The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
clevelandfed.org/research/workpaper/index.cfm Working Paper 92 18 THE DETERMINANTS OF AIRPORT HUB LOCATIONS, SERVICE, AND COMPETITION v by Neil Bania, Paul W. Bauer, and Thomas J. Zlatoper Neil Bania is the research director of the Mandel Center for Non-Profit Organizations at Case Western Reserve University, Cleveland; Paul W. Bauer is an econo~nistat the Federal Reserve Bank of Cleveland; and Thomas J. Zlatoper is an associate professor of economics at John Carroll University, University Heights, Ohio, and a research associate at the Center for Regional Economic Issues at Case Western Reserve University. Working papers of the Federal Reserve Bank of Cleveland are preliminary materials circulated to stimulate discussion and critical comment. The views stated herein are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. December 1992 clevelandfed.org/research/workpaper/index.cfm ABSTRACT Although the airline industry has been studied extensively since passage of the Airline Deregulation Act of 1978, relatively little effort has gone into examining how hub location affects the level of service and degree of competition found at airports in the system. To help close'this gap, we investigate the geographic distribution of airline hub operations, the level of service, and the extent of competition at 112 major U.S.airports, extending previous work by Bauer (1987) and Butler and Huston (1989). Our key innovation is that we derive our measures of service and competition from indicator matrices that describe each airline's route system. clevelandfed.org/research/workpaper/index.cfm Introduction Many of the changes that have rocked the airline industry since passage of the Airline Deregulation Act of 1978 have received a great deal of attention from researchers.' The emphasis has been on the effect of deregulation on airline fares, mergers, and the development of hub-andspoke route systems. Airlines have adopted hub-and-spoke networks to make more efficient use 9 of their equipment--a trend that is exemplified by modification of United Airlines' route structure between 1965 and 1989 (see figure 1). Our focus in this paper is somewhat different. We investigate the geographic distribution of airline hub operations, the level of service, and the extent of competition at major U.S. airports, extending previous work by Bauer (1987) and Butler and Huston (1989). Instead of using an aggregate measure of airline service, we utilize a new, comprehensive measure derived from individual airline route data. We then employ these data to develop and analyze new measures of competition at individual airports. The first section of this paper utilizes information on nonstop service from the nation's 112 largest airports to examine the route structure of the 13 major U.S. airlines, to identify the Section I1 then location of airline hubs, and to measure the extent of competition at each fa~ility.~ develops a model of hub location, airline service, and concentration. Estimates of this model are presented in section III, and section IV summarizes our findings. I. Characteristics of U.S. Airline Service In this section, we use data on nonstop flights from airports in the nation's 100 largest 'See, for example, Bailey and Williams (1988). Bailey, Graham, and Kaplan (1985). Borenstein (1992). Meyer and Oster (1987). and Morrison and Winston (1987). 2 ~ h 13 e airlines included in our sample are Alaskan Airlines, American, America West, Braniff, Continental, Delta, Eastern, Midway, Northwest, Southwest, TWA, United, and USAir. According to the Air Transport Association (1990). U.S. passenger airlines with 1989 revenues in excess of $100 million per year included these 13 plus Pan American and Piedmont We excluded Pan American because its route structure is primarily international, while Piedmont's routes were included in USAir's schedule. clevelandfed.org/research/workpaper/index.cfm metropolitan areas to determine the location of airport hubs.3 We choose to rely on our data rather than statements from the airlines because this allows us to impose uniform standards across carriers. In addition, we develop new airport- and route-based measures of industry concentration, which are used as dependent variables in the model discussed in section II. Data Our sample consists.ofthe airports, served by the major carriers, in the 100 most.populous Metropolitan Statistical Areas (MSAs) in 1987.4 Because some of these regions contain more than one airport, a total of 112 facilities are incl~ded.~ Our data set indicates whether an airline serves a particular route, but provides no Thus, neither the level of actual activity nor passenger information about flight freq~ency.~ enplanements are captured. Still, we do have detail on routes and airlines not available in other data sets.7 We exploit the service data by airline and destination to compute measures of competition based on both overall service and route-by-route information. While it is well known that most airlines have adopted some form of hub-and-spoke 3~ more extensive description of the data and a detailed analysis of each airline's route stnrcture'k be found in Bania, Bauer, and Zlatoper (1992). 4 ~ used e the Office of Management and Budget's 1988 deftnition of MSAs to form the list of qu.alifying regions. SManyof the nation's largest MSAs are adjacent to another MSA (such as New York City and Newark, NJ). In such a case, the second MSA may contain another airport that is a potential substitute; however, even without a second airport, the combined economic activity of the two MSAs creates a greater demand for air service. Thus, we combined MSAs into larger metropolitan areas according to the Officeof Management and Budget's 1988 definition of Consolidated Metropolitan Statistical Areas (CMSAs). This resulted in 10 metropolitan areas with multiple airports (a total of 26 airports). See table 1 for a complete listing. 6The sample includes a total of 12,432 possible routes. However. we collected data for only half of these and assumed that service was symmetric. For example, we held that if an airline serviced the Portland-Atlanta route, then it also serviced the Atlanta-Portland route. To check this, we selected one airline (American) and collected data for routes in both directions. The symmetry assumption was valid in all but one case. o or example, Bauer (1987) includes data on passenger enplanements by airport, but contains no destination or airline-specific information. clevelandfed.org/research/workpaper/index.cfm system, the determination of what constitutes a hub is not straightforward.8 Our approach is to construct, for each airline and airport combination, an index of hub activity that measures the degree to which that airport is connected to the rest of an airline's network. For an airport-airline combination, the index is the percentage of other airports in the airline's route system that can be reached via nonstop service. Hub locations are well connected to an airline's network, while spoke airports are not. In a hub-and-spoke network, we would expect to find only a small number of airports that are well connected, many that are not well connected, and few in between. Thus, the distribution of the hub index should be bimodal, with a large spike at low service levels (low hub-index values) and a much smaller spike at higher levels (high hub-index values). On the j other hand, if an airline does not use a hub-and-spoke system, we would expect to find a relatively steady decline in the distribution of the hub index. Hub Locations To determine hub locations, we examined the hub-index distributions for each airline (displayed in figure 2). We found that in almost all cases, the hub locations were easily identified, since, as expected, very few had high service levels, a large number had extremely low levels, and few fell in the middle. The exceptions were the relatively diffuse carriers, USAir, Alaskan Airlines, and Southwest. These airlines do concentrate their activity in a small number of airports, but there is a relatively steady decline in the hub-index distribution. Thus, determining the lower bound of what constitutes hub service for them is somewhat more difficult. For these airlines, we arbitrarily designated airports with higher levels of service as hubs.9 Table 2 reports the 44 airport-airline combinations that we classified as hub locations. *~esearchershave taken several approaches to defining hubs. The Federal Aviation Administration (FAA) looks at total passenger hoardings. while Butler and Huston (1989) use a functional definition of a hub as an "airponat which large blocks of incoming and outgoing flights are coordinated to create numerous potential connections." Our definition is also a functional one, based on an analysis of each airline's route structure. g ~ hlower e bound varied across airlines primarily because of airline size differences. In small route networks, high hub-index values are easier to obtain, larger airlines showed much greater variety in the size of their hubs. clevelandfed.org/research/workpaper/index.cfm This list represents only 35 airports, since some of these have more than one airline with hub activity. Column 3 reports the total number of airports in the sample served by a given airline, column 4 is the number of those airports that can be reached with a nonstop flight, and column 5 is the number that can be reached with a one-stop flight. a Most airports served by a given airline can be reached via a nonstop or a one-stop flight from the hub airports. This can be seen by comparing the sum of columns 4 and 5 with column 3. For example, from Cleveland, passengers have nonstop service to 25 of the 71 other airports served by Continental. Another 44 airports can be reached with one-stop service. The key variable that we used to classify hubs--the hub index--is contained in column 6. High values correspond to the relatively small number of well-connected airports in the frequency distributions displayed in figure 2. The hub index ranges from a high of 100 percent for Midway Airlines at Chicago idw way airport to a low of 17 percent for United Airlines at Los Angeles International. Measures of Competition If the airline industry were perfectly contestable, there would be no point in calculating any measures of the extegt of competition, since such measures would have no meaning. Because no one has found that the airline industry meets these conditions--in fact, most studies show that the more competitors there are on a route, the lower fares tend to be--we construct various measures of the extent of competition based on the number of carriers offering seivice on a route or from an airport.10 Our measures do assign a large role for potential competition by treating infrequent service on a route in the same way as more frequent service. We computed a measure of the overall degree of competition at each airport by loseeBailey, Graham, and Kaplan (1985). Bauer and Zlatoper (1989). Borenstein (1989). Call and Keeler (1985), Hurdle et al. (1989).and Momson and Winston (1987). clevelandfed.org/research/workpaper/index.cfm calculating two versions of the Herfiindahl index for both nonstop and one-stop service." In table 3, we report the nonstop and one-stop service levels at each airport in the sample (columns 2 and 3), as well as the nonstop and one-stop Herfindahl index computed on the basis of overall service from an airport (columns 4 and 5). These measures are sensitive only to the level of service, not to the destination. The formula is where nroutesijis the number of nonstop routes from airport i for the jm airline.12 A similar measure (Hl,) was calculated for one-stop routes. The main limitation of these measures is that they are not destination sensitive. For example, suppose an airport has 10 airlines each serving a different nonstop route. The Herfindahl index for this airport will be equal to its theoretical minimum for 10 carriers (1,000), even though there is no nonstop competition at the route level. Although these airlines are not competing directly at the route level, the presence of other airlines at a given airport represents potential competition in that providing new service on a given route is easier if an airline already has gate space. Thus, while this measure is not sensitive to the actual destinations of flights departing from a given airport, it does measure the potential competition posed by other airlines serving the same facility. This is an important distinction, because while deregulation has freed airlines to provide service on any route, acquiring gate space may be difficult or impossible at some airports. An alternative measure of airport-level competition that is more sensitive to the actual l ~ h one-stop e calculation involved an aggregation of the nonstop and one-stop data, since we consider nonstop flights to be competition for one-stop flights. We applied this same principle to all of the one-stop measures of competition discussed herein. 12'T'he Herfindahl index is a measure of concentration, with larger values corresponding to greater concentration and therefore less competition. For a more detailed description of this measure, see Koch (1980). pp. 179-80. clevelandfed.org/research/workpaper/index.cfm level of competition on a route-by-route basis can be computed using another version of the Herfindahl index, calculated as where dserviceij, is one if the jh airline flies the route from i to k, and zero otherwise. HHO, is the nonstop Herfindahl index for the route between airport i and airport k. To get an overall measure for each airport (HHO,), we used the unweighted average of HHO, computed over all routes k. A similar measure (HH1,) was calculated for one-stop connections. The results are reported in table 3, columns 6 and 7. This route-by-route Herfindahl measure has two main limitations. First, while it is sensitive to route patterns of competition, it is not sensitive to the actual level of service (as measured by the number of airports that can be reached with a nonstop connection). Thus, an airport with 10 carriers all serving the same nonstop route would have an HE30 value of 1,000-indicating a great deal of competition--even though the facility is not well connected to other cities. A second problem is that this measure misses potential competition from other carriers currently serving different routes at the same airport. For example, an airport having 10 carriers each serving a different nonstop route would have a Herfindahl index equal to its maximum value (10,000), indicating the absence of competition. Although a Herfindahl index of 3,200 would be considered very high in most industries (i.e., the Department of Justice's antimerger guidelines would be violated), there is reason to treat this as a somewhat moderate level for the airlines. For example, one study finds that air fares cease to fall once three caniers are serving a route--equivalent to a Herfindahl index of about 3,200 using our definitions.13 13SeeBauer and Zlatoper (1989). I I clevelandfed.org/research/workpaper/index.cfm In general, three patterns emerge from the Herfindahl indexes. First, one-stop competition is much greater than nonstop competition, whether airport- or route-based measures are employed.14 Second, the route-by-route measures indicate much less competition than do the overall indexes. Finally, the coefficients of variation indicate that there is much more fluctuation - in the level of competition for one-stop routes than for nonstop routes. 11. Model of Hub Location, Airport Service, and Competition Here, we investigate what factors influence hub location, the level of service provided to an airport, and the degree of competition at each facility. A three-equation model of activity at an airport can be written as where H is a measure indicating whether an airport serves as a hub, S is the level of sewice, and C is the level of competition. Equation (1) is similar to the hub equation specified in B-auer (1987) and Huston and Butler (1990), while equations (2) and (3) are introduced here. The presence of a hub carrier is likely to affect the level of service (S) and concentration (C) independently from the effect of regional economic activity (R), distance @), airport characteristics (A), and weather (W). Therefore, equations (2) and (3) are not part of a structural model and should be viewed as reduced-form equations. 15 14Strictlyspeaking, the one-stop Herfidahl index is bounded from above by the nonstop index, since we treat nonstop flights as competition for one-stop flights. See footnote 10. 151hereare two possible approaches to this problem. One would be to use the fitted values from the estimation of equation (1) in equations (2) and (3) (Maddala [1983]). The drawback to this is that the calculation of the standard e m r s is not straightfomard,due to the nonlinearity of equation (1). An alternative approach is to derive maximum likelihood estimates. We intend to pursue both of these methods in future work. clevelandfed.org/research/workpaper/index.cfm A statistical summary of the variables used in the analysis, with definitions and data sources for each, appears in table 4. The measure used to approximate S is SERVICE, the number of airports that can be reached via a nonstop flight on any airline from any given airport H is represented by HUB, a variable equal to one if an airport has at least one hub carrier. (These airports are listed in table 2.) Finally, C is approximated by several measures of concentration: HO, HI, HHO, and HH1 (the Herf'indahl indexes described in section I). The values of these indexes are presented for all airports in table 3. The likelihood that an airport will have a hub carrier depends in part on R, a vector of regional economic activity. Factors such as a larger population (POP), higher per capita income (INCOME), more business- and tourist-related travel (BUSTOUR), and a greater number of large corporate headquarters (CORP) increase the demand for air travel and thus should raise the level of service (S), as well as make the airport a more likely candidate for hub operations. Our measure of business- and tourist-related travel (BUSTOUR) is constructed by regressing the log of the sum of employment in hotels (SIC 70) and amusement parks and recreational services (SIC 79) on the log of population and of per capita income. The residual from this regression, which captures the extent to which local economic activity is insufficient to support employment in SICS70 and 79, can therefore be viewed as a gauge of business and tourist travel to a given airport.16 16Theregression is log (EMW(kEMP79) = 15.4 + 0.89 log(P0P) + 1.27 log(INCOME), (2.92) (0.04) (0.34) where EMP70 is employment in hotels (SIC 70) and EMF79 is employment in amusement parks and recreational services (SIC 79). The adjusted r-squared is 0.89, and the standard errors appear in parentheses. All three coefficients are significant at the 1 percent level. The three airports with the largest residual from this regression are Las Vegas, Orlando, and Daytona Beach The three with the smallest residuals are Toledo, Fremo, and Dayton. By construction, the residual represents the portion of business and tourist travel that is unrelated to either population or income. For example, some portion of tourist travel to New Yo* City is related to characteristics of the city that stem in part from its large population and high income (such as myriad restaurants and culWal events). This stands in contrast to tourist travel to Orlando or Las Vegas, most of which is probably not related to population or income. clevelandfed.org/research/workpaper/index.cfm In the absence of hubbing activity, concentration should fall with increases in POP, INCOME, BUSTOUR, and COW. However, hub networks by their very nature increase the concentration of nonstop flights, since most airports do not have a broad enough economic base . to support even one hub carrier with only local traffic. As a result, hub carriers tend to dominate these airports' nonstop flights. One-stop flights should be much less concentrated, hecause passengers can use one leg of their flight to reach a competing hub.17 The distance variable, D, is a measure of the central location of an airport. For each airline, we measured the sum of 2ir miles from each airport to every other airport in that .airline's route network.18 Airports in favorable locations (smaller D values) are more likely to have hub carriers and to receive more service. Concentration could be affected by hubbing activity, as discussed above. In the absence of hubbing activity, a better location would be expected to support more competition. However, if an airport has a hub airline, its presence may intimidate other carriers, since they would find it harder to compete with the hub carrier's more frequent nonstop flights. A is a vector of regional factors that differentiate airports. Specific components include SLOT, OTHER, MINOR, and GATEWAY. SLOT is equal to one if an airport faces FAA restrictions on the number of takeoff and landing slots. Only four facilities have a value of one: John F. Kennedy International, La Guardia, Chicago O'Hare, and Washington National. If access to these airports were not limited, carriers would offer more service and would be more likely to set up hub-and-spoke operations. Concentration might then be higher because of the barrier to entry, or lower because regulators act to discourage concentration. OTHER indicates the presence of mother airport in a given airport's economic region. 17wedo not present results for two-or-more-stopflights because they closely mirror those for the one-stop routes. 18wealso tried three other measures of distance: the sum of miles between a given airport and every other destination, weighted by the population of each destination; the sum of the natural log of miles between airports; and the sum of the natural log of miles between abports, weighted by the population of the destination. Each of these measures performed similarly to those reported here. clevelandfed.org/research/workpaper/index.cfm For example, Cleveland Hopkins and Akron Canton Regional would both have a value of OTHER equal to one. MINOR, on the other hand, indicates that the airport has significantly less capacity in terms of ground and flight facilities than others in its region. To continue our above example, Cleveland Hopkins would have a MINOR value equal to zero, while Akron Canton Regional would have a value equal to one.I9 Finally, GATEWAY indikates whether an airport has international nonstop connections to Europe, Asia, or the South Pacific. W is a vector of weather-related variables. Good flying conditions should result in more service and thus a greater probability of having a hub carrier. To the extent that the weather is worse for flying, concentration may be higher. To control for these possible effects, we obtained data on the average number of days per year during which snowfall exceeded one inch (SNOW) for each airport, as well as on the number of days per year that fog reduced visibility to less than one-quarter mile (FOG).20 111. Estimation Using the data discussed above, we estimated equation (1) using logit rather than pr0bit.~1 The two techniques yield similar results, but the disturbance in the logit model allows for more outliers in the error term. Equations (2) and (3) were estimated in log linear form. 19,41thoughwe defined this variable in a rather ad hoc way, our approach is equivalent to estimating the service equation with individual airport dummies for airports in regions having more than one facility, and then assigning MINOR to equal one when the coefficienton the airport dummy is significantly less than the coefficients for other airports in the region. The values of OTHER and MINOR for metropolitan areas with multiple airports are listed in table 1. 20rhe weather variables were divided by 365 so that they represented the portion of the year affected by these two conditions. 21~ecausethe determinants of hub location, air service, and competition in Alaska and Hawaii are likely to differ from those for airports in the continental United States, all of the equations were estimated both with and without the Honolulu and Anchorage airports. We report regressions only for the sample excluding these two cities. since the results are similar. clevelandfed.org/research/workpaper/index.cfm Hub Determinants Table 5 presents the regression results for the various models. We found that four factors increase the likelihood of an airport's having a hub carrier: a larger regional population, a better location (a lower D value), gateway connections to Europe, Asia, or the South Pacific, and more v business and tourist travel. The effect of each of the remaining variables was statistically insignificant. Table 6 ranks the airports by their estimated likelihood of having a hub carrier. The most likely new hubs based on these results are Miami International, Boston's Logan International, New York's La Guardia, New Orleans International, and Louisville. The least likely existing hubs are Washington National, Charlotte/Douglas International, Dayton International, Dallas Love Field, and El Paso International. It is worth noting that two of these unlikely hubs are associated with Southwest, a relatively small regional carrier. Southwestis the only airline operating out of . . Dallas Love Field and is the dominant carrier operating out of El Paso International. Another of the unlikely hubs, Dayton International, has since lost hub service from USAir. Service Determinants Ordinary least squares (OLS) estimates of.the service regression are presented in table 5. The results indicate that SERVICE rises less than proportionally with population and falls as location worsens (distance to other airports increases). The effect of per capita income is positive but statistically insignificant. Both OTHER and MINOR have a negative and statistically significant effect, with their magnitudes implying that the presence of another airport in the region lowers SERVICE 34 percent for nonMINOR airports and 72 percent for MINOR airports. International connections (GATEWAY) have a positive and significant effect, increasing SERVICE by 34 percent. Finally, the effect of business and tourist travel (BUSTOUR) is positive and statistically significant. With the exception of SNOW, the remaining variables have the expected sign, although none is statistically significant. clevelandfed.org/research/workpaper/index.cfm i Concentration Determinants Table 5 also presents OLS estimates of the determinants of concentration at both the airport and route levels for nonstop and one-stop flights, using measures derived earlier. For nonstop flights and the airport-level concentration measures, the results indicate that a less central '1 location reduces concentration. While somewhat counterintuitive, this could be a result of airlines' reluctance to compete head to head with nonstop flights. Under these circumstances, the more distant airports, which are less likely to be hubs, will have lower nonstop measures of concentration. Two 'other statistically significant factors are h4INOR and GATEWAY, which result in higher concentrations (84 percent and 52 percent, respe~tively).~~ < For route-level measures of concentration, the results for nonstop flights are qualitatively consistent with those for the airport-level measures. A worse location is associated with lower concentration levels, although the magnitude of the effect is smaller. MINOR airports have higher concentration levels, but the effect is only marginally significant. The effect of the presence of gateway connections is not statistically significant. We find much more explanatory power, using either measure of concentration, for the one-stop equations. For such service, the results using airport-level measures indicate that concentration falls with population and business- and tourist-related activity, but rises for MINOR airports. Unexpectedly, FOG is associated with higher concentration levels, although the effect is only marginally sigmficant. Using route-level measures, we find that population, income, and a better location decrease concentration, while the presence of another airport in the region and status as a MINOR airport tend to be associated with higher concentration levels. An apparent paradox is that central location lowers concentration for one-stop routes, but raises it for nonstop routes. If an airport has a favorable location, it is more likely to be a hub and %f course, concentration should be measured at the regional level if one is interested in determining how much control over fares carriers might have. I I clevelandfed.org/research/workpaper/index.cfm to have highly concentrated nonstop service, because hub caniers tend to dominate service at their airports. But having a favorable location also means that other airlines (with hubs at other airports) will offer at least some service. Consequently, one-stop concentrations tend to be lower as a result of interhub competition. \ IV. Summary We use route-level data to develop measures of the degree to which airlines employ a huband-spoke route structure, and explicitly identify the location of airline hub activity using a new approach. Our data set allows us to develop airport- and route-specific measures of concentration that indicate a great deal of variance, particularly at the nonstop level. This is true even among airports having hub carriers. We find that the location of airline hub activity is positively related to population and negatively related to distance from other airports. Regions that have access to international flights and that are desirable business and tourist destinations are also more likely to have hub caniers. On the other hand, weather conditions, the presence of large corporate headquarters, per capita income, and airport slot restrictions play a very small role. Our findings also show that service (as measured by the number of nonstop connections from a given airport) increases with population, favorable location, business- and tourist-related activities, and access to international flights. The presence of multiple airports in a metropolitan region tends to have just the opposite effect, as do weather, corporate headquarters, per capita income, and airport slot restrictions. The results concerning the degree of competition are mixed, depending on the particular measure employed and whether the unit of analysis is nonstop or one-stop connections. The only consistent result is that concentration is higher at MINOR airports. Airports in more-populous regions that are frequented by business travelers and tourists have lower one-stop concentration measures; however, these factors do not appear to affect nonstop concentration. A favorable location lowers one-stop concentration measures, but raises nonstop concentration measures. clevelandfed.org/research/workpaper/index.cfm 1 One explanation for this phenomenon is that while an airport in a favorable location has a higher probability of attracting a hub carria that will dominate its nonstop service, it is also more likely I clevelandfed.org/research/workpaper/index.cfm References Air Transport Association. Air Transport 199Q,Annual Report of the U.S. Scheduled Air Industry, Washington, D.C., 1990. Bailey, E.E., Graham, David R., and Kaplan, Daniel P. B r e d a t i n ? the Airlines. Cambridge, Mass.: MIT Press, 1985. - . Bailey, E.E. and Williams, J.R. "Sources of Economic Rent in the Deregulated Airfie Industry," Journal of Law and Economics, 31,1988, pp. 173-203. Bania, Neil, Bauer, Paul W., and Zlatoper, Thomas J. "U.S. Air Passenger Service: A Taxonomy of Route Networks, Hub Locations, and Competition," Federal Reserve Bank of Cleveland,Working Paper 9216, December 1992. Bauer, Paul W. "Airline Hubs: A Study of Determining Factors and Effects," Federal Reserve Bank of Cleveland, Economic Review, Quarter 4, 1987, pp. 13-19. Bauer, Paul, W. and Zlatoper, Thomas J. "The Determinants of Direct Air Fares to Cleveland: How Competitive?" Federal Reserve Bank of Cleveland, Economic Review, Quarter 1, 1989, pp. 2-9. Borenstein, S. "Hubs and High Fares: Airport Dominance and Market Power in the U.S. Airline Industry," Rand Journal of Economics, 20, Autumn 1989, pp. 344-65. Borenstein, S. "The Evolution of U.S. Airline Competition," Journal of Economic Persuective~, 6:2, Spring 1992, pp. 45-73. Boyer, Richard and Savageau, David. Places Rated Almana~.New York: Prentice-Hall, 1989. Butler, Richard V. and Huston, John H. "The Location of Airline Hubs," paper presented at the Southern Economic Association Meetings, Orlando, Fla, 1989. Call, G.D. and Keeler, T.E. "Airline Deregulation, Fares, and Market Behavior: Some Empirical Evidence," in Daugherty, A.H., ed., Analytic Studies in Trans~ortEconornici. Cambridge: Cambridge University Press, 1985, pp. 221-47. Hurdle, G.J., Johnson, R.L., Joskow, A.S., Werden, G.J., and Williams, M.A. "Concentration, Potential Entry, and Performance in the Airline Industry," Journal of Industrial Economics, 38, December 1989, pp. 119-39. Huston, J.H. and Butler, R.V. "The Location of Airline Hubs," working paper, Trinity University, May 15,1990. Koch, J.V., Jndustrial Organization and Prices, 2d ed. Englewood Cliffs, N.J.: Prentice-Hall, 1980. . . Maddala, G.S. Limited Deuendent and Oudtanve Variables in Econometri~.Cambridge: Cambridge University Press, 1983. clevelandfed.org/research/workpaper/index.cfm Meyer, J.R. and Oster, C.V. Jr. D e r e g u l a t i o n the Future of Cambridge, Mass.: MIT Press, 1987. Travel. Momson, Steven A. and Winston, Clifford. "Empirical Implications and Tests of the Contestability Hypothesis," Journalofcono&, 30:1, April 1987, pp. 53-66. National Oceanic and Atmospheric Administration, Local Climatological Data, 1988. \ Standard & Poor's Corporation. w Executives. Volume 1, 1989. d & Poor's R a t e r of C-ons. Directors. and U.S. Department of Commerce, Bureau of Economic Analysis. County personal income computer tape file, 1987. U.S. Department of Commerce, Bureau of Economic Analysis. County business patterns computer tape file, 1987. U. S. Department of Transportation. Air Carrier S Summary, Data Bank 6, 1986. u . Origin and Destination City Pair clevelandfed.org/research/workpaper/index.cfm Figure 1. United Airlines Route Structure, 1965 and 1989 1965 Source: United Airlines schedule guides, 1965 and summer 1989. J clevelandfed.org/research/workpaper/index.cfm Figure 2. Distribution of Nonstop Service by Airline and A i r p o ~ t AU Airlines Air Alaska American America West Braniff Continental Delta Eastern Midway Northwest Southwest 'IWA United USAir Source: Various airline schedule guides, 1989, and authors' calculations. Note: The maximum value on the vertical axis is 20 airports. and the height of the bars represenls the number of airports with values of the hub index falling into the following eight categories: less than 0.1. 0.1-0.2,0.2-0.3.0.3-0.4,0.4-0.5,0.5-0.6,0.6-0.7, andlarger than 0.7. Each of the airlines had very large number of airpons with a hub index of less than 0.1: however, the height of the first bar in each graph was truncated at 20 airports to improve the resolution of the data for airpons with higher values of the hub index. Data in the first panel, which represents the composite of all airlines, is truncated for the f i t two categories. There are 37 airpow with a hub index less than 0.1 and 35 airports with a hub index between 0.1 and 0.2. clevelandfed.org/research/workpaper/index.cfm Table 1. Metropolitan Areas with Multiple Airports Metropolitan Area Airport Chicago-Gary-Lake County, IL-IN-WI CMSA Chicago Midway Chicago O'Hare Cleveland-Akron-Lorain, OH CMSA Akron Canton Regional Cleveland Hopkins International Dallas-Fort Worth, TX CMSA OTHER MINOR Dallas Love Field Dallas Ft. Worth International 1 1 1 0 Houston-Galveston-Brazoria,TX CMSA W i a m P. Hobby Houston Intercontinental 1 1 0 0 CA CMSA Los ~n~eles-~naheim-~iverside, Burbank-GlendalePasadena Los Angeles International Long Beach Ontario International John Wayne Airport 1 1 1 1 1 1 0 1 1 1 Miami-Fort Lauderdale, FL CMSA Fort Lauderdale Miami International New York-N. New Jersey-Long Island, NY-NJ-CT CMSA Long Island MacArthur Newark International John F. Ke~nedyInternational La Guardia 1 1 1 1 San Francisco-Oakland-San Jose, CA CMSA Metropolitan Oakland San Francisco International San Jose International 1 1 1 Tampa-St. Petersburg-Clearwater. FL MSA St. Petersburg-Clearwater Tampa International .. 1 Washington, DC-MPVA MSA Washington National Airport Washington Dulles Airport - \ Source: Authors' assignments. 1 " 1 1 clevelandfed.org/research/workpaper/index.cfm Table 2. Selected Statistics for Hub Airport-Airline Combinations '\ 4 Metropolitan Area [I1 Airline Atlanta, GA MSA Atlanta,GAMSA Baltimore. MD MSA Charlotte-Gastonia-Rock Hill,NC-SC MSA Chicago-Gary-Lake County, L I N - W I CMSA (Midway) Chicago-Gary-Lake County. L I N - W I CMSA (O'Hare) Chicago-Gary-Lake County. L I N - W I CMSA (O'Hare) Cincinnati-Hamilton, OH-KY-IN CMSA Cleveland-Alcron-Lorain, OH CMSA ( H o p k h International) Cleveland-Akron-Lorain, OH CMSA ( H o p k h International) Dallas-Ft. Worth, TX CMSA (International) Dallas-Ft. Worth, TX CMSA (International) Dallas-Ft. Worth, TX CMSA (Love Field) Dayton-Springfield, OH MSA Denver-Boulder, CO CMSA Denver-Boulder, CO CMSA Detroit-AM Arbor, MI CMSA El Paso, TX MSA Houston-Galveston-Brazoria,TX CMSA (Hobby) Houston-Galveston-Brazoria,TX CMSA (Intercontinental) Indianapolis, IN MSA Kansas City, MO-KS MSA Las Vegas. NV MSA Los Angeles-Anaheim-Riverside,CA CMSA (LA International) Los Angeles-Anaheim-Riverside, CA CMSA (LA International) Los Angeles-Anaheim-Riverside,CA CMSA (LA International) Memphis, TN-AR-MS MSA Minneapolis-St. Paul MN-WI MSA Nashville, TN MSA New York-N. NJ-Long Island, NY-NJ-CT CMSA (JFK) New York-N. NJ-Long Island, NY-NJ-CT CMSA (Newark) Orlando. FL MSA Philadelphia-Wilmington-Trenton.PA-NJ-DE-MD CMSA Phoenix, AZ MSA Phoenix, AZ MSA Pittsburgh-Beaver Valley CMSA Portland-Vancouver. OR-WA CMSA Raleigh-Durham, NC MSA Salt Lake City-Ogden, UT MSA San Francisco-Oakland-San Jose, CA CMSA (SF International) Seattle-Tacoma, WA CMSA St. Louis. MO-IL MSA Washington. DC-MD-VA MSA (Dunes) Washington, DC-MD-VA MSA (National) Delta Eastern usAir USAir Midway American United Delta Continental USAir American Delta Southwest USAir Continental United Northwest Southwest Southwest Continental usAir Braniff America West Delta United USAir Northwest Northwest American TWA Continental Braniff USAir America West Southwest USAir Air Alaska American Delta united Air Alaska TWA United USAir [21 Source: Various airline service guides, summer 1989, and authors' calculations. Airports Hub Service Served by Hub Airline Nonstop One-stop Index PI [41 151 [61 Percent Regional -- 1 7 66 80 76 75 79 60 63 78 72 71 44 50 100 78 72 67 75 70 100 48 81 63 81 75 59 50 68 47 84 69 74 83 86 82 88 64 100 92 80 60 100 50 75 81 I 1 clevelandfed.org/research/workpaper/index.cfm Table 3. Concentration Statistics lor Airports in Large Metropolitan Areas, 1989 AwJServed by AU Airlines Metropolltan Area 111 Albany-Schendy-Troy, NY MSA Albuquerque. NM MSA Allentown-Bethlehem, PA-NJ MSA Anchorage. AK MSA Atlanta. GA MSA Augusta, GA-SC MSA Austin. TX MSA Baltimore, MD MSA Baton Rouge. LA MSA B u r g h a m , AL MSA Boston-Lawrence-Salm-Lowell-Brohn. MA NECMA Buffalo-Niagarq NY CMSA Charleston. SC MSA Charleston. WV MSA Charlotte-Gastonia-Rod; Hill, NCSC MSA Chananooga. TN MSA Chicagdjary-Lake County, IL-IN-WI CMSA (Midway) Chicago-Gary-Lake County, L I N - W I CMSA (O'Hare) Cincinnati-Hamilton. OH-KY-IN CMSA Cleveland-&on-Lorain, OH CMSA (Akron-Canton) Cleveland-Won-Lorain, OH CMSA (Hopkins) Colorado Springs. CO MSA Columbia, SC MSA Columbus, OH MSA Corpus Christi, TX MSA Dallas-Fort Worth. TX CMSA (Intmdonal) Dallas-Fort Worth. TX CMSA (Love Field) Daytona Beach, FL MSA Dayton-Springfield. OH MSA Denver-Boulder. CO CMSA Des Moines. IA MSA Detroit-Ann Arbor. MI CMSA El Paso. TX MSA Evansville. IN-KY MSA Fayetteville, NC MSA Fort Myers-Cape Coral, FL MSA Fort Wayne. IN MSA Fresno. CA MSA Grand Rapids. MI MSA Greensboro-Winston-Salem-High PoinI. NC MSA Greenville-Spartanburg. SC MSA Harrisburg-Lebanon-Carlisle. PA MSA Honolulu. HI MSA Houston-Galveston-Brazoria,TX CMSA (Hobby) Houston-Galveston-BrazoriqTX CMSA (Intcmat~onal) Indianapolis. IN MSA Jackson. MS MSA Jacksonville, FL MSA Kansas City, MO-KS MSA Knoxville, TN MSA Las Vegas, NV MSA Lexington-Fayem, KY MSA Little Rock-North L i e R o d ; AR MSA Los Angeles-Anaheim-Riverside, CA CMSA (Burbank) Los Angeles-Anaheim-Riverside,CA CMSA (John Wayne) Los Angelcs-Anaheim-Riverside,CA CMSA (LA Intmdonal) Los Angeles-Anaheim-Riverside,CA CMSA (Long Beach) Los Angeles-Anaheim-Riverside,CA CMSA (Ontario) Louisville. KY-IN MSA Madison. WI MSA -- Herfindah1 Index (overall service) Hufindahl Index (airport pairs) Nonstop 121 Onestop 131 Nonstop 141 Onestop 151 Nonstop [a1 One-stop 171 16 16 8 4 84 1 13 47 7 17 43 19 14 9 51 6 38 90 53 7 43 9 9 21 4 76 10 7 30 56 8 60 14 8 8 23 10 12 9 15 4 12 13 28 52 35 8 20 45 14 40 13 11 13 13 53 13 18 18 6 92 90 99 88 27 83 98 64 95 89 68 90 93 83 60 84 72 21 58 99 68 93 98 88 74 35 49 81 80 55 94 51 90 100 95 87 1.872 1,479 3.333 3.061 3.810 5,000 1.247 3,905 3.125 1,818 1,418 2.986 3.772 4.074 1.398 1.266 2,637 4.056 1.104 5,232 1.417 1.191 2,690 1.710 1,025 1,515 2,166 3.540 7.938 7,292 9.375 8.125 6.538 5.000 7,436 8.652 9,286 8,725 8.109 8.860 8.929 10.000 9.248 9.167 8,969 6,056 9.088 9,286 7.345 8.889 8.889 8.095 10,000 6,086 10.000 8,333 9.833 6,711 8.750 8.056 . 8,810 10.000 10.000 9,022 9.000 8.083 8.889 9.000 6.667 9.583 7.692 9.286 8.750 8.238 9.375 8.667 7,796 9.167 7.4% 10.000 8.485 9,103 8,846 6,670 8.141 7.037 8.889 9,167 3.500 2,877 4.924 7.509 1,912 7,073 3.302 2,152 4,828 3,212 1.783 3.300 4,332 5,489 3.055 4,372 3.915 1,822 3.011 3.929 2,020 4.102 4,552 2,438 6.644 2,011 10,000 4.688 3,066 1.852 3,776 1.916 5.372 4,399 4.832 2,272 4.259 5.299 3.525 3.856 4.309 3.472 3.289 2.650 2,413 1.959 5,433 2,513 1.933 ----- % 57 88 94 85 88 97 83 59 76 95 90 66 95 71 95 95 76 89 58 89 91 91 90 2.a~ 2,225 3.939 4,204 5,337 4.W 1,762 4.5211 3.122 2,544 4.078 clevelandfed.org/research/workpaper/index.cfm Tabk 3. Concentration Statistics for Airports in Large Metropolitan Areas, 1989, continued AirPo- Suvcd by All Airlines Hcrfindahl Index (overall service) - Metropolitan Area 111 \ - Melbourne-Tiblsvill~PhBay. FL MSA Memphis. TN-AR-MS MSA Miami-Fort Lauderdale, FL CMSA (Fort Laudcrdale) Miami-Fort Lauderdale, FL CMSA (Miami International) MilwaukccRacine. WI CMSA Minneapolis-St Paul. MN-WI MSA Mobile, AL MSA Nashv~lle.TN MSA New Orleans, LA MSA New York-N. NJ-Long Island, NY-NJ-CT CMSA (IFK) New York-N. NJ-Long Island, NY-NJ-CT m S A (L~Guardia) New York-N. NJ-Long Island, NY-NJ-CT CMSA (Long Island) New York-N. NJ-Long Island, NY-NJ-CT CMSA (New&) Oklahoma City. OK MSA Omaha,NE-IA MSA Orlando, FL MSA Pensacola. FL MSA Phladelphra-W1Imington-Tru1ton, PA-NJ-DEMD CMSA Phoenix. AZ MSA Plusburgh-Beaver Valley. PA CMSA Portland-Vancouver,OR-WA CMSA ProvidenccPawtucket-Woomke&RI MSA Raleigh-Durham, NC MSA bchmond-Petenburg. VA MSA Roanoke, VA USA Rochester, NY MSA Rock Island, lL MSA Sacramento, CA MSA Salt Lake City-Ogden, UT MSA San Antonio. TX MSA S n Diego, CA MSA San Francisco-OaklandSanJose, CA CMSA (Oakland) San Franasfo-Oakland-San Jose. CA CMSA (San Fmcim) Sari Francisco-OaklandSanJose. CA CMSA (San Jose) Santa Barbara-Santa Maria-Lompoc, CA MSA S m t a , FL MSA Savannah. GA MSA Seattle-Tacoma, WA CMSA Shreveporf LA MSA South Bend-Mishawaka. IN MSA Spokane,WA MSA S t Louis, MO-lL MSA Syracuse. NY MSA Tampa-& Petenburg-Clcarwatcr. FL MSA (St Petusburg) Tampa-St Pctcnburg-Clcarwater, FL MSA (Tampa) Toledo. OH MSA Tucson. AZ MSA Tulsa, OK MSA Washing(oq DC-MD-VA MSA (Nstioaal) Washington, DC-MD-VA MSA (Dullu) West Palm Beach-BocaRaton-Delray Beach,FL MSA Wichita, KS MSA Mean Standard Deviation Coefficient of Variation Source: Variour airline service guldes, summr 1989, and autbors' ulculatioah - - - Nonstop 141 Orre-*p Nonstop 161 1.852 6,206 1.689 1 m 2628 5.184 2.099 4.186 1.358 28-47 1.460 4,200 2,710 1,690 1.327 1,552 2593 3,217 1.918 6.778 1,717 1.953 5,233 3.010 3,400 3,950 1.804 1.510 1.152 1.099 1.203 1.120 2.097 1,379 1,139 1.565 1.159 3,156 1,091 1.415 1,264 1.088 2089 1,096 937 1,217 1.439 1,569 1.734 2,171 2,695 1.793 8,056 9,259 7,300 7,848 8,636 8.469 9.375 8,674 9,111 8,034 7.849 10,000 7,500 8,667 9,615 7.244 9.375 8.438 7,280 9.010 7,528 8.333 8,690 8.462 10,000 9,118 9.286 7,639 8.714 7.917 7,427 7.255 6.498 6.583 8,333 8,393 8,750 7,125 9.375 9,500 -- On~stop 17l. 1x1 - -- -. 4,444 2677 2,189 2.027 2,604 2286 4,502 2.417 2.084 3.078 2,060 5.643 1.903 3.129 2876 1.892 4,313 1.968 1,807 2446 3.181 3.788 2.865 4.212 4.631 3,843 3.595 3,855 3.246 2,558 2,350 5,016 1,831 3,422 6,212 2873 4.480 1.803 5.000 3.729 clevelandfed.org/research/workpaper/index.cfm Table 4. Vuiable D81nitionsand Data Sourca Vuiable Nune Definition Data Source HUB Equalto 1 ifairportha a hub &a (see table 2); zero othenvise Authors' ulculations SERVICE N u m b of airports in sample reachable by nonstop airse~iceFmmrirportin 1989 Airline flight scbcdules INCOME 1987 per crpita p a s o d inwme for M A cootlining rirpofi U.S. Department of Commerce. county persod bmme comp-1.pefle POP 1987 population for M A wnaining airport U.S. Department of Commera. county personal bmme comp-wfile COW N u m b of Standard & Poor's 500 companies headquar(ered in hBA or CMSA containing airport Standud & Poor's Register of Corporations. Dktclors, md Executives. Volume 1. 1989 Sum of air miles Frofn U.S. Department of Tmmporcatioq Air M a Staristics. Origin md Datirution City Paii Summary airport to each of other airports in sample Mun StMdud Deviation hkimmn k i m m 0.31 0.47 0.00 1.00 15,896.06 2,624,541.96 2,469.18 4,171,725.32 122,403.68 47.058.39 1987 total employment in hotels md ofhalodging places (SIC 70) for M A containing airport U.S. Deputment of Commerce. m t y business paDtans comp-*file 13.805.34 18.311.88 1987 total anploymmt iu murancut md -tion services (SIC 79) for MSA containing lirport U.S. Department of Commerce. county business paotanr computer crpe Ne 10.666.14 17.969.31 BUSlWlR Business-touristrctivity proxy: residual fmm regression of log(E7WE79) on log(P0P) and log(INC0ME) Authors' calculations 0.00 0.41 SNOW Avenge number of days snowfall exceeded one iDch Loul ClimatologicalData, National Ocunic and Atmospheric Administration FOG Average number of days visibility WJS 1/4 mile or less LoulClirnatologiul Data, GATEWAY Equal to 1 if airport bas service to Europe, Asia. or South Pacific; zcro otherwise OTHER Equal to 1 if mdropolitan Natiod Ocunic md Atmospheric Administration Airline flight schedules Authors'crlculatiom uea hrs mother rirport; zero otherwiw MINOR Equalto1 ifairportisnot the meeopolitan u u ' s major airport (see text for details); zem othuwise Authors' crlculations SLOT Equdtolifrirportis subject to FAA l ~ d h r g& takeoff raeictions: zM othuwiw FAA ' clevelandfed.org/research/workpaper/index.cfm Table 5. Regression Results Variable Logic Hub Equation Dep. Var: HUB OLS, Service Equation Dep. Var: SERVICE OLS, Nonstop, Airport Dep. Var: HO Estimated Wald Coefficient Chi-square Estimated Coefficient T-Ratio Estimated Coefficient T-Ratio 0.61 0.70 0.01 -0.55 -0.41 -1.28 -0.13 0.29. 3.57 -1.83 0.56 -5.73 log(poP) log(INC0ME) log(CORP+ 1) log(D) OTHER MINOR SLOT GATEWAY SNOW/365 FOG/365 BUSTOUR CONSTANT N -2 log L R-Squared Note: 5.97 *** 1.60 0.08 -2.23 ** -2.06 ** -5.20 *** -0.45 1.79 1.45 -1.30 4.48 *** -1.32 -0.01 -0.22 0.00 -0.85 -0.11 0.61 0.08 0.42 2.67 0.46 0.01 19.80 110 139.091 *, **, and *** denote 10,5, and 1 percent significance levels, respectively. Source: Authors' calculations. -0.08 -0.50 -0.05 -3.45 *** -0.56 2.46 ** 0.25 2.56 ** 1.08 0.33 0.08 4.56 *** OLS, Nonstop, Route Dep. Var: HHO Estimated Coefficient T-Ratio -0.02 0.02 -0.01 -0.19 0.00 0.11 -0.05 -0.06 0.54 0.44 0.00 11.29 -0.93 0.22 -0.32 -3.33 ** 0.10 1.98* -0.69 -1.47 0.96 1.36 0.01 11.35 *** OLS, One Stop, Airport Dep. Var: H1 OLS, One Stop, Route Dep. Var: HHl Estimated Coefficient T-Ratio Estimated Coefficient T-Ratio -0.24 -0.31 -0.06 -0.07 0.18 1.06 0.30 0.13 -2.64 1.66 -0.18 14.44 -3.65 *** -1.08 -0.93 -0.41 1.42 6.55 *** 1.51 1.17 -1.63 1.79 * -2.14 ** 5.05 *** -0.28 -0.41 -0.02 0.24 0.16 0.98 0.22 0.01 -1.30 1.06 -0.25 12.95 -5.70 *** -1.95 -0.32 2.04** 1.71 8.22 *** 1.54 0.11 -1.10 1.58 -4.16 *** 6.21 *** clevelandfed.org/research/workpaper/index.cfm Table 6. Actual and Predicted H u b Values Metropolitan Area Chicago-Gaty-Lake County IL-IN-WI CMSA (O'Hare) St. Louis, MO-IL New York-N. New Jersey-Long Island. NY-NJ-CT (Newark) b e a p o l i s - S t . Paul, MN-WI Madelphia-Wilminptm-Trenton. PA-NJ-DE-MD Cmcinnati-Hamilton OH-KY-IN CMSA Atlanta. GA Detroit-Ann Arbor. MI CMSA Denver-Boulder. CO CMSA New York-N. New Jersey-Long Island, NY-NJ-(JT(JFK) Dallas-Fort Worth, TX CMSA (International) HoustonGalveston-Bmria, TX CMSA (Intercontinental) Los Angeles-Anaheim-Riverside. CA CMSA (LA International) Miami-Fort Lauderdale, FL CMSA (MiamiInternational) Boston-Lawrence-Salem,MA-NH CMSA Las Vegas, NV New York-N. New Jersey-Long Island, NY-NJ-CT (La Guardia) Washington. DC-MD-VA (Dulles) Plusburgh-Beaver Valley. PA CMSA New Orleans. LA San Francisco-Oakland-SanJose. CA CMSA (San Francisco) Orlando. FL ChicagoGary-Lake County. IL-IN-WI CMSA (Midway) Raleigh-Durham. NC Portland-Vancouver.OR-WA CMSA Kansas City, MO-KS Louisville, KY-IN Seanle-Tacoma, WA CMSA Memphis, TN-AR-MS Nashville, TN Cleveland-Akron-Lorain,OH CMSA (Hopkins) Houston-Galveston-Braria, TX CMSA (Hobby) Columbus. OH sari h l 0 ~ 0TX . Phoenix, AZ Salt Lake City-Ogden, UT New York-N. New Jersey-Long Island. NY-NJ-CT (Long Island) Indianapolis, IN Miami-Fort Lauderdale, FL CMSA (Fort Lauderdale) Milwaukee-Racine,WI CMSA Buffalo-Niagara Falls CMSA Tulsa, OK Baltimore. MD Oklahoma City, OK Hamsburg-Lebanon-Carlisle,PA Charleston. SC Birmingham. AL Tampa-SLPetersburg-Clearwater,FL (Tampa) El Paso. TX Daytona Beach. FL Dallas-Foxt Worth. TX CMSA (Love Field) Dayton-Springf~eld, OH Knoxville. TN (3harloneGastoNa-RockW, NC-SC Washington. DC-MD-VA (National) Los hgeles-Anaheim-Riverside. CA CMSA (Burbank) Syracuse. NY Hub Predicted Value 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 0 1 1 0 1 1 1. 1 1 1 0 1 1 1 1 1 0 0 1 1 0 1 0 0 0 0 1 0 0 0 0 0 1 0 1 1 0 1 1 0 0 0.998 0.995 0.992 0.992 0.991 0.990 0.990 0.988 0.985 0.984 0.980 0.975 0.969 0.950 0.91 1 0.852 0.842 0.834 0.824 0.769 0.730 0.684 0.656 0.602 0.600 0.586 0.547 0.544 0.537 0.536 0.527 0.489 0.453 0.425 0.365 0.355 0.352 0.340 0.327 0.313 0.308 0.288 0.278 0.265 0.255 0.248 0.241 0.226 0.224 0.207 0.197 0.193 0.179 0.176 0.176 0.168 0.151 , -0 * clevelandfed.org/research/workpaper/index.cfm Table 6. Actual and Predicted Hub Value, continued Meaopolitan Area Jacksonmlle, FL Los Angeles-Anaheim-Riverside.CA CMSA (Ontario) Los Angeles-AMheim-Riverside, CA CMSA (John Wayne) Cleveland-Akron-Lorain,OH CMSA (Akron-Canton) Rochester, NY Jackson. MS GreemilleSpartanburg. SC Los Angeles-Anahem-fiverslde, CA CMSA (Long Beach) Omaha. NE-IA Chat&anooga. TN-GA Grand Rapids, M San Diego, CA Mobile. AL Tucson. AZ Toledo. OH Shreveport, LA LtUe Rock-No& Linle Rock, AR Richmond-Petenburg, VA Lexington-Fayeite. KY Wichita. KS Colorado Springs. CO Greensboro-WinstonSalem-HighPoint. NC Albuquerque. NM Davenport-Rock Island-Moline. IA-IL Austin. ?X West Palm Beach-Boca Raton-Delray Beach, FL Tarnpaat. Petenburg-Clearwater. FL (St. Pelersburg) Des Moines. lA Pensacolq FL C o r p u s ~ ? X Melbourne-Titusville-Palm Bay. FL AlbanySchenectady-Troy,NY Augusta, GASC Fayetteville. NC Baton Rouge. LA Evansville, IN-KY Roanoke. VA Columbiq SC Fort Wayne, IN South Bend-Mishawaka. IN Allentown-Bethlehem.PA-NJ Fort Myers-Cape Coral, FL Savannah. GA Providence-Pawtucket-FallRiver, RI-MA CMSA Madison. WI San Francisco-OaLland-San Jose. CA CMSA (San Jose) San Francisco-Oakland-SanJose, CA CMSA (Oakland) W e s t o n . WV Sacramento. CA Sarasota. FL Fresno. CA Spokane. WA Santa BarbamSanta Maria-Lompoc, CA Source: Authors' calculations. Hub Predicted Value 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.148 0.145 0.139 0.137 0.135 0.127 0.127 0.124 0.120 0.1 18 0.115 0.114 0.112 0.112 0.090 0.090 0.087 0.083 0.082 0.076 0.073 0.07 1 0.067 0.067 0.064 0.060 0.056 0.052 0.052 0.047 0.045 0.044 0.044 0.043 0.041 0.036 0.036 0.036 0.032 0.03 1 0.027 0. m 0.023 0.021 0.021 0.014 0.014 0.007 0.005 0.005 0.002 0.002 0.001 ' ,
@FedFRASER