1 / 37

Estimating the Gains from Liberalizing Services Trade: The Case of Passenger Aviation

Estimating the Gains from Liberalizing Services Trade: The Case of Passenger Aviation. Anca Cristea University of Oregon and David Hummels Purdue University, NBER September 2011. Services trade is important. Services represent Large and growing share of employment

lavi
Download Presentation

Estimating the Gains from Liberalizing Services Trade: The Case of Passenger Aviation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Estimating the Gains from Liberalizing Services Trade: The Case of Passenger Aviation Anca Cristea University of Oregon and David Hummels Purdue University, NBER September 2011

  2. Services trade is important • Services represent • Large and growing share of employment • 20% of international transactions (by value) • Central priority in recent trade negotiations • A key input into production of manufacturing and other services • We know very little about how services trade is affected by regulation, or by efforts at liberalization. • Few papers, mostly aggregate gravity equations exploiting cross-country variation in indices of regulation.

  3. Data Constraints • Service flows are poorly measured • No detailed classification of services trade • transport, travel, communications, construction, insurance, finance, royalties, other business services • Values, but not quantities or prices • Flows are not always observed especially if flows are intra-firm. • Policy change is difficult to quantify • Cross country comparisons are difficult because the underlying services environment varies widely • Liberalization often targets domestic regulations • Existing rules in force are complex. What is changing? • Overlaps with other, often sweeping, reforms • Contrast • Cutting tariff on Mexican steel ball bearings from 15 to 5% • guaranteeing “market access in business services”

  4. We focus on passenger aviation. Why? • International passenger aviation is important • Big (US + EU = $190bn) • A key input into • merchandise trade (Poole 2009, Cristea 2011) • knowledge flows (Hovhannisyan and Keller 2011), • Other services (GATS mode 2, mode 4) • The data are a thing of beauty.

  5. Data on aviation and policy change • We have a nice policy experiment: • From 1992-2007, the US signs 87 bilateral “Open Skies Agreements” that liberalize trade in passenger aviation. • We have detailed firm-level transactions data on US passenger aviation, 1993-2008. • The service provided is precisely defined. • E.g. a coach ticket from Chicago to Copenhagen • We observe quantities and prices • We observe entry/exit of firms into markets and the provision of “new goods”

  6. Air Passenger Traffic Nonstop Routes: grow from 870 to 1444 True OD Routes: grow from 28k to 40k

  7. Trends in Airfares Dotted line: US BTS price index (Fisher) - exact match of ticket Solid line: all DB1B data; estimate period dummies including “true” origin-dest FE

  8. What we do… • Measure growth in air traffic: O-D route x carrier • Decompose traffic growth into changes in existing routes (intensive margin) and addition of new routes (extensive margin) for each country • Measure net entry/exit of carriers on particular routes • Estimate differential growth rates pre- and post-liberalization for OSA countries • Identified relative to non-OSA countries in same period so we can control for year-specific shocks

  9. What we do (cont) • Estimate system of price, quantity equations for “true” origin-destinations • Hold fixed detailed route characteristics, cost shocks • Identify OSA effect on price, quantity with diff-in-diff • Also cut on basis of route characteristics (net entry/exit) • Calculate a quality- and variety-adjusted price index • Using routes as varieties or route x carriers as varieties • This is a much better measure of “new goods” than is typical of most calculations based on merchandise trade

  10. Past Regulatory Regime • Chicago Convention (1944): failed attempt to set multilateral agreements on air services • Countries negotiated air service agreements (ASA) on a bilateral basis. These are typically characterized by • Market access restrictions: pre-defined points of origin and destination • Limits on entry: fixed number of designated airlines • Price control: double-approval required for all airfares • Capacity: limited number of flights per period of time

  11. Eg: US-China Aviation Treaty 1980 • Only 2 carriers per country can offer service • Flights allowed only between • LA, SF, NY, Honolulu • Beijing, Shanghai • Tokyo is only 3rd country city from which airlines can operate in serving market • Carriers can offer two flights per week for a given route • Price changes must be submitted to DC, Beijing for approval two months in advance.

  12. Liberalization in Passenger Aviation Policy changes instituted by the bilateral Open Skies Agreements (OSA) 1) Remove the restrictions imposed by previous agreements • Drop restrictions on entry • Any carrier can enter any route with unrestricted capacity • No price controls 2) Grant new benefits • Allow inter-airline cooperation agreements • (e.g., alliances, code-sharing) • Extensive “beyond” market rights • not just gateway to gateway (e.g. Chicago to Copenhagen; can also offer Chicago to Copenhagen to Rome)

  13. Gateway to gateway Chicago Copenhagen beyond Indianapolis Rome

  14. Timing of open skies agreements Agreements are signed sequentially but with no obvious order. (e.g. Europe spread throughout sample.) See Table A1 Demark Sweden Portugal UK, Spain Italy France Germany

  15. Data: International Passenger Aviation • Traffic databy route (city-pair) x carrier: T100 International Segment data • Firmlevel data: all air traffic for domestic and foreign air carriers • Route Coverage: all non-stop flight segments crossing the US border • data on number of passengers, departures operated, available seats • Doesn’t track connecting flights or have price data • My ticket: Copenhagen to Chicago to Indy… last leg omitted. • Price and quantity data : Origin-Destination Passenger Survey • Transaction level data: 10% sample of international airline tickets • air fare paid • service characteristics (dist, # segments, transit airports, class) • actual origin & destination of the itinerary and carrier(s) • All of my CPH -> ORD -> IND ticket surveyed. • Note: a typical ticket involves joint production of several carriers • Does not cover “non-immunized carriers”

  16. T100 data : only Chicago <-> Copenhagen O-D Ticket data: all segments of flights with US, and beyond markets abroad Gateway to gateway Chicago Copenhagen beyond Indianapolis Rome

  17. Estimate the impact of OSA on traffic Difference-in-difference estimation method for the number of U.S. passengers abroad: Z is growth (relative to 1993) in a measure of passenger traffic Year effects: absorb common cost shocks, trend growth in air travel; Country variation • Country x quarter FE: allow differences in traffic for country j – season q • Income, population growth: absorb differential change in traffic demand for country j OSA = 1 for any year that agreement is in effect Alternative Specification interact OSA dummy with vector D(-3) to D(+5) for the age of the OSA agreement allows us to identify pre-existing trends, lagged effects of signing Index j = country, q = qtr t = year

  18. Decompose changes in traffic Write: Intensive margin = air traffic on existing city-pair routes (continuing service) Extensive margin = flight service on routes never offered before 1. simple counts of routes 2. Passenger weighted counts of routes (in manner of Feenstra 1994) based on t-3 weights. 3. Could also count carriers as distinct “varieties” (UA CPH -> IND is a different service than SAS from CPH-> IND) Replace Z in estimating equation with components above Recall: pre-existing bilateral ASAs specifically restrict entry to particular routes, carriers

  19. Total traffic Growth in New Routes Traffic on Existing Routes Extensive margin is much larger when using simple counts, much smaller if we use route x carriers

  20. Entry and exit Use T100 data to examine the distribution of entry/exit across routes. Carriers enter routes with sparse competition, exit routes with many firms competing

  21. Understanding the channels • OSA could raise or lower prices • Reduced unit costs from rationalized operations, economies of route density; and lower markups generated by net entry v. • Consolidation creates collusion, higher markups • OSA could relax capacity constraints • OSA could raise or lower service quality • Better flight frequency, connectivity, use of preferred carriers • Reduced incentive for capacity constrained firms to compete by overinvesting in quality

  22. Estimating price equation • Use O-D ticket data to estimate changes in prices for a given “true” origin-destination route r. • Starting from about 40 million tickets: Aggregate all tickets within a given route r at time t • We might have 10 different ways to get from Indy to CPH, on 4 different carriers • We create a (passenger weighted) average price for route r, from country j, time t. • Cost shocks • Control for route FE, ticket characteristics (distance, number of segments) – rjt varying • Economies of route density (population & number of possible destinations reached by each airport) • Include time FE (costs common to all routes in a time period) • Route-time varying cost shocks (fuel*dist, insurance*geographic region) • Include OSA, and OSA connect dummy • OSA: direct effect on traffic originating/terminating in OSA ctry • OSA connect: indirect effects for traffic connecting through an OSA country but originating or terminating elsewhere (note: “beyond rights”

  23. All possible routings to get from CPH to IND are aggregated for a given year, but we keep track of average characteristics (distance, number of segments) Gateway to gateway Chicago Copenhagen beyond Indianapolis Rome DCA

  24. Cost shifters: ATA data Fuel ranges from 9-27% of total cost in this period Route x Time varying Time varying

  25. Price Regressions: (DB1B) Control Variables: Cost shifters: Ticket Distance Fuel*Distance Aircraft Insurance*World Region Trip characteristics: One-way Avg. Number of Connections Outbound Traffic Density: US state Population Foreign Country Population Total Departures at Origin Total Departures at Destination Total Direct Routes (country) Other: Partial Liberalization

  26. Entry and exit Use T100 data to examine the distribution of entry/exit across routes. Carriers enter routes with sparse competition, exit routes with many firms competing

  27. Price effects by entry/exit(outbound) Sample: only tickets that match gateway-gateway routes; outbound flows only

  28. Estimating Quantity equation • Use O-D ticket data to estimate changes in demand on a given “true” origin-destination pair “r”. • More general than T-100, has all segments and all destinations (not just gateways); can control for prices • Include all tickets with same origin-destination • Prices instrumented with fuel*distance, insurance costs*region interactions • Demand shifters: • Population, income; bilateral trade; number of segments • OSA variable measures increase in traffic conditional on prices, other demand shifters.

  29. Quantity Regressions (DB1B) Instruments for Airfare: Ticket Distance Fuel*Distance Insurance*World Region Control Variables: Trip characteristics: Direct (non-stop) Avg. Number of Connections Outbound Market size: US State Population US State Income Foreign Country Population Foreign Country Income Total Exports Total Direct Routes (country) Other: Partial Liberalization Caribbean Trend

  30. Quantity Regressions (DB1B) Instruments for Airfare: Ticket Distance Fuel*Distance Insurance*World Region Control Variables: Trip characteristics: Direct (non-stop) Avg. Number of Connections Outbound Market size: US State Population US State Income Foreign Country Population Foreign Country Income Total Exports Total Direct Routes (country) Other: Partial Liberalization Caribbean Trend

  31. Is D(qty) due to relaxed capacity constraint? • Check load factors (passengers/seats) pre/post OSA. • Pre-OSA Load factor never exceeds 85%. Median 63.6% A regression of load factor on OSA has elasticity = 0.026. Suppose carriers enter until P = AC. AC = cost per plane / passengers If cost per plane changes little as we add passengers, then AC is just the inverse of the load factor

  32. D(qty) as a quality effect • OSA’s improve flight frequency, connectivity, use of preferred carriers; may also induce competition through better amenities on planes. • To extract this… This attributes none of other sources of quality change (e.g. reducing number of flight segments) to the OSA

  33. Number of passengers by entry/exit (outbound flows)

  34. Summarizing to this point: OSAs • Lower prices • The effect is stronger for inbound routes, and for routes with net entry • Raise quantities • Likely not a capacity effect, more like quality (higher quantity conditional on price and observable characteristics) • The effect is stronger for outbound routes, for routes with net entry, and over longer horizons • Raise load factors (capacity utilization) • The magnitude is very close to matching the inverse change in price • To combine these into a net effect on consumer welfare we can calculate a quality and variety-adjusted relative price index: OSA countries, OSA connect countries, everyone else.

  35. Welfare calculation To measure OSA relative to non-OSA, we capture 1. Relative price movements from OSA price regression 2. Construct quality adjusted prices by netting off the effects of OSA on “quality” (measured as OSA effect on quantity net of prices from quantity regressions.) • Use quality adjusted prices to form relative price series; then apply Feenstra 1994 to get variety adjusted price index Variety adjusted price index Price index for common set Variety adjustment

  36. Applying this to the policy experiment Sigma estimated using either variation across routes; or variation within routes across carriers

  37. Implications • Think about manufacturing operations outside of major hub cities • Aviation is a critical input into moving managers, engineers, parts & components • Estimates suggest regulatory intervention is a critical barrier to services trade • Reducing regulation affects entry, prices, and the capillarity of the aviation system

More Related