads to higher economic growth, although it is not significant at the 10 percent level. The Crisis variable shows that the economic growth of the middle income countries is lower from 2007 until 2009 than in the other years investigated. Furthermore, the binary variable UMI was created to see if there are any differences in the performances between upper middle income and lower middle income countries. The dummy will be equal to unity if the country belongs to the upper middle income group or zero otherwise. The regression looks like this:
GDPg = β0 + β2POPgct + β3Techct + β4Savct + β5UMIct + γ1calbania + … + γ104cyemen
+ δ1t1995 + … + δ18t2012 + uct (5)
where GDPg is the annual change in GDP per capita, POPg is the annual change in population, Tech is the technological progress measured by the annual change in scientific and technical research
articles published, Sav is gross domestic savings as a percentage of GDP, UMI is a dummy variable that equals one if it is a upper middle income country, γ is the coefficient for the binary variables for each nation, c is the country it corresponds to, δ is the coefficient for the binary time variables, t is the year it corresponds to and u is an error term with mean zero and constant variance. The regression uses the random effects model, because the fixed effects model omits the variable UMI from the regression due to collinearity, as can be seen in appendix 2. The results of the random effects model are: Random effects GDPg Coef. Std. Err. t POPg ‐.8375299 .1768817 ‐4.73* Sav .1004624 .0126182 7.96* Tech ‐.0001781 .001375 ‐0.13 UMI ‐.1244487 .4536353 ‐0.27 _cons 2.027195 .4913235 4.13*
*=1 percent significance level **=5 percent significance level ***=10 percent significance level The parameters of the population growth, savings and technological progress are negative, positive and negative respectively. The dummy variable for upper middle income countries is negative. That means that lower income countries have a higher GDP growth rate per capita. However, this binary variable is not statistically significant at the 10 percent level. Although the two groups have different stages of development, there are no differences in performances by separating countries into these classes. 5.4 Validity threats Stock and Watson (2012) state that there are four assumptions underlying the fixed effects model of panel data. First, there is no omitted variable bias. This implies that the mean of the error term has to be zero. That means that the error term cannot be correlated with the regressors. The use of panel data makes sure that any omitted variables can be left out if they are constant over the years but change across the countries observed or if they are constant across the nations but change over the years examined. The second assumption is that the variables are identically chosen and independent across the countries. The data of this paper was not collected by random sampling from a population, but it used yearly averages of all variables. However, some countries lacked data of one of the variables for some time periods. The variables examined are correlated with itself, because the
economic performance of one year is likely to affect the results in the next year. However, the variables are also interdependent across countries. If one country experiences an economic downturn it is likely to influence other countries which it trades with. Thirdly, large outliers are absent because they can make the regression results biased. There may be any outliers in the data. When the economic circumstances turn bad, the variables may greatly deviate from their normal values. This was especially true during the Great Recession from 2007 until 2009. That is why this paper accounted for this period. However, countries could have experienced an extreme outcome in one of the other years. The last assumption is no perfect multicollinearity. The statistical software used will deal with this by eliminating any variable that is perfectly multicollinear. The results of this paper have no external validity. It only counts for the middle income countries observed. High income countries are much more developed while low income countries are less developed. Hence, these countries have different characteristics and thus they can benefit from different exchange rate regimes. 6. Conclusion This paper examined the effect of exchange rate regimes on economic growth measured as annual change in GDP per capita. The exchange rate regimes are divided into a floating and a fixed currency to the U.S. dollar. Research was conducted on middle income countries which are classified into two groups: upper and lower middle income countries based on their GNI per capita. Regressions were constructed based on the Solow growth model which states that the annual change in GDP per capita depends on savings, technological progress and population growth. The first regression looked at the annual change in the U.S. dollar exchange rate and the annual change in foreign currency reserves. A negative relationship was found between both variables and economic growth. However, none of these outcomes turned out to be statistically significant. Therefore a binary variable was created that equaled one if the country had a fixed exchange rate and zero otherwise. The parameter of this dummy was a negative value but not significant at either the 5 percent or 10 percent level. To investigate if the outcomes were influenced by the Great Recession from 2007 until 2009 a crisis dummy was added. It turned out that the parameter of this variable was negative and statistically significant at the 1 percent level. At last, any differences between upper middle and lower middle income countries were examined, but the coefficient of the dummy variable was not statistically significant. There are two arguments that can explain the insignificance of the results. First, the group of middle income countries can be too broad. The nations at the top and the bottom of this class based on GNI per capita are too dissimilar. The top countries are at a different stage of development than the bottom countries. Second, this paper only controlled for the global crisis from 2007 until 2009.
However, there may be other bad economic circumstances in the remaining years that influenced the outcomes. Hence, there seems to be no single exchange rate regime that benefits middle income countries from higher economic growth in the long‐run. That means that the choice for either exchange rate regime depends on country‐specific factors. Further research can be conducted that accounts for these factors. 7. References Abbott, A., Cushman, D.O. & De Vita, G. (2012). Exchange Rate Regimes and Foreign Direct Investment Flows to Developing Countries. Review of International Economics, Vol. 20 (1), pp. 95‐107. Aizenman, J. (1992). Exchange Rate Flexibility, Volatility, and the Patterns of Domestic and Foreign Direct Investment. NBER working paper, no. 3953 Broda, C. (2001). Coping with Terms‐of‐trade Shocks: Pegs Versus Floats. American Economic Review, Vol. (91), pp. 376‐380 Broda, C. (2006). Exchange Rate Regimes and National Price Levels. Journal of International Economics, Vol. 70 (1), pp. 52‐81 Fisher, S. (2001). Exchange Rate Regimes: Is the Bipolar View Correct? Journal of Economic Perspectives, Vol. 15 (2), pp. 3‐24 Frankel, J. & Rose, A. (2002). An estimate of the Effect of Currencies on Trade and Income. Quarterly Journal of Economics, Vol. 117 (2), pp. 437‐466 Frankel, J., Schmukler, S. & Serven, L. (2000). Verifiability and Vanishing Intermediate Exchange Rate Regimes. NBER working paper, no. 7901 Ghosh, A., Gulde, A. & Wolf, H. (2002). Exchange Rate Regimes: Classification and Consequences. Center for Economic Performance Husain, A.M., Mody, A. & Rogoff, K.S. (2005). Exchange Rate Regime Durability and Performance in Developing versus Advanced Countries. Journal of Monetary Economics, Vol. 52 (1), pp. 35‐64
Klein, M.W. & Shambaugh, J.C. (2006). Fixed Exchange Rates and Trade. Journal of International Economics, Vol. 70 (2), pp. 359‐383 Levy‐Yeyati, E. & Sturzenegger, F. (2003). To Float or to Fix: Evidence on the Impact of Exchange Rate Regimes on Growth. American Economic Review, Vol. 93 (4), pp. 1173‐1193 Mankiw, N.G. (2013). Macroeconomics. Houndmills, Basingstoke: Palgrave Macmillan. Mundell, R.A. (1997). Exchange‐Rate Systems and Economic Growth. Monetary Standards and Exchange Rates, Routledge, London, pp. 13‐38 Pilbeam, K. (2013). International Economics. Houndmills, Basingstoke: Palgrave Macmillan. Rogoff, K.S., Husain, A.M., Mody, A., Brooks, R. & Oomes, N. (2003) Evolution and Performance of Exchange Rate Regimes. IMF working paper WP/03/243 Stock, J.H. & Watson, M.M. (2012). Introduction to Econometrics. Edinburgh Gate, Harlow: Pearson Education Limited. Williamson, J. (2000). Exchange Rate Regimes for Emerging Markets: Reviving the Intermediate Option. Washington: Institute for International Economics.
Appendix 1 Countries in data set The following 105 countries were used in the data set: ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ Lower middle income country Continent ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ Armenia Europe Bhutan Asia Bolivia South America Cape Verde Africa Cameroon Africa Congo Rep. Africa Cote d’Ivoire Africa Djibouti Africa Egypt Africa El Salvador Middle America Georgia Europe Ghana Africa Guatemala Middle America Guyana South America Honduras Middle America India Asia Indonesia Asia Kiribati Oceania Kosovo Europe Kyrgyz Republic Asia Laos Asia Lesotho Africa Mauritania Africa Micronesia Fed. Sts. Oceania Moldova Europe Mongolia Asia Morocco Africa Nicaragua Middle America Nigeria Africa Pakistan Asia
Papua New Guinea Asia Paraguay South America Philippines Asia Samoa Oceania Sao Tome and Principe Africa Senegal Africa Solomon Islands Oceania South Sudan Africa Sri Lanka Asia Sudan Africa Swaziland Africa Syria Asia Timor‐Leste Asia Ukraine Europe Uzbekistan Asia Vanuatu Oceania Vietnam Asia West Bank and Gaza Asia Yemen Asia Zambia Africa ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ Upper middle income country Continent ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ Albania Europe Algeria Africa American Samoa Oceania Angola Africa Argentina South America Azerbaijan Europe Belarus Europe Belize Middle America Bosnia and Herzegovina Europe Botswana Africa Brazil South America Bulgaria Europe
