Camila Lucredi S2542587 clucredi@student.rug.nl
Impacts of Exchange Rate Volatility on Brazilian Agricultural Exports to the
Mercosur
University of Groningen Faculty of Economics and Business
2014
Abstract
This thesis aims to analyze the impact of exchange rate volatility on the Brazilian agricultural exports to other Mercosur countries. This study is focus on reasons why exchange rate volatility decreases trade. I will show some data regarding the trade relationship between Brazil and other Mercosur countries. The methodology applied will be a gravity model. The empirical show that indeed exchange rate volatility has a negative impact on Brazilian agricultural exports to other Mercosur countries.
Index
1. Introduction ... 5
2. Literature Review... 7
2.1. Exchange Rate Volatility and Trade... 7
2.2. Empirical Studies on Exchange Rate in Brazil ... 11
3. Methodology ... 13
3.1. Gravity model ... 13
3.2. Measuring Exchange Rate Volatility ... 15
3.3. Static Panel Data ... 16
4. Empirics ... 18
4.1. Bilateral Trade in Mercosur ... 18
4
1. Introduction
After the collapse of the Bretton Woods system in the early 1970s, following the U.S dollar devaluation in 1973, many countries adopted the floating exchange rate system, instead of the fixed exchange rate system. Following this collapse, researchers have become more interested in the effects of exchange rate volatility on trade in general, and particularly on exports.
There has been a significant disagreement regarding evidence on both positive and negative impacts of volatility on a nation`s exports. Most of studies find negative impacts of the exchange rate fluctuation on international trade (e.g. Thursby and Thursby, 1987). Other works, on the opposite, find both small negative and positive effects (Klein, 1990; Frankel and Wei, 1993; Eichengreen and Irwin, 1995; Frankel, 1997). Additionally, Cho, Sheldon, and McCorriston (2002) use panel data on bilateral trade and exchange rate volatility for G-10 countries, to investigate the effects of long-run real exchange rate volatility on agricultural trade in comparison to other sectors. They came to the conclusion that real exchange rate volatility has a significant negative effect on agricultural trade. This impact is much larger than the estimated impact on trade in other sectors and on aggregate trade.
According to Barros (2009), the coordination of the economic policies, especially the exchange rate policy on the Mercosur countries was always a matter of concern since the beginning in 1991. Bittencourt et. al. (2005) says that the lack of a stable exchange rate system can be a source of misalignment, especially for those countries that pegged their currencies to the US dollar, bringing substantial and persistent deviation of exchange rates from their macroeconomic fundamentals. The size of exchange rate misalignments, in the long run affects substancially international trade. The Mercosur`s countries use flexible exchange rate system in the early 1990`s and at the same time they created an integrated regional block, in order to facilitate trading activities. The creation of Mercosur was an important factor that helped consolidate the economic opening of Brazil. The countries belonging to the Mercosur (Argentina, Uruguay, Paraguay and Brazil) reduced gradually their international trade tariff between 1991 to 1994, resulting in a great increase of trade among members. At the same time, the two biggest countries from this regional block, Brazil and Argentina, suffered several domestic crises.
Mercosur nations presented a great trade growth intra block until 1998, after that they experienced a decrease in their trade activities. The decrease after this period is due to the exchange rate crisis in 1999 in Argentina, which affected all other Mercosur countries, including Brazil.
5 The divergence and fluctuation of the prices and exchange rates, which affects trades and investments has, as one of their causes, the lack of macroeconomic coordination. This affects international trade through two channels, namely risk of international transaction and risk of economic policy.
Referring to the first channel, an increase in the risk of international transactions influences trade decisions and leads to different allocations of resources. A rise in the real exchange rate forces risk-averse exporters and importers to diminish their supply and demand of goods, since now they face an additional risk with their external profits. The second channel is the existence of a lobbying to protect domestic markets when there is a probability of increase in imported goods, in other words, countries will create obstacles (e.g. protectionism) to receive other nations` goods.
A consequence from what has been explained at the previous paragraphs can be found in the Mercosur countries. Every member gradually eliminated most of their trade barriers between 1991 and 1995. However, the tariffs were not completely eliminated; as a result, each member was able to identify the sensitive products to competition, protecting them until 1999 for Argentina and Brazil and until 2001 for Paraguay and Uruguay. Hence, it is interesting to verify the consequences of the exchange rate instability on the trade, especially on the agricultural sector, which is the least protected sector in Latin America. This is due to the fact that demand for commodities keeps changing, because it is more income inelastic compared to other sectors; as a result, there is macroeconomic instability and commodity price fluctuation. Also, agricultural production plays an important role for these nations.
Brazil, after a major currency crisis in the early 1990s, used flexible exchange rates. The flexible exchange rate system has its benefits. First, the adjustment on the balance of payments are continuous, small and do not lead to fundamental macroeconomic disequilibrium. Second, under a flexible exchange rate, in order to adjust the balance of payment, it is only necessary to fluctuate the exchange rate and not the internal prices of the economy. Third, when the exchange rate, which is flexible, is at equilibrium or close to its fundamental value, the comparative advantage of the country is evident and your trade pattern does not distort. Fourth, flexible exchange rate system allows every country to be on their desired part of the Phillips curve, the trade-off between inflation and unemployment. Finally, a flexible exchange rate regime prevents the need of interventions on the exchange rate market and avoids the risks of making mistakes regarding the true exchange rate equilibrium.
6 One of the limitations of the previous studies is that most of the literature focuses only on the impact of the exchange rate volatility on the aggregate trade, ignoring the impact on specific sectors. According to Cho et al. (2002), a few studies evaluate the impact of the fluctuation of the exchange rate and on the agricultural trade. These studies include Schuh (1974), Batten and Belongia (1986), Haley and Krissoff (1987), and Bessler and Babula (1987). Pick (1990), did not find evidences that the risk of the exchange rate affects the trade of the USA with other developed countries. However, he found a negative effect for the trade flow from the USA to developing countries.
This thesis contributes to the literature by analyzing the impact of exchange rate volatility on Brazilian agricultural exports to other Mercosur countries. Exchange rate volatility can lead to a devastation of the international trade of Brazil, especially in the agricultural sector, which is the least protected sector in Latin America. Risk-averse exporters, that make exporting decision before observing the realization of the real exchange rate, choose to export less the more volatile the rate is. If exchange rate volatility diminishes the international trade, the policy makers should implement measures that decrease this volatility. Furthermore, this volatility has an impact not only on the trade volume, but also in the diversification of the products, making the country more dependent to certain type of goods, increasing its vulnerability to exogenous shocks, or affecting negatively economic growth. Knowing how exchange- rate volatility affects exports is important for the design of both exchange-rate and trade policies.
The thesis will focus on the period from 1994 to 2012. Gravity model will be used to analyze the impact of exchange rate volatility on Brazilian agricultural exports to other Mercosur countries. I chose this model because it is the best one to analyze trade flow, since it is easy to use, the required data is generally readily available, it is properly specified and it has proved to be a very reliable model for explaining existing trade patterns. The dependent variable is the Brazilian agricultural exports and independent ones include the product of countries` GDP levels, population, distance and exchange rate volatility. I use a static panel data, and apply fixed and random effects models.
I find that exchange rate volatility has a significant negative impact on Brazilian agricultural exports to other Mercosur countries, in line with the hypothesis of this study.
7
2. Literature Review
2.1. Exchange Rate Volatility and Trade
Exchange rates have been fluctuating since the collapse of the Bretton Woods system of fixed exchange rates. The end of the gold-dollar system led to a greater development of the financial markets and gave an opportunity for developing countries to increase their investments on their firms by attracting foreign capital. Due to increasing reliance of countries on foreign capital inflows, economists have been debating about the impact of the exchange rate uncertainty on international trade. The type of exchange rate system can affect economic growth directly via its effects on the adjustment of the economy to economic shocks and indirectly via its impacts on determinants of growth, which includes international trade. The crucial determinant affected by this exchange rate uncertainty is the exports. The possibility of exporting more goods helps an economy to grow (Mukhar and Malik, 2010). The technological development and the globalization made the movements of the capital more dynamic, increasing the uncertainties over the currencies from the development countries. In other words, the exchange rate can suffer great alteration when there is an internal or external crisis. Hence, there is uncertainty about the exchange rate behavior, which can affect the exports of a country, given that the valuation or devaluation of the national currency can make the goods more or less expensive in the international market.
Exchange rate volatility is associated with about the amount of uncertainty about the size of changes in currency`s value. According to Mukhar and Malik (2010), higher volatility means that a currency`s value can be spread out over a larger range of values, in other words, the price of the currency can change a lot over a short period of time in any direction. On the other hand, a lower volatility means that a currency`s value does not fluctuate a lot. What most of the literature argues is that a greater exchange rate volatility generates uncertainty, increasing the riskiness of trading, decreasing the trading volume. Particularly, in less developed countries, where the forward markets are not that well developed, exchange rate volatility coupled with protectionism, could have a major impact on trade and income. The problem of less developed economies is an underdeveloped financial market. Because of this, developing economies have higher transaction costs. If a firm uses the forward markets to hedge against the exchange rate risk, it can reduce the uncertainties in a short run (Mukhar and Malik, 2010).
8 Baer (2002) shows that the Mercosur countries are more affected by exchange rate uncertainties because their governors make decisions based mostly on internal issues. Hence, during a crisis, the countries make decisions considering only the effects from the crisis on their own national residents and not on the rest of the block. This would be a factor that causes instability of the exchange rate system in the Mercosur countries.
Carranza et. al. (2003) show that during a crisis, the firms that have debts in foreign currency suffer due to the exchange rate floating. With the outflow of foreign capital, there is a devaluation of the national currency, making these firms produce more in order to pay their debts. However, because they finance themselves with scarce external resources, there is no way to increase production. This is mostly likely to happen even more in developing countries as these countries are more affected by exchange rate volatility.
Kandilov (2008) affirms that exchange rate volatility has negative impact on trade and an even bigger impact on agricultural trade, if compared to other sectors. Also, he found empirical evidence that suggests that the impact is bigger for developing countries, fitting Mercosur`s case, than developed ones. This is due to the fact that advanced nations have better access to credit and hedging opportunities in order to try to reduce the risk that exchange rate uncertainty brings.
Clushman (1988) and Thurby and Thurby (1987) found that exchange rate markets have become more sensitive and that exchange rate volatility has negative impact on exports. It is known that exchange rate volatility has a negative effect on exports directly due to uncertainty and adjustment costs or indirectly via its effects on the allocation of resources and government policies. The point is that the volatility leads traders that are risk averse to reduce their trade. This has been empirically confirmed by Hooper and Kohlhagen (1978) who show that exchange rate risk has a negative impact on trade flow, when traders are risk averse. Also, if exports bear the exchange risk, the price goes up as exporters have higher risk premium.
Clark (1973) was one of the first to develop a theoretical model that analyzes the impact of exchange rate volatility on international trade flows. In this model, the firms are averse to risk. However, he imposes some restrictions. The market is perfectly competitive, firms only produce export goods, they possess limit possibilities of hedging, the contracts are on foreign currency and there is no imported input. The model states that increases of exchange rate volatility leads to increases of the uncertainty of profiting from the exports, expressed in domestic currency. Because firms are averse to risk, they tend to reduce the supply of goods until the point where marginal revenue exceeds marginal cost on the amount that compensates the additional risk. Therefore, in this model, exchange rate volatility has a negative effect on the international trade flows.
9 it should pay a permanency cost. Assuming that the firm is neutral in relation with risk, the profit maximization will depend on the exchange rate behavior. In this case, there will be a threshold value in which the exchange rate will induce the entry or the exit of the firm on the domestic market. On the other hand, big shocks (depreciation or valuation) on the exchange rate will induce a market entry or exit and alter the international trade flow. To sum up, only big shocks on the exchange rate will interfere with international trade.
Dellas and Zilberfarb (1993) develop a model in which there is one economic agent that exports, imports and consumes two goods in two periods of time in a small open economy. The stock market is incomplete and the agent makes his decisions of trade with an incomplete knowledge of price risk. The analysis of uncertainty of the exchange rate takes into consideration the lack of future markets, and hedging opportunities can be complete or incomplete. The authors show that the effect of exchange rate volatility is ambiguous on the trade, due to the fact that it depends on the risk aversion. When there is complete hedging possibility and with no cost, the agents will protect themselves from exchange rate risk, consequently, there will be no harm on the trade volume.
More recently, Lin (2012) develops a theoretical model with heterogeneous firms, in which it is possible to evaluate the effect of exchange rate volatility over the extensive margin (number of exported products) and intensive margin (exported monetary volume by product) from international trade. The model begins with one structure of two countries (symmetric ones), there are trade costs and the domestic firms have different levels of productivity, where only the most productive firms are able to overcome the trade costs and sell their products on the partner market. Furthermore, the exchange rate uncertainty comes from monetary shocks, affecting firms from both countries. Because the firms must define their goods ‘prices and decide whether they should export the same as before of knowing the world economy conditions, the uncertainty (flow) of the exchange rate should influence both entry on an international market and quantity that will be exported. The model affirms that when the uncertainty of the exchange rate is low, the trade costs will also going to be low, creating the possibility of less productive firms to enter the international market. On the other hand, when the exchange rate uncertainty is high, the trade costs will also going to be high, and hence, only the most productive firms will be able to enter the international market.
There are many models developed in the literature that explain impact of exchange rate volatility on trade. As one can see from the table below, due to contradictory predictions, researchers tried to examine as much as both of these effects, which led to mixed results as overall evidence. These results are sensitive to sample period, model specification, proxies for exchange rate volatility and countries considered, developed or developing ones (Oztruk, 2006).
Table 1: Exchange Rate Volatility and Trade: Literature Survey.
Study Sample period Research method used
Main results Countries analyzed Akhtar and Hilton
(1984)
1974‐1981Q OLS Negative effect Germany,
10 Gotur (1985) 1974‐1982Q OLS Little to no effect Chile, New
Zealand Bailey et al. (1986) 1973‐1984Q OLS Not significant,
mixed effect
Big 7
industrial economies Bailey et al. (1987) 1962‐1985Q OLS Little to no effect USA
Mann (1989) 1977-87Q OLS Few significant
results on exports
Japan, USA, Germany Pere and Steinherr
(1989)
1960‐1985A OLS Negative effect Industrial
countries Bini and Smaghi
(1991)
1976-84Q OLS Significant
negative effect on export
LDC
Savvides (1992) 1973-1986A Cross section Negative effect Germany
Chowdhury (1993) 1973‐1990Q VAR Significant
negative effect G-7 countries Mckenzie and Brooks (1997) 1973‐1990Q ARCH Generally positive effect Germany, USA Aristotelous (2001) 1989‐1999A Gravity model No effect on
export
UK, USA Vergil (2002) 1990‐2000Q Standard deviation Negative effect
on export Turkey Das (2003) 1980‐2001Q ADF,ECM, Cointegration Significant negative effect on export USA
Baak (2004) 1980‐2002A OLS Significant
negative effect on export East Asian Countries Kasman and Kasman (2005) 1982‐2001Q Cointegration, ECM Significant negative effect on export Turkey
Arize et al. (2005) 1973‐2004Q Cointegration, ECM Significant negative effect on export LDC Lee K.S., Saucier P., (2005)
1990-2000M GARCH-M Positive effect on import and insignificant effect on export Asia M.V.L. Bittencourt, D.W. Larson, S.R. Thompson (2007)
11
Bittencourt (2013) on exports. partners
Note: A= annual, Q = quarterly, M = monthly Source: Oztruk (2006, p 88‐92)
2.2 Empirical Studies on Exchange Rate in Brazil
The number of empirical studies that analyze the effect of exchange rate volatility on the trade is large. However, I focus on the work in which the authors examine the impact of exchange rate volatility on Brazilian international trade. The main studies are by Gonzaga and Terra (1997), Esquivel and Larraín (2002), Aguirre et al. (2007) and Bittencourt et al. (2007).
Gonzaga and Terra (1997) analyze the effect of exchange rate volatility on Brazilian trade flow. The authors estimated eight different export functions, in which the dependent variable is the exports volume and the explanatory variables are the real exchange rate, the exchange rate volatility and GDP. The coefficient associated with exchange rate volatility measure is negative and, most of the time, not significant.
Esquivel and Larrain (2002) analyze the effect of exchange rate volatility of the countries from the G-3 (Germany, USA, and Japan) on some macroeconomic variables (FDI, international trade flow and probability of occurring financial crisis) from developing countries, including Brazil. As the dependent variable the authors use the country`s exports and include two measures of exchange rate volatility: One for the German Mark/Dollar and the other for the Yen/Dollar. The authors show that the Yen/Dollar volatility has a negative effect, but not important over the Brazilian exports. The German Mark/Dollar volatility presented a statistical significant only on one estimation and with a positive sign.
Aguirre el al. (2007) evaluate the effect of exchange rate volatility on Brazilian exports of manufactured products. The authors use as explanatory variables, besides the real exchange rate volatility, the effective exchange rate, the world imports level and a national industry capacity rate. The model is estimated by ARDL (Auto-Regressive Distributed Lag) from 1986 to 2002. The results show a negative and significant coefficient associated with the real exchange rate volatility, in which an 1% increase of real exchange rate volatility reduces the Brazilian exports of manufactured goods by 0.77%.
12 From the discussed studies, I can observe that they tried to analyze the effects of exchange rate volatility on Brazilian international trade and most of them come to the conclusion that exchange rate volatility affects negatively international trade. None of these works analyzed the impact of exchange rate volatility on agricultural exports. That way, this work will try to fill the gap about the effect of exchange rate volatility on Brazilian agricultural exports to the other Mercosur nations. The hypothesis of this thesis is to see if exchange rate volatility has a negative impact on Brazilian agricultural exports to other Mercosur`s nations.
13
3. Methodology
This chapter presents the methodology of the gravity model used in this study. This model has evolved over time; however, the basic variables used in this model, such as a proxy for the countries size and distance still remain. This section is developed as follows. First, I describe the evolution of the gravity model, and then I explain the methodology to calculate exchange rate volatility. Finally, I discuss the use of panel data.
3.1 Gravity Model
The gravity model is considered a suitable model to explain trade relations between countries. This model for economics was developed through the Physics model. One of the first authors to study the gravity model in Economics was Tinbergen (1962). Later on, Anderson (1979) tried to verify if there is any theoretical foundation in applying gravity equation while studying commodities trade. Anderson assumes that countries produce tradable and non-tradable goods. The assumption is also present on the expenditure and consumption functions, in which the consumers are aware of it.
The Anderson`s model (1979) was later improved. For instance, Bergstrand (1985) applies Heckscher-Ohlin assumptions with factor allocation, introducing trade inter and intra firms. Additionally, Bergstrand (1989) developed theoretical foundations that explain the behavior of importers and exporters in each good. As a result, 40% to 80% of the trade variations between countries can be explained by the gravity model.
Several authors criticize the simplicity of the model when it is applied to real values obtained through historical series and part of the effect was not captured by the model. Trefler (1995) shows that considerable amount of observable trade volume between the countries could not be explained by the gravity model of Anderson (1979). There was the need to improve this model and include more variables.
This study establishes the connection between exchange rate volatility and its impact on trade flow for Brazil. A gravity model is specified and estimated in order to evaluate the impacts of exchange rate volatility on agricultural export sector from Brazil to other Mercosur countries. The gravity model not only considers trade flows, but also transportation costs, trade barriers, location, population, national income and exchange rate. The gravity model is a general equilibrium system from final goods in the international trade. This equation shows that trading activities between two countries depends on their respective sizes, economic development, opening of their market and proximity. Trade, in other words, is directly proportional to the country size and inversely correlated with the distance between the two countries. Trade flow is a function of income, distance and other variables.
3.1.1. General Model Specification
14 distance between Brazil and all the other Mercosur countries. In the gravity model, bilateral trade flows between countries (Brazil) and country (other regional block member) at time t is represented as follows:
(1)
So:
(2)
Where is the bilateral trade flows between countries and at time . is the product of the GDP levels of both countries at time . is the product of population of countries and at time . is the geographical distance between countries and country . Finally, is the exchange rate volatility between country and country at time
(3)
As far as the constant is concerned, using an exponentiated version of in place of is equivalent in the sense that both of them are arbitrary constants.
Most studies using panel data estimate the multiplicative gravity model after the model is log-linearly transformed. This technique allows the use of classical estimation methods, such as fixed effects and random effects (Bobková, 2012). As a result, in this thesis, I will choose to transform the equation using natural logarithm.
Taking the natural logarithm of the equation (3), except from the exchange rate volatility, since it is already normally distributed I obtain the following econometric model:
( ) = + + + + + (4)
In equation (4), I expect that coefficients and have negative signs. Furthermore, I will include a dummy variable, in which for crisis period, it takes the value of 1 and no crisis 0. As a result, the dummy variable will be 1 for the year of 2001 (Argentinian crisis) and 2008 and 2009 (subprime crisis).
3.1.2 The Gravity Model Specification for Exports
15 equation captures the volume of exports between the two trading partners as a function of their GDPs and the distance between them.
( ) = + + + + + (5)
Countries usually trade more as they increase in size and this can be measured either by the product of the two variables of population or the product of the two countries`GDPs. The GDP of the domestic country, in this case Brazil, is believed to reflect the capacity to supply exports. Similarly, the GDPs of the other Mercosur countries, such as Argentina, Paraguay and Uruguay, the importing nations, are believed to show its demand for exports. In the equation (5), I will differentiate between recipient country (Mercosur partners) and export country (Brazil). The coefficient for the recipient country population is expected to be a positive one, due to the fact that bigger market demands more goods. Moreover, population in the export country is also expected to have positive effects on the variable of exports, since the export nation is able to provide more as the population grows in size.
Distance also explains the trade flow between Brazil and the other countries. An increase in distance between economies is expected to increase transportation costs, and,
consequently, reduce exports.
Studies using gravity model include dummies to study the effect of certain variables, such as language, common border or membership in same trade block. However, in this study, it does not make sense to include those variables because all countries of the analysis have common border, Brazil has a different language from the others and the study evaluate only one block, which is the Mercosur. So final equation (6) is going to be:
( ) = + + + + + (6)
Where are the agricultural exports from Brazil to other countries at time , is exchange rate volatility at time , is the product of the GDP from Brazil and the GDP of country at time , is the product of the Brazilian population and population of country population at time is the distance between these two nations, is the dummy variable, which is 1 in the presence of a crisis and 0 if not, and finally is the error at time . The error term has mean zero and variance .
3.2. Measuring Exchange Rate Volatility
Exchange rate volatility takes place due to the uncertainty of agents regarding the floating exchange rate. Previous studies used different variables to measure volatility. Esquivel et. al. (2002) use as a volatility measure the standard deviation of the exchange rate growth and the coefficient of the real exchange rate variation over a period of 12 months.
16 exchange rate flexibility, the standard deviation of the logarithms from the annual differences of the effective exchange rate, and an average deviation of the predicted level of the effective exchange rate considering a five years period.
I will use for this study the moving standard deviation, which in all previous literature was the most used one.
3.2.1. Moving Standard Deviation (MOVSD)
First, the bilateral exchange rate between countries i and j at time t is calculated as:
(7),
Where and represent, respectively, the real exchange rate of countries against US dollars and the real exchange rate of countries against US dollars at time t.
The moving standard deviation method of measuring exchange rate volatility is the most widely used method in the previous literature. As a result I used this measure of volatility on our study. I use a moving standard deviation of the first difference in the real exchange rates to compute an ex ante measure of volatility. This measure gives a larger weight to extreme observations, showing the behavior of risk – averse traders better. Mathematically:
√ ∑ ̅̅̅̅̅̅
(8)
Where,
̅̅̅̅̅̅ is ∑ , the average of for the last k years.
Furthermore, when I use moving standard deviation, I have a serious autocorrelation problem in the error terms. As a result, I need to test for it and correct it, if it is significant.
3.3. Static Panel Data
For this study, the econometric method that I will use is the panel data. According to Pindyck (2004), the panel data lets the authors of the studies have cross-section and temporal effects of the dataset. The studies with panel data use cross-section data from the units’ analysis during a time period, in other words, evaluate the impact of variables at same time on the units of the studies. Due to that, it increases the observation numbers and the degrees of freedom from the sample.
17 We can estimate the static panel data using two different methodologies: Fixed effects and random effects.
In the estimation by fixed effects, we consider that the non-included variables on the model are correlated with the included ones. When we have panel data, the fixed effects model is estimated as fixed effects regression. The use of fixed effects is whenever we are only interested in analyzing the impact of variables over time. Furthermore, unobserved cross-country characteristics are correlated with the error term.
18
4. Empirics
4.1. Bilateral Trade in Mercosur
The Mercosur started in 1991 after the Asunción Treaty, in which it tried to constitute a common market in order to reduce fees and barriers between member countries. The countries of the block are Argentina, Brazil, Paraguay and Uruguay. In 2012, Paraguay was suspended from the block due to the destitution of the president Fernando Lugo, while Venezuela got accepted to the Mercosur. The member countries agreed about a transition period from 1991 to 1994. During this period, import tariff reductions goals were implemented, in which these tariffs should be reduced gradually until their completely elimination. Trying to achieve a common market, the first paragraph of the Asunción Treaty states the free circulation of goods, services and capitals among their members, being these the pre-requisites to make a common market.
At the beginning, there was a significant growth on exports among the trade partners from Mercosur until 1998. However, after 1988 there was an increase of extra block and the Mercosur started to loose representativeness on the intra block trade. In 2002, the volume of intra block exports was reduced to 11.5%, the smallest percentage since the creation of the block. After 2002, there was a small growth on the intra block exports, arriving to 15.8% (Graf and Azevedo, 2013).
In 2006, there is still an increase on the growth of the intra block exports for all member countries. Brazil, presents as the country with less dependence on Mercosur exports, sending on average only 10% of its exports to the regional block. The second position in terms of less dependence on intra block exports is Argentina, sending 25% of its exports to Mercosur. After comes Uruguay with 29% and Paraguay with 49% (Graf and Azevedo, 2013).
In Latin America, Mercosur is an agricultural export region due to natural resources and investments in agrobusiness. According to Chailoult and Hilcoat (1997), from the total of Latin American agricultural and livestock exports in 1990, 53% comes from the Mercosur countries. It was from the Mercosur most of the exports of meat (85%), soy (99%) and wheat (93%). During the 1980`s, the agricultural production contributed to 50% of the total exports from that regional block, representing 75% of the total of Argentinian exports and 40% of the Brazilian surplus. Brazil and Argentina are the two most important countries in the world in terms of exporting oilseed and similar. However, the way that these two countries is inserted in the international market is different. Brazil chose to develop its agribusiness and Argentina is the biggest exporter of crude products. These specializations are shown on their trade (Chailoult and Hillcoat, 1997).
19 Secondly, if I observe the total evolution between 1990 and 1993, the exports growth (+400%) is larger than the imports growth (only +8%). This tendency changes in 1993, year in which the imports increase a lot (+30%) and even more during 1994 and 1995. This is the period in which Brazil has a deficit of US$ 565 million. Comparing this period with 1994, I see an increase of 121% of the imports and only 14% of the exports. These numbers show the potential of Brazilian market regarding the Mercosur partners. Mostly, it is the Argentinian exports that explains this evolution and shows the importance of the macroeconomic and monetary context on the trade flows and bilateral balance. While the Argentinian economy slows down since half of 1994, the Brazilian demands for imports increase due to the new adjustment plan (the “Real Plan”) and the reevaluation of its currency.
Chaloult and Hillcoat (1997) affirm that the results from the agricultural/livestock and agrifood trade from the Mercosur countries allow us to conclude that there is a tendency towards a trade regionalization, in particular in terms of exportation. Moreover, the opening and trade liberalization was positive for all the Mercosur.
4.2 Data Analysis
This research will focus on the years from 1994 to 2012. The Brazilian agricultural exports to the other Mercosur countries data were collect from the website of the Brazilian Ministry of Development, Industry and Foreign Trade, SECEX, the Secretary from Foreign Trade. This dataset is available on US$ F.O.B and shows the aggregate agricultural exports from Brazil to all the other Mercosur countries. Data is available from 1989 to April 2014. With this dataset, I can see the behavior of Brazilian agricultural exports to Mercosur nations. I can see that the tendency for the Brazilian agricultural exports was to grow as the years pass by; however, in 2009 there was a decrease, probably because the countries got hit by the financial crisis. The same happened for the year of 2012, mostly due to the fact that Brazil now faces the increase competition from other countries, mostly China.
20 Source: Secex
Graph 2 shows the Brazilian agricultural exports in US$ F.O.B for all the Mercosur`s nations from 1994 to 2012. Data are in millions. Left axis is for Paraguay and Uruguay and the right one for Argentina
Source: Secex
The GDP data was taken out from The World Bank database. The data are in current U.S dollars and in millions. Dollar figures for GDP are converted from domestic currencies using single year official exchange rates. According to the gravity model theory mentioned before, countries that have bigger GDP tend to trade more. The same happens with the population,
0 200 400 600 800 1000 1200 1400 1600 1800
Mercosur Argentina Paraguay Uruguay
Graph 1: Brazilian Exports in Millions of US$ for 2012
0 500 1000 1500 2000 2500 3000 3500 4000 0 100 200 300 400 500 600 700 800 900 1000 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Graph 2: Brazilian Agricultural Exports in US$ F.O.B
21 the bigger the population the more the nations will trade more. Regarding the GDP, I can see a tendency towards growth, with a slow down during the financial crisis.
The population data for all the countries has also as a source The World Bank and are in millions. In general, the populations have been increasing throughout the years, leading possibly to an increase in exports, since the country will tend to produce more goods.
For the distance variable I will consider, as in Sologa and Winters (2001), Bittencourt (2004) and Bittencourt et. Al. (2007), the distance between the main economic centers from the countries and obtained in Sologa and Winters (2001). According to previous studies in gravity model, the distance also contributed for the reduction of the trade flow among countries. This hypothesis was tested and proven by Samuelson (1962). He tested what happened when the countries were trading with other ones that were more geographically distant from their respective borders. The outcome is that the greater the distance, the higher the cost for the products, since higher distances imply higher transportation costs.
Table 2: Distance between Brazil and Mercosur`s countries (in kilometers)
Argentina 1670
Uruguay 1127
Paraguay 1977
22 Source: USDA
4.3. Empirical Results
I use STATA to run the statistical analysis. First of all, I present descriptive statistics of all the variables.
Table 3: Descriptive Statistics of the Used Database.
Variable Number of
observations
Mean Minimum Maximum Standard
Deviation NaturalLogExports 58 18.592 16.258 21.274 1.2199 Exchange Rate Volatility 58 .502 .0302 3.950 1.088 NaturalLogGDP 58 52.003 49.514 55.361 1.618 NaturalLogPopulation 58 35.012 33.866 36.638 1.062 NaturalLogDistance 58 7.345 7.027 7.589 .237
Table 4: Correlation Between Variables
Variables NaturalLog Exports Exchange Rate Volatility NaturalL ogGDP NaturalLogPo-pulation NaturalLog Distance NaturalLogEx-ports 1 Exchange Rate Volatility -0.073 1 NaturalLogGDP 0.964 -0.178 1 NaturalLogPo-pulation 0.780 -0.324 0.797 1 NaturalLogDis- 0.429 0.368 0.369 0.016 1 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
Graph 3: Exchange Rate Volatility
23 tance
Table 4 shows the correlation between the variables used in my gravity model. GDP is highly positive correlated with population and even more with exports (the bigger the GDP the more a nation will produce and, consequently, export more, according to the gravity model theory). Exports are also positive correlated with population. Finally, distance is correlated with exports and exchange rate volatility.
Doing a visual inspection from the dataset and also from the graphs, I can see that the variables might present a certain tendency, due to that, I will conduct a panel data unit–root test to check whether the variables are stationary to ensure the consistency of the estimated results. This is important to avoid spurious regressions, which is when there is a high squared R value without a significant relationship between the variables. This happens due to the presence of a tendency.
The unit–root test that I chose was the Levin-Lin-Chu unit-root test. For this test, the null hypothesis is that there is unit root, and it is conducted individually for each variable. 1 lag was used and trend is not included. After conducting the test, I see that all p–values are higher than 0.05, except for exchange rate volatility, ranging from 0.273 for population to 0.981 for exports. This way, I cannot reject the null hypothesis about the existence of unit root, in other words, the variables are nonstationary and present non consistent results.
Table 5: Levin–Lin – Chu Unit –Root test
Variables p-values
Exchange rate volatility 0.002
Natural Log Exports 0.981
Natural Log GDP 0.107
Natural Log Population 0.273
Since I detected the presence of unit root, I need to take out the first difference, in other words, the tendency needs to be removed, in order to ensure consistency of my estimated results, and then run the regressions, random and fixed effect models. First of all, I will show the results with all the variables and using no lags as instrument for exchange rate volatility:
Table 6: Results for Fixed and Random Effects Model
24 Exchange rate volatility -.998 (.7383) -.782 (0.277) Crisis .063 (.121) .065 (0.590) Constant .413 (.261) .125 (0.371)
***Significant at 1%; **Significant at 5%; *Significant at 10%.
The numbers in parenthesis represent always the respective standard errors. The squared R for the fixed effects is 0.690 and for the random effects is 0.950. Analyzing the results, I can see that, despite distance and crisis, all the variables present the expected coefficient sign. However, only GDP can be considered significant at 10% level. I also performed the Hausman specification test, which compares the consistent fixed – effects model with the efficient random – effects model and concluded that the most appropriate results are the ones from the fixed effects model.
Now, since there is a potential problem for endogeneity regarding exchange rate volatility, that is, exchange rates can become more volatile as countries trade more, in order to cope with this, I use lagged values at t-1 and t-2 of the same endogeneous variable. Furthermore, I use the Akaike Criteria to check for the number of lags that are more appropriate and conclude that lag at t-1 is more appropriate to be used as an instrument.
Table 7: Results t-1 for Exchange Rate Volatility
Variables Fixed Effects Random Effects
Natural Log GDP .0137
(.007)*
.0153 (.007)** Natural Log Population 10.052
(11.468)
.120 (0.984)
Natural Log Distance - -12.488
(0.852) Exchange rate volatility (lagged
value) -.048 (.589) -.019 (.575) Crisis .075 (0.545) .074 (0.544) Constant .360 (.2679) .128 (.047)
***Significant at 1%; **Significant at 5%; *Significant at 10%.
The squared R for fixed effects model is 0.679 and for the random effects model is 0.988.
Table 8: Results t-2 for Exchange Rate Volatility
Variables Fixed effects Random Effects
25
(.008)* (.008)*
Natural Log Population 11.222
(11.90957)
.054
(5.953)
Natural Log Distance - -13.212
(0.894)
Exchange rate volatility (lagged value) .141 (.652) .208 (.641) Crisis .080 (.127) .082 (.126545) Constant .388 (.276) .131 (.1450)
***Significant at 1%; **Significant at 5%; *Significant at 10%.
The squared R for the fixed effects model is 0.587 and for the random effects one is 0.493. For t-1, except for crisis, all the variables have the expected sign, however, only GDP can be considered significant at 10% level. Looking at t-2, crisis and exchange rate volatility have the opposite expected sign, and once again only GDP can be considered significant.
Next, I will perform the same regressions, with lagged value of exchange rate volatility at t- 1, but excluding GDP, since population might be already doing the work as a control variable and GDP could be interfering, since both of them are highly correlated to with each other. As a result I get:
Table 9: Results at t-1 with no GDP
Variables Fixed effects Random Effects
Natural Log Population 15.130
(0.092)*
.405
(0.047)**
Natural Log Distance - -11.777
(0,67)
Exchange rate volatility -1.174
(0,150)** -.116 (0.046)** Crisis .0156 (0.898) .021 (0.862) Constant .436 (0.108) .113 (0.441)
***Significant at 1%; **Significant at 5%; *Significant at 10%.
The squared R for the fixed effects model is 0.690 and for the random effects model is 0.934. Doing again a Hausman test to check which model is better, I conclude that at t-1 for exchange rate volatility and excluding GDP, the fixed effects model is better, since I got a significant p-value (0.329).
Just to ensure the previous results, I will perform the regression with exchange rate volatility in t-1, but this time, excluding population. The results are:
26
Variables Fixed effects Random Effects
Natural Log GDP -.014
(.007)
-.0143 (.007)
Natural Log Distance - -2.33
(.311) Exchange rate volatility -.043
(.588) -.018 (.575) Crisis .074 (.124) .074 (.122) Constant .129 (.048) .129 (.047)
***Significant at 1%; **Significant at 5%; *Significant at 10%.
However, none of the above results are significant and the GDP, in this case, even has a negative coefficient sign. As a result, I can assume for sure that the best model that explains the relationship between Brazilian agricultural exports to other Mercosur`s countries and exchange rate volatility is fixed effect model with lagged value (t-1) of exchange rate volatility and excluding GDP.
Analyzing this last regression, I can see that GDP, because it was highly correlated with population, could be indeed interfering with the final results. Once I exclude GDP, I got the results that I was expecting. First of all, population is significant and has a positive impact on Brazilian agricultural exports to Mercosur (the bigger the population of a nation, the more it will produce and, consequently, export). Furthermore, exchange rate volatility is significant at 5% level and has a negative impact on commodities exports, corroborating my hypothesis that exchange rate volatility of the Brazilian Real against the other Mercosur countries’ currencies has a negative impact on Brazilian agricultural exports to the same nations. Finally, analyzing the dummy variable for crisis, it presents a positive sign; however it is not significant enough to take into account, which makes us conclude that there is the necessity of more studies about the crisis and exports in the Mercosur area.
The residuals of a regression should be random and must not present a certain pattern. If they present a certain pattern there is autocorrelation or serial correlation. Autocorrelation leads to an underestimation of the standard errors and too high t - statistics. If the researcher is not aware of the fact that there is serial correlation, there will be some coefficients statistically significant when they are actually not.
To check for autocorrelation in panel-data models, Wooldridge (2002) derived a simple model in which the null hypothesis is no serial correlation, as a result I am going to use this test on STATA. Since the p – value of the test was significant (0.298), I failed to reject the null and conclude that the data have no autocorrelation. That way the models were estimated with robust standard errors. Furthermore, since the test indicated the presence of heteroskedasticity, the use of robust standard errors was the best option.
27 As a robustness check measurement I included some more additional variables that could affect the Brazilian agricultural exports to the other Mercosur countries. I am going to include trade agreements dummy for whether countries have trade agreements with the United States of America, in order to check for some trade divergence effects. Brazil and USA implemented a trade agreement between each other in 1935 to try to reduce the tariffs of the goods exchanged. Same for Paraguay in 2003. In 2007, Uruguay also implemented an agreement with the United States. Argentina is the only country in the Mercosur area that does not present a trade agreement with the USA. After running the regression using t-1 as an instrument for exchange rate volatility, without including GDP and using the fixed effects model (the model that presented most reliable results), the coefficients are still plausible and robust. I arrived to the following results:
Table 11: Results at t-1 with no GDP and with USA Trade Agreements
Variables Fixed effects
Natural Log Population 14.399
(11.551)**
Natural Log Distance - Exchange rate volatility -1.148
(.752)**
Crisis .015
(.121)
USA trade agreement .176
(.250)
Constant .413
(.270)
28 5.
Limitations
This thesis only takes into account the moving standard deviation to measure the exchange rate volatility, however, several other authors, such as Cho et. al. (2002), Bittencourt (2004) and Bittencourt et. al. (2007) use both measure of volatility: besides the moving standard deviation, also the Peree and Steinherr measure, in which the authors consider the maximum and minimum value of exchange rate over a given time interval of size k up to time t. These two measures use different perspectives to show the exchange rate variability during a certain period of time.
Furthermore, the volatility that affects the trade flow between two countries is not only the volatility of these two countries, the only fact that I considered for this thesis. According to Dell`ariccia (1999), the variables of the other trade partners also affect the trade flow of these two countries, this is what the author calls “the third – country effect”. As a result, Bittencourt (2004) and Bittencourt et. al. (2007) also use this in their work.
29 6.
Conclusions
In this study, I estimate the impact of exchange rate volatility on Brazilian agricultural exports to the other Mercosur countries through a gravity model. The result proves our hypothesis that the volatility has a negative impact on the exports. This is a consequence of the uncertainty that makes the entrepreneurs to invest on other sectors, the ones that they have full market knowledge, since they are risk – averse. The impact of exchange rate volatility is stronger in developing economies due to the fact that these countries do not possess highly developed financial markets and do not have as many hedging options as the developed countries.
The time period of this study (1994 to 2012) is small, but, even though I was able to find the expected and significant results. However, for further studies I could increase the time range, use besides the moving standard deviation the Peree and Steinherr measure for exchange rate volatility as well, consider the “third-country effect” and, most important, use the dynamic panel data, in which some studies, obtained significant coefficients for the trade history, as a result, for future studies, the hypothesis of the trade history as one of the explanatory variables should be considered.
The Mercosur countries after the integration increased their bargain power with the other regional blocks. However, during the years, we observed that the actions from these countries governors are not coherent with the ones from a block, since they only think about their own internal interest. Protectionist policies and actions that increase the uncertainty over the exchange rate movements have negative impact over the economy as a whole and especially on the agricultural sector.
30
7. Bibliography
P. Aghion, P. Baccheta, R. Ranciere, and K. Rogoff (2009), “Exchange Rate Volatility and Productivity Growth: The Role of Financial Development” Journal of Monetary Economics 56, pp 494–513.
A. Aguirre, A. Ferreita and H. Notini (2007), “The Impact of Exchange Rate Volatility on Brazilian Manufactured Exports”. Económica, La Plata, V LIII, nº 1-2.
J. E. Anderson (1970), “A Theoretical Foundation for the Gravity Equation”. The American Economic Review, V. 69, n. 1, pp. 106-116.
M. Arellano and O. Bover (1995), “Another Look at the Instrumental-Variable Estimation of Error-Components Model”. Journal of Econometrics, v. 68, pp. 29-52.
M. Arellano and S. Bond (1991), “Some Tests of Specification for Panel Data: Monte Carlo Evidence and Application to Employment Equations”. The Review of Economic Studies, v. 58, n. 2, pp. 277-297.
A. C. Arize (1998), “The Effects of Exchange Rate Volatility on U.S. Imports: an Empirical Investigation”. International Economic Journal, Volume 12, Number 3, Autumn 1998, pp. 31-38.
Asuncion Treaty, Asuncion, 26/03/1991. Available at: <www.mercosul.gov.br>. Acessed on May, 2014.
W. Baer, T. Cavalcanti and P. Silva (2002), “Economic Integration without Policy Coordination: the Case of Mercosur”. Emerging Markets Review 3(2002), pp 269-291. R. Baldwin and P. Krugman (1989), “The Effects of Large Exchange Rate Shocks”. The Quarterly Journal of Economics, vol. 104, nº 4, pp. 635-654.
D. Barros (2009), “Volatilidade Cambial: Um Obstáculo à Coordenação de Políticas Macroeconômicas no Mercosur”. A Economia em Revista, V. 17 No 1, pp. 31-40.
J.H Bergstrand (1985), “The Gravity Equation in International Trade: Some Microeconomic Foundation and Empirical Evidence.” Review of Economics and Statistics, August 1985, 67(3), pp. 474–81.
J.H Bergstrand (1989), ”The Generalized Gravity Equation, Monopolistic Competition, and the Factor-Proportions Theory in International Trade”. The Review of Economic and Statistics, Vol. 71, No. 1 (Feb., 1989), pp. 143-153.
D.A. Bessleer and R.A Babula (1987), “Forecasting Wheat Exports: Do Exchange Rates Matter?”.Journal of Business, Economics and Statistics 5, pp. 397- 406.
31 M.V.L. Bittencourt (2004), “The Impacts of Trade Liberalization and Macroeconomic Instability on the Brazilian Economy”.Ph.D. Dissertation, The Ohio State University.
M.V.L. Bittencourt, D.W Larson and S.R. Thompson (2007), “Impactos da Volatilidade da Taxa de Câmbio no Comércio Setorial do Mercosul.” Estudos Econômicos, São Paulo, V. 37, N. 4, pp. 791-816.
A.S.S. Carmo and M.V.L. Bittencourt (2013), “O Efeito da Volatilidade da Taxa Real de Câmbio Sobre a Diversificacao da Pauta de Exportação do Brasil: Uma Investigação Empírica”, pp. 1-23.
L.J Carranza, J.M Cayo and J.E Galdon-Sanchez (2003), “Exchange Rate Volatility and Economic Performance in Peru: a Firm Level Analysis”. Emerging Economic Review 4 (2003), pp. 472-496.
T.I. Chaloult and G. Hillcoat (1997),” Mercosul e o comercio agropecuario” .
G.D. Cho (2001), “Real Exchange Rate Movements and Agricultural Trade”. Ph.D. Dissertation, The Ohio State University.
G.Cho, I.M Sheldon and S. Mccorriston (2002), “Exchange Rate Uncertainty and Agricultural Trade”. American Journal of Agricultural Economics, 84(4) (November 2002), pp. 931–942.
P. Clark, N. Tamirisa and S. Wei (2004), “Exchange Rate Volatility and Trade Flows - Some New Evidence,” IMF report.
G. Dell‘Ariccia (1999), “Exchange Rate Fluctuations and Trade Flows: Evidence from the European Union”. IMF Staff Papers, Vol. 46, Nº 3(September/December, 1999).
H. Dellas and B. Zilberfarb (1993), “Real Exchange Rate Volatility and International trade: a Reexamination of the Theory”. Southern Economic Journal, V 59, pp. 651-657.
B. Eichengreen and D.A Irwin (1995), “Trade Blocs, Currency Blocs and the Reorientation of Trade in the 1930s.” Journal of International Economics, 38 (1995), pp. 1-24.
G. Esquivel and B.F. Larrain (2002), “The Impact of G-3 Exchange Rate Volatility on Developing Countries”. Centre for International Development Harvard University and United Nations Conference on Trade and Development, G-24 Discussion Paper Series, No. 16. R.P. Flood and A.K Rose (1995), “Fixes: Of the Forward Discount Puzzle” Review of Economics and Statistics.
R.P. Flood and A.K Rose (1997), “Fixing Exchange Rates: A Virtual Quest for Fundamentals” Journal of Monetary Economics 36 , pp.3-37.
32 J. Frankel and S. Wei (1993), “Trade Blocs and Currency Blocs,” NBER Working Paper No. 4335.
E.W.Frees (2004), “Longitudinal and Panel Data: Analysis and Applications in the Social Sciences”. Londres: Cambridge University Press.
G. Gonzaga and C. Terra (1997), “Equilibrium real exchange rate, volatility and stabilization’. Journal of Development Economics, vol. 54, pp 77-100.
C.O Graf and A.F.Z. Azevedo (2013), “Comercio Bilateral entre os Paised Membros do Mercosur: Uma visao do Bloco atraves do Modelo Gravitacional”. Economia Aplicada, V 17, No 1, 2013, pp. 135-158.
W. Greene (2005), “Reconsidering Heterogeneity in Panel Data Estimators of the Stochastic Frontier Model”. Journal of Econometrics 126, Pages 269–303.
S.L. Haley and B. Krissoff (1987), “U.S. grain exports and the value of the U.S. dollar”. Journal of Agricultural Economics Research 39, pp.12-21.
P. Hooper and S.W. Kohlhagen (1978), HOOPER, “The Effect of Exchange Rate Uncertainty on the Prices and Volume of International Trade.” Journal of International Economics, 8 (1978), pp. 483-511.
T.I. Kandilov (2002), “The Effects of Exchange Rate Volatility on Agricultural Trade”. Agricultural and Resource Economics, NC State, pp. 1-34.
M.W. Klein (1990), “Sectoral Effects of Exchange Rate Volatility on United States Exports.” Journal of International Money and Finance, 9, pp. 299-308.
C. Lin (2012), “Exchange Rate Uncertainty and Trade”. The B.E. Journal of Macroeconomics, v. 12 (1), pp. 1-35.
L.D. Marques (2000), “Modelos Dinâmicos com Dados em Painel: Revisão de Literatura”. Faculty of Economics of Porto, working paper.
I. Martinez-Zarzoso, N.L.D. Felicitas and N. Horsewood (2009), “Are Regional Trading Agreements Beneficial? Static and Dynamic Panel Gravity Models”. North American Journal of Economics and Finance 20 (2009) pp. 46–65.
T. Mukhtar and S.J. Malik (2010), “Exchange Rate Volatility and Export Growth: Evidence from Selected South Asian Countries”. University of Piraeus, 60(3-4), pp. 58-68.
C. Hsiao (2003), “Analysis of Panel Data”. Econometric Society Monographs: Cambridge University Press, Second Edition.
33 R.S. Pindyck and D.L. Rubinfeld (2004), “Econometria: Modelos e Previsões”. Translation: Honório Kume e Hamilton Massataki Kai. Tradução de: Econometric Models and Economic Forecasts. Rio de Janeiro: Elsevier Ltda, 4ª. Edition.
G.E. Schuh (1974), “The Exchange Rate and U.S. Agriculture”. American Journal of Agricultural Economics 56, pp.1-13.
I. Soloaga and A. Winters (2001), “Regionalism in the Nineties: What Effect on Trade?” The North American Journal of Economics and Finance 12, pp.1-29.
J.G Thursby and M.C Thursby (1987), “Bilateral Trade Flows, the Linder Hypothesis and Exchange Rate Risk”. Review of Economics and Statistics 69, pp. 488-95.
J. Tinbergen (1962), “An Analysis of World Trade Flows.” Shaping the world economy, New York, NY, Twentieth Century Fund.
D. Trefler (1995), TREFLER, D., “The Case of Missing Trade and Other Mysteries.” The American Economic Review, December 1995, pp.1029-1046.
