论文部分内容阅读
Abstract ： This paper analyzes the data consisting of European Union international trade information with difference-in-difference method
and fixed-effect estimator. The volume of international trade of countries in a currency union is increased greatly overall. But the size of benefit varies from country to country.
Introduction
For many years， a single currency has been a dream of international merchants and investors. Since 1999， with the introduction of the euro， the European Union has taken the lead on this trial. Trade value grew so significantly after the introduction of the euro that trade flows inside Europe are now the largest in the world， and the Euro has surpassed the US dollar as the highest-valued currency in the world. The most obvious benefit of a currency union is the reduction of transaction costs and exchange rate risks. In addition， sharing a common currency partly eliminate the "home bias" in international trade， which is the phenomenon of more intense trade inside countries than between， as indicated by Rose （2000）. And as the gains in trade benefits the member nations， more countries sharing a similar background （e.g.In the EU） will consider joining the currency union and further magnifying the benefit， leading to a growing increase in trade. Although the effect of a currency union seems to be empirically obvious， this paper intends to confirm the effect econometrically with panel data and more importantly， distinguish between the effect of a common currency and the effect of a stable relative exchange rate.
The European Union provides a promising data source on this topic. Before the introduction of the Euro in 1999， members of the European Union have controlled their relative exchange rates in a close range to reduce exchange rate volatility. Currently， 17 out of 27 countries in the European Union have adopted the Euro. However the remaining 10 countries including United Kingdom， Denmark， Sweden and Poland are each still using their customary currency as official currency. Thus， the comparison between these countries and the countries which have adopted euro will do a good job differentiating the effect of common currency. This difference is estimated to be large according to Rose（2000） with pooled data and OLS estimation. In this paper， I will estimate the size of effect through difference-in-difference method in two-period panel data specifically on countries in EU.
Statement of the empirical problem and main regression equation of interest The increase in trade in European Union is obvious after the introduction of the Euro， especially between 1999 and 2000. All countries in European Union experienced a sudden increase in trade volume， no matter if they have adopted euro or not. This is a reasonable phenomenon since in a closely incorporated market such as Europe， exports and imports of even a single country have effect on trade all other countries. Increased demand for import in Euro zone countries will cause increase in export in all other countries in the EU， and the thus results in higher trade volume. So the introduction of the Euro in 11 countries including large economies like Germany and France changed the overall trade situation in Europe. However my hypothesis is the increase in trade in euro zone countries are higher than the other countries， i.e. the effect is more significant for countries which have adopted the Euro. To test this hypothesis， I construct regression function as follows：
■ = β0 +β1 GDPit +β2poplationit
+β3outofeuroi+β4yearafterit+β5outofeuroi
*yearafterit+αi+uit，
Where ■ is the exports and imports as percentage of country i's GDP in year t，GDPit is the GDP of country i in year t， is the population of country i in year t， is a binary variable which is unity if country i is still using their own currency instead of euro in the year of interest（1999 for Group 1， 2000 for Group 2， 2007 for Group 3， 2008 for Group 4， and 2011 for Group 5）.yearafterit is a binary variable which is unity if t is equal to or larger than 1999 for Group 1， 2000 for Group 2， 2007 for Group 3， 2008 for Group 4， and 2011 for Group 5.
ai is the country-specific effect which do not change over time.
uit represents other influences on exports
and imports.
β3 represents the effect of not adopting euro on trade and represents the effect of shocks over the time period.
The difference-in-difference estimator is β5. It represents the difference of changes in countries which did not adopt euro and countries which did before and after the introduction of euro.
In the regression process， I separated the countries adopted euro to 5 groups by time. The first group is the eleven countries first introduced euro in 1999. Although euro did not replace all their official currency until 2002， the trade between these countries has already been affected after 1999. So I take 1999 as the line of separation. The second group is Greece which joined euro zone in 2000. The third is Slovenia in 2007 and fourth is Cyprus and Malta in 2008. The fifth group is Slovakia which joined in 2009. Afterwards Estonia joined euro zone in 2011. However insufficient data source makes it impossible to test the effect on Estonia. Besides， the later a country introduced euro， the data are more likely to be affected by the European sovereign debt crisis since late 2009. Thus we should take the economic shocks over time into consideration and analyze the results cautiously. Countries in 5 treatment Groups：
Group 1（1999）： Austria Belgium Finland France Germany Ireland Italy Luxembourg Netherlands Portugal Spanish
Group 2（2000）： Greece
Group 3（2007）： Slovenia
Group 4（2008）： Cyprus and Malta
Group 5（2009）： Slovakia
The five groups of countries are the treatment group in this research， and the countries which have not adopted the Euro at the time are the control group. （eg. For Group 3， control group includes Cyprus， Malta， Estonia and the 10 countries which never adopted the Euro. So control group varies by the corresponding treatment group. ） By running regression on the respective sample data between them we can get the difference-in-difference estimator.
Aside from the difference-in-difference approach， I tried the fixed-effect estimator to confirm the size of effect of adopting euro. The regression function is as follows：
+β2 （poplationit-populationi） +β3（euroit-euroi）+uit-ui
euroit is a binary variable which is unity if country i has adopted euro in year t.
ai has been eliminated in the subtraction of average so is no longer in the regression function.
Data sources
I choose the World Development Indicators （WDI） from world databank. The data include all 27 countries in European Union from 1993 to 2010. 17 of them have already adopted euro and the other 10 have not. The original data provide exports as percentage of GDP and imports as percentage of GDP separately. By suggestion I take the sum of them to represent the international trade volume of a country. The data also includes the total population of the 27 countries through 1993 to 2010 and GDP in current US dollars. The data source is very reliable.
Unlike most researches on the relevant topic， I prefer the exports and imports as percentage of total GDP to be the dependent variable， since countries like Germany and Luxemburg have very different GDP because of the size of the nations. Therefore only by comparing the value of trade we may get very misleading results. By using the percentage of GDP this bias is eliminated.
Possible econometric problems
Before running the difference-indifference regression I analyzed the data source and find it possible that some coefficients may be difficult to estimate due to collinearity and limited number of observations. But regression results do not show those problems. However， some of the estimates have particularly large standard deviations. This greatly limits the reliability of the estimate. One possible reason of this is that the binary independent variables lack variability in sample data. Thus it is difficult to fix this while still having binary difference-in-difference variables. Instrument variable is not necessary for DID estimates. The fixed effect estimate is economically
significant as well as convincible.
Results
As shown in Table 2 in Appendix， the difference-in-difference estimates of the five groups include both positive and negative results. For group 1（11 countries in 1999）， the DID estimate is -6.95， which means the sum of exports and imports as percentage of GDP of countries which did not adopt euro is 6.95 lower than the countries which did adopt euro. This shows that 11 countries in the first group did benefit from the currency union more than the non-eurozone countries. However， group 2， 4 and 5 has positive DIDs （2.22， 8.29， 2.45）， which means Greece， Cyprus， Malta and Slovakia did not receive as much benefit as the countries which did not join the currency union during same time period. Slovenia in group 3 did receive more benefit than the non-eurozone countries but the difference is relatively insignificant economically （-3.34）. All five DID estimates are economically significant but none of them are statistically significant. The t-stats are all smaller than 0.5.
As shown in Table 3 in Appendix， the fixed effect estimates show that both difference in population and GDP have insignificant effect in international trade. By contrast， euro has a very large coefficient estimate （18.83）， which means the difference in joining euro zone or not can make an 18.83 gap in the sum of exports and imports as percentage of GDP. （eg. From 10% to 28.83%）
Conclusion
The result from difference-in-difference regression shows the effects of introducing euro on different groups of countries are different. Austria， Belgium， Finland， France， Germany， Ireland， Italy， Luxembourg， Netherlands， Portugal， and Spain are the first to adopt euro in 1999 and they receive significant higher benefit than the non-eurozone countries. However， Slovenia receives a smaller benefit over the non-eurozone countries and Greece， Cyprus， Malta and Slovakia did not receive as much benefit as the countries which never adopted euro as their official currency. This difference has two possible explanations：
I.The difference in the economy structure cause the difference in benefit received through currency union. We can easily see that the 11 countries which first adopted euro include 3 of the largest economies in Europe： Germany， France and Italy. These countries are relatively more developed than Greece， Cyprus， etc. Thus they are more likely to have absolute advantage in production. Most models in international trade suggest both sides can benefit from trade， but the one with absolute advantage in production usually benefit more. There are quite a few powerful economies among the countries never joined euro zone as well， eg.UK， Denmark， Sweden. So the countries with little absolute advantage in production can receive smaller benefit than those countries. II.The size of impact is different as well. In 1999， 11 countries adopted euro together. And after that only one or two countries joined euro zone in a year. So the first group has an enormously larger impact on international trade than the latter ones. So the adoption of euro in Slovenia gives the country smaller benefit than in the 11 countries in 1999.
Besides， the data on Cyprus， Malta and Slovakia are already affected by the European sovereign debt crisis， which may cause a lot of change to the international trade of these countries. However the currency union and its inflation policy is a trigger of the crisis as well. This crisis should affect whole euro zone and the tradeoff between risk and benefit from currency union is still debatable.
As shown in Table 3， the fixed-effect estimate show very obvious result on contrast. The coefficient of GDP and population are both insignificant. Only euro makes a significant difference in international trade. This confirms the benefit from a currency union is large overall.
From all above I conclude the total effect of currency union on international trade in positive and significant. But the effect of common currency differs from country to country and time to time. So it is possible that a country which did not join the currency union gain higher benefit than a country which did. Thus there is no definite optimum response for any country. Many articles， both academic and non-academic， have discussed the possible solution and outcome. There is a lot of interest focusing on various aspects of currency union. Frankel （1999） discusses the currency regime for optimum currency area （OCA）. The conclusion is intermediate solution is best for it and this provides a good way out for monetary management in Europe. Gruben， Koo and Millis （2002） suggest that countries in a currency union are more likely to have synchronized or inverse business cycles based on different situations. Lots of news reviews have commented the pros and cons in currency union and most of them take euro zone as the main example. A currency union like euro zone has benefit as well as risks， and the overall valuation cannot be done until European Union figure out their solution for the current crisis.
Table 1： Country list
1 Austria；2 Belgium；3 Bulgaria； 4 Cyprus；
5 Czech Republic；6 Denmark；7 Estonia（dropped because of insufficient data）； 8Finland；9 France；10 Germany；11Greece；12Hungary； 13 Ireland；14 Italy； 15 Latvia； 16 Lithuania； 17 Luxembourg； 18 Malta；19 Netherlands； 20 Poland； 21 Portugal； 22 Romania； 23 Slovakia；24 Slovenia； 25 Spain； 26 Sweden； 27 United Kingdom
Table 2： Result of Difference-in-Difference estimates.
Dependent Variable： ex_importperc
Note： The quantities in the parentheses below the estimates are the standard errors.
Table 3： Result of fixed-effect estimates.
Dependent Variable： diff_ex_importperc
Note： The quantities in the parentheses below the estimates are the standard errors.
References
[1] Rose， A. （2000）. One Money， One Market： Estimating The Effect of Common Currencies on Trade. California Management Review， Vol. 42， No. 2 （Winter 2000）， pp. 52-62.
[2] Frankel， J. （1999）. No Single Currency Regime is Right for All Countries or All Times. National Bureau of Economic Research Working paper.
[3] Gruben， W.， Koo ， J. and Millis， E.
（2002） How Much Does International Trade Affect Business Cycle Synchronization. Federal Reserve Bank of Dallas Working paper.
（Instructor： Professor Junichi Suzuki University of Toronto）
and fixed-effect estimator. The volume of international trade of countries in a currency union is increased greatly overall. But the size of benefit varies from country to country.
Introduction
For many years， a single currency has been a dream of international merchants and investors. Since 1999， with the introduction of the euro， the European Union has taken the lead on this trial. Trade value grew so significantly after the introduction of the euro that trade flows inside Europe are now the largest in the world， and the Euro has surpassed the US dollar as the highest-valued currency in the world. The most obvious benefit of a currency union is the reduction of transaction costs and exchange rate risks. In addition， sharing a common currency partly eliminate the "home bias" in international trade， which is the phenomenon of more intense trade inside countries than between， as indicated by Rose （2000）. And as the gains in trade benefits the member nations， more countries sharing a similar background （e.g.In the EU） will consider joining the currency union and further magnifying the benefit， leading to a growing increase in trade. Although the effect of a currency union seems to be empirically obvious， this paper intends to confirm the effect econometrically with panel data and more importantly， distinguish between the effect of a common currency and the effect of a stable relative exchange rate.
The European Union provides a promising data source on this topic. Before the introduction of the Euro in 1999， members of the European Union have controlled their relative exchange rates in a close range to reduce exchange rate volatility. Currently， 17 out of 27 countries in the European Union have adopted the Euro. However the remaining 10 countries including United Kingdom， Denmark， Sweden and Poland are each still using their customary currency as official currency. Thus， the comparison between these countries and the countries which have adopted euro will do a good job differentiating the effect of common currency. This difference is estimated to be large according to Rose（2000） with pooled data and OLS estimation. In this paper， I will estimate the size of effect through difference-in-difference method in two-period panel data specifically on countries in EU.
Statement of the empirical problem and main regression equation of interest The increase in trade in European Union is obvious after the introduction of the Euro， especially between 1999 and 2000. All countries in European Union experienced a sudden increase in trade volume， no matter if they have adopted euro or not. This is a reasonable phenomenon since in a closely incorporated market such as Europe， exports and imports of even a single country have effect on trade all other countries. Increased demand for import in Euro zone countries will cause increase in export in all other countries in the EU， and the thus results in higher trade volume. So the introduction of the Euro in 11 countries including large economies like Germany and France changed the overall trade situation in Europe. However my hypothesis is the increase in trade in euro zone countries are higher than the other countries， i.e. the effect is more significant for countries which have adopted the Euro. To test this hypothesis， I construct regression function as follows：
■ = β0 +β1 GDPit +β2poplationit
+β3outofeuroi+β4yearafterit+β5outofeuroi
*yearafterit+αi+uit，
Where ■ is the exports and imports as percentage of country i's GDP in year t，GDPit is the GDP of country i in year t， is the population of country i in year t， is a binary variable which is unity if country i is still using their own currency instead of euro in the year of interest（1999 for Group 1， 2000 for Group 2， 2007 for Group 3， 2008 for Group 4， and 2011 for Group 5）.yearafterit is a binary variable which is unity if t is equal to or larger than 1999 for Group 1， 2000 for Group 2， 2007 for Group 3， 2008 for Group 4， and 2011 for Group 5.
ai is the country-specific effect which do not change over time.
uit represents other influences on exports
and imports.
β3 represents the effect of not adopting euro on trade and represents the effect of shocks over the time period.
The difference-in-difference estimator is β5. It represents the difference of changes in countries which did not adopt euro and countries which did before and after the introduction of euro.
In the regression process， I separated the countries adopted euro to 5 groups by time. The first group is the eleven countries first introduced euro in 1999. Although euro did not replace all their official currency until 2002， the trade between these countries has already been affected after 1999. So I take 1999 as the line of separation. The second group is Greece which joined euro zone in 2000. The third is Slovenia in 2007 and fourth is Cyprus and Malta in 2008. The fifth group is Slovakia which joined in 2009. Afterwards Estonia joined euro zone in 2011. However insufficient data source makes it impossible to test the effect on Estonia. Besides， the later a country introduced euro， the data are more likely to be affected by the European sovereign debt crisis since late 2009. Thus we should take the economic shocks over time into consideration and analyze the results cautiously. Countries in 5 treatment Groups：
Group 1（1999）： Austria Belgium Finland France Germany Ireland Italy Luxembourg Netherlands Portugal Spanish
Group 2（2000）： Greece
Group 3（2007）： Slovenia
Group 4（2008）： Cyprus and Malta
Group 5（2009）： Slovakia
The five groups of countries are the treatment group in this research， and the countries which have not adopted the Euro at the time are the control group. （eg. For Group 3， control group includes Cyprus， Malta， Estonia and the 10 countries which never adopted the Euro. So control group varies by the corresponding treatment group. ） By running regression on the respective sample data between them we can get the difference-in-difference estimator.
Aside from the difference-in-difference approach， I tried the fixed-effect estimator to confirm the size of effect of adopting euro. The regression function is as follows：
+β2 （poplationit-populationi） +β3（euroit-euroi）+uit-ui
euroit is a binary variable which is unity if country i has adopted euro in year t.
ai has been eliminated in the subtraction of average so is no longer in the regression function.
Data sources
I choose the World Development Indicators （WDI） from world databank. The data include all 27 countries in European Union from 1993 to 2010. 17 of them have already adopted euro and the other 10 have not. The original data provide exports as percentage of GDP and imports as percentage of GDP separately. By suggestion I take the sum of them to represent the international trade volume of a country. The data also includes the total population of the 27 countries through 1993 to 2010 and GDP in current US dollars. The data source is very reliable.
Unlike most researches on the relevant topic， I prefer the exports and imports as percentage of total GDP to be the dependent variable， since countries like Germany and Luxemburg have very different GDP because of the size of the nations. Therefore only by comparing the value of trade we may get very misleading results. By using the percentage of GDP this bias is eliminated.
Possible econometric problems
Before running the difference-indifference regression I analyzed the data source and find it possible that some coefficients may be difficult to estimate due to collinearity and limited number of observations. But regression results do not show those problems. However， some of the estimates have particularly large standard deviations. This greatly limits the reliability of the estimate. One possible reason of this is that the binary independent variables lack variability in sample data. Thus it is difficult to fix this while still having binary difference-in-difference variables. Instrument variable is not necessary for DID estimates. The fixed effect estimate is economically
significant as well as convincible.
Results
As shown in Table 2 in Appendix， the difference-in-difference estimates of the five groups include both positive and negative results. For group 1（11 countries in 1999）， the DID estimate is -6.95， which means the sum of exports and imports as percentage of GDP of countries which did not adopt euro is 6.95 lower than the countries which did adopt euro. This shows that 11 countries in the first group did benefit from the currency union more than the non-eurozone countries. However， group 2， 4 and 5 has positive DIDs （2.22， 8.29， 2.45）， which means Greece， Cyprus， Malta and Slovakia did not receive as much benefit as the countries which did not join the currency union during same time period. Slovenia in group 3 did receive more benefit than the non-eurozone countries but the difference is relatively insignificant economically （-3.34）. All five DID estimates are economically significant but none of them are statistically significant. The t-stats are all smaller than 0.5.
As shown in Table 3 in Appendix， the fixed effect estimates show that both difference in population and GDP have insignificant effect in international trade. By contrast， euro has a very large coefficient estimate （18.83）， which means the difference in joining euro zone or not can make an 18.83 gap in the sum of exports and imports as percentage of GDP. （eg. From 10% to 28.83%）
Conclusion
The result from difference-in-difference regression shows the effects of introducing euro on different groups of countries are different. Austria， Belgium， Finland， France， Germany， Ireland， Italy， Luxembourg， Netherlands， Portugal， and Spain are the first to adopt euro in 1999 and they receive significant higher benefit than the non-eurozone countries. However， Slovenia receives a smaller benefit over the non-eurozone countries and Greece， Cyprus， Malta and Slovakia did not receive as much benefit as the countries which never adopted euro as their official currency. This difference has two possible explanations：
I.The difference in the economy structure cause the difference in benefit received through currency union. We can easily see that the 11 countries which first adopted euro include 3 of the largest economies in Europe： Germany， France and Italy. These countries are relatively more developed than Greece， Cyprus， etc. Thus they are more likely to have absolute advantage in production. Most models in international trade suggest both sides can benefit from trade， but the one with absolute advantage in production usually benefit more. There are quite a few powerful economies among the countries never joined euro zone as well， eg.UK， Denmark， Sweden. So the countries with little absolute advantage in production can receive smaller benefit than those countries. II.The size of impact is different as well. In 1999， 11 countries adopted euro together. And after that only one or two countries joined euro zone in a year. So the first group has an enormously larger impact on international trade than the latter ones. So the adoption of euro in Slovenia gives the country smaller benefit than in the 11 countries in 1999.
Besides， the data on Cyprus， Malta and Slovakia are already affected by the European sovereign debt crisis， which may cause a lot of change to the international trade of these countries. However the currency union and its inflation policy is a trigger of the crisis as well. This crisis should affect whole euro zone and the tradeoff between risk and benefit from currency union is still debatable.
As shown in Table 3， the fixed-effect estimate show very obvious result on contrast. The coefficient of GDP and population are both insignificant. Only euro makes a significant difference in international trade. This confirms the benefit from a currency union is large overall.
From all above I conclude the total effect of currency union on international trade in positive and significant. But the effect of common currency differs from country to country and time to time. So it is possible that a country which did not join the currency union gain higher benefit than a country which did. Thus there is no definite optimum response for any country. Many articles， both academic and non-academic， have discussed the possible solution and outcome. There is a lot of interest focusing on various aspects of currency union. Frankel （1999） discusses the currency regime for optimum currency area （OCA）. The conclusion is intermediate solution is best for it and this provides a good way out for monetary management in Europe. Gruben， Koo and Millis （2002） suggest that countries in a currency union are more likely to have synchronized or inverse business cycles based on different situations. Lots of news reviews have commented the pros and cons in currency union and most of them take euro zone as the main example. A currency union like euro zone has benefit as well as risks， and the overall valuation cannot be done until European Union figure out their solution for the current crisis.
Table 1： Country list
1 Austria；2 Belgium；3 Bulgaria； 4 Cyprus；
5 Czech Republic；6 Denmark；7 Estonia（dropped because of insufficient data）； 8Finland；9 France；10 Germany；11Greece；12Hungary； 13 Ireland；14 Italy； 15 Latvia； 16 Lithuania； 17 Luxembourg； 18 Malta；19 Netherlands； 20 Poland； 21 Portugal； 22 Romania； 23 Slovakia；24 Slovenia； 25 Spain； 26 Sweden； 27 United Kingdom
Table 2： Result of Difference-in-Difference estimates.
Dependent Variable： ex_importperc
Note： The quantities in the parentheses below the estimates are the standard errors.
Table 3： Result of fixed-effect estimates.
Dependent Variable： diff_ex_importperc
Note： The quantities in the parentheses below the estimates are the standard errors.
References
[1] Rose， A. （2000）. One Money， One Market： Estimating The Effect of Common Currencies on Trade. California Management Review， Vol. 42， No. 2 （Winter 2000）， pp. 52-62.
[2] Frankel， J. （1999）. No Single Currency Regime is Right for All Countries or All Times. National Bureau of Economic Research Working paper.
[3] Gruben， W.， Koo ， J. and Millis， E.
（2002） How Much Does International Trade Affect Business Cycle Synchronization. Federal Reserve Bank of Dallas Working paper.
（Instructor： Professor Junichi Suzuki University of Toronto）