@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Business, Sauder School of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Kei, Wendy Wai Yee"@en ; dcterms:issued "2011-01-25T19:14:31Z"@en, "2011"@en ; vivo:relatedDegree "Master of Science in Business - MScB"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """This paper uses panel regression techniques and trade gravity models to explore the linkages between Open Skies agreement (OSA) signed by the United States and recent bilateral trade development. US bilateral trade in services data is not available; thus only US merchandise trade by air data is used as dependent variable in this paper’s econometric analysis. US merchandise trade by air series has not experienced significant growth since the implementation of numerous Open Skies agreements in 2007. Few studies have analyzed the relationship between OSAs and trade. These provide the motivation for exploring if signing more Open Skies agreements continues to benefit recent US merchandise trade by air development, and if the performance of these policies depends on other macroeconomic factors and on the properties of the agreement itself. Using data between years 2004 and 2009, panel regression models suggest that the performance of Open Skies agreements are not robust to market volatilities. Reductions in air cargo costs and expansions of air markets resulting from OSAs are not strong enough to combat trade declines when the recession hits. On the other hand, free trade agreements exert large, positive influences to US trade by air even during times of economic slowdown. Yet, the duration of the Open Skies agreements and the economic power of the trading partners do influence the performance of these policies. The preferred model specifications are different for exports and imports by air data, which confirms that the performance of OSAs on exports is different from that on imports. Finally, model results indicate that the impact of Open Skies policies on passenger traffic flows indirectly improves US trade by air figures. OSAs have stimulated passenger traffic growth, and model results suggest that lagged passenger traffic is positively related to trade value. Increased business travel opportunities conducted prior to the delivery of the goods help lower information asymmetries and develop trust among the supply chain partners. Combination of these effects aids expansion of trade by air, as well as trade by other modes of transportation."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/30817?expand=metadata"@en ; skos:note "EXPLORING THE LINKAGES BETWEEN OPEN SKIES AGREEMENTS SIGNED BY THE UNITED STATES AND INTERNATIONAL TRADE DEVELOPMENT by WENDY WAI YEE KEI B.Com. (Hon.), The University of British Columbia, 2007 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN BUSINESS ADMINISTRATION in The Faculty of Graduate Studies THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) January 2011 © Wendy Wai Yee Kei, 2011 ii Abstract This paper uses panel regression techniques and trade gravity models to explore the linkages between Open Skies agreement (OSA) signed by the United States and recent bilateral trade development. US bilateral trade in services data is not available; thus only US merchandise trade by air data is used as dependent variable in this paper’s econometric analysis. US merchandise trade by air series has not experienced significant growth since the implementation of numerous Open Skies agreements in 2007. Few studies have analyzed the relationship between OSAs and trade. These provide the motivation for exploring if signing more Open Skies agreements continues to benefit recent US merchandise trade by air development, and if the performance of these policies depends on other macroeconomic factors and on the properties of the agreement itself. Using data between years 2004 and 2009, panel regression models suggest that the performance of Open Skies agreements are not robust to market volatilities. Reductions in air cargo costs and expansions of air markets resulting from OSAs are not strong enough to combat trade declines when the recession hits. On the other hand, free trade agreements exert large, positive influences to US trade by air even during times of economic slowdown. Yet, the duration of the Open Skies agreements and the economic power of the trading partners do influence the performance of these policies. The preferred model specifications are different for exports and imports by air data, which confirms that the performance of OSAs on exports is different from that on imports. Finally, model results indicate that the impact of Open Skies policies on passenger traffic flows indirectly improves US trade by air figures. OSAs have stimulated passenger traffic growth, and model results suggest that lagged passenger traffic is positively related to trade value. Increased business travel opportunities conducted prior to the delivery of the goods help lower information asymmetries and develop trust among the supply chain partners. Combination of these effects aids expansion of trade by air, as well as trade by other modes of transportation. iii Table of contents Abstract .................................................................................................................................................. ii Table of contents.................................................................................................................................... iii List of tables ........................................................................................................................................... iv List of figures ........................................................................................................................................... v List of abbreviations ............................................................................................................................... vi Acknowledgements ............................................................................................................................... vii Chapter 1 Introduction ..................................................................................................................... 1 1.1 Literature review ........................................................................................................................... 2 1.2 Gaps in Open Skies agreements research and objectives of this paper ........................................... 5 Chapter 2 Background information on Open Skies agreements ........................................................ 9 2.1 Open Skies agreements in Canada and in the United States ........................................................... 9 2.2 Basic trade statistics comparison between Canada and U.S. ......................................................... 18 2.3 Normalized plot of US merchandise trade by air transport ........................................................... 21 2.4 Concluding remarks ..................................................................................................................... 23 Chapter 3 The model ...................................................................................................................... 25 3.1 The basic gravity equation ........................................................................................................... 26 3.2 The augmented gravity model ..................................................................................................... 32 3.3 Model issues and considerations.................................................................................................. 39 3.4 Data sources ................................................................................................................................ 40 Chapter 4 Econometric results and economic implications ............................................................ 46 4.1 Open Skies agreement variables .................................................................................................. 46 4.2 Alternative views of the augmented gravity models ..................................................................... 61 4.3 Preferred model specification ...................................................................................................... 66 4.4 Concluding remarks on model results .......................................................................................... 67 Chapter 5 Conclusion .................................................................................................................... 69 References ............................................................................................................................................ 73 iv List of tables Table 1: Open Skies and Open Skies-type partners for Canada ............................................................... 11 Table 2: Imports by merchandise: top 10 markets ................................................................................. 13 Table 3: Exports by merchandise: top 10 markets .................................................................................. 15 Table 4: Conventional gravity model results for US merchandise exports by air ..................................... 30 Table 5: Conventional gravity model results for US merchandise imports by air ..................................... 31 Table 6: Variables and regression models used for augmented gravity equation analysis ....................... 38 Table 7: Countries / regions in panel regression model .......................................................................... 42 Table 8: Country income groups based on World Bank’s classification ................................................... 43 Table 9: Free trade agreements signed with the United States .............................................................. 44 Table 10: Currency unions with the United States .................................................................................. 45 Table 11: Open Skies effects on US merchandise exports by air ............................................................. 48 Table 12: Open Skies effects on US merchandise imports by air ............................................................. 49 Table 13: Impact of other Open Skies-related variables on US merchandise exports by air .................... 50 Table 14: Impact of other Open Skies-related variables on US merchandise imports by air .................... 50 Table 15: Impact of other control variables on US merchandise exports by air ....................................... 51 Table 16: Impact of other control variables on US merchandise imports by air ...................................... 52 Table 17: US merchandise exports by air: top 10 markets ...................................................................... 55 Table 18: US merchandise imports by air: top 10 markets...................................................................... 56 Table 19: US trading partners contributing most to the GDP interaction term ....................................... 61 Table 20: Davidson and MacKinnon (1981) J-Test results ....................................................................... 67 Table 21: Akaike Information Criteria results ......................................................................................... 67 v List of figures Figure 1: US Open Skies agreements concluded per year, segregated by income groups........................ 18 Figure 2: Exports (normalized at time of first Open Skies agreements) .................................................. 20 Figure 3: Imports (normalized at time of first Open Skies agreements) .................................................. 20 Figure 4: Exports (normalized at time of new Open Skies agreements / Blue Sky policies initiation) ....... 21 Figure 5: Imports (normalized at time of new Open Skies agreements / Blue Sky policies initiation) ...... 21 Figure 6: US exports by air versus aggregate exports (normalized) ......................................................... 22 Figure 7: US imports by air versus aggregate imports (normalized) ........................................................ 22 Figure 8: Share of trade in commercial services ..................................................................................... 32 Figure 9: US net exports of goods and services ...................................................................................... 60 Figure 10: Residual and normal quantile-quantile plots ......................................................................... 64 vi List of abbreviations ASA Air Service Agreements CU Currency Unions FC Foreign Currency FTA Free Trade Agreements GDP Gross Domestic Product GNI Gross National Income HHI Herfindahl-Hirschman Index IMF International Monetary Funds IMF WEO International Monetary Funds’ World Economic Outlook database NAFTA North America Free Trade Agreement OLS Ordinary Least Squares OSA Open Skies Agreements US United States USD US Dollar WTO World Trade Organization vii Acknowledgements I thank Dr. David Gillen, Dr. Tae Oum, and Dr. Anming Zhang for their helpful comments and suggestions. Special thanks to the Social Sciences and Humanities Research Council of Canada (SSHRC) for selecting me as the winner of the Joseph Armand Bombardier Canada Graduate Scholarship (Master’s level). This thesis was supported by this scholarship. Most importantly, I thank my parents for their love and support. 1 Chapter 1 Introduction1 Since deregulation in 1978, the United States has been active in facilitating liberalized air service agreements to stimulate economic growth. Liberalized air service agreement contains less restrictive terms and conditions for air transport services between and beyond the country pair that engaged in this policy (Clougherty, Dresner, & Oum, 2001). Currently, the U.S. has reached 100 “full” Open Skies agreements (U.S. Department of State, 2010). These Open Skies policies contain unrestricted first through sixth freedoms, with some country pairs having all cargo seventh freedom (Gillen, 2009). Contrary to its neighbor, Canada has been slow in facilitating free air transport and has classified twelve “Open Skies-type” agreements to date (Foreign Affairs and International Trade Canada, 2010). The significant difference between the two nations in terms of the number of Open Skies agreements (OSA) signed to date raises the question of whether signing more of these agreements will necessarily bring more economic benefits. Numerous studies have suggested that Open Skies agreements are generally beneficial. However, research that supports Open Skies agreements or liberalized air service agreements has focused primarily on passenger traffic volumes in North America. Recent papers evaluating the impact of US Open Skies agreements on air cargo growth or trade volumes are generally inconclusive or produce contradictory results. As such, this paper, using more up-to-date information, explores the mechanisms that would underlie linkages between Open Skies agreements and bilateral trade developments. This note also examines if US Open Skies policies are robust to other economic factors, since trade plays an important role to economic growth. Furthermore, this research determines whether the effectiveness of Open Skies agreements changes depending on the number of years from which an OSA was signed, the distance between the pair of countries, and the economic size of the trading partner. This chapter reviews research conducted on Open Skies agreements, discusses gaps in this field of research, and ends with objectives of this paper. 1 Data and literatures cited in this paper were retrieved on or before December 11, 2010. 2 1.1 Literature review Overall, most Open Skies research has focused primarily on passenger traffic volumes for both Canadian and US experiences and these studies have shown that Open Skies policies tend to be beneficial. However, these studies seem to suggest one side of Open Skies agreements. Minimal studies have been conducted relating liberalized air service agreements or OSAs to trade development, and past papers have tended to find either contradictory or inconclusive results. As such, it is important to review these Open Skies literatures conducted using passenger traffic data and using merchandise trade data to determine if these studies together can help improve models relating Open Skies agreements and bilateral trade development. Several researchers have illustrated that both liberalized air service agreements and Open Skies agreements help increase competition and stimulate passenger traffic growth. According to Clougherty, Dresner, and Oum (2001), full Open Skies agreements (as seen in the US) and partial liberalization policies (as seen in Canada) tend to increase passenger traffic volumes and market shares for both domestic and foreign carriers, based on panel data between years 1982 and 1994. InterVISTAS Consulting Inc. (2006), focusing primarily on passenger traffic, shows that traffic growth would increase by 12 – 35% after countries engage in liberalized agreements. Dresner and Oum (1998) find that more passengers would travel directly to Canada, rather than passing through the US if Canadian air service agreements were more liberalized. Their study illustrates using simulations that annual traffic travelling directly to Canada would increase by roughly 2-3% (which translates to annual increases of 7100 – 9000 passengers) if Canada had facilitating bilateral agreements with a foreign country (regardless if US had liberalized agreements with the same foreign country). However, the same study also shows that Canada would incur roughly 4-5% annual loss in passenger traffic growth (about 2800-3500 passengers) travelling directly to Canada if US had liberalized ASA with another foreign country of which Canada did not have (Dresner & Oum, 1998). Moreover, Endo (2007) uses the gravity equation and panel estimation techniques to show that Open Skies policies tend to increase US import of air services. In other words, full liberalized agreements encourage more US citizens taking foreign carriers to travel between the US and other countries, and Endo’s (2007) results are consistent with the findings from Clougherty, Dresner, and Oum (2001). This also suggests that liberalized agreements have promoted more competition in the airline industry. 3 Past research has also suggested that Open Skies agreements promote economic growth and profitability. Gillen, Harris, and Oum (2002) show that even under partial liberalization framework, consumer surplus and overall profitability for Canadian and Japanese carriers expect to increase when pricing, entry, and frequency regulations are removed. Governmental agencies, such as BC Ministry of Transportation and Infrastructure (2009) have also published reports suggesting that Open Skies policies bring positive impact to the tourism industry. Furthermore, InterVISTAS Consulting Inc. (2006)2 suggests that 24.1 million full-time job positions and an additional 490 billion USD in gross domestic product would be generated if 320 bilateral agreements of any pairs of countries were signed. The magnitude of additional GDP resulting from liberalized air service agreements corresponds to roughly the size of Brazil’s GDP (InterVISTAS Consulting Inc., 2006). In contrast to the numerous studies relating Open Skies agreements or liberalized air service agreements to passenger traffic volumes, few number of research and consulting reports has been done to assess the impact OSAs have on international trade or cargo traffic, which is one of the major gaps in this area of research. Consulting reports such as InterVISTAS (2006) have near zero analysis on finding the relationship between international trade (or cargo) and liberalized air service agreements. The “Air Freight Model” used in InterVISTAS (2006) does not relate trade or cargo traffic directly to ASAs. Instead, the model determines the additional cargo capacity that the belly space of the passenger flight could carry given increases in passenger volumes resulting from Open Skies agreements (InterVISTAS Consulting Inc., 2006). Results linking trade to Open Skies policy as determined by econometrics techniques also tend to be different from expectations. For example, Yamaguchi (2008) shows using US Exports of Merchandise Trade data between years 1998 and 2002 for 21 trading partners that Open Skies agreements tend to lower air transport unit cost. The model also suggests that these policies increase market concentration as measured by the Herfindahl-Hirschman Index, which is contrary to the expectation that Open Skies increases competition (i.e. lower HHI value) (Yamaguchi, 2008). However, Yamaguchi’s (2008) study analyzes exports data. Imports data should also be analyzed, as these data may yield different results. 2 For the rest of this note, the citation, “InterVISTAS (2006)” is used interchangeably with “InterVISTAS Consulting Inc. (2006)”. 4 Country income effects may contribute to contrary results for research examining the linkages between OSA and trade, and these effects may introduce endogeneity issues into the model. In other words, aggregated panel results may lead to “aggregation bias”. As such, Micco and Serebrisky (2006) have different regression models for different set of country income groups to examine if the effectiveness of OSA changes depending on the trading partner’s economic power. Using U.S. Imports of Merchandise Trade data between years 1990 and 2003, Micco and Serebrisky’s (2006) models show that only developed nations benefit from Open Skies through lower import charges. The effect of Open Skies agreements on import charges for developing nations is small and is not statistically significant. Furthermore, the share of US merchandise imports by air increases when Open Skies policies are implemented for developed and upper-middle income countries. Open Skies effect is not conclusive when developing nations are included into their models. Results calculated by Micco and Serebrisky (2006) seem to contradict with the motivation for signing more Open Skies agreements with developing nations. Furthermore, the same paper also finds that Open Skies agreements signed for longer duration tend to lower transport cost more than those with shorter duration. However, Micco and Serebrisky’s (2006) models do not analyze exports data or trade in services data and it is unclear whether incorporating exports data and/or trade in services data would produce similar results. To summarize, studies analyzing the relationship between Open Skies agreements and international trade are quite preliminary. Improvement opportunities exist in this field of research. 5 1.2 Gaps in Open Skies agreements research and objectives of this paper Numerous gaps exist in research exploring the linkages between Open Skies agreements and international trade. Using econometrics approaches, this note narrows the following gaps in order to improve this area of research:  Analyze both exports and imports. Past research (Micco and Serebrisky, 2006; Yamaguchi, 2008) has analyzed exports or imports separately and tended to conclude OSAs-related results from either imports or exports data. However, movements in imports do not imply the same trend for exports and similarly, results for exports cannot be inferred to imports. Also, understanding the movement of net exports (exports less imports) is important to economic welfare, as it influences the growth in gross domestic product. If the impact resulting from Open Skies is higher for imports than for exports for the same time period, it may be possible to imply that the net export is negative, holding all else constant. This would suggest decline in economic growth (measured in terms of gross domestic product in expenditure terms).  Incorporate other factors that may influence merchandise trade into the model. Despite that Yamaguchi (2008) includes NAFTA as a dummy variable into the model, literature exploring the linkages between air-borne merchandise trade (air cargo traffic) and liberalized air service agreements fails to account for other factors, such as trade in services, currency unions, or other free trade agreements. These components could influence merchandise export and import by air values. Failing to account for other influential variables in the model may lead to “omitted variables bias”3. These “other” factors could influence merchandise trade substantially. Baier and Bergstrand (2009) suggest that free trade agreements in general impact trade movements (for both exports and imports), whether log-linear gravity model or matching econometrics techniques are used in the analysis. Using data between 1948 and 1997, Glick and Rose (2002) suggest that currency unions nearly double trade volumes. Certain country groups tend to have more trade in services than others, and merchandise trade and trade in services interact with each other. For example, higher merchandise trade volumes would induce higher demand for labour, such as consulting 3 Omitted variable bias (under-specification) tends to produce biased estimators, small standard errors for the coefficients, and thus lead to invalid conclusions in regression and hypothesis testing analyses (Davidson and MacKinnon, 2004). 6 services, which would count as an increase in trade in services; and increase in consulting services would help produce more goods efficiently, which would stimulate merchandise trade growth. Furthermore, analogous to the results presented in Dresner and Oum (1998) for passenger traffic, trade volumes for the domestic country would reduce if the adjacent nation signed liberalized air service agreements with another foreign country of which the domestic country does not have any facilitating bilateral arrangements with. Given that changes in merchandise trade values may be influenced by other factors, the model should be augmented by including trade in services as another explanatory variable and by adding dummy variables that capture effects from currency union, from free trade agreements, and from other OSAs signed by adjacent nations to determine if any or all of these variables add any value to the model.  Use more recent data. Past studies, as described in the Literature Review section, have used data mostly from 1990s to early 2000s, which would focus on when Open Skies agreements first initiated. However, more recent data should be analyzed to assess whether signing more Open Skies agreements in recent years and having agreements signed from decades ago continue to improve economic development. Furthermore, air transport industry changes dramatically in response to various shocks, such as September 11th and the global economic crisis that started in 2008. Past data cannot help explain these shocks, which are influential to both the economy and the air transportation industry. It is also important to observe the performance of OSAs at times of economic uncertainty, to determine if this effect is robust to other market volatilities.  Analyze all markets. Numerous papers (Endo, 2007; Gillen, 2009; Yamaguchi, 2008) have used “N” markets instead of all markets to conduct panel regression analysis. Including the top markets would introduce survivorship bias, and the effect of OSA would tend to be over-rated. As such, the model should include as many markets as possible in order to reduce this bias. In addition to narrowing gaps in this area of research, this paper uses bivariate relationship, panel regression techniques, and trade gravity models to explore the linkages between US Open Skies agreements and its bilateral trade development in order to evaluate if recent Open Skies agreements remain beneficial. The objectives of this research include the following: 7  Compare and contrast the properties of US Open Skies agreements with Canadian liberalized air service agreements, including examining if the definitions of these policies between both regions vary, if Canada and the United States conclude agreements with countries of the same income group(s), and whether both regions sign agreements with their respective major trading partners. This comparison is important in providing the motivation of assessing whether the properties of the Open Skies agreements affect the performance of these policies.  Analyze and compare US aggregate trade movements with Canada’s before and after implementations of major rounds of liberalized ASAs and/or Open Skies agreements. This analysis uses normalized plots. The motivation for this analysis is to assess if trade growth for the US is much superior to Canada’s. In other words, the focus is to examine whether a country that signed numerous OSAs tend to experience much greater trade growth. Furthermore, if the trade movements after recent rounds of OSAs are relatively flat, as shown by normalized plots, this flat trend may imply that new rounds of Open Skies agreements may not be influential to trade, OSAs may take longer time to flow through, and/or trade may be more heavily influenced by other macroeconomic factors other than OSAs.  Use econometric techniques and the conventional gravity model to illustrate the average effect that Open Skies agreements have on merchandise trade by air. Discuss the shortcomings of the conventional gravity model.  Develop new, augmented gravity models to try to explain how Open Skies agreements influence merchandise trade by air, taking into account of other macroeconomic factors, and to treat the Open Skies agreement for each country pair as distinct as possible. In particular, the augmented gravity model helps explore the possible mechanisms that would underlie a linkage between Open Skies agreements and merchandise trade by air. For example, OSAs is expected to introduce new markets and create more convenient routes, which would reduce air cargo costs. These effects expect to increase air-borne merchandise trade. The implementation of OSAs would also facilitate more business travel, which helps firms acquire more information and develop trust among the supply chain partners. The facilitation of more business travels would lead to expansion of trade opportunities. Furthermore, the models examine if the number of 8 years from which an OSA was signed, the distance between the pair of countries, and the economic size of the trading partner affect the effectiveness of the Open Skies agreements.  Determine if the effect of Open Skies agreements on merchandise exports and imports by air differ. It is expected that the effect of OSA on merchandise exports and imports by air to differ, since the major trading partners for exports and for imports are generally different. The difference in results between these two sets of data confirms the need to analyze both exports and imports data in order to properly assess the value of Open Skies agreements on trade statistics. To address the objectives above, models analyzed in this note take into account the interactions of the trading partner’s income level and Open Skies agreements. As noted previously, Micco and Serebrisky (2006) show that the effects of US Open Skies agreements on air transport costs and shares of imports tend to be different across income groups of its trading partners. As such, this research investigates if the effects of US Open Skies agreements on its recent merchandise export by air values and on its merchandise import by air values would differ depending on US trading partners’ income group classifications as defined by the World Bank4. This paper contains five chapters. Chapter 2 compares and contrasts US Open Skies policies with the Canadian Blue Sky policies. This introductory chapter uses normalized plots to see if Open Skies policies have generally encouraged more trade for Canada and for the US. Chapter 3 discusses the methodology and data used to conduct this study. This section starts by explaining the underlying assumptions and motivations for using the trade gravity model. Chapter 3 also discusses the panel regression models and data series used for the analysis. Chapter 4 discusses the results and their economic implications. Finally, Chapter 5 ends with some concluding remarks, discusses possible future work that can be done in this area of research, and suggests the contribution this research will make to the advancement of knowledge. 4 The US Foreign Trade Division (personal communication, September 7, 2010) confirmed that bilateral trade by services data is not available. Thus, econometric models that are presented in subsequent chapters use merchandise trade by air statistics as dependent variable. 9 Chapter 2 Background information on Open Skies agreements As noted above, the United States has reached 100 Open Skies agreements to date, which is more than eight times the number of Canadian “open skies-type” agreements. As such, one would expect substantial discrepancies in how both countries define Open Skies or liberalized air service agreements. Open Skies policies are also expected to contribute to increases in traffic volumes. Thus, divergence in trade data between the two nations may result from how air services agreements are defined and from the number of these agreements signed for each nation. Furthermore, it is vital to compare the US trade data against Canada’s, to determine if US trade growth is much superior to Canada’s due to the number of Open Skies agreements signed and if the effectiveness of these agreements influence the number of them signed to date. This chapter compares and contrasts the properties of Open Skies agreements in Canada and in the United States. This section also notes interesting trade trends for the two nations. Normalized plots are illustrated in Sections 2.2 and 2.3 to explore if real trade volumes have experienced positive changes resulting from Open Skies policies. 2.1 Open Skies agreements in Canada and in the United States International air services are generally governed by bilateral air service agreements that impose restrictions on the airports that can be served, on flight frequencies and routes, on airfares, on seat allocations, and on air carrier designations, in order to protect each nation’s airline industry (Clougherty, Dresner, & Oum, 2001; Dresner & Oum, 1998). Liberalized air service agreements contain more liberalized terms and conditions for air transport services between and beyond each pair of countries that signed the agreement (Clougherty, Dresner, & Oum, 2001), and Open Skies agreements are intended to contain terms and conditions that are as liberal as possible (Gillen, 2009). InterVISTAS Consulting Inc. (2006) suggests that “full” Open Skies agreements follow market structure, in that any airlines from either signatory nation are flexible in setting airfares, routes and frequencies, and capacity arrangements between and beyond city pairs based on market forces. Another report by InterVISTAS Consulting Inc. (2005) defines “full” Open Skies agreements as those that include fifth freedom rights for 10 passengers, both fifth and seventh freedom rights for cargo carriers5, as well as “unrestricted sixth freedom fare setting, sixth freedom through flight numbers and cargo co-terminalization” (p.7). However, most ASAs especially in Canada are not classified as “Open Skies” (Clougherty, Dresner, & Oum, 2001). Air service agreements in the United States have generally been more liberal than those in Canada and are generally considered as “Open Skies” agreements. U.S. Department of State (2010) specifies that the goal of US ASAs is to eliminate government interference over airline pricing, routing, and capacity decisions, in order to stimulate economic growth and consumer welfare. U.S. Department of State (2010) outlines eight key provisions that are applicable to both passenger and cargo traffic 6 . Generally, these provisions encourage market competition, competitive pricing, and flexible business models including code-sharing and charter arrangements. Over sixty percent of the US OSAs signed also contain “seventh freedom all-cargo rights”, which means that an airline of one country can operate all- cargo services between the foreign country and another third country without the home country as flight origin or destination (U.S. Department of State, 2010). On the other hand, Canadian air service agreements tend to protect domestic air carriers rather than encourage consumer welfare despite that Foreign Affairs and International Trade Canada’s (2010) website appears to suggest Canada being more demand (market) driven. At this point, twelve out of 45 Canadian air service agreements are considered as Open Skies or Open Skies-type agreements (Foreign Affairs and International Trade Canada, 2010). This number is far from what the US has, as noted above. Table 1 below lists the countries that have engaged in liberalized air service agreements with Canada. 5 Fifth freedom of rights allows domestic carriers to carry traffic from one foreign country to another foreign country as an extension of service from the domestic country. Seventh freedom of rights allows domestic carriers to carry traffic between two foreign countries, without home country as flight origin or destination (InterVISTAS Consulting Inc., 2005). 6 Refer to (U.S. Department of State, 2010) for details. 11 Table 1: Open Skies and Open Skies-type partners for Canada Countries Year Concluded All-cargo 7th? Country Income Group (as of the year that OSA was concluded) United Kingdom 2007 Yes High income group United States 2007 Yes High income group Ireland 2007 Yes High income group Iceland 2007 Yes High income group New Zealand7 2007 Yes High income group Barbados 2008 Yes High income group Dominican Republic 2008 Yes Upper middle income group Costa Rica 2009 N/A 8 Upper middle income group European Union 2009 N/A9 All members belong to high income group except for Bulgaria, Lithuania, and Romania, which are part of the upper-middle income economies. Republic of Korea 2009 Yes High income group El Salvador 2010 Yes Lower middle income group Switzerland 2010 Yes High income group Sources: Foreign Affairs and International Trade Canada (2010), InterVISTAS Consulting Inc. (2008), World Bank Analytical Classifications (2010) Canadian air service agreements are generally initiated by consulting with domestic carriers and airports, before any negotiations with the foreign country takes place (Foreign Affairs and International Trade Canada, 2010). However, this process tends to protect the airline industry as infants, rather than encouraging the Canadian airline industry to compete and grow. The views of carriers and airports tend to be restrictive and limit the economic gains (such as increases in business and employment opportunities and in gross domestic product) that can be derived from full Open Skies agreements. Air carriers are afraid of foreign competition, which would limit the amount of monopoly profits that they could earn. As such, the views expressed by air carriers and airports would not be in the interest of consumers and businesses. Pricing, routing, and capacity arrangement decisions would not be determined purely by market forces. These arrangements would be contrary to the government’s original goal of encouraging competition and expanding capacity to benefit Canadians. Oum (2007) also 7 Agreement was further updated on July 21, 2009 (Foreign Affairs and International Trade Canada, 2010). 8 The press release on Foreign Affairs and International Trade Canada’s (2010) website notes that air service rights between Costa Rica and Canada contained in this agreement are not immediately available. As such, there are no details on whether the agreement with Costa Rica contains all-cargo seventh freedom. 9 The European Union member states include: Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and the United Kingdom. Except for United Kingdom and Ireland, the effect of the agreement is assumed to commence in 2009, as the agreement was concluded on December 9, 2008 (Foreign Affairs and International Trade Canada, 2010; Gillen, 2009). Gillen (2009) suggests that the details of the agreement are not available to public; thus it is unclear if the all cargo seventh freedoms are included in the agreement. 12 notes that Canadian policies tend to allow capacity expansion only if domestic carriers gain. To summarize, this type of negotiation process is slow and passive, limiting trade volumes and tourism figures, which could lead to stagnant economic growth. The above discussion seems to suggest that the United States would incur higher trade growth than Canada’s given that the US encourages signing more Open Skies agreements. However, it is interesting to note from Tables 2 and 310 that the US has not signed OSA with some of its major trading partners. This questions whether the US could extract even more values from OSA arrangements if it concluded agreements with some of these countries. The US top ten trading partners have composed roughly two- third of US total trade values (for both exports and imports) and these major exporters and importers in recent years are not dramatically different from those of year 1992. China, Mexico, and Japan have not concluded any Open Skies agreements with the US prior to year 200911, but recent data show that these three countries have contributed to roughly 35-40% of US’s total merchandise import values and 25% of US’s total merchandise export values. Similarly, Canada has not signed Open Skies agreements with most of its major trading partners. However, these countries without OSAs have contributed to less than 10% of total Canadian merchandise exports and roughly 20-25% of total Canadian merchandise imports. As such, for both exports and imports, the loss in trade opportunities from not signing OSAs with the major trading partners would be smaller for Canada than for the US. This loss in trade opportunities, especially for US merchandise imports, would most likely be captured under the all-cargo seventh freedom variable. For example, if US signed OSAs with Mexico and China, it would be possible that more OSA-related benefits be derived from the all-cargo seventh freedom12. 10 The tables are constructed similar to those presented in Gillen (2009). The author thanks Dr. David Gillen for sharing his paper, “Canadian International Aviation: Policy and Challenges”. 11 Japan has just reached Open Skies agreement with the United States on October 25, 2010 (U.S. Department of State, 2010). 12 According to Fu, Lei, and Zhang (2010), UPS and FedEx have already implemented hub operations in China due to the all-cargo seventh freedom of the Singapore-US OSA. These cargo hub operations expect to stimulate trade growth between the US and China. Yet, an official China-US Open Skies agreement is expected to further increase (free) trade opportunities. 13 Table 2: Imports by merchandise: top 10 markets United States Year 1992 Year 2007 Year 2009 Rank Country Nominal CIF Import Value Billions of USD Percent of Imports from World Country Nominal CIF Import Value Billions of USD Percent of Imports from World Country Nominal CIF Import Value Billions of USD Percent of Imports from World 1 Canada 101.29 18.3% China 340.12 16.9% China 309.56 19.3% 2 Japan 99.48 18.0% Canada 317.60 15.7% Canada 228.38 14.2% 3 Mexico 35.89 6.5% Mexico 212.89 10.6% Mexico 178.34 11.1% 4 Germany 29.60 5.4% Japan 149.42 7.4% Japan 98.40 6.1% 5 China 27.41 5.0% Germany 96.64 4.8% Germany 72.64 4.5% 6 United Kingdom 20.69 3.7% United Kingdom 58.10 2.9% United Kingdom 48.32 3.0% 7 Korea, Republic of 17.36 3.1% Korea, Republic of 49.32 2.4% Korea, Republic of 40.54 2.5% 8 France 15.26 2.8% France 42.50 2.1% France 34.67 2.2% 9 Italy 12.84 2.3% Venezuela 41.01 2.0% Venezuela 28.79 1.8% 10 Singapore 11.56 2.1% Saudi Arabia 37.16 1.8% Ireland 28.19 1.8% Non-OSA Countries (Grey cells) 780.61 38.7% Non-OSA Countries (Grey cells) 615.08 38.4% * Grey cells denote countries that did not sign Open Skies agreements with the United States at year t, where t = {2007, 2009}. Sources: International Monetary Funds, Direction of Trade Statistics (Retrieved on December 11, 2010); U.S. Department of State (2010) 14 Table 2: Imports by merchandise: top 10 markets (continued) Canada Year 1992 Year 2007 Year 2009 Rank Country Nominal CIF Import Value Billions of USD Percent of Imports from World Country Nominal CIF Import Value Billions of USD Percent of Imports from World Country Nominal CIF Import Value Billions of USD Percent of Imports from World 1 United States 87.22 63.5% United States 226.41 54.1% United States 180.39 51.1% 2 Japan 9.80 7.1% China 39.49 9.4% China 38.40 10.9% 3 United Kingdom 3.75 2.7% Mexico 17.69 4.2% Mexico 16.08 4.6% 4 Germany 3.19 2.3% Japan 15.88 3.8% Japan 11.92 3.4% 5 France 2.44 1.8% Germany 11.86 2.8% Germany 10.28 2.9% 6 Mexico 2.43 1.8% United Kingdom 11.81 2.8% United Kingdom 9.09 2.6% 7 China 2.25 1.6% Korea, Republic of 5.54 1.3% Korea, Republic of 5.75 1.6% 8 Korea, Republic of 1.83 1.3% Norway 5.51 1.3% France 5.43 1.5% 9 Italy 1.59 1.2% Algeria 5.25 1.3% Italy 4.29 1.2% 10 Norway 1.34 1.0% France 5.24 1.3% Algeria 3.68 1.0% Non-OSA Countries (Grey cells) 106.46 25.4% Non-OSA Countries (Grey cells) 70.08 19.9% * Grey cells denote countries that did not sign Open Skies agreements with Canada at year t, where t = {2007, 2009}. Sources: Foreign Affairs and International Trade Canada (2010); International Monetary Funds, Direction of Trade Statistics (Retrieved on December 11, 2010); InterVISTAS Consulting Inc. (2008) 15 Table 3: Exports by merchandise: top 10 markets United States Year 1992 Year 2007 Year 2009 Rank Country Nominal FOB Export Value Billions of USD Percent of Exports from World Country Nominal FOB Export Value Billions of USD Percent of Exports from World Country Nominal FOB Export Value Billions of USD Percent of Exports from World 1 Canada 90.16 20.1% Canada 248.44 21.4% Canada 204.73 19.4% 2 Japan 47.76 10.7% Mexico 136.54 11.7% Mexico 129.00 12.2% 3 Mexico 40.60 9.1% China 65.24 5.6% China 69.58 6.6% 4 United Kingdom 22.81 5.1% Japan 62.66 5.4% Japan 51.18 4.8% 5 Germany 21.24 4.7% United Kingdom 50.30 4.3% United Kingdom 45.71 4.3% 6 Korea Republic of 14.63 3.3% Germany 49.65 4.3% Germany 43.30 4.1% 7 France 14.58 3.3% Korea Republic of 34.70 3.0% Netherlands 32.35 3.1% 8 Netherlands 13.74 3.1% Netherlands 32.99 2.8% Korea Republic of 28.64 2.7% 9 Belgium- Luxembourg 10.05 2.2% France 27.82 2.4% France 27.04 2.6% 10 Singapore 9.62 2.1% Singapore 26.28 2.3% Brazil 26.18 2.5% Non-OSA Countries (Grey cells) 264.44 22.7% Non-OSA Countries (Grey cells) 275.93 26.1% * Grey cells denote countries that did not sign Open Skies agreements with the United States at year t, where t = {2007, 2009}. Sources: International Monetary Funds, Direction of Trade Statistics (Retrieved on December 11, 2010); U.S. Department of State (2010) 16 Table 3: Exports by merchandise: top 10 markets (continued) Canada Year 1992 Year 2007 Year 2009 Rank Country Nominal FOB Export Value Billions of USD Percent of Exports from World Country Nominal FOB Export Value Billions of USD Percent of Exports from World Country Nominal FOB Export Value Billions of USD Percent of Exports from World 1 United States 103.86 77.8% United States 332.02 78.9% United States 220.83 75.1% 2 Japan 6.07 4.6% United Kingdom 11.93 2.8% United Kingdom 10.11 3.4% 3 United Kingdom 2.51 1.9% China 8.91 2.1% China 9.57 3.3% 4 Germany 1.85 1.4% Japan 8.60 2.0% Japan 7.07 2.4% 5 China 1.81 1.4% Mexico 4.64 1.1% Mexico 3.39 1.2% 6 Netherlands 1.18 0.9% Netherlands 3.75 0.9% Germany 2.99 1.0% 7 Korea, Republic of 1.15 0.9% Germany 3.63 0.9% Korea, Republic of 2.90 1.0% 8 France 1.14 0.9% Norway 3.43 0.8% Netherlands 2.26 0.8% 9 U.S.S.R 1.06 0.8% France 2.92 0.7% France 2.11 0.7% 10 Italy 0.98 0.7% Korea, Republic of 2.81 0.7% India 1.82 0.6% Non-OSA Countries (Grey cells) 38.69 9.2% Non-OSA Countries (Grey cells) 21.84 7.4% * Grey cells denote countries that did not sign Open Skies agreements with Canada at year t, where t = {2007, 2009}. Sources: Foreign Affairs and International Trade Canada (2010); International Monetary Funds, Direction of Trade Statistics (Retrieved on December 11, 2010); InterVISTAS Consulting Inc. (2008) 17 The effectiveness of Open Skies agreements on trade could also depend on the economic power of the trading partners. The properties of the US and Canadian air service agreements could thus differ depending on which set of countries these two regions concluded OSAs with. As such, the value of these policies on trade statistics may influence the number of Open Skies agreements signed to date, and the number of OSAs signed to date would also affect overall trade statistics. Figure 1 shows that the United States has not signed Open Skies agreements with one particular income group13,14 and the income group of these OSA countries varies substantially across the years. On the other hand, Canada has restrictively signed agreements with mostly upper-middle or high income countries (refer to Table 1 on page 11). Interestingly, since 2001, roughly 60% of the US Open Skies agreements signed has been with low or lower-middle income nations; whereas prior to 2001, roughly 40% of the agreements were with countries of these lower income groups. The first sets of US Open Skies policies were primarily concluded with higher-income nations, as shown in Figure 1. As noted previously, Micco and Serebrisky (2006) find that Open Skies tend to be more beneficial if agreements were signed with upper-middle or high-income countries. As such, it is questionable whether recent US agreements, especially those signed with poorer nations stimulate economic growth. If these countries do not add much value to trade, then the number of OSAs signed should have been smaller. The marginal cost of reaching these agreements could be greater than the marginal benefit derived. Thus, analyzing the data by country income groups becomes essential in order to better understand the performance of OSAs on merchandise trade. 13 Country income groups used in this note are based on the World Bank’s classifications. 14 Refer to (U.S. Department of State, 2010) for the list of countries that signed Open Skies agreements with the US. Note that Armenia, Aruba, Bosnia-Herzegovina, Brunei, Bulgaria, Burkina Faso, Cameroon, Chad, Cook Islands, Estonia, Gabon, Kenya, Lithuania, Madagascar, Maldives, Mali, Namibia, Netherland Antilles, Rwanda, Slovenia, Sri Lanka, Tanzania, and Uganda are excluded from the panel regression models (Chapters 3 and 4) due to lack to available gross domestic product, trade, and/or air passenger traffic data. 18 Figure 1: US Open Skies agreements concluded per year, segregated by income groups 2.2 Basic trade statistics comparison between Canada and U.S.15 Normalized plots of US exports and imports data series are expected to increase due to Open Skies policies. However, it is vital that some comparisons be made to the US data, to determine if US trade growth is much superior when more Open Skies agreements are initiated. Given that there are significant differences between Canada and the United States in terms of the number of Open Skies agreements signed to date, this section examines using various normalized plots to determine if exports and imports figures have improved after major round of implementations of Open Skies agreements. Note that the normalized plots present preliminary analysis of trade statistics between Canada and US, and provide the motivation for further investigation on whether signing more Open Skies agreements is necessarily better or if Open Skies agreements exhibit some forms of diminishing marginal returns. Exports and imports values could be influenced by other factors, such as free trade agreements, currency unions, and other governmental regulations. 15 Trade data were retrieved on or before December 8, 2010, with both Canada and US observations end in third quarter of year 2010. Normalized plots are constructed similar to the methodology used in Chart 2-B of the Bank of Canada’s (January 2010) Monetary Policy Report, using exports and imports data taken from Statistics Canada’s National Income and Expenditure Accounts and from U.S. Bureau of Economic Analysis’ National Income and Product Accounts. 19 The data used in this section are real aggregate exports and imports of goods and services16, so that increases in exports and imports are not attributed to inflationary effects. Ideally, the real exports and imports data segregated by transport mode should be used. The total exports and imports data include trade by other transport modes that may not be related to or non-complementary to Open Skies agreements. However, Statistics Canada does not publish real exports and imports data broken down by transport mode. As such, the aggregate exports and imports data based on National Accounts basis are used to compare trade movements before and after Open Skies took place for the two neighboring countries17. Fortunately, for the US data, the correlation between merchandise trade by air value and aggregate trade value is quite high, ranging from 0.6 to 0.9. This suggests that aggregate trade values may not be a poor choice for normalized plots comparison purposes. Note that normalized plots of exports and imports for US merchandise trade by air transport mode are presented in Section 2.3. Figure 2 and Figure 3 suggest that exports and imports for both Canada and US increased at a steady rate after the first Open Skies or liberalized agreements were signed (1992Q4 for US and 1995Q1 for Canada)18. The growth in exports and imports appear more persistent, suggesting that in the long-run, these early Open Skies agreements may be more beneficial for the economy. Interestingly, in both figures, the US series is above the Canadian series. Yet, the gap between the two series is small and the US line appears to be parallel to the Canadian line. These imply that the US trade growth was not substantially greater than Canada’s after the introduction of Open Skies policies. 16 The data should include trade in services, if possible. Open Skies agreements play major role in tourism, and tourism is embedded in trade in services data (World Trade Organization, 2010). 17 It should be noted that the comparison between the US and Canadian trade trends using the respective nation’s National Accounts data is not exact. National Accounts data for the US is in chained year 2005 currency and data for Canada is in chained year 2002 currency. However, National Accounts data is more preferable in this context since it includes both goods and services data in a timely manner. 18 Gillen (2009) mentions that Canada signed its first Open Skies agreement with the US in 1995. 20 Figure 2: Exports (normalized at time of first Open Skies agreements) Figure 3: Imports (normalized at time of first Open Skies agreements) Relative to Canada, the US has been experiencing higher export volumes after new rounds of Open Skies agreements were signed in late 2000s (Figure 4). Canadian exports have been declining after initiating the new Blue Sky policies in 2006. Consistent with Gillen (2009), the Canadian Blue Sky policy is not that effective. The US export volumes have not increased at a steady pace similar to when the first set of Open Skies agreements were signed. In terms of imports, Canadian import growth appears to exceed the US’s after new sets of Open Skies agreements were signed, as shown in Figure 5. Since the Blue Sky policies, Canadian import growth has exceeded export growth, suggesting that if all else were held constant, then GDP growth would decrease due to this new air service policies (i.e. net export is negative). Ignoring the trough in the series that commence in late 2008 (which most likely is related to the global financial meltdown), growth for exports and for imports for both countries have been much more modest compared to when the first set of Open Skies agreements were signed. Combination of these observations question whether signing more Open Skies agreements tend to provide much higher trade growth, whether the effectiveness of these agreements take longer time to flow through, and/or if the effectiveness of OSAs are vulnerable to movements in other macroeconomic variables. 21 Figure 4: Exports (normalized at time of new Open Skies agreements / Blue Sky policies initiation) Figure 5: Imports (normalized at time of new Open Skies agreements / Blue Sky policies initiation) 2.3 Normalized plot of US merchandise trade by air transport19 This section examines how the US merchandise trade by air series compares with the aggregate real trade series, to determine if the aggregate series convey similar message as the trade by goods series segregated by transport mode. Two “merchandise trade by air” series are analyzed in particular – “value” and “value per weight” – and both series were converted to real terms using the implicit GDP price index for trade taken from the National Income and Product Accounts. The real term is used to eliminate inflationary effects. The value per weight series is important for analysis, in addition to the “value” series, since weight is important for transport flows and cost (Yamaguchi, 2008). “Value per weight” figures suggest the type of goods that may be transported by air and whether freight cost is high. For example, higher “value per weight” figure implies that more expensive (valuable) goods are being shipped at small quantities. In this case, freight cost may be high for low profit margin products. Higher air freight cost may decrease trade flows by air. Note that the value by air and value per weight by air series start in 2005Q1. As such, only the 2007Q2 episode is analyzed in this section. 19 The U.S. Foreign Trade Division (personal communication, September 7, 2010) confirmed that bilateral trade in services data is not available. Thus, the econometric models that are presented in subsequent chapters use merchandise trade by air statistics as dependent variable. This section examines specifically on merchandise trade by air series to better understand the relationship between aggregate trade (including trade in services) and merchandise trade by air data. 22 Both normalized plots (Figure 6 and Figure 7) convey interesting messages. Figure 6 shows that aggregate real export values and exports by air measured in terms of values both declined in late 2008, a result of the global financial crisis. Both of these series follow similar movements two years before and after new sets of US Open Skies agreements were signed in 2007Q2. However, the value per weight series exhibits somewhat opposite trend. When the two export values series increase prior to 2007Q2, the value per weight series decreases. In contrary, when the two export values series dip in late 2008, as noted above, the value per weight series picks up around the same period of time. This suggests that weight transported by air significantly decreased by late 200820. Overall, all three exports series have been relatively stable and do not experience any major growth or decline after new rounds of Open Skies agreements were signed. Figure 7 conveys similar message as Figure 6. The import value series is relatively flat for points prior to 2010. Conversely, the value per weight measure for the imports by air series increased significantly between late 2008 and early 2009. Figure 7 also suggests that the weight transported by air decreased significantly during the financial turbulence period. Figure 6: US exports by air versus aggregate exports (normalized) Figure 7: US imports by air versus aggregate imports (normalized) 20 International Air Transport Association (2010) suggests that high US manufacturing inventories to sales ratio in late 2008 is one of the contributing factors to this trend. 23 Graphs in this section convey the following information:  Values for trade by air follow closely the movements of aggregate trade values. Between 2005Q1 and 2010Q3, the correlation between these two series is roughly 0.9 for exports and 0.6 for imports.  The flat movements in the normalized plot suggest that latest rounds of Open Skies agreements may not be as beneficial as the earlier rounds of agreements.  The upswings for the value per weight series and the dips for the two value series imply that Open Skies agreements may not be robust to other macroeconomic variables.  Trends in exports by air and in imports by air series differ. This strengthens the objective of analyzing both exports and imports data in order to properly assess the effectiveness of Open Skies agreements. 2.4 Concluding remarks The preliminary analysis in this chapter provides the motivation for investigating the effect of Open Skies agreement on international merchandise trade development. Properties of air service agreements between the US and Canada vary, which may explain the discrepancy in the number of Open Skies agreements signed and the performance of these policies. Despite that the US encourages signing more liberal air service agreements than Canada does, and the progress in ASA development for the US is quicker and more active than for Canada, the US has not signed Open Skies agreements with some of its major trading partners such as China, Japan, Mexico, and Brazil. This questions whether the US could achieve greater trade growth if it concluded agreements with some of its major trading partners. Normalized plots from Sections 2.2 and 2.3 also confirm the need to model the effects that Open Skies agreements have on recent merchandise trade figures (in real terms). Given that these series may be influenced by other macroeconomic variables, as apparent in the graphs capturing the 2007Q2 episode, the plots also confirm the need to include other variables, such as free trade agreements, currency unions, liberalized ASAs from adjacent countries, which may influence trade data. The graphs also question whether signing more Open Skies policies is necessarily better, if benefits derived from OSAs happen immediately or take time to pass through, and/or if the performance of these agreements depends on other market volatilities. 24 The analysis in this chapter focuses primarily on bi-variate relationships, which would not fully capture effects from other market shocks. As such, econometric techniques are presented in subsequent chapters to explore if Open Skies agreements have any significant impact on recent merchandise trade development. The use of econometric models allows one to analyze merchandise trade data from a more multi-dimensional perspective. 25 Chapter 3 The model This chapter discusses the motivation for using augmented gravity models 21 to explain the relationships between trade movements and Open Skies agreements. Several researchers have used the gravity equation to explain international trade flows. Past air transport literature has also used the gravity equation to analyze relationships between Open Skies agreements and bilateral trade developments. However, as noted previously, improvement opportunities exist in this area of research. Minimal number of studies has used updated data to evaluate the possible benefits derived from recent Open Skies agreements. In addition to studying the effects from the basic gravity model, this study augments the gravity equation to account for other macroeconomic variables, to measure time evolution of OSAs, and to differentiate each OSA, in order to better understand the linkages between Open Skies agreements and recent merchandise trade development. Panel regression techniques are used in this study. This note does not conduct case-study based comparisons similar to what InterVISTAS Consulting Inc used in its papers (InterVISTAS Consulting Inc., 2005; InterVISTAS Consulting Inc., 2006), as inference can not be made logically from one particular case to another. Each country-pair may exhibit different characteristics due to differences in economic conditions. Overall, this study aims to develop parsimonious models for policy making implications. Finally, this chapter also describes the data used to carry out this research. The U.S. Foreign Trade Division (Census Bureau) confirmed (personal communication, September 7, 2010) that bilateral trade in services data is not available. As such, only merchandise trade by air data is used as dependent variable in the econometric analysis. 21 The terms “gravity model” and “gravity equations” are used interchangeably in this note. 26 3.1 The basic gravity equation Numerous studies on international trade flows, including those that are air transport policy-related have used the gravity equation. The traditional form of the gravity equation as described by Bergstrand (1985, p.474) is as follows: Equation 1: Where, Variables Descriptions Tij = U.S. dollar value of the trade flow from country i to country j Yi = U.S. dollar value of nominal GDP in country i Yj = U.S. dollar value of nominal GDP in country j Dij = Distance between country i and country j’s economic centres Aij = Any other factors either aiding or resisting trade between country i and country j uij = Log-normally distributed error term, assuming E(ln uij) = 0 According to Anderson (1979), the basic (traditional) gravity model is based on Cobb-Douglas expenditure system, assuming that each country specializes in producing one good. Equation 1 is a reduced form equation derived from a four-equation partial equilibrium model of export supply and import demands, with prices excluded (Bergstrand, 1985). Bergstrand (1985) suggests that the reduced- form model assumes perfect substitutability, perfect commodity arbitrage, zero tariffs, and zero transport costs. Both Anderson (1979) and Bergstrand (1985) note that the usual estimators of the gravity equation may be biased, since the underlying assumptions have been refuted by empirical research. For example, it is unrealistic to assume that there are no tariffs, and no price and exchange rate volatilities. Furthermore, products are not perfect substitutes, as these products are differentiated by country origins (Bergstrand, 1985). Bergstrand (1985) also shows using the F-test that the reduced form model as illustrated in Equation 1 is inadequate; thus, augmented models should be developed to capture the actual trends in international trade. On the other hand, studies such as Baier and Bergstrand (2009) argue that matching econometrics should be used to analyze trade flows instead of using the gravity equation. Their findings conclude that matching econometrics provide more consistent results regarding effects of free trade agreements. Gravity equations performed on cross-sectional basis produce extreme instability across years. The gravity equation shows that free trade agreement yields positive effects using data from 1960, 1965, 1970, and 1975. Conversely, the gravity equation using data from 1980 illustrates that free trade 27 agreements lead to negative treatment effects (Baier & Bergstrand, 2009). Baier and Bergstrand (2009) conclude that omitted variable bias, nonlinearity, and non-randomness are contributing factors to inconsistent results from the gravity equation. However, the matching econometrics technique as illustrated in Baier and Bergstrand (2009) is applicable for cross-sectional analysis and for analyzing average treatment effects. The matching econometrics technique is also subject to selection bias, as it is generally challenging to find country pairs without OSAs that are identical in all other respects to a pair with OSAs (Baier & Bergstrand, 2009). Limited studies have been done to relate free trade agreement to Open Skies agreements; thus it is unclear whether Open Skies agreements would incur similar findings as Baier and Bergstrand (2009) if the gravity equation or the matching econometrics technique is used. To be consistent with past transport research analyzing relationships between Open Skies agreement and trade flows, the gravity equation is used in this note. Yamaguchi (2008) suggests that the gravity equation is the more appropriate model for air cargo analysis, since merchandise trade by air is composed of differentiated goods. This appears to conform to the underlying assumptions of the gravity equation (Yamaguchi, 2008). One of the conventional gravity models that has been used by past transport researchers such as Endo (2007) is shown in Equation 2. Table 4 and Table 5 (pages 30-31) show the results for this conventional model using data between years 2004 and 2009 (inclusive)22. Similar to Micco and Serebrisky (2006), the panel is segregated by the trading partner’s income groups for illustration purposes. Year dummy variables are also incorporated into the model. This conventional model suggests that at the aggregate level, Open Skies agreements on average have insignificant impact on merchandise trade. At the more micro-level, OSAs have negative, significant impact (at the 5% significance level) on imports from low-income countries and positive, significant influences on imports from high-income regions. 22 Time frame depends on the sample period of the merchandise trade by air series retrieved from the U.S. Merchandise of Trade website. See (U.S. Census Bureau, 2010) for details. 28 Equation 2 Where, Variables Descriptions TFijt Merchandise trade by air values between countries i (US) and j (US trading partner) at time t (in USD) Yi US real GDP (in USD) Yj Real GDP of US trading partner (in USD) Yi / Pi US Real GDP per capita (in USD) Yj / Pj Real GDP per capita of US trading partner (in USD) Dij Distance between US and its trading partner Langij Language dummy variable, where it equals to “1” if both countries share common official language OSAijt The dummy variable will equal to “1” if the Open Skies agreement was signed between the US and its trading partner at or before time t. Year dummies Year dummy variables to control for shifts in the intercept over time (Head & Ries, 2009) εijt Error term (assumed to be log-normal) However, the results for this conventional model are most likely biased. Gillen (2009) suggests that Equation 2 is not a good specification, as it assumes that the effect from all Open Skies agreements is homogeneous. In other words, Open Skies agreements for all country pairs are the same, and this is not true. Each Open Skies agreement is quite distinct and the effect is expected to differ between each country pair depending on when the OSA for each country pair was signed, on the distance between the countries, and/or on other conditions in the agreement. Micco and Serebrisky (2006) also suggest that the conventional gravity model measures the average effect of OSAs on merchandise trade, but the model does not take into account that the effect of OSAs take time to flow through. Furthermore, analyzing country income effects by segregating the panel may not be the most desirable choice. Segregating the panel by trading partner income group would exclude any income interaction effects in the model. Panel segregation is also vulnerable to small sample bias issues, as well as problems with non-linearity, non-randomness, and non-normality. As such, alternative methods, such as the incorporation of interaction terms of OSAs and income groups should be used instead. 29 Furthermore, trade in services data is missing in this formula. As noted previously, trade in services should have positive impact on merchandise trade. As shown in Figure 8 (page 32), the share of trade in services varies across the country income groups23. Failing to account for trade in services may bias the results. Furthermore, air passenger traffic flow data is missing in the model. Incorporating this variable is important since it is expected that negotiations, information acquisitions, and trust building would take place prior to the delivery of the good. As such, lagged passenger flow is expected to be positively related to merchandise trade by air. Other macroeconomic variables are also not included into this conventional model, as these other variables may also influence trade movements. As such, this basic model may be underspecified. Given these observations, the basic gravity model should be augmented to account for other macroeconomic factors, and to consider each OSA as distinct. 23 The share of trade in services may be under-estimated, since government services are excluded from the WTO data (World Trade Organization, 2010). 30 Table 4: Conventional gravity model results for US merchandise exports by air Estimation method: Ordinary least squares with year dummy variables Sample: 2004 - 2009 Specification Aggregate Low-income Lower-middle income Upper-middle income High-income Periods included (T) 5-6 1-6 1-6 1-6 1-6 Cross-sections included (N) 122 23 43 42 47 Total panel (unbalanced) observations 731 116 186 182 247 R-squared 0.891 0.860 0.882 0.867 0.873 Root mean squared error 0.860 0.666 0.890 0.811 0.738 F-statistic 533.435 58.662 118.791 101.447 148.128 ln(GDP product) 0.989a 0.956a 1.056a 0.900a 0.991a (0.033) (0.086) (0.058) (0.058) (0.047) ln(GDP per capita product) 0.244a 0.139 0.261 0.167 0.344c (0.050) (0.221) (0.224) (0.154) (0.200) ln(Distance) -0.565a -1.141a -0.836a -0.742a 0.278 (0.159) (0.188) (0.230) (0.267) (0.176) Language Dummy Variable 0.641a 0.973a 0.382 0.466 1.191a (0.162) (0.222) (0.377) (0.309) (0.274) OSA Dummy Variable 0.027 -0.150 -0.257 0.429 0.105 (0.137) (0.198) (0.238) (0.337) (0.239) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. Year dummies are not reported. 31 Table 5: Conventional gravity model results for US merchandise imports by air Estimation method: Ordinary least squares with year dummy variables Sample: 2004 - 2009 Specification Aggregate Low-income Lower-middle income Upper-middle income High-income Periods included (T) 5-6 1-6 1-6 1-6 1-6 Cross-sections included (N) 122 23 43 42 47 Total panel (unbalanced) observations 731 115 186 182 248 R-squared 0.785 0.710 0.786 0.741 0.772 Root mean squared error 1.476 1.412 1.411 1.469 1.273 F-statistic 238.696 23.109 58.266 44.509 73.009 ln(GDP product) 1.132a 1.317a 1.215a 1.131a 1.153a (0.064) (0.125) (0.101) (0.121) (0.090) ln(GDP per capita product) 0.263a -0.097 0.469 0.248 0.489 (0.101) (0.451) (0.521) (0.397) (0.383) ln(Distance) -0.335 -0.363 -0.795 -0.205 0.191 (0.264) (0.529) (0.486) (0.431) (0.362) Language Dummy Variable 0.650b -0.049 1.146b 1.427b 1.452a (0.306) (0.500) (0.572) (0.675) (0.517) OSA Dummy Variable 0.194 -1.287a 0.043 1.018c 0.918b (0.262) (0.478) (0.394) (0.567) (0.455) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. Year dummies are not reported. 32 Figure 8: Share of trade in commercial services Trade in commercial services over the sum of trade in commercial services and total merchandise trade 3.2 The augmented gravity model New, augmented gravity models are constructed to better explain the impact of Open Skies agreements on merchandise trade by air. Two scenarios are analyzed – one with merchandise exports by air as dependent variable and another with merchandise imports by air as dependent variable. As noted before, analyzing both exports and imports is important, in order to see if Open Skies agreements affect exports and imports in similar fashion. Micco and Serebrisky (2006) show that the effects of US Open Skies agreements on air transport costs and shares of US imports tend to differ across country income groups. The performance of the Open Skies agreements on merchandise trade by air could depend on the economic power of the trading partners. As such, the analysis also includes dummy variables that capture the interaction of the trading partner’s income level and the effect of Open Skies agreements. Most of the following explanatory variables featured in the conventional gravity model (from Section 3.1) are also used in the augmented gravity models:  Product of the gross domestic product of home and foreign countries: captures the economic size of the home and foreign countries and it is expected that trade increases as economic size grows (Endo, 2007). Thus, this variable is expected to be positive.  Product of the gross domestic product per capita of home and foreign countries: a proxy to capture the income level and fraction of income that one would spend on home and foreign goods (Endo, 2007). The expected sign for this coefficient is positive. 33  Distance between home and foreign country: a proxy for transport cost (Bergstrand, 1985). Disdier and Head (2006) suggest that the distance coefficient is the product of the elasticity of trade costs with respect to distance and the elasticity of trade with respect to trade costs. It is expected that merchandise trade would increase if transport cost were to fall. As such, one would expect the distance variable to be negative in the gravity equation.  Language dummy variable: a proxy to capture transaction cost, as one expects transaction cost to reduce if pairs of countries share common official language (Endo, 2007). As such, the sign for this coefficient should be positive. Furthermore, dummy variable capturing the adjacency country pair is also included into the augmented gravity models. Disdier and Head (2006) note that failing to account for the adjacency effect would cause overestimation of the distance effect, which is another form of the omitted variable bias. In Disdier and Head (2006), the adjacency effect is negatively correlated with distance (or positively related to trade flows). However, in this paper, this effect is expected to be negatively related to trade by air, since cargo could be transported by ground due to the short proximity. The cost to transport the goods by air could be greater than that to transport by ground, which would lower the merchandise trade by air. Air passenger traffic flows between the US and its trading partner, which is the sum of the flow from the US to the foreign country and the flow from the foreign country to the US, are incorporated into the augmented gravity model to help explore one of the mechanisms that would underlie the linkages between Open Skies agreements and merchandise trade development. The passenger traffic flow variable assumes that the individual would return to the head office (his or her home country) to complete the transaction. As noted previously, trade usually contains a lag, as contracts are typically negotiated prior to the delivery of the good. Furthermore, firms would arrange visits to their supply chain partners in order to facilitate trade between the country pairs and to minimize information asymmetries. For simplicity, the paper assumes that these visits are conducted one year prior to the delivery of the good. The lagged air passenger flow is expected to be positively related to trade by air values, implying that business travel does facilitate expansion of trade that depends on information and trust. The increase in business travel would be a byproduct of Open Skies agreements, since past studies have 34 shown positive relationship between OSAs and passenger traffic flows. OSAs thus would create more convenient routing options, which would make site visits easier. The effect of OSAs on passenger traffic would help improve information acquisition and trust building opportunities. As such, these mechanisms would encourage more trade by air, which would also expand trade opportunities in other modes of transportation. For example, surface transportation would complement air transportation, since goods would need to be transported from airports to the final destinations. The augmented models also include additional explanatory variables to account for other market forces. Merchandise trade figures are also influenced by trade in services, exchange rate volatilities, and free trade agreements. As noted in Chapter 1, Baier and Bergstrand (2009) and Glick and Rose (2002) show respectively that free trade agreements and currency unions are positively related to merchandise trade figures. Trade in services is expected to have positive influence on merchandise trade. InterVISTAS (2006) suggests that sale of services, such as consulting services, financial services, management, and insurance, helps determine the demand for travel. It is also reasonable to assume that the sale of services (trade in services), such as insurance, transportation, and communication services, is important to predict the demand for merchandise trade. Although this set of data is not available on a country pair basis, the augmented models use an interaction term to capture trade in services. InterVISTAS (2006) suggests using the product of the trade in services data for the two countries as the interaction term. More details on the trade in services data are presented in Section 3.4. Additional explanatory variables need to be incorporated to account for differences in each Open Skies agreement. Some OSAs include all cargo seventh agreements; thus a dummy variable should capture this difference. Furthermore, similar to passenger traffic volumes, trade may decrease if the adjacent nation signed liberalized air service agreements with another foreign country that the domestic country did not, as proposed by Dresner and Oum (1998). As such, another dummy variable should be added to capture this effect. For simplicity, this ASA adjacency dummy variable looks at the Open Skies and Open Skies-type agreements that Canada has signed (excluding Mexico). This dummy variable also assesses whether Canadian Blue Sky policies pose any threat to the effectiveness of US OSAs and to US trade statistics. 35 Moreover, other variables need to be incorporated into the augmented model to account for the fact that OSA for each country pair is distinct. Firstly, the augmented model needs to explore the time evolution of OSA, in order to understand if consistent with InterVISTAS (2006) that the impact of OSA on merchandise trade may take longer duration to flow through. Similar to Micco and Serebrisky (2006), one methodology is to incorporate time dummy variables, where “1” is recorded if the trading partner has signed Open Skies agreement with the US for “n” number of years. These cumulative time dummy variables are then multiplied by the income group of the trading partners in order to examine if the effect of Open Skies agreements differ depending on the partner’s income level. For simplicity, two income groups are used for this interaction term – “lower income group” includes both low-income and lower-middle income countries; and “higher income group” includes high-income and upper-middle income countries. This methodology is captured under Model A. Dummy variables that are not statistically significant with low coefficients for shorter time frame, but are statistically significant with large coefficients for longer time frame imply that the effects from US Open Skies agreements do not appear immediately after the agreements were signed. Equation 3 shows the proposed model: Equation 3 36 Where, Variables Descriptions TFijt Merchandise trade by air values between countries i (US) and j (US trading partner) at time t (in USD) Yi US real GDP (in USD) Yj Real GDP of US trading partner (in USD) Yi / Pi US Real GDP per capita (in USD) Yj / Pj Real GDP per capita of US trading partner (in USD) Dij Distance between US and its trading partner Langij Language dummy variable, where it equals to “1” if both countries share common official language OSA0 Dummy variable for US Open Skies agreements signed in current year OSA1 Dummy variable for US Open Skies agreements signed for one year OSA2 Dummy variable for US Open Skies agreements signed for two years OSA3 Dummy variable for US Open Skies agreements signed for three years OSA4 Dummy variable for US Open Skies agreements signed for four years OSA5 Dummy variable for US Open Skies agreements signed for more than or equal to five years Lower income group Trading partners that belong to low-income group or lower-middle income group, based on the World Bank’s income classifications Higher income group Trading partners that belong to upper-middle income group or high-income group, based on the World Bank’s income classifications TSijt Interaction term capturing trade in services between countries i (US) and j (US trading partner) at time t Factors Vector containing other influential variables, such as currency union, free trade agreement, liberalized ASAs signed by Canada, and all cargo seventh freedom of rights Adjij Adjacency dummy variable, where it equals to “1”if both countries are adjacent to each other (Paxij)t-1 Passenger traffic flows (sum of flows from US to trading partner and flows from trading partner to US), with one year lag Year dummies Year dummy variables to control for shifts in the intercept over time (Head & Ries, 2009) εijt Error term (assumed to be log-normal) Incorporating the distance interaction term in addition to time dummy variables also helps assess how the impact of Open Skies agreements differs depending on distance. The interaction term used in Micco and Serebrisky (2006) is the product between the Open Sky agreement variable and distance (in natural log form), where the Open Sky agreement variable equals to “1” if the agreement between the United States and the trading partner was signed on or before year “t”. Micco and Serebrisky (2006) suggest that this interaction term can help assess whether countries farther from the United States benefit from Open Skies agreements. They find that at the aggregate level, using share of US imports data as dependent variable, the interaction term is positive. This implies that the countries that are more distant from the US benefit more from OSAs (Micco & Serebrisky, 2006). As such, this interaction term is used in this paper under Model B to examine if the same story holds for recent US merchandise exports and imports data. Note that when the distance interaction term is included into the model, the 37 distance variable (lnDij) is dropped from the equation to prevent collinearity issues, since the distance and interaction terms would be identical for countries that have concluded OSAs. However, Model B has a shortcoming. The interaction term of OSA and distance used in Micco and Serebrisky (2006) assumes zero weight on country pairs that did not engage in Open Skies agreements. This implicitly assumes that the transport cost for these country pairs is zero and the distance for the non-OSA country pairs is homogenous. These are invalid assumptions. As such, Model C includes a new interaction term that aims to improve Model B. This new term is the product of time (the number of years of which OSA was signed for plus one extra year) and distance, in natural logarithm terms. When the country pair does not have OSA or the country just initiated OSA, the time would equal to zero. However, the log of zero is undefined. Taking this into consideration, an extra year is added, so that it is assumed that the starting point is time 1 instead of time 0. At the initiation point (time 1), the interaction term would equal to natural logarithm of distance term, which is what the conventional model has. The distance term is re-weighted by the time aspect of the OSA in order to make OSA for each country pair as distinct as possible. Overall, this alternative model helps assess whether the combination of time and distance, in addition to the individual time dummy variables, helps explain the impact of OSA on merchandise trade movement. OSAs would help create more convenient routes, which would implicitly lower the distances between the trading partners. OSAs take time to flow through. Thus, this term is expected to be negative since the number of years of which OSA was signed for and distance should be inversely related. Finally, year dummy variables are included into the model to control for shifts in intercepts over time (Head & Ries, 2009). This is necessary since trade collapsed starting in late 2008. The recession effect would influence the 2009 data, which implies that the intercept term for year 2009 should be different. Failing to include the year dummies would positively skew the results. With the implementation of year dummy variables into the model, the intercept term is dropped to prevent collinearity issues. 38 Table 6 summarizes the three augmented gravity equations that are analyzed in this paper – Models A, B, and C. Model B is included in the analysis for illustration purposes, to see if the results are dramatically different from those of Models A and C. The econometric results for these new models are discussed in detail in Chapter 4. Table 6: Variables and regression models used for augmented gravity equation analysis Model: A B C Dependent Variable: Merchandise trade by air values (exports or imports) X X X Explanatory Variables: Natural logarithm of the GDP product X X X Natural logarithm of the GDP per capita product X X X Natural logarithm of distance X Language X X X Adjacency dummy variable X X X Natural logarithm of trade in services X X X Natural logarithm of air passenger flows, lagged one period X X X Currency union24 X X X Free trade agreement X X X Country J having liberalized agreement with Canada X X X All cargo seventh freedom of right X X X OSAn*income group of trading partnerj where n = [0,5] and j =[lower, higher] X X X OSA * natural logarithm of distance, where OSA equals to “1” if the agreement has been signed between the US and its trading partner X Natural logarithm of ((time+1)*distance), where time equals to the number of years for which an OSA has been signed for X Year dummies X X X ** “X” denotes the variables used for that particular regression model 24 The definition for “currency union” is based on Glick and Rose (2002). Currency union exists when nominal exchange rate are at 1:1 par for an extended period of time. In other words, prices do not need to be converted between a pair of countries (Glick & Rose, 2002). In this study, the author assumes that currency union between the foreign country and US exists if the nominal exchange rate (FC/USD) equals to one. 39 3.3 Model issues and considerations Panel regression models are used since both time-series and cross-sectional effects are important for the analysis. InterVISTAS (2006) argues that identifying “before” and “after” periods for liberalization is challenging, since the effect of Open Skies policies take time to flow through. Thus, some time-series aspect needs to be incorporated into the model (InterVISTAS Consulting Inc., 2006). Differences in trade volumes may also be due to country-specific effects. Prior work, such as Endo (2007), has used the ordinary least squares (OLS) with year dummies technique to model the gravity equation. Consistent with past research, this paper uses the same regression technique for the analysis. Although the Hausman test statistics suggest using the fixed (within) effect model25, the OLS model is preferred since it allows analysis of time-invariant variables. Time-invariant variables, such as distance, are important in understanding the mechanism underlying the linkages between Open Skies agreements and international trade development. For example, distance acts as proxies for transport costs and also helps determine the effect of OSAs on international trade. The augmented gravity models exhibit heteroskedasticity and serial correlation. As such, the standard errors for models used in this note are adjusted based on Arellano’s method. Croissant and Millo (2008) suggest that Arellano’s method is a general structure to account for both heteroskedasticity and serial correlation. This method is preferable for panels with large numbers of observations and with short time frame (i.e. small T) (Croissant & Millo, 2008). As such, the Arellano’s method is selected for this paper, since the panel has 122 observations (country pairs) and 6 years of data. 25 The within / fixed effect approach with two-way effects has been tested, but this approach did not produce statistically significant findings. 40 3.4 Data sources Data featured in this paper came from multiple sources. Open Skies-related data were retrieved from several websites, including U.S. Department of State26, Foreign Affairs and International Trade Canada, Transport Canada, and InterVISTAS. The air passenger traffic data were taken from the US Bureau of Transportation Statistics’ Air Carrier: T-100 Market (All Carriers) database. This study uses annual data between years 2004 and 2009 to model the relationship between recent bilateral trade by air and Open Skies agreements. The time frame is based on the trade data availability from the US Merchandise of Trade website (U.S. Census Bureau, 2010). US merchandise trade data were taken from the “FT 920 U.S. Merchandise of Trade Selected Highlights” webpage (U.S. Census Bureau, 2010), containing annual data from 2004 to 2009. The merchandise trade by air values for both exports and imports are taken into consideration. These data are in nominal terms, where the U.S. Census Bureau reports FAS export values and custom import values (i.e. import duties, freight, insurance, and other charges incurred in bringing merchandise to the United States are excluded) (U.S. Census Bureau, 2010). The implicit GDP price index for exports and imports of goods taken from the U.S. National Income and Product Accounts were used to deflate nominal trade data into real terms. Trade in services data were taken from the World Trade Organization’s (WTO) Statistical Data Set. These data published by the WTO exclude government services (i.e. these data capture trade in commercial services) and are booked consistent with the classification, definition, and concepts of the IMF’s Balance of Payment manual (World Trade Organization, 2010). As such, to be more consistent, these nominal data were converted to real terms using the IMF’s GDP deflators. As noted in the World Trade Organization (2010), significant distortions exist in the trade in services data because some countries do not report certain service categories. Furthermore, certain services are difficult to track, such as those transmitted electronically and/or are part of multinational corporations (World Trade Organization, 2010). As noted before, the U.S. Foreign Trade Division also confirmed (personal communication, September 7, 2010) that bilateral trade in services data is not available. As such, InterVISTAS Consulting Inc. (2006) suggests the following interaction term as explanatory variable to capture the trade in service flows: 26 The OSA dummy variable is coded as 0 for Croatia. According to the November 22, 2010 version of the US Department of State website, the OSA agreement with Croatia has not been signed or applied (U.S. Department of State, 2010). 41 Equation 4 [Service exports by home country (US) * Service imports by foreign country] + [Service exports by foreign country * Service imports by home country (US)] Real gross domestic product (GDP) and real GDP per capita figures are used in the panel regression, in order to eliminate inflationary effects that may distort the analysis. Real GDP and real GDP per capita data chained to the common year (2000) and common currency (US dollar) were taken from the World Bank’s World Development Indicator database. However, for some of the countries, the World Bank series go up to 2008. The IMF World Economic Outlook database provides GDP information up to 2015 (where years 2010 to 2015 data are IMF’s forecasts). However, these IMF values are inconsistent with the World Bank’s. In order to obtain the 2009 real GDP and real GDP per capita figures chained to common currency and base year as the World Bank’s, the study took the World Bank’s 2008 figures and linked by a ratio. For real GDP, for example, the ratio was calculated by taking the 2009 real GDP figure (chained to the country’s base year) over the 2008 real GDP figure (chained to the country’s base year), where both of these figures were taken from the IMF’s World Economic Outlook (WEO) database. By doing so, this note assumes that the 2009 World Bank GDP figures would grow at the same path as the IMF’s figures27. Some US trading partners are eliminated from the panel datasets if merchandise trade, trade in services, gross domestic product, gross national income (GNI) per capita, and/or passenger traffic flow data are missing. For example, countries having missing values for real GDP for the whole sample period from both the World Bank and IMF databases are dropped from the analysis. Eliminating these observations with no data for the whole sample period does not significantly skew the results, as these countries form less than 2% of the US trade by air data. On the other hand, countries that have missing data for some years are not eliminated, to minimize survivorship bias. Table 7 lists the regions that are included in the panel regression models. 27 Between years 2005 and 2008, annual growth rates for World Bank’s real GDP figures for most of the countries are very similar to those of the IMF’s. Thus, the IMF’s GDP figures are used as a proxy to project the World Bank’s 2009 GDP values. 42 Table 7: Countries / regions in panel regression model Albania Cambodia Egypt Haiti Kuwait Nigeria Senegal Trinidad and Tobago Angola Canada El Salvador Honduras Kyrgyzstan Norway Serbia and Montenegro Turkey Antigua and Barbuda Cape Verde Equatorial Guinea Hong Kong Laos Oman Seychelles Ukraine Argentina Central African Republic Ethiopia Hungary Latvia Pakistan Singapore United Arab Emirates Australia Chile Fiji Iceland Lebanon Panama Slovakia United Kingdom Austria China Finland India Liberia Papua New Guinea South Africa Uruguay Azerbaijan Colombia France Indonesia Luxembourg Paraguay Spain Uzbekistan Bahamas Costa Rica Gambia Ireland Macao Peru St Kitts and Nevis Venezuela Bahrain Cote d'Ivoire Georgia Israel Malaysia Philippines St Lucia Vietnam Bangladesh Croatia Germany Italy Malta Poland St Vincent and the Grenadines Zambia Barbados Cyprus Ghana Jamaica Mexico Portugal Suriname Belgium Czech Republic Greece Japan Mongolia Qatar Sweden Belize Denmark Grenada Jordan Morocco Romania Switzerland Benin Dominica Guatemala Kazakhstan Netherlands Russia Taiwan Bolivia Dominican Republic Guinea Kiribati New Zealand Samoa Thailand Brazil Ecuador Guyana Korea, South Nicaragua Saudi Arabia Tonga 43 There are also countries that have GDP data from the IMF WEO database, but not from the World Bank database. However, the IMF GDP value cannot be automatically inserted into the panel dataset to replace the missing values. The IMF GDP figures are based on the respective country’s base year, which implies that the basket of goods for one country indexed to certain time period would be different from another nation’s basket indexed to another time period. One would expect the basket of goods to change as time flies. The World Bank sets all nations’ real GDP figures to year 2000 index, putting GDP figures on a more comparable basis. In this case, for certain regions such as Taiwan, the estimated GDP figures for panel datasets were linked by a ratio using data from the IMF WEO multiplied by the World Bank’s figures. This linking was calculated as the share of the country’s nominal GDP over the “World” nominal GDP taken from the IMF WEO database multiplied by the World Bank’s “World” real GDP figure (in US Dollars). Similarly, the estimated real GDP per capita figures were calculated by taking the population number calculated from the nominal WEO number and divided that by the estimated real GDP number as explained above. The country income groups used to perform panel regressions by income sectors are based on the World Bank’s classifications. Table 8 lists the country income groups: Table 8: Country income groups based on World Bank’s classification Gross national income per capita (in current USD) Income group classifications used in this note World Bank’s income group classifications Year 2004 Year 2005 Year 2006 Year 2007 Year 2008 Year 2009 Lower income group Low income $825 or less $875 or less $905 or less $935 or less $975 or less $995 or less Lower middle income $826 - $3,255 $876 - $3,465 $906 - $3,595 $936 - $3,705 $976 - $3,855 $996 - $3,945 Higher income group Upper middle income $3,256 - $10,065 $3,466 - $10,725 $3,596 - $11,115 $3,706 - $11,455 $3,856 - $11,905 $3,946 - $12,195 High income $10,066 or more $10,726 or more $11,116 or more $11,456 or more $11,906 or more $12,196 or more Source: World Bank Analytical Classifications (2010) 44 The language and distance data used for the gravity equations were taken from the CEPII distance database28. The distance is based on the population-weighted great circle distance between the largest cities of the country pair, which Head and Mayer (2002) refer to as the “effective distance”29. The effective distance measure is more preferable for the gravity model because existing distance measures (using coordinates of capital city or centre point of the nation) tend to overestimate the effective distances, especially for country pairs with close proximities (Head & Mayer, 2002). Distances calculated using the coordinates of the capital city may not be appropriate, since the capital city may not be hubs for major air cargo traffic. It is expected that more air cargo traffic would route through larger cities, which may not necessarily be the capital city of the region. However, in cases where the bilateral effective distance measure is not available for certain regions, such as Macao, the simple distance calculated using the coordinates of the capital city was used instead. Free trade agreements data were taken from the Regional Trade Agreement database from the WTO. Table 9 presents the countries that have free trade agreements with the United States. Table 9: Free trade agreements signed with the United States Countries Date of entry into force Country Income Group (as of year 2009) Israel August 19, 1985 High-income group Canada (NAFTA) January 1, 1994 High-income group Mexico (NAFTA) January 1, 1994 Upper-middle income group Jordan December 17, 2001 Lower-middle income group Chile January 1, 2004 Upper-middle income group Singapore January 1, 2004 High-income group Australia January 1, 2005 High-income group Morocco January 1, 2006 Lower-middle income group El Salvador (CAFTA-DR) March 1, 2006 Lower-middle income group Honduras (CAFTA-DR) April 1, 2006 Lower-middle income group Nicaragua (CAFTA-DR) April 1, 2006 Lower-middle income group Guatemala (CAFTA-DR) July 1, 2006 Lower-middle income group Bahrain August 1, 2006 High-income group Dominican Republic (CAFTA-DR) March 1, 2007 Upper-middle income group Costa Rica (CAFTA-DR) January 1, 2009 Upper-middle income group Oman January 1, 2009 High-income group Peru February 1, 2009 Upper-middle income group Sources: World Trade Organization, Regional Trade Agreement (2010); World Bank Analytical Classifications (2010) 28 Refer to http://www.cepii.fr/anglaisgraph/bdd/distances.htm for more information. 29 Regression models using distances calculated with the latitudes and longitudes coordinates taken from the CIA’s World Factbook (2010) yield results similar to those using population-weighted distances. 45 Finally, the currency union dummy variable was constructed based on the definition by Glick and Rose (2002). Glick and Rose (2002) suggest that countries belong to the same “currency union” when the nominal exchange rate between the domestic and foreign economies is at 1:1 par for an extended period of time. In other words, prices do not need to be converted between the pair of countries. This study assumes that currency union between the United States and the foreign country exists if the nominal exchange rate (FC/USD) equals to “1”. Table 10 displays the countries that have currency union with the United States (i.e. nominal exchange rate (FC/USD) of 1). Table 10: Currency unions with the United States30 Countries 1980 – 1984 1985 1986 – 2008 2009 The Bahamas X X X X Bermuda X X X Dominican Republic X Guatemala X X Micronesia, Fed Sts. X X X X Panama X X X X Virgin Islands (U.S.) X X X X * “X” denotes the countries that have currency unions with the United States at different time periods since 1980s. Source: World Bank, World Development Indicators (2010) 30 Data for gross national income per capita are missing for Virgin Islands (U.S.) and trade in services data is not available for Bermuda and Micronesia. These countries are thus excluded from the econometric analysis. 46 Chapter 4 Econometric results and economic implications31 This section presents the panel regression results and their economic implications, based on the three augmented gravity models described in Chapter 3. The first section discusses the Open Skies- related variables from all model results and evaluates if adding new variables that help distinguish each OSA conveys more information about the performance of these agreements. This section explores if any linkages exist between Open Skies agreements and US merchandise trade by air and if the effect of Open Skies agreements on merchandise exports and on imports differ. The second section provides alternative views of the augmented gravity models and how these views relate to trade and Open Skies agreements. Section 4.3 describes the criteria for selecting the preferred model specification and Section 4.4 ends with some concluding remarks based on the results from the preferred model specification. 4.1 Open Skies agreement variables To summarize, all three augmented gravity model specifications suggest that Open Skies agreements do not have strong impact on recent US merchandise trade by air statistics. This is consistent with the normalized plots featured in Chapter 2, which illustrate that trade growth collapsed during the financial turbulence period between late 2008 and early 2009. Majority of the OSA coefficients, especially for the imports data, are insignificant. The insignificance of these coefficients implies that the performance of these policies is not robust to market volatilities. As expected, Open Skies agreements are not expected to bring substantial effect to merchandise trade. Open Skies policies are primarily designed for passenger traffic, rather than for cargo. Moreover, all cargo seventh freedom tends to be implemented prior to the introduction of OSAs (InterVISTAS Consulting Inc., 2006). In other words, countries that have not signed OSAs with the US would have already benefited from the all-cargo seventh freedom, which have helped expand more trade opportunities. For example, Fu, Lei, and Zhang (2010) note that the Singapore-US Open Skies agreement contains provisions for the US cargo carriers, such as UPS and FedEx, to set up hub operations in China. 31 The package “plm” (version 1.2-6) from the program R was used to conduct panel regressions and relevant tests, including the Breusch-Godfrey Test of serial correlation of errors, Durbin Watson Test, and tests for cross-sectional dependence and contemporaneous correlation. These tests found that the panels do not have cross-sectional dependence. However, the models have serial correlation problems. 47 The implementation of cargo hub operations in China is expected to stimulate more trade growth between the US and Asian countries. Given that the hub operations and expansion of trade opportunities happened prior to the implementation of Open Skies agreements, the OSA-related variables would not show additional effects on trade. No additional opportunities are expected to be exploited from the OSAs. Nonetheless, the augmented gravity models do provide alternative views about the performance of Open Skies agreements. The following subsections describes how the new variables that help distinguish OSA for each country pair provide more information on how Open Skies agreements affect US merchandise trade by air statistics. Tables 11 to 16 list the econometric results. 48 Table 11: Open Skies effects on US merchandise exports by air Sample period: 2004 – 2009 Specifications A B C Total panel (unbalanced) observations 577 577 577 R-squared 0.912 0.911 0.911 Root mean squared error 0.770 0.773 0.772 F-statistic 194.627 192.580 193.124 Interaction of Open Skies agreements with lower-income trading partners: OSA signed in current year 0.439 1.633 0.473 (0.512) (2.184) (0.511) OSA signed for one year -0.275c -0.321c -0.189 (0.157) (0.181) (0.223) OSA signed for two years -0.486b -0.494b -0.420c (0.218) (0.231) (0.221) OSA signed for three years -0.310 -0.319 -0.271 (0.364) (0.371) (0.368) OSA signed for four years 0.159 0.181 0.210 (0.222) (0.231) (0.236) OSA signed for at least five years -0.124 -0.103 -0.025 (0.204) (0.206) (0.201) Interaction of Open Skies agreements with higher-income trading partners: OSA signed in current year -0.457b 0.670 -0.463b (0.224) (1.991) (0.223) OSA signed for one year 0.028 0.031 0.138 (0.055) (0.055) (0.131) OSA signed for two years 0.276b 0.266b 0.330a (0.110) (0.112) (0.123) OSA signed for three years -0.421 -0.394 -0.356 (0.271) (0.290) (0.296) OSA signed for four years -0.145 -0.136 -0.100 (0.170) (0.172) (0.166) OSA signed for at least five years 0.777a 0.757a 0.878a (0.225) (0.222) (0.240) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. 49 Table 12: Open Skies effects on US merchandise imports by air Sample period: 2004 – 2009 Specifications A B C Total panel (unbalanced) observations 577 577 577 R-squared 0.840 0.840 0.841 Root mean squared error 1.269 1.271 1.265 F-statistic 99.531 99.175 100.293 Interaction of Open Skies agreements with lower-income trading partners: OSA signed in current year 0.355 0.843 0.337 (0.658) (3.554) (0.655) OSA signed for one year 0.363b 0.413b 0.091 (0.176) (0.195) (0.291) OSA signed for two years -0.451 -0.472 -0.581 (0.332) (0.329) (0.364) OSA signed for three years 0.104 0.116 -0.001 (0.133) (0.136) (0.149) OSA signed for four years -0.994b -1.027b -1.062b (0.421) (0.431) (0.414) OSA signed for at least five years 0.336 0.280 0.184 (0.452) (0.459) (0.479) Interaction of Open Skies agreements with higher-income trading partners: OSA signed in current year -0.267 0.236 -0.245 (0.410) (3.396) (0.410) OSA signed for at least one year 0.155 0.147 -0.083 (0.126) (0.123) (0.225) OSA signed for two years 0.016 0.015 -0.109 (0.166) (0.165) (0.194) OSA signed for three years -0.590 -0.567 -0.750 (0.522) (0.532) (0.546) OSA signed for four years 0.468 0.451 0.389 (0.537) (0.526) (0.546) OSA signed for at least five years 0.929a 0.939a 0.681b (0.291) (0.289) (0.333) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. 50 Table 13: Impact of other Open Skies-related variables on US merchandise exports by air Specifications A B C All Cargo Seventh Freedom of Rights Dummy Variable 0.189 (0.245) 0.148 (0.235) 0.159 (0.242) Liberalized ASA signed by Canada with US Trading Partners -0.294c (0.158) -0.265 (0.165) -0.262 (0.164) ln(Distance) -0.266 (0.203) OSA*ln(Distance) -0.127 (0.223) ln((time+1)*distance) -0.160 (0.195) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. Table 14: Impact of other Open Skies-related variables on US merchandise imports by air Specifications A B C All Cargo Seventh Freedom of Rights Dummy Variable -0.327 (0.302) -0.262 (0.293) -0.351 (0.298) Liberalized ASA signed by Canada with US Trading Partners 0.269 (0.273) 0.238 (0.274) 0.244 (0.279) ln(Distance) 0.225 (0.327) OSA*ln(Distance) -0.058 (0.373) ln((time+1)*distance) 0.362 (0.299) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. 51 Table 15: Impact of other control variables on US merchandise exports by air Sample period: 2004 – 2009 Specifications A B C Total panel (unbalanced) observations 577 577 577 R-squared 0.912 0.911 0.911 Root mean squared error 0.770 0.773 0.772 F-statistic 194.627 192.580 193.124 ln(GDP product) 0.920a 0.894a 0.904a (0.077) (0.078) (0.078) ln(GDP per capita product) 0.188b 0.207b 0.198b (0.080) (0.082) (0.080) Language dummy variable 0.401b 0.357b 0.362b (0.169) (0.175) (0.168) Adjacency dummy variable -1.392a -0.986a -1.222a (0.382) (0.251) (0.405) ln(Trade in services) -0.049 -0.063 -0.053 (0.072) (0.074) (0.073) CU dummy variable -0.338 -0.194 -0.264 (0.470) (0.412) (0.461) FTA dummy variable 0.648a 0.638a 0.667a (0.213) (0.211) (0.213) ln(Lagged passenger flow) 0.130a 0.151a 0.141a (0.031) (0.024) (0.030) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. Year dummies are not reported. 52 Table 16: Impact of other control variables on US merchandise imports by air Sample period: 2004 – 2009 Specifications A B C Total panel (unbalanced) observations 577 577 577 R-squared 0.840 0.840 0.841 Root mean squared error 1.269 1.271 1.265 F-statistic 99.531 99.175 100.293 ln(GDP product) 0.744a 0.776a 0.733a (0.140) (0.141) (0.137) ln(GDP per capita product) -0.038 -0.053 -0.034 (0.143) (0.141) (0.141) Language dummy variable 0.122 0.189 0.114 (0.314) (0.321) (0.299) Adjacency dummy variable -0.964 -1.377a -0.698 (0.689) (0.475) (0.682) ln(Trade in services) 0.198 0.209 0.188 (0.138) (0.136) (0.140) CU dummy variable -1.204b -1.382b -1.108c (0.599) (0.559) (0.580) FTA dummy variable 1.242a 1.233a 1.204a (0.344) (0.344) (0.343) ln(Lagged passenger flow) 0.232a 0.208a 0.244a (0.043) (0.039) (0.041) The numbers in parentheses are standard errors constructed based on Arellano's method to account for heteroskedasticity and serial correlation. Superscripts \"a\", \"b\", and \"c\" denote significance at the 1%, 5%, and 10% levels, respectively. Year dummies are not reported. 53 Interaction of Open Skies agreements and partner country income levels The performance of Open Skies agreements do change depending on the trading partners’ income level. The sign and the magnitude of the interaction terms vary between the lower and higher income economies. This observation holds for both exports and imports data and also in all three model specifications. Firstly, engaging OSAs with lower income economies does not improve US merchandise trade by air statistics. Models A and B show that agreements signed for two years would lower US merchandise exports by air to these economies, and OSAs signed for at least three years would have minimal impact on exports. Somewhat similar, Model C suggests that having OSAs with this income group do not show any significant impact on US exports by air data, at the 5% significance level. On the other hand, Models A and B indicate that OSAs help increase US merchandise imports by air immediately after the implementation of these policies with the lower income trading partners, as shown by the positive and statistically significant coefficient for agreements signed for one year. However, Model C suggests that short-term OSAs have insignificant impact. Yet, all three model specifications also show that US merchandise imports from this income group would decline substantially for OSAs signed for four years. Nonetheless, it is not surprising to find that OSAs engaged with lower income countries to be non- beneficial to US merchandise trade by air statistics. Although Micco and Serebrisky (2006) find that OSAs lower air transport cost, which would encourage more trade by air opportunities, the same authors also find that lower income economies do not benefit from OSAs. They conjecture that limited market size and barriers to competition impede the gain that could be derived from Open Skies agreements. The free trade agreement coefficient (as shown in Tables 15 and 16) indirectly supports the conjecture posed by Micco and Serebrisky (2006). The strong FTA variable complemented by the weakly significant OSA variables implies that Open Skies policies are not strong enough to stimulate trade by air growth, especially for the lower income economies. In all three model specifications (for both exports and imports data), the free trade agreement variables are positive and strongly, statistically significant. The FTA coefficients in the imports panel is nearly doubled the FTA figures in the exports panel. The strong FTA variable suggests that the elimination of trade imperfections, such as tariffs, quotas, and subsidies, would have larger influence on trade than reductions in air transport cost. Majority of the US FTAs have been engaged with the higher income economies. As such, the lower income economies 54 would experience more restrictions on trade to the US. The US would impose restrictions on imports from the developing nations in order to protect its domestic industries and to minimize its current account deficit. Given these restrictions, some of these economies would retaliate, which would limit the amount of US exports to these countries. As such, the implementation of the Open Skies agreements would not be powerful enough to eliminate the trade imperfections between these regions. Thus, the dummy variables for OSAs do not show much impact on merchandise trade. On the other hand, the free trade agreement variable appears to be robust to the volatile economic conditions, and the positive terms in all three models suggest that the elimination of trade imperfections between the US and its trading partners would stimulate significantly more exports to the US (which is reflected in the US imports data) and more imports from the US (which is reflected in the US exports data). Furthermore, US merchandise trade by air has been primarily conducted with the higher income regions (Tables 17 and 18). In 2009, the lower-income regions composed of 12% of US merchandise exports by air and roughly 25% of US merchandise imports by air32. These figures imply that OSA effects should be stronger for exports to the higher income economies than to the lower income regions. The exports panel show that the higher income economies have more significant and positive coefficients. In all models, agreements signed for two years and for more than five years with this income group, holding all else constant, would stimulate more US merchandise exports. The imports data, however, show weak relationship between short-term OSAs signed with higher income countries and merchandise trade by air. Open Skies agreements appear effective only for agreements signed for five years or more for the imports data. The performance of OSAs for the higher income nations appears to conform to Micco and Serebrisky (2006) that OSAs signed for longer duration tend to perform better than those signed for shorter duration. 32 If China were excluded from the calculations, the lower income regions contribute to roughly 6% of 2009 US merchandise trade by air statistics (for both exports and imports). 55 Finally, the effect of Open Skies agreements does depend on time. The interaction terms within each income group do not exhibit similar magnitudes. This suggests that the impact of OSAs on trade would change depending on the duration of the contract. As noted above, within the higher-income economies, the interaction term for OSAs signed for at least five years show the largest positive magnitude. This is expected since countries need to adjust before fully assimilating the effect of Open Skies agreement. Table 17: US merchandise exports by air: top 10 markets Rankings 2004 2005 2006 2007 2008 2009 1 Japan Japan Japan United Kingdom United Kingdom United Kingdom 2 United Kingdom United Kingdom United Kingdom Japan Germany Germany 3 Germany Germany Germany Germany Japan Japan 4 Canada Netherlands China China China China 5 Netherlands France Netherlands Canada Switzerland France 6 France Canada Canada France France Netherlands 7 Korea, South Korea, South Korea, South Netherlands Netherlands Canada 8 Singapore China Singapore Singapore Canada Switzerland 9 Taiwan Singapore France Korea, South Singapore Hong Kong 10 China Taiwan Taiwan Switzerland Hong Kong Singapore Sum of top 10 trading partners’ (excluding China) exports by air value as a share of total US exports by air value 54% 52% 51% 50% 50% 50% *Grey cells denote regions that do not belong to higher income group (upper-middle income + high-income countries). Source: U.S. Census Bureau (2010) 56 Table 18: US merchandise imports by air: top 10 markets Rankings 2004 2005 2006 2007 2008 2009 1 China China China China China China 2 Japan Japan Japan Japan Japan Japan 3 Germany Germany Germany Germany Germany United Kingdom 4 Ireland Ireland Malaysia United Kingdom United Kingdom Germany 5 United Kingdom Malaysia Ireland Ireland Ireland Ireland 6 Korea, South United Kingdom United Kingdom Malaysia France France 7 Malaysia France France France Malaysia Korea, South 8 France Korea, South Israel Israel Israel Malaysia 9 Taiwan Israel Taiwan Korea, South Korea, South Israel 10 Singapore Taiwan Singapore Taiwan Italy Switzerland Sum of top 10 trading partners’ (excluding China) imports by air value as a share of total US imports by air value 54% 51% 50% 49% 49% 46% * Grey cells denote regions that do not belong to higher income group (upper-middle income + high-income countries). Source: U.S. Census Bureau (2010) All cargo seventh freedom variable For both exports and imports data, all three model specifications show that all cargo seventh freedom is insignificant. However, the effect of the all-cargo seventh freedom in this paper may be underestimated, since this freedom would have spill-over effects to non-OSA countries. The all-cargo seventh freedom covers the right of an airline to transport goods between two foreign countries without the home country as flight origin or destination (InterVISTAS Consulting Inc., 2005). As such, this seventh freedom acts as a precursor for the US to set up hub operations in countries of which the US has not concluded OSAs with (Zhang & Zhang, 2002). In this paper, the model assumes that all-cargo seventh freedom affects only countries that have concluded OSAs (with all-cargo seventh freedom provision) with the US. As such, the all-cargo seventh freedom coefficient in this model would not capture the spill-over effect to the non-OSA countries, since it is unclear which countries would benefit from this freedom without actually engaging Open Skies agreements with the US. As noted previously, prior to the implementation of Open Skies agreements, the all-cargo seventh freedom tend to be incorporated in the bilateral air service agreements (InterVISTAS Consulting Inc., 2006). Thus, it is not 57 expected that the all-cargo seventh freedom to show significant benefit resulting from the introduction of the Open Skies agreements. Liberalized air service agreements signed by adjacent nation and the adjacency effect: As noted in the Literature Review section, Dresner and Oum (1998) show that liberalized air service agreements signed by the adjacent nation tend to lower passenger traffic flows for the domestic economy, if the domestic economy did not sign OSA with countries that the adjacent nation did. All model specifications, however, cannot conclude the same effect for the US merchandise trade by air data. In other words, the Canadian Blue Sky policies do not have much effect on US merchandise trade by air values and thus, do not appear to pose any threats to the effectiveness of US OSAs. The poor performance of the Canadian air service agreements may be due to the restrictive terms of the agreement and the small number of agreements signed to date. Given that the US has signed substantially more Open Skies agreements than Canada, most air cargo traffic would transport between the US and the foreign country, instead of transporting directly by air between Canada and the foreign country. Cargo could then be transported by ground between the US and Canada. The adjacency dummy variable further confirms that trade may be re-routed by ground between the US and Canada, instead of transporting by air. The adjacency dummy variable is negative and statistically significant for all model specifications for the exports data, and negative and statistically significant for Model B for the imports data. Given that Canada is US’s top trading partner, the adjacency coefficient further confirms that trade by air would not increase given the close proximity between the two nations. Ground transportation would substitute this cargo flow. In this case, Canadian liberalized air service agreements would be ineffective and would not pose much threat to the US merchandise trade by air statistics. Linkages between Open Skies agreements and trade: As suggested in previous sections, two mechanisms provide linkages between Open Skies agreement and merchandise trade by air – lowering air cargo costs and facilitation of more business travel opportunities. The distance variable is a proxy for lower air cargo costs, since the distance coefficient is the product of elasticity of trade with respect to trade costs and elasticity of trade costs 58 with respect to distance (Disdier & Head, 2006). As such, it is expected that all three model specifications to show negative and statistically significant distance coefficient. Contrary to expectation, all variations of the distance term are insignificant in all model specifications. For example, the distance between the pair of countries does not appear to influence the performance of Open Skies agreements. This unexpected result appears to suggest that the implementation of recent OSAs do not significantly lower air cargo cost during the economic slowdown. This finding, however, is not unrealistic. Consistent with the interaction dummy variables, the OSAs do not appear to have strong effect on merchandise trade during the economic slowdown. Furthermore, the International Air Transport Association (2010) suggests that although air freight rates did fall when the recession hit, ocean freight rates were much lower than air freight rates. This has encouraged firms to shift from air to ocean transport (International Air Transport Association, 2010). As such, the lower air cargo rates, which would be captured under the distance term, do not have any apparent effect on US merchandise trade statistics. The models thus do not show any apparent linkages between lower air cargo costs through the implementation of Open Skies agreements and recent merchandise trade by air. Moreover, the incorporation of the natural logarithm of lagged passenger flow term contributes to the insignificant distance variable. The lagged passenger flow term is positive and statistically significant in all model specifications for both imports and exports data. However, the distance term would be negative and statistically significant at the 10% significance level if the lagged passenger variable were eliminated from the model. These observations suggest that the lagged passenger variable captures some of the effects from distance. The positive relationship between merchandise trade and lagged passenger traffic flow may be a byproduct of Open Skies agreement. As noted in the Literature Review section, passenger flow is expected to increase with the implementation of OSAs, as more markets are opened up. Opening up more aviation markets would imply that distances between the pair of countries would implicitly be lowered. As such, including the distance term would duplicate the effect from the passenger flow variable. The distance term thus would not add additional benefit to the model. The positive and statistically significant lagged passenger flow variable in all model specifications provides additional insight in understanding the linkages between Open Skies agreements and merchandise trade. The coefficient suggests that merchandise trade by air would increase if travel between the US and its trading partner were made one year ago. This result is reasonable. Trade by air 59 usually contains a lag, since contracts are typically negotiated and signed prior to the delivery of the goods. Business travels that are intended for acquiring more information and for building trust between the supply chain partners would help expand trade opportunities. The expansion of trade opportunities would also spill over to surface transportation, as surface transportation is required to transport the good from the airports to the final destinations. As such, Open Skies agreements would facilitate more business travel opportunities, as these policies intend to open up more aviation markets. Increase in business travel opportunities, which is embedded in the lagged passenger flow variable, is thus positively related to US merchandise trade by air. Exports versus imports Majority of the OSA-related variables for exports and imports data panels do exhibit similar trends. For example, both panels show that policies that have been engaged with higher income trading partners for at least five years show large, positive, and significant coefficients. The all cargo seventh freedom, Canadian liberalized ASAs, and all variations of the distance variables are insignificant for both exports and imports. Yet, some of the interaction variables within these two panels do show somewhat different trends, which support the hypothesis that both exports and imports by air data should be analyzed to appropriately conclude the effect of OSAs on merchandise trade. Holding all else constant, all model specifications show relatively larger declines in US imports by air than exports by air if the US engaged OSAs with the lower income economies. As noted in previous section, the larger decline in imports may be due to trade restrictions imposed by the US in order to protect domestic industries and to minimize its current account deficit. It appears that the significant reduction in US imports from having OSAs with the lower income economies has helped lower the US current account deficit. Figure 9 shows that the US’s negative net exports, which is a proxy for the current account deficit, was lower between years 2006 and 2009, even when the recession hit. 60 Figure 9: US net exports of goods and services Having OSAs with lower income partners that have not contributed much to the US trade data may also negatively affect the growth of US imports by air. Among all the trading partners, the US has imported most from China by air. Currently, China has contributed to roughly 20% of the total US merchandise imports by air data. On the other hand, the other lower income nations have composed of less than 6% of the US trade statistics. Yet, the US has not concluded any official Open Skies agreements with China. As of 2009, the US has concluded OSAs with 25 out of 45 lower income nations (U.S. Department of State, 2010). Given the substantial contributions that China has made to the US imports figure, failing to engage in OSAs with China would lead to lower trade expansion opportunities. All model specifications also show that US starts to export more after engaging in Open Skies agreement for two years with the higher income economies. However, this effect cannot be found from the imports data. As noted previously, the free trade agreement variable is substantially larger for the imports data than for the exports data. As such, free trade agreement appears to exert more influence on imports data than on exports data. As mentioned, the US may impose restrictions on imports in order to protect its industries and these restrictions would limit the benefit that could be derived from Open Skies agreements. For example, the US may include terms and conditions in the OSAs that would indirectly allow them to export more to foreign countries in short term, but limit the amount of imports that could enter the US immediately after the initiation of these air service policies. 61 4.2 Alternative views of the augmented gravity models This section analyzes other explanatory variables, such as the GDP interaction term, GDP per capita interaction term, and other macroeconomic variables, to see how these variables influence merchandise trade by air and Open Skies agreements. The section also presents the residual and normal quantile- quantile plot analyses. Gross domestic product interaction term: The gross domestic product (GDP) interaction term, which is the product of US’s GDP and its trading partner’s GDP, is very close to the value of one in all model specifications for the exports data. However, the GDP interaction term is 0.75, on average for the imports panel. The relatively large interaction term suggests that the economic size of the country pair has close, positive relationship with US merchandise trade by air statistics. Under this view, one would expect that the countries with the largest interaction term, such as Japan, China, and Germany, to contribute most to the US merchandise trade by air figures (refer to Table 19). This is consistent with the discussion in Section 2.1 and with the discussion in Section 4.1, since these countries are some of the top ten US trading partners. Table 19: US trading partners contributing most to the GDP interaction term Rankings 2004 2005 2006 2007 2008 2009 1 Japan Japan Japan Japan Japan Japan 2 Germany Germany China China China China 3 China China Germany Germany Germany Germany 4 United Kingdom United Kingdom United Kingdom United Kingdom United Kingdom United Kingdom 5 France France France France France France 6 Italy Italy Italy Italy Italy Italy 7 Canada Canada Canada Canada Canada India 8 Brazil Brazil Brazil Brazil Brazil Canada 9 Spain Spain Spain India India Brazil 10 Korea, South Korea, South India Spain Korea, South Korea, South Source: World Bank, World Development Indicator (2010); and author’s calculations The difference in income elasticities between the imports and exports data could be due to the composition of the trading partners for US merchandise trade by air statistics. The top 20 major trading partners for exports have been primarily composed of high-income countries. Among these major export trading partners, China (one of the lower-middle income economies) has contributed to less than 62 6% of US’s total merchandise exports by air. On the other hand, the composition of the trading partners for the imports by air data has been somewhat different. The top 20 trading partners for the imports data are a combination of low-income countries, as well as high-income countries. For example, India, Thailand, and Philippines belong to the lower income group categories, but these nations are also US’s major import partners. It appears that these major import partners do not necessarily have large economic powers. Furthermore, in 2009, lower income regions contributed to over 25% of US’s total merchandise imports by air figure, but these economies contributed to 10% of US’s total merchandise exports by air data. Given these observations, the income elasticity for the imports data is not expected to be near the value of one. Obtaining value of close to one for the GDP interaction term suggests that other (possibly unknown) market frictions may also be influencing merchandise exports by air statistics. Although the work of Calza, Manrique, and Sousa (2006) is on finding the relationship between aggregate credit demand and real output, they find the income elasticity (coefficient of natural log of real GDP) to be greater than one. These researchers hypothesize that income elasticity greater than one implies that other market frictions, such as wealth of non-GDP financial transactions, are not fully incorporated into the model to explain credit demand (Calza, Manrique, & Sousa, 2006). Under similar views, it may be possible to hypothesize that the elasticity value of close to one for the GDP interaction term may also reflect the existence of other, possibly unknown market frictions that affects merchandise exports by air movements. The relatively large and significant year dummy variables, which also acts as intercepts for all three models, confirms the existence of other market frictions that are not properly captured by the gross domestic product. The insignificant OSA coefficients reflect the possibility that other market frictions may be collinear with the OSA terms. For example, Gillen (2009) suggests that airline alliances are an imperfect substitute of Open Skies agreements. As such, effects from airline alliances may be part of these non-GDP related components. Gross domestic product per capita product: The gross domestic product per capita product term is a proxy for the fraction of income that one would spend on home and foreign goods. Higher income level is generally associated with higher merchandise trade levels (Endo, 2007). This term is statistically significant for exports by air data, yet it is inconclusive for the imports by air data. These results imply that the wealth effect does not play influential role in merchandise imports. It is possible that households prefer to travel to foreign 63 countries to purchase goods, rather than transacting by trade flows. The choice of traveling to foreign countries directly may be a byproduct of Open Skies agreements on passenger travel. As noted in the Literature Review section, passenger traffic between the domestic and foreign countries is expected to increase with the implementation of OSAs. Trade in services: The trade in services coefficient is not statistically significant in all of the models. This does not mean that trade in services has no impact on trade in goods. Instead, the insignificant term implies that it is collinear with other explanatory variables in all three augmented models. The GDP coefficient is the top candidate to be collinear with trade in services, since the correlation between the two coefficients is near 0.90 and the net exports component of the GDP measure would have included trade in services. Furthermore, the correlation between trade in services and the natural log of lagged passenger traffic flow is near 0.50. Thus, trade in services coefficient does not add much value to the models. Currency Union: The currency union variable, which is a proxy for exchange rate stability, is not statistically significant for merchandise exports by air data. This suggests that exchange rate volatility would influence the movement in merchandise exports by air. Surprisingly, the currency union variable is negative and statistically significant at the 5% level for the US merchandise imports by air data for Models A and B (but not for Model C). Technically, one would expect currency union to improve overall trade figures. However, the number of currency unions that the US has engaged in has been declining over the past years (based on Table 10). Trade flows between the US and these countries, such as Bahrain and Panama, have been low. Thus, the currency union variable would not show the same effect as what Glick and Rose (2002) find, since their data were chosen to be between years 1948 and 1997. 64 Residual Analysis: Residual plots and normal quantile-quantile plots of Models A and C33 produced by the program “R” (Figure 10) suggest that Malaysia and Paraguay are outliers for the US merchandise exports by air panel, and that Cambodia, Ethiopia, Malaysia, and Thailand are outliers for the imports by air panel. The program considers outliers to be points that deviate from the norm for the residual plots and/or points that are at the tails and not on the straight line for the normal quantile-quantile plot. Figure 10: Residual and normal quantile-quantile plots *Country codes shown in the residual and normal quantile-quantile plots are outliers and they refer to the following countries: “17” refers to Cambodia for year 2004; “36” refers to Ethiopia for year 2004; “73” and “317” refer to Malaysia for years 2004 and 2006, respectively; “331” and “453” refer to Paraguay for years 2006 and 2007, respectively; and “721” refers to Thailand for year 2009. 33 Model B is not used for the residual analysis. As noted in Chapter 3, Model B is not a valid specification as it inappropriately assumes that the distance between the non-OSA countries is zero. 65 The actual merchandise exports by air values for Malaysia in years 2004 and 2006, and for Paraguay in years 2006 and 2007 are greater than the predicted values from both Models A and C, which lead to relatively large residual values. Low passenger traffic flows and low GDP interaction terms are the primary drivers for the low predicted values for these two nations. In particular, Malaysia has been the top 20 partners for US merchandise exports by air and has signed Open Skies agreements with the US since 1997. One would thus expect Malaysia to have greater economic power and larger passenger traffic flows between the US and Malaysia. Comparing to other top US merchandise trading partners such as Israel, US has exported less goods to Israel by air than to Malaysia. However, the GDP interaction term and the passenger traffic flow numbers for Israel are larger than those for Malaysia, despite that Israel has not concluded OSAs with the US until April 23, 2010. The numbers for Malaysia are inconsistent and therefore, it is not surprising to find Malaysia to be one of the outliers in the exports data panel. On the other hand, US merchandise imports by air from Cambodia in year 2004, from Ethiopia in year 2004, from Malaysia in year 2006, and from Thailand in year 2009 are outliers in the imports data panel. Similar to the exports data panel, relatively low passenger traffic flows and GDP interaction term put Malaysia, Thailand, and Cambodia as outliers in both Models A and C. On the other hand, Ethiopia’s predicted value is larger than the actual value by over 20%. This suggests that the US imported very small number of goods from Ethiopia and the amount is less than expectations. Overall, the residual and normal quantile-quantile plot analyses do not find the economies that have contributed most to the GDP interaction term, such as Japan, Germany, and China, to be outliers in the models. The actual exports and imports by air figures for these larger economies appear to match with expectations (i.e. predicted values). For example, one would find that these nations tend to be associated with more trade volumes and greater expansion opportunities that resulted from increasing number of business travels. These effects are by-products of the Open Skies agreements (on passenger traffic flows). 66 4.3 Preferred model specification Despite that each model specification conveys very similar messages about the linkages of Open Skies agreement and merchandise trade development, the best model still needs to be chosen. Using both quantitative and qualitative criteria, this paper finds that Model A and Model C are the preferred specifications for the US merchandise exports by air and US merchandise imports by air data, respectively. As such, conclusions regarding the effects of Open Skies agreements on merchandise trade by air data should be based on Model A for exports and Model C for imports. By qualitative criteria, Model B should be eliminated among the three model specifications. The results from Model B are most likely biased, since the product of OSA and distance term attributes zero weight for the distance between the US and non-OSA trading partner. Thus, it is unreasonable to assume that the distance for non-OSA countries are homogeneous. Furthermore, one should recall that distance is a proxy for transport cost and recording zero weight on distance for non-OSA countries would not be valid. This is similar to assuming that the air cargo cost for the non-OSA countries is zero and the cost for the OSA countries would be greater than zero. As such, quantitative criteria are used to choose the preferred model specifications between Models A and C. Normal quantile-quantile plots (Figure 10) suggest that both model specifications for the exports and imports data appear to be normal. At the 5% significance level, the Rainbow test34 suggests that Models A and C for the exports data and Model A for the imports data exhibit linearity. Most of the models appear to be normal and linear. Thus, the J-test (Davidson and MacKinnon (1981) test), which is intended for testing non-nested linear regression models, is used to determine the appropriate model specification. As a cross check, the Akaike’s Information Criteria is also used to select the best model. 34 The Rainbow test suggests that a good linear fit can be achieved on a subsample at the middle of the data, since the true relationship may not necessarily exhibit linearity due to outliers. The null hypothesis assumes that the overall fit for the true relationship is not significantly worst than the fit for the subsample at the middle of the data. If that is true, then the overall relationship exhibits linear relationship (Hothorn, Zeileis, Millo, & Mitchell, 2009). In this context, the Rainbow test for linearity is more suitable since outliers exist in the model, and they could distort the overall relationship. 67 Table 20: Davidson and MacKinnon (1981) J-Test results Coefficient and standard error in parenthesis P-value Exports Model A + fitted (Model C) -2.436 (1.455) 0.095 Model C + fitted (Model A) 2.440 (0.943) 0.010 Imports Model A + fitted (Model C) 2.820 (1.059) 0.008 Model C + fitted (Model A) -3.451 (1.832) 0.060 *Rejects test at 5% significance level Table 21: Akaike Information Criteria results Exports Imports Model A 1365.40 1942.99 Model C 1369.47 1939.29 *Grey shadings denote the best model specification The J-test results (Table 20) suggest that Model A is more preferable for the exports data, since including the fitted values of Model C into Model A does not produce statistically significant result. Instead, the inclusion of the fitted values of Model A into Model C is statistically significant, which implies that Model C is inadequate. The Akaike Information Criteria also shows that Model A is preferred over Model C for the exports data. On the other hand, the J-test and the Akaike Information Criteria suggest that Model C is better for the imports data. 4.4 Concluding remarks on model results To summarize, the preferred model specifications for exports and imports data are different. This further confirms that exports and imports data exhibit different trends. As such, it is not unreasonable to find that the effect of OSAs on merchandise exports by air to be different from that on merchandise imports by air. The difference in results and model specifications justify the need to analyze both exports and imports separately in order to properly assess the value of Open Skies agreements on trade statistics. The preferred model specifications convey several messages about the effect of Open Skies agreement on merchandise trade by air. Firstly, the cumulative effect of Open Skies agreements on exports to higher income countries is smaller than that for imports from trading partners of this income group. The free trade agreement also shows the coefficients for the exports panel to be smaller than those for the imports panel. Given that most of the trade by air has been conducted with higher income 68 economies, combination of OSA and FTA results is consistent with the current US net exports trend (Figure 9), where imports have been greater than exports. Concurrently, the preferred model specifications also show the negative OSA impact on imports by air from the lower income countries to be greater than that for exports to the same regions. The lower import levels from these economies help reduce some of the US’s current account deficit. Although the preferred model specifications for exports and imports show varying results for the Open Skies interaction dummy variables and the free trade agreement coefficient, Model A for exports and Model C for imports do exhibit the following similarities:  No additional benefit can be derived from the all cargo seventh freedom.  Exports and imports by air do not expect to increase substantially due to lower air cargo costs and due to expansion of air markets resulting from Open Skies agreements.  Instead, increase in business travel prior to the delivery of the good (in order to lower information asymmetries) helps expand trade opportunities. Increase in business travel is a byproduct of Open Skies agreements.  At the 5% significance level, currency unions do not appear to interfere with the effect of Open Skies agreements.  Finally, Canadian Blue Sky policies do not pose much threat to the US merchandise trade by air statistics. 69 Chapter 5 Conclusion This paper finds that the performance of Open Skies agreements is not robust to market volatilities. Normalized plots do not show US trade growth to be substantially different from the Canadian trade growth after the implementation of major rounds of OSAs. Using data between years 2004 and 2009, which attempts to measure the impact of OSAs on US merchandise trade by air during times of economic uncertainties, the models show that majority of the OSA-related coefficients are insignificant. On the other hand, the free trade agreement variable is large and statistically significant, which implies that reductions in air cargo costs and expansions of air markets resulting from Open Skies agreements is not strong enough to combat the trade declines when the recession hits. Yet, the difference in sign and magnitude of the OSA dummy variables do show that the number of years from which the Open Skies agreement was signed for, as well as the economic power of the trading partner do affect the effectiveness of these air service policies on US air-borne merchandise trade. Higher income economies appear to benefit more from OSAs than lower income economies, and agreements signed for more than five years appear to be better. The preferred model specifications are different for exports and imports by air statistics. Model results also show that the effect of Open Skies agreements to vary between the two sets of data. These findings suggest the need to analyze both exports and imports by air data rather than use one or the other to explain the behaviour of OSAs. Within the lower income group, the decline in imports is substantially greater than exports after the US engaged in OSAs for four years. The cumulative effect for US exports by air to the higher income economies is less than the effect for imports, which is consistent with the current US net exports trend. Trade imperfections used to protect the US domestic industries and to reduce the US current account deficit may also be contributing factors to the differences in performance of OSAs on merchandise exports and imports by air. The model results show that the impact of Open Skies policies on passenger traffic flows indirectly improves US trade by air figures. Previous studies have suggested that OSAs stimulate passenger traffic growth and passenger traffic growth would also implicitly assume greater number of business travels. Model results show that lagged passenger traffic flow variable is positively related to trade value. As noted previously, increased business travel opportunities conducted prior to the delivery of the goods 70 would help lower information asymmetries, develop trust between supply chain partners, and thereby aid expansion of trade of all kinds (both air and surface transportation). This research, however suffers from shortcomings. Low-income and lower-middle income countries do not have data for some of the sample periods. This causes an unbalanced panel, which may distort the results. This problem is further magnified with a short sample period. Furthermore, the model also does not consider dynamic properties. As such, future work should consider improving the model using Generalized Method of Moments (GMM) techniques. Failing to incorporate alliances into the model is another weakness in this research. Including alliances into the model is not easy, since airline alliances for passenger traffic do not necessarily imply that cargo (merchandise trade) would follow the same structure. Zhang and Zhang (2002) note that alliances have worked well for passenger traffic. However, the effect of alliances on cargo traffic is not clear, since air cargo alliances are more likely to occur for certain markets rather than on a global basis (Zhang & Zhang, 2002). This paper assumes that the date concluded for the Open Skies agreements equals to when the agreement was actually initiated. However, this is not necessarily true. The date concluded could reflect the day the agreement was revised, not when the agreement was initially signed. For example, the U.S. Department of State (2010) indicated on August 3, 2010 that the US concluded agreement with Switzerland this year; but on March 17, 2010, the same agency stated that the agreement with Switzerland was concluded in year 1995. It is generally unclear when the agreement officially started since contracts do revise from time to time and the U.S. Department of State does not provide dates of when the first version of the contract was concluded. As such, given short sample period, different years inserted into the model for when the Open Skies agreements was signed could skew the results. The model also assumes that the initiation date of the all-cargo seventh freedom is the same as when the Open Skies policies were implemented and only countries that signed OSA with the US with the all-cargo seventh freedom provision would feel the effect of this air service right. However, this may not necessarily be the case. As noted previously, the all-cargo seventh freedom would have spill-over effects to non-OSA countries. For example, Fu, Lei, and Zhang (2010) note that US cargo carriers have hub operations in China given the Singapore-US Open Skies agreement (which contains the all-cargo 71 seventh freedom provision). It is also possible that the US has engaged in some form of all-cargo seventh freedom with other non-OSA countries, without having liberalized agreements for passenger traffic. For example, Kasarda and Green (2005) note that the July 2004 China-US Air Service Agreement focused primarily on all-cargo air service rights. Fu, Lei, and Zhang (2010) also mention that the 2007 version of the China-US ASA stated that both countries would engage in Open Skies for cargo flights by 2011. Yet, these two versions of the agreement did not mention engaging Open Skies for passenger traffic. In this case, the US would not consider itself having Open Skies agreement with China. The OSA dummy variables used in this paper would thus not include China. As such, holding all else constant, the calculation would have underestimated the effect of the all-cargo seventh freedom and Open Skies agreements given the large contributions that China has made to the US imports by air data. Furthermore, cargo and passenger traffic operate in different structures. Zhang and Zhang (2002) suggest the possibility that cargo having air service rights different from those for passenger traffic. The current Open Skies agreements are primarily based on passenger traffic trends. However, Kasarda and Green (2005) suggest that air cargo is an important lead indicator of economic development. As such, future work should consider simulating the effect of having air service rights for cargo traffic only, by examining the effect of all-cargo seventh freedom on OSA and non-OSA countries. Despite inevitable shortcomings with the models, this thesis would like to encourage Canadian policymakers to explore linkages between Canadian international trade and Canadian Open Skies policies. This is particularly important when the model results show that Canadian Blue Sky policies are ineffective and are not a threat to US trade by air statistics. The small number of Canadian Open Skies agreements signed to date, as well as restrictive terms of the agreements could contribute to the Blue Sky policies’ poor performance. Overall, understanding international trade development is important, as both imports and exports play major roles in the Canadian economy. More beneficial Open Skies policies would help stimulate Canadian business opportunities, which would help achieve desirable economic growth especially during times of economic slowdown. As such, future work should focus on Canada if publicly available Canadian trade micro-data are available for analysis. Similar analysis can also be done using commodity group breakdowns to evaluate if certain commodities are more responsive to Open Skies agreements. 72 At times of global economic slowdown, it would also be interesting to do cross country comparisons using similar models if sufficient data are available. The experience of one nation, such as the United States, cannot be logically inferred to other countries with different economic and social characteristics. As such, cross country comparisons allow policy makers and researchers to understand how different countries react to implementations of Open Skies agreements. 73 References Anderson, J. E. (1979). A theoretical foundation for the gravity equation. The American Economic Review, 69 (1), 106-116. Baier, S. L., & Bergstrand, J. H. (2009). Estimating the effects of free trade agreements on international trade flows using matching econometrics. Journal of International Economics, 77, 63-76. Bank of Canada. (2010, January 21). Monetary Policy Report. Retrieved from http://www.bankofcanada.ca/en/mpr/mpr_previous.html BC Ministry of Transportation and Infrastructure. (2009). Benefits of Open Skies for British Columbia. Retrieved September 12, 2009, from http://www.th.gov.bc.ca/openskies Bergstrand, J. (1985). The gravity equation in international trade: Some microeconomic foundations and empirical evidence. Review of Economics and Statistics, 67 (3), 474-481. Calza, A., Manrique, M., & Sousa, J. (2006). Credit in the euro area: An empirical investigation using aggregate data. The Quarterly Review of Economics and Finance, 46, 211-226. Central Intelligence Agency (CIA). (2010). The World Factbook. Retrieved June 5, 2010, from http://www.cia.gov/cia/publications/factbook/ Clougherty, J. A., Dresner, M., & Oum, T. H. (2001). An empirical analysis of Canadian international air policy: Effects of dual carrier designation and partial liberalization. Transport Policy, 8, 219-230. Croissant, Y., & Millo, G. (2008). Panel data econometrics in R: The plm package. Journal of Statistical Software, 27 (2), 1-43. Davidson, R., & MacKinnon, J. G. (2004). Econometric theory and methods. New York: Oxford University Press. Disdier, A.-C., & Head, K. (2006, July 29). The puzzling persistence of the distance effect on bilateral trade. Retrieved from UBC Sauder School of Business website: http://strategy.sauder.ubc.ca/head/pubs.html Dresner, M., & Oum, T. H. (1998). The effect of liberalised air transport bilaterals on foreign traffic diversion. Journal of Transport Economics and Policy, 32 (3), 317-329. 74 Endo, N. (2007). International trade in air transport services: Penetration of foreign airlines into Japan under the bilateral aviation policies of the US and Japan. Journal of Air Transport Management , 13, 285-292. Foreign Affairs and International Trade Canada. (2010). Air agreements recently signed or updated by Canada. Retrieved December 7, 2010, from http://www.international.gc.ca/trade-agreements- accords-commerciaux/agr-acc/facts-air-eclair.aspx?lang=eng Fu, X., Lei, Z., & Zhang, A. (2010, May). Development, prospect and challenges in China's air transport. Paper presented at Cheng-du. 1-6. Gillen, D. (2009, February 27). Canadian international aviation: Policy and challenges. Paper presented at the Calgary Chamber of Commerce and The Van Horne Institute. 1-20. Gillen, D., Harris, R., & Oum, T. H. (2002). Measuring the economic effects of bilateral liberalization air transport. Transportation Research Part E, 38, 155-174. Glick, R., & Rose, A. K. (2002). Does a currency union affect trade? The time-series evidence. European Economic Review, 46, 1125-1151. Head, K., & Mayer, T. (2002). Illusory border effects: Distance measurement inflates estimates of home bias in trade. Retrieved from CEPII Working Papers website: http://www.cepii.fr/anglaisgraph/workpap/wp2002.asp Head, K., & Ries, J. (2009, December 9). Do trade missions increase trade. Retrieved from UBC Sauder School of Business website: http://strategy.sauder.ubc.ca/head/pubs.html Hothorn, T., Zeileis, A., Millo, G., & Mitchell, D. (2009, December 8). Package \"lmtest\": Testing linear regression models. Retrieved June 13, 2010, from Comprehensive R Archive Network website: http://cran.at.r-project.org International Air Transport Association. (2010, March). Cargo E-Chartbook Q1 2010. Retrieved from http://www.iata.org/economics InterVISTAS Consulting Inc. (2008, February 25). Comparison of US Open Skies agreements with Canadian Blue Sky arrangements. Retrieved from http://www.intervistas.com/publications/publications.php 75 InterVISTAS Consulting Inc. (2005, June 30). The economic impacts of the Open Skies initiative -- Past and future. Retrieved from http://www.intervistas.com/publications/publications.php InterVISTAS Consulting, Inc. (2006, June). The economic impact of air service liberalization -- Final report. Retrieved from http://www.intervistas.com/publications/publications.php Kasarda, J. D., & Green, J. D. (2005). Air cargo as an economic development engine: A note on opportunities and constraints. Journal of Air Transport Management, 11, 459-462. Micco, A., & Serebrisky, T. (2006). Competition regimes and air transport costs: The effects of open skies agreements. Journal of International Economics, 70, 25-51. Oum, T. (2007, February). International air services and agreements [Lecture notes]. Retrieved from http://www.elearning.ubc.ca U.S. Census Bureau. (2010). U.S. merchandise of trade: Selected highlights. Retrieved December 9, 2010, from http://www.census.gov/foreign-trade/Press-Release/ft920_index.html U.S. Department of State. (2010). Open Skies partners. Retrieved December 7, 2010, from http://www.state.gov/e/eeb/tra/ata/ World Trade Organization. (2010). WTO Statistical data sets - Metadata. Retrieved from http//www.stat.wto.org/TechnicalNotes%5CDataSetTechnicalNotes_E.htm Yamaguchi, K. (2008). International trade and air cargo: Analysis of US export and air transport policy. Transportation Research Part E, 44, 653-663. Zhang, A., & Zhang, Y. (2002). Issues on liberalization of air cargo services in international aviation. Journal of Air Transport Management, 8, 275-287. "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2011-05"@en ; edm:isShownAt "10.14288/1.0071593"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Business Administration"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivatives 4.0 International"@en ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/4.0/"@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Exploring the linkages between Open Skies agreements signed by the United States and international trade development"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/30817"@en .