Infrastructure, Competition Regimes, and Air Transport Costs: Cross-Country Evidence Alejandro Micco and Tomás Serebrisky12 Abstract: The relevance of transport costs has increased as liberalization continues to reduce artificial barriers to trade. Countries need to adopt policies to "get closer" to global markets. Can improvements in infrastructure and regulation reduce transport costs? Is it worthwhile to implement policies designed to increase competition in transport markets? Focusing on air transport, which has increased its share in US imports from 24% in 1990 to 35% in 2000, this paper quantifies the effects of infrastructure, regulatory quality and liberalization of air cargo markets on transport costs. During the 1990s, the United States implemented a series of Open Skies Agreements, providing a unique opportunity to assess the effect that a change in the competition regime has on prices. We find that infrastructure, quality of regulation and competition matter. In our sample, an improvement in airport infrastructure from the 25th to 75th percentiles reduces air transport costs by 15 percent. A similar improvement in the quality of regulation reduces air transport costs by 14 percent. Open Skies Agreements further reduce air transport costs by 8 percent. Keywords: Infrastructure, transport costs, air transport liberalization, regulatory quality. JEL Classification: [F4], [L4], [L9] World Bank Policy Research Working Paper 3355, July 2004 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ. worldbank.org. 1 Inter-American Development Bank and World Bank respectively. Contact information: alejandromi@iadb.org and tserebrisky@worldbank.org. 2 We thank Danielken Molina for valuable research assistance and the participants of the Trade Seminar of the World Bank for their valuable comments. 1 I. Introduction There is a close relationship between trade costs and the capacity of a country to increase its exports and to integrate in the world economy. The relevance of transport costs, as a component of trade costs, has been increasing as liberalization continues to reduce artificial barriers to trade. In many cases, the effective rate of protection provided by transport costs is higher than the one provided by tariffs (Clark et al., 2004; Hummels, 1999). One of the most important and evident components of transport costs is distance. In its simplest formulation, the gravity model for trade, introduced by Linnemmann (1966), states that bilateral trade flows depend positively on the product of the GDPs of both economies and negatively on the distance between them, which stands for bilateral transport costs. The impact of distance on countries' volume of trade is significant: recent estimates of the elasticity of trade volumes with respect to distance indicate that when distance increases by 10 percent, the volume of trade is reduced between 9 and 15 percent (Overman et al., 2003).3 However, in addition to distance, many other elements influence transport costs. As explained by Limão and Venables (2001), transport costs and trade volumes depend on many complex details of geography, infrastructure, administrative barriers and the state of competition in the transport industry. Provided that distance and infrastructure related costs are major determinants of the success of a country's export sector, immediate questions arise: what can governments do to "get closer" to markets with high import demand? Can improvements in infrastructure and regulation reduce transport costs? Is it worthwhile to implement policies designed to increase competition in transport markets? Do these policies have a quantifiable impact on transport costs? Not many papers have tried to estimate the impact on transport costs of policies that improve the quality of regulation and infrastructure or implement new competition regimes. Focusing on infrastructure and using data from maritime shipping companies, Limão and Venables (2001) show that poor infrastructure accounts for more than 40% of predicted transport costs. In a study specific to the port sector, Clark et al. (2004) show that an improvement in port efficiency from the 25th to the 75th percentile reduces 2 shipping costs by more than 12 percent.4 Fink et al. (2002) argue that both public policies, like restrictions on the provision of port services and private practices-- collusive carrier arrangements--exercise a significant influence on maritime transport costs. A policy implication derived from their estimation is the need to pursue attempts to break up international cartels in that market.5 Because their data do not include a change in the intensity of competition in the market (i.e., from cartel to non-collusive behavior), they cannot estimate the effects on transport costs of a change in the competition regime. The aim of this paper is to address this gap in the literature. In particular, we estimate the effects of infrastructure, quality of regulation and changes in the competition regime on air transport costs. We focus exclusively on the costs of air transport due to its increasing importance as a transport mode, the availability of detailed micro data for US imports and the recent change in competition regimes introduced by Open Skies Agreements. The advent of wide-body aircrafts in the 1970s made available large volumes of belly space. Being able to accept palletized or containerized freight, airlines began addressing the air cargo market more aggressively. As the evolution and design of aircrafts made it possible to carry more cargo in an efficient manner, dedicated cargo airlines entered this market.6 The size of the airfreight and express market worldwide is approximately USD 75 billion, and during the 1990s this market grew at an average rate of 6% per year.7 The geographic distribution of the revenue generated in the air cargo and air passenger markets are similar, explaining the US almost 40 % of total revenue. In the United States, as indicated by Figure 1, the value of air shipments relative to the aggregate value of air and vessel shipments increased from 24% in 1990 to 35% in 2000. The drastic drop in air 3 Deardorff (1984) surveys the early work on this subject. 4Clark et al (2004) show that reductions in country inefficiencies associated to transport costs from the 25th to 75th percentiles imply an increase in bilateral trade of around 25 percent. 5 Clark et al. (2004) show that their results are not robust to the inclusion of additional control variables (e.g., unit value of the shipped merchandise). 6 As Walker (1999) explains, until recently, the majority of airlines considered cargo a by-product of their passenger activities. The consequence was a pricing strategy that regarded cargo as a low additional cost product. In recent times, airlines significantly improved the way common costs are allocated and implemented a yield management strategy, which involves the creation of a stand- alone company with full profit responsibility. 7 Data obtained from Air Cargo Management Group (www.cargofacts.com). 3 shipments in 2001 may have been caused by the restrictions that the United States imposed to air traffic after September 11th 2001.8 Figure 1. Share of Air Shipments in US Imports 36.00 35.63 35.30 34.38 33.98 34.00 ) 32.68 32.68 %( 32.09 stne 32.00 mpihSriAfo 30.00 29.16 e 28.00 27.57 harS 26.58 26.00 25.68 24.37 24.00 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Year Share: Value of Air Shipments / (Value of Air Shipments and Vessel Shipments) SOURCE: U.S. Imports of Merchandise 1990 - 2001 ­ U.S. Department of Commerce. In 1992 the United States signed the first open skies agreement with the Netherlands. Since then, the United States has signed more than 55 Open Skies Agreements with developed and developing countries on all continents. These agreements give us a unique opportunity to estimate the effect that a liberalized air cargo market has on transport costs. Given that there is no estimation in the literature of the effects of Open Skies Agreements, or any other change in competition regime, on cargo rates, this paper adds a new dimension to the literature that estimates the determinants of transport costs. 8 According to data provided by the Air Transport Association, the recent evolution of international cargo transport for US airlines (measured in million of ton miles) is as follows: 1999: 23,501; 2000: 25,121; 2001: 22,421; 2002:23,627; 2003: 24198 (estimated data). 4 Bilateral Open-Skies Agreements. United States. Year Country Year Country Year Country 2002 Jamaica 1999 Dominican Republic 1997 El Salvador 2002 Cape Verde 1999 Tanzania 1997 Guatemala 2002 Uganda 1999 Qatar 1997 Panama 2001 Sri Lanka 1999 Argentina 1997 Taiwan 2001 France 1999 Bahrain 1997 Brunei 2001 Oman 1999 United Arab Emirates 1997 Singapore 2001 Poland 1999 Pakistan 1996 Jordan 2000 Senegal 1998 Italy 1996 Germany 2000 Benin 1998 Peru 1995 Czech Republic 2000 Malta 1998 Korea 1995 Austria 2000 Rwanda 1998 Uzbekistan 1995 Belgium 2000 Morocco 1997 Netherlands Antilles 1995 Denmark 2000 Nigeria 1997 Romania 1995 Finland 2000 The Gambia 1997 Chile 1995 Iceland 2000 Turkey 1997 Aruba 1995 Luxembourg 2000 Ghana 1997 Malaysia 1995 Norway 2000 Burkina Faso 1997 New Zealand 1995 Sweden 2000 Namibia 1997 Nicaragua 1995 Switzerland 2000 Slovak Republic 1997 Costa Rica 1992 Netherlands 1999 Portugal 1997 Honduras Source: Aviation and International Affairs of the U.S Department of Transportation. http://ostpxweb.dot.gov/aviation/ The results obtained have important policy implications. We find strong evidence that investments in airport infrastructure and improvements in the quality of regulation reduce air transport costs. In our sample, improvements in both variables from the 25th percentile to the 75th percentile reduce transport costs by more than 20 percent. In addition, we find that a more competitive air transport market--through Open Skies Agreements-- reduces air transport costs by around 8 percent.9 The paper is organized as follows. Section II briefly summarizes the economics of air cargo. Section III presents the empirical framework while Section IV shows the results for the cross-section and panel data estimation. Finally, Section V concludes. II. The Determinants of Air Transport Costs, the Economics of Air Cargo and Open Skies Agreements This section addresses, based on a qualitative description, the main determinants 9 This result depends on the econometric specification used. 5 of air transport costs, emphasizing recent developments in the economics of air cargo. The nature of the services provided by air cargo airlines forces them to be both capital intensive and transnational companies serving more than one country. In general, these companies have access to international capital markets and they are able to hire a fraction of their workers from all over the world.10 Thus we should not expect differences in capital or labor costs to be the main factors explaining differences in transport costs across countries. However there are other important specific factors affecting transport costs across countries. The most studied determinant of transport cost is geography, particularly distance. The greater the distance between two markets, the higher the expected transport costs. For air carriers, the cost variable most affected by distance is fuel cost, which during most of the 1990s represented between 12 and 15 percent of airlines' total operating costs (Doganis, 2001). Dedicated freight airlines pay special attention to airport use-related fees. They have more flexibility than passenger airlines, as they do not have to operate from airports with the best location for business passengers and they can avoid slot-constrained airports or operate during off-peak hours. Consequently, dedicated freight airlines usually have a broader selection of competing airports to choose from. The available airport infrastructure and the quality of regulation, which have a direct impact on airport use fees, are important variables for cargo airlines when deciding which airports they serve, and thus are relevant variables in the determination of air transport costs. There is not a unique model for tariff regulation in airports. However, in the majority of airports, tariffs for "aeronautical services" (runway and taxiway, air control, aircraft parking, security) are regulated by a government agency or sector-specific regulator (Serebrisky et al., 2002). The quality of regulation, i.e., the level and structure of tariffs for aeronautical services, how they are set and the regulatory process to modify them, are a key factor that directly impacts airlines' operating costs and transport costs. The creation of cargo dedicated airlines and the implementation of yield management strategies in passenger-cargo airlines allowed these companies to adopt a 10 Doganis (2001) shows that for passenger airlines labor costs explain between 25 to 35 percent of total operating costs. Undoubtedly, for dedicated freight carriers, this percentage is much lower because they do not need to employ flight attendants and 6 flexible approach to the selling of cargo space. As a consequence, there are not only directional differences in rates, mainly due to bilateral trade imbalance,11 but there are also wide variations in the rates offered to bulk contract shippers as opposed to one-off clients.12 Trade composition additionally helps to explain other differences in transport costs. Due to the insurance component of transport costs, products with higher unit value have higher charges per unit of weight. On average, insurance fees are around 1.75 percent of the traded value and represent around 15 percent of total air charges. Therefore, high value added exporting countries should have higher charges per unit of weight due to this insurance component.13 Besides, some products require special transport features and therefore have different freight rates.14 Finally, competition regimens also matter for air transport costs. Since 1992, the United States has signed more than 50 bilateral Open Skies Agreements. The main objective sought by these agreements is the promotion of an international aviation system based on competition among airlines with minimum government regulation. As stated in the introduction of these agreements,15 the motivation governments have to support them is the desire to facilitate the expansion of air transport opportunities, making it possible for airlines to offer the traveling and shipping public a variety of service options at the lowest possible prices. The substance of all Open Skies Agreements signed by the United States is alike, and apply to passenger, all-cargo and combination air transportation and encompass both scheduled and charter services. Key provisions include16: Free market competition: no restrictions on international route rights; number of designated airlines; capacity; frequencies; and types of aircrafts. Pricing determined by market forces: a fare can be disallowed only if both other personnel that work in passenger related services (VIP lounges, ckeck-in counters, customer service personnel in airports, etc.). 11 Directional imbalance in trade between countries implies that many air-carriers are forced to haul empty space back. As a result, either imports or exports become more expensive. 12 Transport is a classic example of an industry that faces increasing return to scale. 13 Clark et al. (2004) show that the insurance component is an important determinant for maritime transport costs as well. 14 For example, in the case of maritime transport costs, LSU-National Ports and Waterways Institute (1998) shows that the average freight rates between Central America and Miami for cooled load merchandise is about twice the transport cost for textiles. 15 The text of open skies agreements can be found at http://ostpxweb.dot.gov/aviation/ 7 governments concur ­"double disapproval pricing"- and only for certain, specified reasons intended to ensure competition. Fair and equal opportunity to compete: all carriers of both countries may establish sales offices in the other country, can convert and remit earnings in hard currencies at any time. Besides, designated airlines are free to provide their own ground handling services and airlines and cargo consolidators may arrange ground transport of air cargo and are guaranteed access to customs services. According to the model text of Open Skies Agreements, user charges can not be discriminatory and should be based on costs. The text also includes procedures for resolving differences that arise under the agreement. Optional 7th Freedom all-cargo rights: provide authority for an airline of one country to operate all-cargo services between the other country and a third country, via flights that are not linked to its homeland. Most of the Open Skies Agreements signed by the United States include 7th freedom for all-cargo services. The inclusion of specific provisions for air cargo in the Open Skies Agreements signed by the United States suggests that, irrespective of the size of the air cargo market, the American government is concerned about entry barriers, competition and ultimately prices of air cargo services. Most of the empirical literature in the area of air transport focuses on the effects that Open Skies Agreements have on passengers. A recent and comprehensive study (Brattle Group, 2002) estimates the effects on passengers of Open Skies Agreements between the United States and countries in the European Union. Button (2002) describes the potential effects that the liberalization of U.S-European air transport market could have on airlines, passengers and labor, but he does not provide any quantitative evidence. The U.S. Department of Transportation (2000) published a report that argues that between 1996 and 1999 average passenger airfares in transatlantic markets declined 10.3% in non-open skies countries and 20.1% in open skies countries. This report neither controls for other factors nor explains the methodology used to estimate the reduction in airfares. Despite some work on the effects of open skies on passenger fares, a survey of the 16 For a detailed explanation and examples see http://www.state.gov/e/eb/rls/fs/208.htm 8 empirical literature shows that there is no estimation of the effects of Open Skies Agreements on cargo rates, a task we pursue in this paper. III. Empirical Framework To estimate the importance of each of the factors that explain air transport costs, we use a standard reduced-form approach. Air transport freight prices are assumed to be equal to the marginal cost multiplied by the air shipping companies' markup. Expressed in logarithms, the reduced form equation takes the following form: pI jkt = mc(I, j,k,t) + µ(I, j,k,t) [1] Where: I : corresponds to a foreign country . j : corresponds to a US district or region . k: corresponds to the product, aggregated at four digits of the HS classification code. pijkt: air transport cost (or charges). mc : marginal cost expressed in logarithm. µ : markup expressed in logarithm. Therefore, pIjkt represents air transport costs17 (freight charges), and it is measured by the logarithm of the freight charges per unit of weight for each of the k products transported between foreign country I to any of the US districts or regions j in period t.18 19 We assume that both marginal cost and markup are functions of factors that depend on the airport or country of origin (I), and the airport or district of destiny in the US (j) for each of the k product types. Specifically, we assume that the marginal cost has 17 We do not take into account the observations that have zero trade with the US. 18 We consider 36 US districts. 19 We divide the districts in three regions. 9 the following functional form: mc(I, j, k, t) = j + k + wvIjkt + dIj + qIjt + unbIjt + APEIj + I jkt [2] Where: j : dummy variable referring to US district j. k : dummy variable referring to product k. wvIjkt : represents the logarithm of the value per unit of weight of product k. dIj : logarithm of the distance between country I and district j in the US. qIjt : logarithm of the value of imports carried by air from country I to the US. unbIjt : unbalance between country I and the US. APEI : state of infrastructure in airports of foreign country I. The first dummy takes into account potential differences in airport efficiencies across US custom regions or districts, and the second one accounts for different marginal transport costs across products. wvijkt represents the value per unit of weight of product k and is used as a proxy for the insurance component of air transport cost (pIjkt)20. Unbalance between country I and the US corresponds to the ratio between US exports minus US imports and the level of bilateral trade between the two countries. The variable APEI is a proxy for the state of infrastructure in foreign airports and corresponds to the ratio between the squared number of airports in the foreign country that have runways over 1,500m long and the product of the country area and total population. In addition, in some specifications we include variables that capture the quality of regulation across countries. With respect to the second term of the reduced form equation, we assume that air shipping companies' markups have the following functional form: 20 We assume that insurance costs increase unit price of transported goods. 10 µ(I, j,k,t) = k +AIJt [3] Where: k : dummy variable per product k AIt: dummy variable for Open Skies Agreements between country I and the US. k is a product-specific dummy variable that captures differences in transport demand elasticities across goods (derived from the final demand of each good k in the US). AIJt is a variable that specifies if a pair of countries has an Open Skies Agreement. Substituting the second and third equations into the first one, we obtain the econometric model to be estimated: p Ijkt = j + k + wv Ijkt + d Ij+ qIjt + unb Ijt + APE Ij + AIjt + [4] Ijkt where + Ijk : error term21 In the estimation, we should expect a positive sign for the coefficients of the proxy for insurance costs, , and distance, , and a negative sign for the coefficients of the volume of trade, , trade unbalance, , the state of airport infrastructure, , and Open Sky Agreements, . We divide our empirical results in two sets. First, we run cross-sectional regressions to identify the effects on transport costs of those variables that rarely or never vary over time. These variables are distance and foreign airport infrastructure. Using cross-sectional regressions we are able to identify whether countries that implemented an Open Skies Agreement face lower or higher transport costs than others that did not. 21 We allow the error term to be correlated among country clusters. 11 Although this information by itself is valuable, it is not precisely the most relevant question for a policymaker. A policymaker would like to know what is the impact of signing Open Skies Agreements; that is, if by implementing this type of agreements, transport costs are reduced over time. To answer this question, our second set of results relies on panel data and includes country fixed effects in order to isolate the time-series dimension of Open Skies Agreements and their impact on air transport costs, leaving out the cross-sectional variation. Thus, time-invariant country specific variables such as distance between a foreign country and the US will be subsumed in these country fixed effects. In addition, to some extent, the inclusion of country dummies addresses potential endogeneity problems that would arise if countries following a cost-benefit analysis tend to invest in an Open Skies Agreement only with partners with which they have high air transport costs, since the potential benefits for these countries derived from more competition will be higher. If this is the case, a cross-sectional analysis would underestimate the effect of Open Skies Agreements on transport costs. Indeed, as will be shown below, a comparison of our panel results with those obtained when we use cross-section regressions suggests that the latter in fact understate the impact of Open Skies Agreements on air transport costs.22 It is important to point out that the use of country dummies does not fully eliminate the endogeneity bias. It is possible that countries decide to sign an open skies agreement following a substantial increase in their air transport costs. The shorter the period used to estimate the effect since the implementation of Open Skies Agreements, the less severe are the remaining concerns about endogeneity.23 In any case, if endogeneity were a problem, one would expect our estimates to underestimate the effect of Open Skies Agreement on transport costs. The relation between transport costs and imported volume causes an additional endogeneity problem. To control for this endogeneity problem, in the empirical section, following the gravity literature on trade, we use the foreign country's GDP as an instrument for the volume of imports. 22 Glick and Rose (2001) and Micco, Stein and Ordoñez (2003) use this same technique to identify the effect of currency union and EMU on trade, respectively. 12 IV. Empirical Results Data on air-transport costs comes from the U.S. Imports of Merchandise Database put together by the U.S. Department of Commerce. The level of data aggregation is HS four-digit, and the period covered is 1990 ­ 2001. Our dependent variable, air-transport costs, is the variable imports charges (by unit of weight), which is defined by the U.S. Bureau of Census as: "...the aggregate cost of all freight, insurance, and other charges (excluding U.S. import duties) incurred in bringing the merchandise from alongside the carrier at the port of exportation--in the country of exportation--and placing it alongside the carrier at the first port of entry in the United States."24 Our explanatory variables are HS four-digit aggregated, and they were obtained from different sources. In the case of value of imports, volume of imports and directional unbalance,25 the source is the U.S. Imports of Merchandise Database gathered by the U.S. Department of Commerce, 1990 - 2001. Population and GDP data were obtained from the World Bank's World Development Indicators (2002). Most of the country-specific variables (coordinates for the calculation of distances, number of airports, etc.) were taken from the CIA's World Factbook.26 Finally, the information for regulatory quality and government efficiency were taken from Kaufmann et al. (2003). Details on the definition of the variables are provided in Appendix A. Cross Sectional Results Cross Sectional results are reported in Table 1 and Table 2. The main difference between these tables is that in Table 2 we control for country development level using a dummy variable for developed countries.27 Both tables report the results for the year 23 For this reason, our short period of time is perfect to estimate the effect of open skies. 24 To avoid a measurement error, in our empirical exercises we only use countries that have 50 or more observations in our sample. Our results are robust to the inclusion of those countries with less than 50 observations. 25 This variable was only available for the year 2000. 26 We obtained the number of airports from the CIA World Fact Book, 1990 ­ 2001. 27 To control for this country specific characteristic, we constructed a dummy variable that takes the value of 1 if the country is classified by the World Bank Income Classification (2002) as a high income country, and zero in all other cases. 13 2000. We chose the year 2000 because it was the only year for which we had information to construct the variable directional trade imbalance. Table 1 reports our estimations for equation [4]. In all the specifications we control for distance, volume (US imports measured in US dollars), product unit value, directional trade unbalance, type of product and district of cargo entry in the US and airport infrastructure in the exporter country. In all the specifications we allow the error term to be correlated among country clusters to avoid misspecifications of the var-cov matrix.28 As we already mentioned, when we introduce import volume, an endogeneity problem arises. It is expected that the bigger the trade volume, the lower transport costs are going to be (due to economies of scale) but, on the other way around, lower transport costs may increase trade. To solve this problem in all specifications shown, we instrumentalize imports volume using countries' GDP.29 Column (1) reports the results obtained for the benchmark regression specified in equation [4] using the square number of large airports in the foreign country normalized by area and population as our measure of airport infrastructure. As shown, distance has a significant (at 1%) and positive effect on air transport costs. For instance, doubling the distance between a country I and the US generates a 20 percent increase in air transport costs. The variable capturing economies of scale is the level of trade that goes through a particular route.30 This variable, calculated in terms of monetary volume, has the expected significant and negative coefficient. This result could be explained by the fact that the more transited routes are served by the biggest airplanes, which have more cargo space available, or have more competition due to the presence of more cargo companies covering the same route. In our sample, an increase in import volume from the level of Zimbabwe (percentile 25) to the level of Denmark (percentile 75) reduces air transport costs by around 11 percent. The value per weight variable is also positive and highly significant. As already mentioned, these regressions include dummy variables for products aggregated at the 28 Clusters control for the fact that, even though we use thousands of air-shipping observations, most of our variables of interest only vary across countries. 29 Following the gravity model, trade between two countries is proportional to the product of countries' GDP divided by the distance between them. 30 Each "foreign country and US region" pair is defined as an air route. We define three regions in the US: East, West and Gulf 14 four-digit HS level. One might think that unit values would be quite similar across countries at that level of disaggregation, but that is not the case. Clark et al. (2004) found the same results for maritime transport costs. Additionally, Feenstra (1996) shows that there is a large variation in unit values even at the 10-digit HS level. He cites the example of cotton shirts for men, which the U.S. imports from almost half of its 162 trading partners. The unit values range from $56 (Japan) to $1 (Senegal). These differences in unit values lead to large differences in insurance costs per kilogram, even for "homogeneous" products. Thus, it is not surprising that we find that the more expensive the product per unit of weight, the higher the insurance and hence the overall transport cost. Directional imbalance in trade between the US and the source country has the expected negative sign and is significant at one percent. If we move from a favorable unbalance (from the point of view of the exporters to the US31) of 50 percent to a negative one of the same amount, air transport costs increase around 16 percent. Finally, the coefficient associated to airport infrastructure is negative and significant (at 1%): the greater the investment in infrastructure, the lower transport costs are. In our sample, an improvement in airport infrastructure from the level of Colombia (p25) to the level of Great Britain (p75) reduces air transport costs in 15 percent. This result may be due to the following reverse causality. Airport infrastructure reduces air transport costs, but at the same time, low air transports costs increases trade and may induce investments in airport infrastructure. To control for this reverse causality in column (2) we instrumentalize the level of foreign airport infrastructure of column (1) with indexes of telephones per capita and paved roads. In addition, as a robustness test, in column (3) we use a five-year lag of our infrastructure variable. In both cases we obtain very significant effects, similar to the results obtained in column (1). In columns (4), (5) and (6), besides the level of airport infrastructure, we include institutional variables to observe their effect on air transport costs. We should expect a better institutional framework -measured by regulatory quality and/or government Coast (see Appendix B, Table B4.). 31 For foreign exporters, the larger is the unbalance of US bilateral trade (exports ­ imports divided by bilateral trade) the lower should be the transport costs they face because of the low capacity utilization of the airplanes coming back to the US. 15 efficiency- to reduce air transport costs.32 In columns (4) to (6) we present our results using regulatory quality and/or government efficiency obtained from Kaufmann et al. (2003).33 As shown, separately, both dimensions have the expected sign and both are statistically significant at conventional levels (Column 4 and 5). But, when we include both institutional variables at the same time, only regulatory quality has the expected sign and is significant at standard levels.34 Not surprisingly, this last result suggests that regulatory quality, which should directly affect airport efficiency, has a significant effect on air transport costs. In our sample, a country that increases its level of regulatory quality from the level of Ecuador (p25) to the level of France (p75) can experience a reduction in its air transport costs of 14 percent. The last two columns of Table 1 report the results when we include our Open Skies Agreement dummy variable35 with and without regulatory quality. In both cases, Open Skies Agreement has the expected negative sign, but it is not statistically different from zero at conventional levels. It is possible that this result is biased to zero due to the endogeneity problem described at the beginning of this section (countries with initial high transport cost are willing to sign Open Skies Agreements). Unreported regressions show that our results are robust to the use of other years in our sample instead of 2000.36 To see if our results are robust to the addition of country-income controls, in Table 2 we include, in all specifications, a dummy variable that takes the value of one if the foreign country is a developed country (high income country), and zero otherwise. We found that all our previous results are robust to the inclusion of this dummy, except for the total volume variable, which keeps its negative sign in all specifications but is not statistically different from zero. The variables of airport infrastructure and regulatory 32 See Kaufmann,Kraay, and Mastruzzi (2003) for a general discussion of the role of the institutional environment in country performance. 33 To check the robustness of these results, we estimated the benchmark equation with Rule of Law and Control of Corruption and we found that the results obtained where very similar in magnitude of coefficients and in level of significance. 34 The lack of significance of government efficiency may come from the fact that both institutional variables are highly correlated in our sample. 35 The dummy variable that accounts for the existence of an Open Skies Agreement between partner countries was obtained from the U.S. Department of Transport. The introduction of this paper presents a table with all Open Skies Agreements and the years they were signed. 36 For all other years we have to use year 2000 unbalance 16 quality keep their negative signs but their effects on air transport costs are slightly smaller. These results are not surprising if we take into account that most of the observed variance experienced by airport infrastructure and regulatory quality occurs between developing and developed countries, which is now captured by our developed country dummy. Controlling by infrastructure and regulatory quality, on average, developed countries have 14 percent lower air transport cost than the other countries in our sample. Panel Data Results In the cross section estimations we mentioned the possibility that the estimated coefficients of Open Skies Agreements might be biased to zero. To solve this problem, Table 3 presents the results of a country-product fixed effect estimation for the period 1990­2001. The inclusion of country-product dummies allows us to focus on the time series effect of Open Skies Agreements. The country-product fixed effect captures the initial level of transport costs as well as those variables that do not change over time (for instance distance).37 As we did in the cross section estimation, we still allow the error term to be correlated within countries in a given year, and we control for US region or district fixed effects. In column (1) we report the results obtained when we allow the average air transport costs to follow a linear trend. As we found in the cross section regressions, volume has a negative sign but is not significant at standard levels. Unit weight values remain highly significant with a similar coefficient. More interestingly, in our panel setup, Open Skies Agreements remains negative and becomes significant at standard levels. This result is robust even when we control with a quadratic trend (column 2), dummies per years (column 3) and US district fixed instead of US region dummies (column 4). The result also holds when we include our measure of airport infrastructure, which has an estimated value of zero and has no statistical power. This low power may be explained by the low time variability of our airport infrastructure measure.38 37 We do not include our institutional variables because they are only available since 1996 and they almost do not change over time. In most specifications we do not include our measure of airport infrastructure because it barely changes over time. 38 Most improvements in airport infrastructure over time are within the same airport (e.g., equipment), and therefore are not captured in our measure. 17 The results in columns (1) to (5) suggest that, even though statistically significant, Open Skies Agreements imply only a small decline in air transport costs of around 2.3%. This is the average effect of open skies agreement on freight rates independent of the number of years the agreement has been standing. Open Skies Agreements may reduce freight rates over time, in which case the total effect of these agreements would be larger than 2.3%. We want to test the hypothesis that air carriers take time to adapt to the new rules in the market and need to go through an underlying "learning by doing" process. A similar hypothesis, which cannot be tested separately with the available data, would be that those firms that survive after Open Skies Agreements are signed, are more efficient and, given the existence of more contestable markets, they set lower tariffs. The process of fighting to survive is not a one-period game and that would explain why freight rates decrease over time. To test the first hypothesis, columns (6) to (8) compute the previous regressions using only the years 1990 and 2001.39 Using this "type of" first difference regression,40 the dummy for Open Sky Agreements captures the fall in freight rates for the whole period since the agreement was signed in each country. For example, Spain signed an Open Skies Agreements with the US in 1995, therefore, in the year 2001 the agreement had been standing for 6 years. In column (6) the agreement dummy implies a large drop in freight rates of around 9 percent. When we include our infrastructure measure in column (7), the fall in freight rates increases to 13 percent. In this specification our infrastructure variable is significant at 10 percent levels and it has a magnitude similar to that obtained in column (2) of Table 1. Finally, to confirm that the effect of agreements increases over time, in columns (8) and (9), besides our Open Sky dummy, we include the number of years since the agreement was signed. Reassuring our previous results, the interaction term is negative and significant at conventional levels. These results, even though they are significant at conventional levels, are estimated imprecisely. To check the results of the previous paragraph, Table 4 presents country-product fixed-effects regressions for the whole sample period (as columns 1-5 in Table 3) but allowing the effect of Open Skies Agreements to differ over time. The dummy "Year of 39 We drop countries that are not in the whole sample to avoid any composition effect (2 countries). Results hold with the whole sample. 18 Signature" captures the change in freight rates the same year the agreement was signed. The dummy "One Year After" captures the change the first year after the agreement was signed. Finally, the dummies "Four or More Years After" captures the average fall in freight rates after four or more years since the agreement was signed. Columns (1) to (4) show similar results. In the year the agreement is signed there is a 1 percent fall in freight rates. This fall increases 1 percent per year after the agreement is signed.41 Three years after the agreement was signed we observe a 3-4 percent fall in rates, which is significant at conventional levels in all the specifications. Focusing on columns (2) and (4), the long run effect of Open Sky Agreement is a fall in air transport costs of around 8 percent. It is important to highlight that this result does not allow us to identify the source of the reduction in air transport costs associated with Open Skies Agreements. It could be the case that costs are lower because a more intense competition induced a lower mark-up. An alternative explanation consistent with the reduction in transport costs would be that airlines became more efficient, and keeping mark-ups constant, were able to reduce freight rates. V. Conclusion During the 1980s and 1990s, many countries engaged in a process of reduction of tariff and non-tariff barriers to trade. As a consequence, the relevance of transport costs as a determinant of the ability of a country to integrate into the global economy increased significantly. At first glance, it could be argued that governments cannot reduce transport costs because they are, to a great extent, determined by exogenous factors, mainly distance. Even though it is true that distance is an important explanatory variable of transport costs, this paper shows that governments can implement policies to reduce transport costs and effectively help their countries "get closer" to high demand markets. This paper concentrates on air transport, the fastest-growing cargo transport mode. Relying on detailed micro data and the opportunity that Open Skies Agreements provide to evaluate the impact of changes in the competition regime, our estimations 40 These would be first difference regressions if there were only one district of entry in the US. 41 In all specifications, the sum of the first three dummies ("year of Signature," "One Year after" and "Two Years After") is 19 show that improvements in infrastructure and the quality of regulation and a more liberal air cargo market significantly reduce transport costs. In our sample, an improvement in airport infrastructure from the 25th to 75th percentile reduces air transport costs 15 percent. A similar improvement in the quality of regulation reduces air transport costs 14 percent. Besides, deregulating the air cargo market--through what are usually called Open Skies Agreements --further reduces air transport costs around 8 percent. Using previous estimates of transport cost - trade elasticity (1.3),42 an Open Sky Agreement that reduces transport costs 8 percent implies an increase in trade of around 10 percent. These results have important policy implications. Efforts aimed at improving the quality of regulation and the state of infrastructure (airports) have definite effects on the ability of local producers to compete in the global economy. Signing Open Skies Agreements has been difficult and strongly resisted, especially by airlines. This paper provides sound evidence that many economic sectors could benefit from a deregulated air cargo market. significant at conventional levels. 42 See Clark et al (2004). 20 References Button, Kenneth.2002. "Toward Truly Open Skies." Regulation. Fall 2002. Boyer, Kenneth. 1998. Principles of Transportation Economics. Addison Wesley. CIA. (2002). World Fact Book. http://www.cia.gov/cia/publications/factbook/ Clark, Ximena, Dollar, David, Micco, Alejandro. 2004. "Port Efficiency, Maritime Transport Costs and Bilateral Trade". National Bureau of Economic Research. Working Paper 10353. Forthcoming in the Journal of Development Economics. Deardorff, A. 1984. "Testing Trade Theories and Predicting Trade Flows." Handbook of International Economics, Vol 1. Amsterdam. Elsevier Science Publishers. Doganis, Rigas. 2001. The Airline Business in the Twenty-first Century. Routledge. London. Fink, Carsten, Aaditya Matoo, and Ileana Cristina Neagu. 2002. Trade in International Maritime Services: How Much does Policy Matter? The World Bank Economic Review. Vol 16, No 1: 81-108. Gillen, D.W., Oum, Tae., Tretheway, M. 1990. "Airline Cost Structure and Policy Implications." Journal of Transport Economics and Policy, 24(2): 9-34. Hummels, David, and Skiba, Alexandre. 2002. "A Virtuous Circle? Regional Tariff Liberalization and Economies of Scale in Transport." Purdue University. Mimeo Hummels, David. 1999. "Have International Transport Costs Declined?" Purdue University. Mimeo. 21 Kaufmann, D., Kraay, A., and Mastruzzi, M. 2003. "Governance Matters III: Governance Indicators for 1996-2002." World Bank Research Working Papers Series, 3106. Limão, Nuno., Venables, Anthony. 2001. "Infrastructure, Geographical Disadvantage, Transport Costs, and Trade." The World Bank Economic Review. Vol 15. No 3: 451-479 Linnemmann, H. 1966. An Econometric Study of International Trade Flows. Amsterdam: North-Holland. Micco, A., Stein E., and Ordoñez, G. 2003. "The Currency Union Effect on Trade: Early Evidence from EMU." Economic Policy. No 37, October, pp:315-43. O'Connor, William. 1985. An Introduction to Airline Economics. Praeger. New York. Overman, Henry, Redding, Stephen, Venables, Anthony. 2003. "The Economic Geography of Trade, Production and Income: A Survey of Empirics." Handbook of International Trade. Edited by Kwan Choi and James Harrigan. Blackwell Publishing. Redding, Stephen., Venables, Anthony. 2003. "Geography and Export Performance: External Market Access and Internal Supply Capacity." National Bureau of Economic Research. Working Paper 9637. Serebrisky, Tomas, Presso, Pablo. 2002. "An Incomplete Regulatory Framework? Vertical Integration in Argentine Airports". XXXVII Meeting of the Argentine Political Economy Association. Salazar de la Cruz, Francisco. 2003. Introducción a la Gestión Económica de Aeropuertos. Fundación Aena, Madrid. The Brattle Group. 2002. "The Economic Impact of an EU-US Open Aviation Area." 22 Available at www.brattle.com U.S. Department of Transportation, Office of the Secretary. 2000. "Transatlantic Deregulation. The Alliance Network Effect." International Aviation Developments. Second Report. http://ostpxweb.dot.gov/aviation/ 23 Appendix A. Data Description Air Transport Costs: Corresponds to the import charge per unit of weight by type of commodity and by foreign country. The variable was constructed with the information of import charges and weight reported by the U.S. Imports of Merchandise Database, of the U.S. Department of Commerce and U.S. Census Bureau , 1990 - 2001. Air transport costs per commodity is calculated at a 4 digit aggregated HS level. Control of Corruption: Measures perceptions of corruption, conventionally defined as the exercise of public power for private gain. Despite this straightforward focus, the particular aspect of corruption measured by the various sources differs somewhat, ranging from the frequency of "additional payments to get things done," to the effects of corruption on the business environment, to measuring "grand corruption" in the political arena or in the tendency of elite forms to engage in "state capture". The presence of corruption is often a manifestation of a lack of respect of both the corrupter (typically a private citizen or firm) and the corrupted (typically a public official or politician) for the rules which govern their interactions, and hence represents a failure of governance according to our definition. This measure is obtained from Kaufmann, Kraay and Mastruzzi (2003). Developed Country Dummy: This dummy variable takes the value of one if the country is classified by the World Bank (2002) as a High Income Country, otherwise zero. Distance: Corresponds to the distance between the foreign airport I and the U.S. customs district J. The geographic coordinates used to calculate the distance between US custom district and the foreign airport were obtained from CIA Factbook 2001. Directional Trade Unbalance: Corresponds to the ratio between the difference of U.S. exports and imports, and bilateral trade. The variable was constructed with information on imports and exports reported by the U.S. Imports of Merchandise Database, of the U.S. Department of Commerce and U.S. Census Bureau for year 2000. Foreign Airport Infrastructure Index: Corresponds to the logarithm of the ratio between the number of airports (square) with runaways of at least 1500m long (ac ) per country, and the product between country surface (surfc ) and country population (popct) . Numberof airports = ln ac 2 surfc * popct where t is year The Number of runaways per country was obtained from CIA World Fact Book, 1990 ­ 2001. Government Efficiency: Measures the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies. 24 The main focus of this index is on "inputs" required for the government to be able to produce and implement good policies and deliver public goods. This variable was obtained from Kaufmann, Kraay and Mastruzzi (2003). Number of Pavimented Roads: This variable was obtained from World Bank World Development Indicators, 2002. Open Sky Agreement Dummy Variable: This dummy variable takes the value of one if there is an Open Skies Agreement between U.S and the foreign country in that specific year. The information was obtained from the U.S. Department of Transportation, 1990 - 2001. Population: This variable was obtained from World Bank World Development Indicators, 2002. Product Unit Value: Corresponds to the total value per unit of weight of U.S. imports calculated from foreign airport to each of the U.S. customs districts. The variable was constructed with the information of import value and weight reported by the U.S. Imports of Merchandise Database, of the U.S. Department of Commerce and U.S. Census Bureau, 1990 - 2001. Product Unit Value is calculated at a 4 digit aggregated HS level. Product Unit Weight: Corresponds to the weight of U.S. imports from foreign airport to each of U.S. customs districts. The variable was obtained from U.S. Imports of Merchandise Database, of the U.S. Department of Commerce and U.S. Census Bureau , 1990 - 2001. It is aggregated at a 4 HS level. Real GDP: This variable was obtained from World Bank World Development Indicators, 2002. Real GDP per capita: This variable was obtained from World Bank World Development Indicators, 2002. Region: Corresponds to the U.S. customs districts classification done by the authors. In Table B4 of appendix B, there is a detailed description of the regions associated to each custom district. Regulatory Quality: Measures the incidence of market-unfriendly policies such as price controls or inadequate bank supervision, as well as perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development. This measure is obtained from Kaufmann, Kraay and Mastruzzi (2003). Rule of Law: This variable was obtained from Kaufmann, D., Kraay, A., and Mastruzzi, M. 2003. "Governance Matters III: Governance Indicators for 1996-2002". World Bank Research Working Papers Series, 3106. Surface: This variable was obtained from World Bank World Development Indicators. 25 Telephones: This variable was obtained from World Bank World Development Indicators, 2002. Total Liner Volume: Corresponds to the total value of imports transported between each foreign country and each U.S. customs district. The variable was obtained from U.S. Imports of Merchandise Database, of the U.S. Department of Commerce and U.S. Census Bureau, 1990 - 2001. It is aggregated at a 4 HS level. Appendix B. Data Used Table B1 Country Infrastructure Phones Roads Control of Government Rule of Regulatory Corruption Efficiency Law Quality Andorra 1.442 n.a. n.a. n.a. 1.358 1.294 1.548 Afghanistan -25.513 0.182 -10.578 -1.469 -1.344 -1.544 -2.699 Angola -1.483 n.a. n.a. n.a. -1.354 -1.119 -1.435 Albania -22.488 3.842 -5.665 -0.587 -0.534 -0.726 -0.232 United Arab Emirates -21.086 6.845 -12.232 0.702 0.584 1.096 0.690 Argentina -25.139 5.931 -7.688 -0.356 0.136 -0.011 0.284 Armenia -24.012 5.053 -6.048 -0.699 -0.525 -0.432 -0.357 Antigua and Barbuda -17.214 6.667 -6.171 0.838 0.558 1.017 0.704 Australia -26.442 6.879 -5.410 1.973 1.771 1.906 1.390 Austria -23.648 7.113 -2.820 1.837 1.660 1.998 1.411 Azerbaijan -23.111 5.070 -7.018 -1.030 -0.876 -0.847 -0.833 Burundi -25.887 1.675 -6.726 -1.057 -1.131 -0.867 -1.121 Belgium -21.264 6.931 -2.729 1.303 1.490 1.484 1.079 Benin -27.265 2.862 -9.620 -0.527 -0.245 -0.320 -0.097 Burkina Faso -26.560 1.874 -9.890 -0.389 -0.403 -0.582 -0.191 Bangladesh -27.695 1.608 -5.982 -0.649 -0.473 -0.705 -0.419 Bulgaria -21.537 6.087 -6.476 -0.360 -0.400 -0.093 0.296 Bahrain, Kingdom of -18.596 6.310 -3.803 0.448 0.563 0.860 0.863 Bahamas, The -19.060 6.173 -6.036 0.815 0.950 1.048 0.997 Bosnia & Herzegovina -23.263 4.887 -6.052 -0.486 -0.764 -0.734 -1.207 Belarus -21.770 5.612 -5.927 -0.577 -0.977 -1.036 -1.830 Belize -22.423 5.388 -6.497 -0.129 -0.344 0.284 0.047 Bermuda 1.442 n.a. n.a. n.a. 1.092 1.294 1.282 Bolivia -25.227 4.868 -8.046 -0.689 -0.366 -0.515 0.528 Brazil -28.466 5.762 -6.183 -0.006 -0.189 -0.195 0.261 Barbados -18.559 6.081 -3.575 1.294 1.358 0.767 0.688 Brunei Darussalam -21.301 6.281 -7.206 0.144 0.788 0.793 0.866 Bhutan -0.282 2.976 -7.930 n.a. 0.585 0.898 -0.417 Botswana 0.746 n.a. n.a. n.a. 0.672 0.674 0.744 Central African Rep. -27.084 1.380 -8.315 -0.852 -1.040 -0.638 -0.572 Canada -26.285 6.868 -5.854 2.208 1.890 1.888 1.322 Switzerland -22.480 7.223 -4.031 2.234 2.230 2.166 1.374 Chile -24.397 6.095 -7.489 1.364 1.223 1.271 1.337 China,P.R.: Mainland -28.167 5.180 -8.697 -0.240 0.179 -0.297 -0.196 Côte d'Ivoire -27.061 3.876 -7.603 -0.367 -0.508 -0.728 -0.136 Cameroon -25.982 2.284 -8.680 -1.082 -0.655 -1.100 -0.417 Congo, Republic of -27.661 3.440 -8.747 -0.904 -1.232 -1.196 -0.917 Colombia -26.444 5.405 -8.144 -0.471 -0.148 -0.625 0.242 Comoros -20.942 2.303 -7.382 -0.711 -0.916 -0.944 -0.796 Cape Verde -21.298 5.145 -7.292 0.068 0.045 0.378 -0.309 Costa Rica -24.608 5.708 -5.018 0.841 0.408 0.738 0.797 Cuba -22.292 3.788 -5.804 -0.186 -0.330 -0.691 -1.117 (Continued) 26 Country Infrastructure Phones Roads Control of Government Rule of Regulatory Corruption Efficiency Law Quality Cayman Islands 1.442 n.a. n.a. n.a. 1.892 1.294 1.548 Cyprus -18.777 6.875 -4.032 1.206 1.118 0.825 1.013 Czech Republic -22.430 6.687 -5.555 0.415 0.681 0.641 0.885 Germany 1.358 n.a. n.a. n.a. 1.753 1.846 1.836 Djibouti -22.021 2.745 -7.470 -0.840 -0.985 -0.443 -0.500 Dominica -0.042 n.a. n.a. n.a. -0.534 0.002 0.008 Denmark -21.751 7.208 -3.788 2.320 1.909 1.964 1.473 Dominican Republic 0.161 n.a. n.a. n.a. -0.398 -0.387 -0.290 Algeria -24.746 4.091 -8.809 -0.590 -0.746 -0.677 -0.799 Ecuador -24.992 4.928 -7.537 -0.867 -0.854 -0.573 -0.175 Egypt -23.961 4.680 -9.652 -0.154 -0.081 0.178 -0.092 Eritrea -22.857 2.067 -10.156 0.150 -0.125 -0.214 -0.447 Spain 1.223 n.a. n.a. n.a. 1.665 1.357 1.260 Estonia -23.395 6.620 -3.086 0.488 0.670 0.601 1.222 Ethiopia -26.014 1.356 -11.075 -0.401 -0.554 -0.347 -0.670 Finland 1.618 n.a. n.a. n.a. 1.870 2.390 2.038 Fiji -23.420 5.157 -7.134 0.279 -0.070 -0.345 -0.501 France -23.541 6.978 -3.702 1.490 1.541 1.457 0.991 Micronesia, Fed. Sts. -0.728 n.a. n.a. n.a. -0.391 -0.319 -0.547 Gabon -25.096 4.864 -8.395 -0.830 -0.661 -0.397 -0.182 United Kingdom -22.775 7.182 -4.645 2.060 2.056 1.906 1.635 Georgia -21.977 5.150 -6.739 -0.846 -0.545 -0.815 -0.735 Ghana -29.111 2.898 -7.948 -0.424 -0.050 -0.120 -0.035 Guinea -26.845 2.575 -7.580 -0.367 -0.527 -0.870 -0.177 Gambia, The -23.291 3.394 -7.489 -0.271 -0.205 -0.200 -0.536 Guinea-Bissau -24.241 2.219 -7.463 -0.625 -1.005 -1.276 -0.852 Equatorial Guinea -23.274 2.595 -7.343 -1.482 -1.565 -1.383 -1.496 Greece -21.757 6.993 -4.600 0.642 0.731 0.737 0.876 Grenada -17.322 5.934 -3.428 0.279 -0.084 0.328 0.178 Guatemala -25.069 4.772 -8.731 -0.744 -0.447 -0.728 0.294 Guyana 0.008 4.834 -7.766 n.a. -0.212 -0.361 -0.120 China,P.R.:Hong Kong 1.662 n.a. n.a. n.a. 1.445 1.506 1.576 Honduras -25.103 4.248 -8.264 -0.767 -0.589 -0.776 0.056 Croatia -22.065 5.442 -5.736 -0.132 0.105 -0.071 0.179 Haiti -26.114 2.186 -9.447 -1.123 -1.352 -1.352 -1.050 Hungary -22.424 6.513 -3.263 0.658 0.708 0.790 0.981 Indonesia -0.206 n.a. n.a. n.a. -0.390 -0.920 -0.753 India -0.178 n.a. n.a. n.a. -0.117 -0.229 0.126 Ireland -24.903 6.982 -3.419 1.772 1.714 1.766 1.542 Iran, I.R. of -1.394 n.a. n.a. n.a. -0.310 -0.599 -0.558 (Continued) 27 Country Infrastructure Phones Roads Control of Government Rule of Regulatory Corruption Efficiency Law Quality Iceland -24.062 7.148 -5.122 2.159 1.813 1.898 1.039 Israel -20.974 7.076 -6.184 1.278 1.009 1.062 0.930 Italy -23.187 7.099 -4.301 0.779 0.863 0.919 0.855 Jamaica -22.687 5.832 -4.401 -0.307 -0.323 -0.240 0.441 Jordan -21.668 5.019 -9.022 0.059 0.375 0.422 0.340 Japan 0.756 n.a. n.a. n.a. 1.099 1.261 1.617 Kazakhstan -27.432 4.830 -8.711 -0.894 -0.700 -0.788 -0.461 Kenya -26.888 2.678 -8.340 -1.007 -0.737 -0.953 -0.317 Kyrgyz Republic -24.799 4.367 -7.921 -0.781 -0.562 -0.766 -0.455 Cambodia -26.997 2.507 -9.545 -0.903 -0.642 -0.797 -0.227 Kiribati -0.841 3.808 -4.994 n.a. -0.194 -0.531 -0.426 St. Kitts and Nevis 0.177 6.397 -4.971 n.a. -0.052 0.175 0.134 Korea -23.180 6.938 -6.425 0.368 0.606 0.779 0.542 Kuwait -21.516 6.200 -7.488 0.901 0.142 0.943 0.048 Lao People's Dem.Rep -1.168 n.a. n.a. n.a. -0.482 -0.933 -1.108 Lebanon 0.117 n.a. n.a. n.a. -0.132 -0.339 -0.114 Liberia -26.432 0.742 -7.895 -1.301 -1.655 -1.692 -2.010 St. Lucia -18.371 5.797 -4.174 0.303 0.074 0.134 0.177 Liechtenstein 1.688 n.a. n.a. n.a. 1.625 1.294 1.548 Sri Lanka 0.386 n.a. n.a. n.a. -0.273 -0.159 0.057 Lesotho -0.319 n.a. n.a. n.a. -0.058 -0.029 -0.129 Lithuania -22.618 6.137 -3.745 0.118 0.303 0.199 0.493 Luxembourg 1.552 n.a. n.a. n.a. 2.067 1.973 1.923 Latvia -21.556 6.150 -3.313 -0.131 0.297 0.244 0.628 Macao, China 0.704 n.a. n.a. n.a. 0.825 -0.074 0.752 Morocco -24.636 4.889 -8.255 0.108 0.083 0.289 0.156 Moldova -0.409 n.a. n.a. n.a. -0.673 -0.615 -0.335 Madagascar -0.232 n.a. n.a. n.a. -0.469 -0.278 -0.662 Maldives 0.325 n.a. n.a. n.a. 0.538 -0.371 -0.293 Mexico 0.600 n.a. n.a. n.a. 0.146 -0.340 -0.267 Marshall Islands -0.637 n.a. n.a. n.a. -0.507 -0.133 -0.438 Macedonia, FYR -23.281 5.743 -6.529 -0.607 -0.375 -0.395 -0.082 Mali -26.994 1.465 -10.968 -0.437 -0.680 -0.633 -0.043 Malta -18.642 6.702 -3.201 0.497 0.984 0.640 0.558 Myanmar -1.377 1.755 -10.584 n.a. -1.287 -1.252 -1.268 Mongolia -23.538 4.614 -7.345 -0.121 -0.079 0.265 -0.085 Mozambique -27.042 1.931 -9.616 -0.671 -0.362 -0.892 -0.493 Mauritania -25.417 2.282 -10.749 -0.254 -0.188 -0.466 -0.315 Martinique 0.950 n.a. n.a. n.a. 0.825 0.838 1.282 Mauritius -21.602 5.956 -6.476 0.426 0.561 0.846 0.420 Malawi -0.167 n.a. n.a. n.a. -0.648 -0.639 -0.386 Malaysia 0.547 n.a. n.a. n.a. 0.794 0.447 0.686 (Continued) 28 Country Infrastructure Phones Roads Control of Government Rule of Regulatory Corruption Efficiency Law Quality Namibia -24.417 4.695 -5.791 0.582 0.334 0.784 0.340 Niger -28.864 0.689 -11.809 -0.796 -0.877 -0.874 -0.548 Nigeria -26.715 1.518 -8.026 -1.149 -1.162 -1.213 -0.743 Nicaragua -25.760 3.892 -7.438 -0.546 -0.662 -0.750 0.027 Netherlands 1.688 n.a. n.a. n.a. 2.165 2.242 1.916 Norway -22.822 7.157 -5.104 2.082 1.814 2.042 1.248 Nepal -0.345 n.a. n.a. n.a. -0.621 -0.392 -0.350 New Zealand -25.460 6.969 -4.797 2.313 1.820 2.011 1.594 Oman -21.826 5.034 -6.159 0.689 0.877 1.104 0.578 Pakistan -0.470 n.a. n.a. n.a. -0.514 -0.773 -0.611 Panama 0.801 n.a. n.a. n.a. -0.185 -0.338 0.051 Peru -24.531 4.713 -8.729 -0.145 -0.179 -0.434 0.554 Philippines -26.352 4.823 -6.314 -0.433 0.086 -0.287 0.373 Papua New Guinea -28.474 2.595 -8.707 -0.738 -0.633 -0.446 -0.558 Poland -23.044 6.123 -4.483 0.433 0.581 0.574 0.612 Korea, Dem. Rep. -22.528 3.822 -7.921 -0.743 -1.041 -1.068 -1.874 Portugal -22.413 6.999 -5.267 1.360 1.117 1.261 1.230 Paraguay -25.639 5.006 -7.828 -0.914 -1.063 -0.800 -0.288 Qatar -21.199 6.152 -8.355 0.598 0.721 1.060 0.307 Romania -24.144 5.658 -4.875 -0.342 -0.519 -0.211 -0.091 Russia -23.864 5.482 -9.068 -0.831 -0.517 -0.804 -0.657 Rwanda -0.840 n.a. n.a. n.a. -0.720 -0.340 -0.807 Saudi Arabia -0.039 n.a. n.a. n.a. -0.106 0.174 0.670 Sudan -27.539 2.576 -13.165 -0.994 -1.385 -1.295 -1.148 Senegal -28.238 3.869 -9.064 -0.344 -0.042 -0.227 -0.267 Singapore -19.423 7.063 -5.563 2.340 2.341 2.030 1.939 Solomon Islands -1.101 3.002 -8.819 n.a. -0.965 -0.543 -0.772 Sierra Leone -26.610 1.850 -7.940 -0.962 -0.966 -0.926 -1.060 El Salvador -25.591 5.385 -7.165 -0.436 -0.263 -0.390 0.794 Somalia -25.178 0.405 -9.330 -1.450 -2.138 -1.791 -2.377 São Tomé & Principe -0.514 3.431 -7.235 n.a. -0.754 -0.293 -0.696 Suriname -24.898 5.591 -8.078 0.036 -0.126 -0.611 -0.772 Slovak Republic -23.064 6.253 -4.958 0.206 0.235 0.240 0.397 Slovenia -23.026 6.906 -4.588 0.942 0.679 0.844 0.645 Sweden -23.648 7.244 -4.393 2.328 1.789 1.943 1.357 Swaziland -0.062 n.a. n.a. n.a. -0.418 -0.222 -0.143 Seychelles -0.960 n.a. n.a. n.a. -0.606 0.131 -0.041 Syrian Arab Republic -0.955 n.a. n.a. n.a. -0.796 -0.568 -0.371 Chad -27.129 -0.330 -9.069 -0.820 -0.495 -0.725 -0.527 Togo -24.843 2.996 -8.379 -0.677 -0.954 -0.863 -0.427 Thailand -25.399 4.960 -8.914 -0.227 0.228 0.398 0.417 (Continued) 29 Country Infrastructure Phones Roads Control of Government Regulatory Corruption Efficiency Rule of Law Quality Tajikistan -24.270 3.578 -7.026 -1.214 -1.310 -1.319 -1.555 Turkmenistan -2.234 n.a. n.a. n.a. -1.384 -1.205 -1.152 Tonga -18.094 4.591 -5.050 -0.319 -0.493 -0.653 -0.689 Trinidad and Tobago -21.235 5.811 -4.569 0.188 0.415 0.385 0.650 Tunisia 0.282 n.a. n.a. n.a. 0.803 0.279 0.314 Turkey -24.026 6.264 -5.821 -0.154 -0.162 0.070 0.391 Taiwan Prov.of China 1.019 n.a. n.a. n.a. 1.222 0.783 0.989 Tanzania -28.252 2.306 -8.250 -0.980 -0.567 -0.416 -0.194 Uganda -26.335 2.415 -8.700 -0.731 -0.239 -0.602 0.156 Ukraine -22.149 2.786 -6.906 -0.881 -0.778 -0.725 -0.819 Uruguay -27.093 6.017 -8.887 0.589 0.594 0.553 0.830 Uzbekistan -27.761 4.238 -7.339 -0.928 -1.029 -1.020 -1.547 St. Vincent & Grens. 0.129 5.482 -3.725 n.a. -0.095 0.175 0.243 Venezuela, Rep. Bol. -25.275 5.785 -7.743 -0.745 -0.886 -0.776 -0.267 Vietnam -26.076 3.730 -7.985 -0.662 -0.236 -0.602 -0.606 Vanuatu -21.599 3.571 -7.649 -0.319 -0.451 -0.426 -0.379 West Bank and Gaza -0.178 n.a. n.a. n.a. -0.183 0.089 0.472 Samoa -19.992 n.a. 3.857 n.a. -6.648 n.a. n.a. Yemen, Republic of -0.525 n.a. n.a. n.a. -0.632 -0.549 -0.946 Serbia and Montenegro -24.739 5.854 -6.999 -0.924 -0.829 -0.993 -1.108 South Africa -26.457 5.716 -5.988 0.466 0.315 0.259 0.313 Congo, Dem. Rep. of -28.488 -0.527 -8.452 -1.563 -1.772 -1.820 -2.387 Zambia -26.427 2.856 -7.428 -0.825 -0.729 -0.414 0.083 Zimbabwe -25.998 3.721 -9.584 -0.586 -0.833 -0.570 -1.148 Source: Infrastructure, Telephones and Phones are the indexes calculated by the authors. Control of Corruption, Government Efficiency, Rule of Law and Regulatory Quality were obtained from Kauffmann, D.(2003). The information used to construct the airport infrastructure index was obtained from CIA World Fact Book, 1990 - 2001, and the World Bank World Development Indicators (2002). The indexes of Roads and Telephones were contructed using the information of phones, roads, surface and population available in the World Development Indicators (2002), World Bank. The first three variables are in logarithms. All the indexes reported are only for year 2000. 30 Tables B2 and B3 present the summary statistics for the variables used for year 2000 and for the whole sample, respectively. Table B2. Summary Statistics Year 2000 Variable: Obs. Mean SD Pct.5 Pct. 95 Log Value of Air Transport Costs per Unit of Weight 116620 0.864 1.298 -1.296 2.736 Total Volume (ln) 116620 21.857 1.865 17.939 23.787 Product Unit Value (ln) 116620 3.765 1.602 1.396 6.604 Unbalance 116620 -0.048 0.350 -0.727 0.548 Distance (ln) 116620 8.916 0.552 7.951 9.578 Open Sky Agreement 116620 0.560 0.496 0.000 1.000 Airport Inf. 116620 -23.937 2.091 -28.167 -20.974 Regulatory Quality 116620 0.896 0.624 -0.196 1.635 Gov. Effectiveness 116620 1.084 0.880 -0.473 2.230 Developed Countries 116620 0.634 0.482 0.000 1.000 Table B3. Summary Statistics Whole Sample (1990 - 2001) Variable: Obs. Mean SD Pct.5 Pct. 95 Log Value of Air transport costs per unit of weight 1830170 0.980 1.226 -0.997 2.788 Total Volume (ln) 1830170 17.418 1.660 14.179 19.296 Product Unit Value (ln) 1830170 3.739 1.557 1.426 6.470 Unbalance 1829563 -0.078 0.331 -0.727 0.509 Distance (ln) 1830170 8.945 0.545 7.879 9.579 Open Sky Agreement 1830170 0.292 0.455 0.000 1.000 Airport Inf. 1605448 -23.926 2.032 -28.060 -21.239 Regulatory Quality 1826134 0.936 0.622 -0.196 1.662 Gov. Effectiveness 1826134 1.133 0.855 -0.447 2.230 Developed Countries 1828762 0.659 0.474 0.000 1.000 31 Table B4. Specifies the regions associated with each U.S. custom district. Table A3. U.S. Regions District of Entry Region District Code Baltimore M.D. 1 13 Boston Mass. 1 4 Buffalo N.Y. 1 9 Chicago ILL. 1 39 Cleveland Ohio 1 41 Detroit Mich. 1 38 New York City N.Y. 1 10 Norfolk Va. 1 14 Ogdensburg N.Y. 1 7 Philadelphia Pa. 1 11 Portland Maine 1 1 Providence R.I. 1 5 St. Albans Vt. 1 2 St. Louis Mo. 1 45 Washington D.C. 1 54 Charleston S.C. 2 16 El Paso 2 24 Houston Tex. 2 53 Laredo Tex. 2 23 Miami Fla. 2 52 Milwaukee Wis. 2 37 Mobile Ala. 2 19 New Orleans La. 2 20 Port Arthur Tex. 2 21 Savannah Ga. 2 17 Tampa Fla. 2 18 Wilmington N.C. 2 15 Worth, Texas 2 55 Columbia-Snake 3 29 Duluth Minn. 3 36 Great Falls 3 33 Los Angeles 3 27 Minneapolis Minn. 3 35 Nogales Ariz. 3 26 Pembina N. Dak. 3 34 San Diego Calif. 3 25 San Francisco Calif. 3 28 Seattle Wash. 3 30 Source: U.S. Imports of Merchandise Database. U.S. Department of Commerce. 32 Table 1: Determinants of Air Transport Costs Dependent Variable: Log Value of Air Transport Costs per Unit of Weight Variables: (1) (2) (3) (4) (5) (6) (7) (8) Distance (ln) 0.197 0.208 0.196 0.191 0.184 0.182 0.199 0.185 (0.035)*** (0.033)*** (0.038)*** (0.037)*** (0.034)*** (0.032)*** (0.033)*** (0.035)*** Total Volume (ln) -0.031 -0.032 -0.032 -0.016 -0.018 -0.025 -0.029 -0.017 (0.011)*** (0.012)*** (0.011)*** (0.015) (0.013) (0.014)* (0.011)*** (0.012) Product Unit Value 0.485 0.491 0.483 0.488 0.489 0.488 0.485 0.489 (0.013)*** (0.013)*** (0.014)*** (0.013)*** (0.013)*** (0.013)*** (0.013)*** (0.013)*** Unbalance -0.156 -0.138 -0.160 -0.108 -0.098 -0.110 -0.164 -0.102 (0.059)*** (0.053)*** (0.064)** (0.056)* (0.042)** (0.043)** (0.057)*** (0.045)** Airport Inf. -0.041 -0.066 -0.030 -0.025 -0.025 -0.037 -0.024 (0.011)*** (0.014)*** (0.012)** (0.009)*** (0.009)*** (0.011)*** (0.009)*** Airport Inf. (Lagged 5 years) -0.037 (0.012)*** Gov. Effectiveness -0.066 0.065 (0.029)** (0.061) Regulatory Quality -0.121 -0.193 -0.118 (0.034)*** (0.075)** (0.037)*** Open Sky Agreement. -0.048 -0.017 (0.042) (0.038) Observations 116620 116620 116620 116620 116620 116620 116620 116620 R-squared 0.339 0.340 0.338 0.340 0.341 0.341 0.339 0.341 IV Tel & Road F Test, Reg. Qual. & Gov. Eff. = 0 7.078 Prob > F 0.001 Clustered Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% 33 Table 2: Determinants of Air Transport Costs Dependent Variable: Log Value of Air Transport Costs per Unit of Weight Variables: (1) (2) (3) (4) (5) (6) (7) Distance (ln) 0.181 0.191 0.179 0.179 0.181 0.179 0.177 (0.034)*** (0.036)*** (0.035)*** (0.034)*** (0.034)*** (0.035)*** (0.036)*** Total Volume (ln) -0.008 -0.014 -0.007 -0.009 -0.008 -0.009 -0.007 (0.011) (0.010) (0.011) (0.011) (0.011) (0.010) (0.011) Product Unit Value 0.490 0.491 0.489 0.490 0.490 0.490 0.490 (0.013)*** (0.013)*** (0.013)*** (0.013)*** (0.013)*** (0.013)*** (0.013)*** Unbalance -0.095 -0.102 -0.092 -0.082 -0.097 -0.083 -0.079 (0.049)* (0.053)* (0.051)* (0.044)* (0.052)* (0.048)* (0.045)* Airport Inf. -0.024 -0.042 -0.020 -0.024 -0.020 (0.010)** (0.016)*** (0.009)** (0.010)** (0.009)** Airport Inf. (Lagged 5 years) -0.020 -0.016 (0.010)** (0.009)* Regulatory Quality -0.076 -0.076 -0.080 (0.043)* (0.043)* (0.044)* Open Sky Agreement -0.005 -0.001 (0.036) (0.034) Dummy Developed Country. -0.176 -0.128 -0.191 -0.109 -0.175 -0.109 -0.120 (0.035)*** (0.046)*** (0.034)*** (0.051)** (0.036)*** (0.051)** (0.051)** Observations 116620 116620 116620 116620 116620 116620 116620 R-squared 0.341 0.341 0.341 0.342 0.341 0.342 0.341 Instrumental Variables Tel & Road Clustered Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% 34 Table 3: Determinants of Air Transport Costs Dependent Variable: Log value of air transport costs per unit of weight Variables: (1) (2) (3) (4) (5) (6) (7) (8) (9) Total Volume(ln) -0.014 -0.023 -0.005 -0.010 -0.062 0.068 -0.076 0.043 -0.097 (0.045) (0.042) (0.044) (0.044) (0.034)* (0.073) (0.048) (0.065) (0.052)* Product Unit Value 0.501 0.501 0.501 0.503 0.505 0.502 0.512 0.502 0.512 (0.003)*** (0.003)*** (0.003)*** (0.003)*** (0.003)*** (0.009)*** (0.010)*** (0.009)*** (0.010)*** Open Sky Agreement -0.027 -0.024 -0.019 -0.018 -0.027 -0.085 -0.132 -0.022 -0.070 (0.013)** (0.012)** (0.010)* (0.010)* (0.011)** (0.037)** (0.042)*** (0.044) (0.048) Open Sky Agreement -0.016 -0.018 *Years since signed1 (0.009)* (0.007)** Airport Inf. 0.003 -0.067 -0.073 (0.011) (0.037)* (0.037)* Year -0.022 0.335 -0.022 -0.004 -0.020 -0.002 (0.006)*** (0.064)*** (0.009)*** (0.007) (0.008)** (0.008) year (square) -0.002 (0.000)*** Observations 1747483 1747483 1747483 1747483 1601247 276858 221903 276858 221903 R-squared 0.422 0.423 0.423 0.433 0.430 0.465 0.487 0.465 0.487 Sample 1990 - 2001 1990 & 20012 Region Fixed Effect Yes Yes Yes Yes Yes Yes Yes Yes District of entry Fixed Effect Yes Year Fixed Effect Yes Yes Yes Clustered Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% 1: Open Sky Dummy interacted with the number of years since the agreement was signed 2: Columns 6 to 9 include data, only for 1990 and 2001 for countries that are in the whole period. 35 Table 4: Determinants of Air Transport Costs Dependent Variable: Log Value of Air transport costs per unit of weight Variables: (1) (2) (3) (4) Total Volume(ln) -0.056 -0.057 -0.084 -0.090 (0.037) (0.037) (0.033)** (0.033)*** Product Unit Value 0.500 0.500 0.505 0.505 (0.003)*** (0.003)*** (0.003)*** (0.003)*** Airport Inf. 0.003 0.002 (0.011) (0.011) Open Sky Agreement -0.009 -0.009 -0.016 -0.017 *Year signed2 (0.009) (0.009) (0.010) (0.010)* One Year After -0.018 -0.019 -0.028 -0.029 (0.013) (0.013) (0.015)* (0.015)* Two Years After -0.029 -0.031 -0.040 -0.042 (0.019) (0.019) (0.023)* (0.023)* Three Years After -0.043 -0.044 -0.053 -0.055 (0.019)** (0.019)** (0.022)** (0.022)** Four Years After -0.058 -0.037 (0.023)** (0.018)** Four of More Years After -0.074 -0.061 (0.020)*** (0.017)*** Five or More Years After -0.088 -0.081 (0.025)*** (0.019)*** Observations 1695944 1695944 1554189 1554189 R-squared 0.420 0.420 0.427 0.427 Sample 1990 - 2001 Region Region Region Region Fixed Effect & & & & Year Year Year Year Clustered Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% 36